![\""\""](\""https://owid.cloud/app/uploads/2023/09/animal-welfare-thumbnail-768x403.png\"")
This page is maintained in Google Docs: https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview
","{""id"": ""wp-58486"", ""slug"": ""state-capacity"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""This page is maintained in Google Docs: "", ""spanType"": ""span-simple-text""}, {""url"": ""https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview"", ""children"": [{""text"": ""https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""State Capacity"", ""authors"": [""Bastian Herre"", ""Pablo Arriagada"", ""Max Roser""], ""dateline"": ""November 30, 2023"", ""sidebar-toc"": false, ""featured-image"": ""state-capacity-featured-image.png""}, ""createdAt"": ""2023-11-28T09:19:26.000Z"", ""published"": false, ""updatedAt"": ""2023-11-30T13:01:41.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-11-30T13:01:40.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 1, ""numErrors"": 0, ""wpTagCounts"": {""paragraph"": 1}, ""htmlTagCounts"": {""p"": 1}}",2023-11-30 13:01:40,2024-02-16 14:22:42,,"[""Bastian Herre"", ""Pablo Arriagada""]",,2023-11-28 09:19:26,2023-11-30 13:01:41,https://ourworldindata.org/wp-content/uploads/2023/11/state-capacity-featured-image.png,{},This page is maintained in Google Docs: [https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview](https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview),"{""id"": 58486, ""date"": ""2023-11-30T13:01:40"", ""guid"": {""rendered"": ""https://owid.cloud/?page_id=58486""}, ""link"": ""https://owid.cloud/state-capacity"", ""meta"": {""owid_publication_context_meta_field"": [], ""owid_key_performance_indicators_meta_field"": []}, ""slug"": ""state-capacity"", ""tags"": [], ""type"": ""page"", ""title"": {""rendered"": ""State Capacity""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/58486""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/page""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=58486"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=58486"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=58486"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=58486""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/58486/revisions"", ""count"": 1}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/58487"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 58488, ""href"": ""https://owid.cloud/wp-json/wp/v2/pages/58486/revisions/58488""}]}, ""author"": 49, ""parent"": 0, ""status"": ""publish"", ""content"": {""rendered"": ""\nThis page is maintained in Google Docs: https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview
\n"", ""protected"": false}, ""excerpt"": {""rendered"": """", ""protected"": false}, ""date_gmt"": ""2023-11-30T13:01:40"", ""modified"": ""2023-11-30T13:01:41"", ""template"": """", ""categories"": [44, 56, 239, 57, 193], ""menu_order"": 6, ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre"", ""Pablo Arriagada""], ""modified_gmt"": ""2023-11-30T13:01:41"", ""comment_status"": ""closed"", ""featured_media"": 58487, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/11/state-capacity-featured-image-150x79.png"", ""medium_large"": ""/app/uploads/2023/11/state-capacity-featured-image-768x403.png""}}" 58184,Animal Welfare Explorer Block,animal-welfare-explorer-block-2,wp_block,publish,"THIS PAGE ONLY EXISTS TO GET ECONOMIC INEQUALITY LISTED IN THE BY-TOPIC MENU.
The inequality page is now in Google Docs: https://docs.google.com/document/d/1yzOrFd6uWvrAl2oFB3S67oOSbhgL1ffPHxjAqS7i-4w/edit
","{""id"": ""wp-57770"", ""slug"": ""economic-inequality"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""THIS PAGE ONLY EXISTS TO GET ECONOMIC INEQUALITY LISTED IN THE BY-TOPIC MENU."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The inequality page is now in Google Docs: https://docs.google.com/document/d/1yzOrFd6uWvrAl2oFB3S67oOSbhgL1ffPHxjAqS7i-4w/edit"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Economic Inequality"", ""authors"": [""Joe Hasell"", ""Max Roser""], ""dateline"": ""July 7, 2023"", ""sidebar-toc"": false, ""featured-image"": ""economic-inequality-featured-image-1.png""}, ""createdAt"": ""2023-07-07T08:48:36.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T12:32:22.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-07T08:00:13.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 2, ""numErrors"": 0, ""wpTagCounts"": {""paragraph"": 2}, ""htmlTagCounts"": {""p"": 2}}",2023-07-07 08:00:13,2024-02-16 14:22:42,,"[""Joe Hasell""]",,2023-07-07 08:48:36,2023-07-10 12:32:22,https://ourworldindata.org/wp-content/uploads/2023/07/economic-inequality-featured-image-1.png,{},"THIS PAGE ONLY EXISTS TO GET ECONOMIC INEQUALITY LISTED IN THE BY-TOPIC MENU. The inequality page is now in Google Docs: https://docs.google.com/document/d/1yzOrFd6uWvrAl2oFB3S67oOSbhgL1ffPHxjAqS7i-4w/edit","{""id"": 57770, ""date"": ""2023-07-07T09:00:13"", ""guid"": {""rendered"": ""https://owid.cloud/?page_id=57770""}, ""link"": ""https://owid.cloud/economic-inequality"", ""meta"": {""owid_publication_context_meta_field"": [], ""owid_key_performance_indicators_meta_field"": {""raw"": ""Many countries have high levels of economic inequality. Inequality across the world as a whole has fallen but remains vast."", ""rendered"": ""Many countries have high levels of economic inequality. Inequality across the world as a whole has fallen but remains vast.
\n""}}, ""slug"": ""economic-inequality"", ""tags"": [], ""type"": ""page"", ""title"": {""rendered"": ""Economic Inequality""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/57770""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/page""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/14"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57770"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57770"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57770"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57770""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/57770/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57778"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57792, ""href"": ""https://owid.cloud/wp-json/wp/v2/pages/57770/revisions/57792""}]}, ""author"": 14, ""parent"": 0, ""status"": ""publish"", ""content"": {""rendered"": ""\nTHIS PAGE ONLY EXISTS TO GET ECONOMIC INEQUALITY LISTED IN THE BY-TOPIC MENU.
\n\n\n\nThe inequality page is now in Google Docs: https://docs.google.com/document/d/1yzOrFd6uWvrAl2oFB3S67oOSbhgL1ffPHxjAqS7i-4w/edit
\n"", ""protected"": false}, ""excerpt"": {""rendered"": """", ""protected"": false}, ""date_gmt"": ""2023-07-07T08:00:13"", ""modified"": ""2023-07-10T13:32:22"", ""template"": """", ""categories"": [44, 51, 189], ""menu_order"": 9, ""ping_status"": ""closed"", ""authors_name"": [""Joe Hasell""], ""modified_gmt"": ""2023-07-10T12:32:22"", ""comment_status"": ""closed"", ""featured_media"": 57778, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/07/economic-inequality-featured-image-1-150x79.png"", ""medium_large"": ""/app/uploads/2023/07/economic-inequality-featured-image-1-768x402.png""}}" 57760,About this data (main inequality explorer),about-this-data-main-inequality-explorer,wp_block,publish,"This data explorer provides a range of inequality indicators measured according to two different definitions of income obtained from different sources.
Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators from different sources:
OWID Data Collection: Inequality and Poverty
","{""id"": ""wp-57760"", ""slug"": ""about-this-data-main-inequality-explorer"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""About this data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data explorer provides a range of inequality indicators measured according to two different definitions of income obtained from different sources."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""text"": ""Data from the World Inequality Database relates to inequality before taxes and benefits."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Data from the World Bank relates to either income after taxes and benefits or consumption, depending on the country and year."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators from different sources:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/owid-data-collection-inequality-and-poverty"", ""children"": [{""text"": ""OWID Data Collection: Inequality and Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""About this data (main inequality explorer)"", ""authors"": [null], ""dateline"": ""July 6, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-07-06T18:18:16.000Z"", ""published"": false, ""updatedAt"": ""2023-07-06T17:30:15.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-06T17:18:16.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 5, ""numErrors"": 1, ""wpTagCounts"": {""list"": 1, ""heading"": 1, ""paragraph"": 3}, ""htmlTagCounts"": {""p"": 3, ""h3"": 1, ""ul"": 1}}",2023-07-06 17:18:16,2024-02-16 14:23:03,,[null],,2023-07-06 18:18:16,2023-07-06 17:30:15,,{},"## About this data This data explorer provides a range of inequality indicators measured according to two different definitions of income obtained from different sources. * Data from the World Inequality Database relates to inequality before taxes and benefits. * Data from the World Bank relates to either income after taxes and benefits or consumption, depending on the country and year. Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators from different sources: [OWID Data Collection: Inequality and Poverty](https://ourworldindata.org/owid-data-collection-inequality-and-poverty)","{""data"": {""wpBlock"": {""content"": ""\nThis data explorer provides a range of inequality indicators measured according to two different definitions of income obtained from different sources.
\n\n\n\nFurther information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators from different sources:
\n\n\n\nOWID Data Collection: Inequality and Poverty
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 57756,About this data (PIP explorers),about-this-data-pip-explorers,wp_block,publish,"This data explorer provides a range of indicators obtained from the World Bank Poverty and Inequality Platform. Depending on the country and year, the data relates to income measured after taxes and benefits, or consumption, per capita.
Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
OWID Data Collection: Inequality and Poverty
","{""id"": ""wp-57756"", ""slug"": ""about-this-data-pip-explorers"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""About this data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data explorer provides a range of indicators obtained from the World Bank Poverty and Inequality Platform. Depending on the country and year, the data relates to income measured after taxes and benefits, or consumption, per capita."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/owid-data-collection-inequality-and-poverty"", ""children"": [{""text"": ""OWID Data Collection: Inequality and Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""About this data (PIP explorers)"", ""authors"": [null], ""dateline"": ""July 6, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-07-06T18:14:45.000Z"", ""published"": false, ""updatedAt"": ""2024-02-02T17:32:21.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-06T17:14:45.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 4, ""numErrors"": 0, ""wpTagCounts"": {""heading"": 1, ""paragraph"": 3}, ""htmlTagCounts"": {""p"": 3, ""h3"": 1}}",2023-07-06 17:14:45,2024-02-16 14:23:03,,[null],,2023-07-06 18:14:45,2024-02-02 17:32:21,,{},"## About this data This data explorer provides a range of indicators obtained from the World Bank Poverty and Inequality Platform. Depending on the country and year, the data relates to income measured after taxes and benefits, or consumption, per capita. Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators: [OWID Data Collection: Inequality and Poverty](https://ourworldindata.org/owid-data-collection-inequality-and-poverty)","{""data"": {""wpBlock"": {""content"": ""\nThis data explorer provides a range of indicators obtained from the World Bank Poverty and Inequality Platform. Depending on the country and year, the data relates to income measured after taxes and benefits, or consumption, per capita.
\n\n\n\nFurther information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
\n\n\n\nOWID Data Collection: Inequality and Poverty
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 57755,About this data (LIS explorers),about-this-data-lis-explorers,wp_block,publish,"This data explorer provides a range of indicators obtained from the Luxemburg Income Study dataset, related to income before and after taxes and benefits.
Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
OWID Data Collection: Inequality and Poverty
","{""id"": ""wp-57755"", ""slug"": ""about-this-data-lis-explorers"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""About this data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data explorer provides a range of indicators obtained from the Luxemburg Income Study dataset, related to income before and after taxes and benefits."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/owid-data-collection-inequality-and-poverty"", ""children"": [{""text"": ""OWID Data Collection: Inequality and Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""About this data (LIS explorers)"", ""authors"": [null], ""dateline"": ""July 6, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-07-06T18:13:07.000Z"", ""published"": false, ""updatedAt"": ""2024-02-02T17:32:28.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-06T17:13:07.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 4, ""numErrors"": 0, ""wpTagCounts"": {""heading"": 1, ""paragraph"": 3}, ""htmlTagCounts"": {""p"": 3, ""h3"": 1}}",2023-07-06 17:13:07,2024-02-16 14:23:03,,[null],,2023-07-06 18:13:07,2024-02-02 17:32:28,,{},"## About this data This data explorer provides a range of indicators obtained from the Luxemburg Income Study dataset, related to income before and after taxes and benefits. Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators: [OWID Data Collection: Inequality and Poverty](https://ourworldindata.org/owid-data-collection-inequality-and-poverty)","{""data"": {""wpBlock"": {""content"": ""\nThis data explorer provides a range of indicators obtained from the Luxemburg Income Study dataset, related to income before and after taxes and benefits.
\n\n\n\nFurther information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
\n\n\n\nOWID Data Collection: Inequality and Poverty
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 57750,About this data (WID explorers),about-this-data-wid-explorers,wp_block,publish,"This data explorer provides a range of indicators obtained from the World Inequality Database, related to income before and after taxes and benefits.
Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
OWID Data Collection: Inequality and Poverty
","{""id"": ""wp-57750"", ""slug"": ""about-this-data-wid-explorers"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""About this data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data explorer provides a range of indicators obtained from the World Inequality Database, related to income before and after taxes and benefits."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/owid-data-collection-inequality-and-poverty"", ""children"": [{""text"": ""OWID Data Collection: Inequality and Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""About this data (WID explorers)"", ""authors"": [null], ""dateline"": ""July 6, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-07-06T18:09:52.000Z"", ""published"": false, ""updatedAt"": ""2023-07-06T17:29:14.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-06T17:09:52.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 4, ""numErrors"": 0, ""wpTagCounts"": {""heading"": 1, ""paragraph"": 3}, ""htmlTagCounts"": {""p"": 3, ""h3"": 1}}",2023-07-06 17:09:52,2024-02-16 14:23:03,,[null],,2023-07-06 18:09:52,2023-07-06 17:29:14,,{},"## About this data This data explorer provides a range of indicators obtained from the World Inequality Database, related to income before and after taxes and benefits. Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators: [OWID Data Collection: Inequality and Poverty](https://ourworldindata.org/owid-data-collection-inequality-and-poverty)","{""data"": {""wpBlock"": {""content"": ""\nThis data explorer provides a range of indicators obtained from the World Inequality Database, related to income before and after taxes and benefits.
\n\n\n\nFurther information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
\n\n\n\nOWID Data Collection: Inequality and Poverty
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 57742,About this data (Inequality and Incomes Across the Distribution comparison explorers),57742,wp_block,publish,"This data explorer provides a range of indicators obtained from different sources.
Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
OWID Data Collection: Inequality and Poverty
","{""id"": ""wp-57742"", ""slug"": ""57742"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""About this data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data explorer provides a range of indicators obtained from different sources."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""text"": ""Data from the World Bank relates to either income per capita after taxes and benefits or consumption, depending on the country and year."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Data from the World Inequality Database and Luxembourg Income Study relate to income before and after taxes and benefits."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/owid-data-collection-inequality-and-poverty"", ""children"": [{""text"": ""OWID Data Collection: Inequality and Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""About this data (Inequality and Incomes Across the Distribution comparison explorers)"", ""authors"": [null], ""dateline"": ""July 6, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-07-06T17:59:59.000Z"", ""published"": false, ""updatedAt"": ""2023-07-06T17:29:02.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-06T17:00:46.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 5, ""numErrors"": 1, ""wpTagCounts"": {""list"": 1, ""heading"": 1, ""paragraph"": 3}, ""htmlTagCounts"": {""p"": 3, ""h3"": 1, ""ul"": 1}}",2023-07-06 17:00:46,2024-02-16 14:23:03,,[null],,2023-07-06 17:59:59,2023-07-06 17:29:02,,{},"## About this data This data explorer provides a range of indicators obtained from different sources. * Data from the World Bank relates to either income per capita after taxes and benefits or consumption, depending on the country and year. * Data from the World Inequality Database and Luxembourg Income Study relate to income before and after taxes and benefits. Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators: [OWID Data Collection: Inequality and Poverty](https://ourworldindata.org/owid-data-collection-inequality-and-poverty)","{""data"": {""wpBlock"": {""content"": ""\nThis data explorer provides a range of indicators obtained from different sources.
\n\n\n\nFurther information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
\n\n\n\nOWID Data Collection: Inequality and Poverty
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 57728,About this data (Poverty comparison explorer),about-this-data,wp_block,publish,"This data explorer provides a range of poverty indicators obtained from different sources.
Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
OWID Data Collection: Inequality and Poverty
","{""id"": ""wp-57728"", ""slug"": ""about-this-data"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""About this data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data explorer provides a range of poverty indicators obtained from different sources."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""text"": ""Data from the World Bank relates to either income per capita after taxes and benefits or consumption per capita, depending on the country and year."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Data from the Luxembourg Income Study relates to income after taxes and benefits per capita."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/owid-data-collection-inequality-and-poverty"", ""children"": [{""text"": ""OWID Data Collection: Inequality and Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""About this data (Poverty comparison explorer)"", ""authors"": [null], ""dateline"": ""July 6, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-07-06T17:36:01.000Z"", ""published"": false, ""updatedAt"": ""2023-07-06T17:29:07.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-06T16:36:08.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 5, ""numErrors"": 1, ""wpTagCounts"": {""list"": 1, ""heading"": 1, ""paragraph"": 5}, ""htmlTagCounts"": {""p"": 5, ""h3"": 1, ""ul"": 1}}",2023-07-06 16:36:08,2024-02-16 14:23:03,,[null],,2023-07-06 17:36:01,2023-07-06 17:29:07,,{},"## About this data This data explorer provides a range of poverty indicators obtained from different sources. * Data from the World Bank relates to either income per capita after taxes and benefits or consumption per capita, depending on the country and year. * Data from the Luxembourg Income Study relates to income after taxes and benefits per capita. Further information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators: [OWID Data Collection: Inequality and Poverty](https://ourworldindata.org/owid-data-collection-inequality-and-poverty)","{""data"": {""wpBlock"": {""content"": ""\nThis data explorer provides a range of poverty indicators obtained from different sources.
\n\n\n\nFurther information about the definitions and methods behind this data can be found in the article below, where you can also explore and compare a much broader range of indicators:
\n\n\n\nOWID Data Collection: Inequality and Poverty
\n\n\n\n\n\n\n\n\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 57720,We published a new topic page on Human Rights,human-rights-redesign,post,publish,,"{""id"": ""wp-57720"", ""slug"": ""human-rights-redesign"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Human Rights"", ""authors"": [""Bastian Herre"", ""Pablo Arriagada""], ""excerpt"": ""Explore data on human rights across the world, and see how these rights have changed over time."", ""dateline"": ""July 5, 2023"", ""subtitle"": ""Explore data on human rights across the world, and see how these rights have changed over time."", ""sidebar-toc"": false, ""featured-image"": ""Human-Rights.png""}, ""createdAt"": ""2023-07-05T12:15:21.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:14:36.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-05T11:15:21.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-07-05 11:15:21,2024-02-16 14:22:55,,"[""Bastian Herre"", ""Pablo Arriagada""]","Explore data on human rights across the world, and see how these rights have changed over time.",2023-07-05 12:15:21,2023-07-27 12:14:36,https://ourworldindata.org/wp-content/uploads/2023/05/Human-Rights.png,{},,"{""id"": 57720, ""date"": ""2023-07-05T12:15:21"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57720""}, ""link"": ""https://owid.cloud/human-rights-redesign"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""human-rights-redesign"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Human Rights""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57720""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57720"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57720"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57720"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57720""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57720/revisions"", ""count"": 3}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57085"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57930, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57720/revisions/57930""}]}, ""author"": 49, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore data on human rights across the world, and see how these rights have changed over time."", ""protected"": false}, ""date_gmt"": ""2023-07-05T11:15:21"", ""modified"": ""2023-07-27T13:14:36"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre"", ""Pablo Arriagada""], ""modified_gmt"": ""2023-07-27T12:14:36"", ""comment_status"": ""closed"", ""featured_media"": 57085, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/05/Human-Rights-150x79.png"", ""medium_large"": ""/app/uploads/2023/05/Human-Rights-768x403.png""}}" 57695,We published a new topic page on Diarrheal Diseases,diarrheal-diseases-redesign,post,publish,,"{""id"": ""wp-57695"", ""slug"": ""diarrheal-diseases-redesign"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Diarrheal Diseases"", ""authors"": [""Saloni Dattani"", ""Fiona Spooner""], ""excerpt"": ""Diarrheal diseases are a leading cause of death, especially among children. Explore global data on deaths and treatment for this killer."", ""dateline"": ""July 3, 2023"", ""subtitle"": ""Diarrheal diseases are a leading cause of death, especially among children. Explore global data on deaths and treatment for this killer."", ""sidebar-toc"": false, ""featured-image"": ""diarrheal-diseases-thumbnail.png""}, ""createdAt"": ""2023-07-03T11:35:18.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:14:56.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-07-03T10:35:18.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-07-03 10:35:18,2024-02-16 14:22:55,,"[""Saloni Dattani"", ""Fiona Spooner""]","Diarrheal diseases are a leading cause of death, especially among children. Explore global data on deaths and treatment for this killer.",2023-07-03 11:35:18,2023-07-27 12:14:56,https://ourworldindata.org/wp-content/uploads/2023/06/diarrheal-diseases-thumbnail.png,{},,"{""id"": 57695, ""date"": ""2023-07-03T11:35:18"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57695""}, ""link"": ""https://owid.cloud/diarrheal-diseases-redesign"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""diarrheal-diseases-redesign"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Diarrheal Diseases""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57695""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/47"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57695"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57695"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57695"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57695""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57695/revisions"", ""count"": 3}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57378"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57931, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57695/revisions/57931""}]}, ""author"": 47, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Diarrheal diseases are a leading cause of death, especially among children. Explore global data on deaths and treatment for this killer."", ""protected"": false}, ""date_gmt"": ""2023-07-03T10:35:18"", ""modified"": ""2023-07-27T13:14:56"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Saloni Dattani"", ""Fiona Spooner""], ""modified_gmt"": ""2023-07-27T12:14:56"", ""comment_status"": ""closed"", ""featured_media"": 57378, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/06/diarrheal-diseases-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2023/06/diarrheal-diseases-thumbnail-768x402.png""}}" 57643,We published a new topic page on Women's Rights,womens-rights-launch,post,publish,,"{""id"": ""wp-57643"", ""slug"": ""womens-rights-launch"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Women's Rights"", ""authors"": [""Bastian Herre"", ""Pablo Arriagada""], ""excerpt"": ""How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data on women’s rights."", ""dateline"": ""June 29, 2023"", ""subtitle"": ""How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data on women’s rights."", ""sidebar-toc"": false, ""featured-image"": ""Womens-Rights.png""}, ""createdAt"": ""2023-06-29T10:57:34.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:15:09.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-06-29T10:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-06-29 10:00:00,2024-02-16 14:22:55,,"[""Bastian Herre"", ""Pablo Arriagada""]",How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data on women’s rights.,2023-06-29 10:57:34,2023-07-27 12:15:09,https://ourworldindata.org/wp-content/uploads/2023/05/Womens-Rights.png,{},,"{""id"": 57643, ""date"": ""2023-06-29T11:00:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57643""}, ""link"": ""https://owid.cloud/womens-rights-launch"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""womens-rights-launch"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Women’s Rights""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57643""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57643"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57643"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57643"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57643""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57643/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57051"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57932, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57643/revisions/57932""}]}, ""author"": 49, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data on women’s rights."", ""protected"": false}, ""date_gmt"": ""2023-06-29T10:00:00"", ""modified"": ""2023-07-27T13:15:09"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre"", ""Pablo Arriagada""], ""modified_gmt"": ""2023-07-27T12:15:09"", ""comment_status"": ""closed"", ""featured_media"": 57051, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/05/Womens-Rights-150x79.png"", ""medium_large"": ""/app/uploads/2023/05/Womens-Rights-768x403.png""}}" 57580,We published a new topic page on Illicit Drug Use,drug-use-redesign,post,publish,,"{""id"": ""wp-57580"", ""slug"": ""drug-use-redesign"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Illicit Drug Use"", ""authors"": [""Hannah Ritchie"", ""Pablo Arriagada""], ""excerpt"": ""Explore global data on addiction and deaths from opioids, cocaine, and other illicit drugs."", ""dateline"": ""June 26, 2023"", ""subtitle"": ""Explore global data on addiction and deaths from opioids, cocaine, and other illicit drugs."", ""sidebar-toc"": false, ""featured-image"": ""Illicit-Drug-Use.png""}, ""createdAt"": ""2023-06-26T10:32:16.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:17:16.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-06-26T09:32:16.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-06-26 09:32:16,2024-02-16 14:22:55,,"[""Hannah Ritchie"", ""Pablo Arriagada""]","Explore global data on addiction and deaths from opioids, cocaine, and other illicit drugs.",2023-06-26 10:32:16,2023-07-27 12:17:16,https://ourworldindata.org/wp-content/uploads/2023/06/Illicit-Drug-Use.png,{},,"{""id"": 57580, ""date"": ""2023-06-26T10:32:16"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57580""}, ""link"": ""https://owid.cloud/drug-use-redesign"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""drug-use-redesign"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Illicit Drug Use""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57580""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57580"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57580"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57580"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57580""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57580/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57412"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57937, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57580/revisions/57937""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore global data on addiction and deaths from opioids, cocaine, and other illicit drugs."", ""protected"": false}, ""date_gmt"": ""2023-06-26T09:32:16"", ""modified"": ""2023-07-27T13:17:16"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Pablo Arriagada""], ""modified_gmt"": ""2023-07-27T12:17:16"", ""comment_status"": ""closed"", ""featured_media"": 57412, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/06/Illicit-Drug-Use-150x79.png"", ""medium_large"": ""/app/uploads/2023/06/Illicit-Drug-Use-768x403.png""}}" 57557,We published a new topic page on LGBT+ Rights,lgbt-rights-launch,post,publish,,"{""id"": ""wp-57557"", ""slug"": ""lgbt-rights-launch"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on LGBT+ Rights"", ""authors"": [""Bastian Herre"", ""Pablo Arriagada""], ""excerpt"": ""Explore global data on the protection of LGBT+ rights across the world and over time."", ""dateline"": ""June 22, 2023"", ""subtitle"": ""Explore global data on the protection of LGBT+ rights across the world and over time."", ""sidebar-toc"": false, ""featured-image"": ""LGBT-Rights.png""}, ""createdAt"": ""2023-06-22T10:57:56.000Z"", ""published"": false, ""updatedAt"": ""2023-06-22T10:03:27.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-06-22T09:57:56.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-06-22 09:57:56,2024-02-16 14:22:55,,"[""Bastian Herre"", ""Pablo Arriagada""]",Explore global data on the protection of LGBT+ rights across the world and over time.,2023-06-22 10:57:56,2023-06-22 10:03:27,https://ourworldindata.org/wp-content/uploads/2023/05/LGBT-Rights.png,{},,"{""id"": 57557, ""date"": ""2023-06-22T10:57:56"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57557""}, ""link"": ""https://owid.cloud/lgbt-rights-launch"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""lgbt-rights-launch"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on LGBT+ Rights""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57557""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57557"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57557"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57557"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57557""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57557/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57050"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57560, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57557/revisions/57560""}]}, ""author"": 49, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore global data on the protection of LGBT+ rights across the world and over time."", ""protected"": false}, ""date_gmt"": ""2023-06-22T09:57:56"", ""modified"": ""2023-06-22T11:03:27"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre"", ""Pablo Arriagada""], ""modified_gmt"": ""2023-06-22T10:03:27"", ""comment_status"": ""closed"", ""featured_media"": 57050, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/05/LGBT-Rights-150x79.png"", ""medium_large"": ""/app/uploads/2023/05/LGBT-Rights-768x403.png""}}" 57534,We published a new topic page on Hunger and Undernourishment,hunger-redesign,post,publish,,"{""id"": ""wp-57534"", ""slug"": ""hunger-redesign"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Hunger and Undernourishment"", ""authors"": [""Hannah Ritchie"", ""Pablo Rosado""], ""excerpt"": ""Explore the global data, visualizations, and writing on hunger and undernourishment. It looks at how many people are undernourished, where they are, and other metrics on food security."", ""dateline"": ""June 19, 2023"", ""subtitle"": ""Explore the global data, visualizations, and writing on hunger and undernourishment. It looks at how many people are undernourished, where they are, and other metrics on food security."", ""sidebar-toc"": false, ""featured-image"": ""Hunger-and-Undernourishment.png""}, ""createdAt"": ""2023-06-20T15:46:34.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:16:29.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-06-19T14:56:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-06-19 14:56:00,2024-02-16 14:22:55,,"[""Hannah Ritchie"", ""Pablo Rosado""]","Explore the global data, visualizations, and writing on hunger and undernourishment. It looks at how many people are undernourished, where they are, and other metrics on food security.",2023-06-20 15:46:34,2023-07-27 12:16:29,https://ourworldindata.org/wp-content/uploads/2023/06/Hunger-and-Undernourishment.png,{},,"{""id"": 57534, ""date"": ""2023-06-19T15:56:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57534""}, ""link"": ""https://owid.cloud/hunger-redesign"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""hunger-redesign"", ""tags"": [127, 128], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Hunger and Undernourishment""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57534""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57534"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57534"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57534"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57534""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57534/revisions"", ""count"": 5}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57516"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57935, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57534/revisions/57935""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore the global data, visualizations, and writing on hunger and undernourishment. It looks at how many people are undernourished, where they are, and other metrics on food security."", ""protected"": false}, ""date_gmt"": ""2023-06-19T14:56:00"", ""modified"": ""2023-07-27T13:16:29"", ""template"": """", ""categories"": [47], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Pablo Rosado""], ""modified_gmt"": ""2023-07-27T12:16:29"", ""comment_status"": ""closed"", ""featured_media"": 57516, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/06/Hunger-and-Undernourishment-150x79.png"", ""medium_large"": ""/app/uploads/2023/06/Hunger-and-Undernourishment-768x403.png""}}" 57528,We published a new topic page on Mental Health,mental-health-redesign,post,publish,,"{""id"": ""wp-57528"", ""slug"": ""mental-health-redesign"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Mental Health"", ""authors"": [""Saloni Dattani"", ""Lucas Rodés-Guirao""], ""excerpt"": ""Explore global data on the prevalence of mental health illnesses, treatment, and public openness to talking about mental health. "", ""dateline"": ""June 20, 2023"", ""subtitle"": ""Explore global data on the prevalence of mental health illnesses, treatment, and public openness to talking about mental health. "", ""sidebar-toc"": false, ""featured-image"": ""mental-health-thumbnail.png""}, ""createdAt"": ""2023-06-20T14:43:22.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:16:48.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-06-20T16:30:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-06-20 16:30:00,2024-02-16 14:22:55,,"[""Saloni Dattani"", ""Lucas Rodés-Guirao""]","Explore global data on the prevalence of mental health illnesses, treatment, and public openness to talking about mental health. ",2023-06-20 14:43:22,2023-07-27 12:16:48,https://ourworldindata.org/wp-content/uploads/2023/06/mental-health-thumbnail.png,{},,"{""id"": 57528, ""date"": ""2023-06-20T17:30:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57528""}, ""link"": ""https://owid.cloud/mental-health-redesign"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""mental-health-redesign"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Mental Health""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57528""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/47"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57528"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57528"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57528"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57528""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57528/revisions"", ""count"": 3}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57321"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57936, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57528/revisions/57936""}]}, ""author"": 47, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore global data on the prevalence of mental health illnesses, treatment, and public openness to talking about mental health. "", ""protected"": false}, ""date_gmt"": ""2023-06-20T16:30:00"", ""modified"": ""2023-07-27T13:16:48"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Saloni Dattani"", ""Lucas Rodés-Guirao""], ""modified_gmt"": ""2023-07-27T12:16:48"", ""comment_status"": ""closed"", ""featured_media"": 57321, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/06/mental-health-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2023/06/mental-health-thumbnail-768x403.png""}}" 57457,We published a new topic page on Oil Spills,oil-spills-redesign,post,publish,,"{""id"": ""wp-57457"", ""slug"": ""oil-spills-redesign"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Oil Spills"", ""authors"": [""Hannah Ritchie"", ""Veronika Samborska""], ""excerpt"": ""Explore global data on oil spills, looking at their frequency, size, and how this has changed over time."", ""dateline"": ""June 15, 2023"", ""subtitle"": ""Explore global data on oil spills, looking at their frequency, size, and how this has changed over time."", ""sidebar-toc"": false, ""featured-image"": ""Oil-Spills.png""}, ""createdAt"": ""2023-06-15T11:26:21.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:16:14.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-06-15T10:26:21.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-06-15 10:26:21,2024-02-16 14:22:55,,"[""Hannah Ritchie"", ""Veronika Samborska""]","Explore global data on oil spills, looking at their frequency, size, and how this has changed over time.",2023-06-15 11:26:21,2023-07-27 12:16:14,https://ourworldindata.org/wp-content/uploads/2022/12/Oil-Spills.png,{},,"{""id"": 57457, ""date"": ""2023-06-15T11:26:21"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57457""}, ""link"": ""https://owid.cloud/oil-spills-redesign"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""oil-spills-redesign"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Oil Spills""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57457""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57457"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57457"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57457"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57457""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57457/revisions"", ""count"": 3}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/54742"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57934, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57457/revisions/57934""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore global data on oil spills, looking at their frequency, size, and how this has changed over time."", ""protected"": false}, ""date_gmt"": ""2023-06-15T10:26:21"", ""modified"": ""2023-07-27T13:16:14"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Veronika Samborska""], ""modified_gmt"": ""2023-07-27T12:16:14"", ""comment_status"": ""closed"", ""featured_media"": 54742, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/12/Oil-Spills-150x79.png"", ""medium_large"": ""/app/uploads/2022/12/Oil-Spills-768x403.png""}}" 57446,We published a new topic page on Crop Yields,crop-yield-redesign,post,publish,,"{""id"": ""wp-57446"", ""slug"": ""crop-yield-redesign"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Crop Yields"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""Explore global data on crop yields, their changes over time, and remaining yield gaps."", ""dateline"": ""June 14, 2023"", ""subtitle"": ""Explore global data on crop yields, their changes over time, and remaining yield gaps."", ""sidebar-toc"": false, ""featured-image"": ""Crop-Yields.png""}, ""createdAt"": ""2023-06-14T16:29:20.000Z"", ""published"": false, ""updatedAt"": ""2023-07-27T12:15:51.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-06-14T15:29:20.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-06-14 15:29:20,2024-02-16 14:22:55,,"[""Hannah Ritchie""]","Explore global data on crop yields, their changes over time, and remaining yield gaps.",2023-06-14 16:29:20,2023-07-27 12:15:51,https://ourworldindata.org/wp-content/uploads/2023/05/Crop-Yields.png,{},,"{""id"": 57446, ""date"": ""2023-06-14T16:29:20"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57446""}, ""link"": ""https://owid.cloud/crop-yield-redesign"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""crop-yield-redesign"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Crop Yields""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57446""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57446"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57446"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57446"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57446""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57446/revisions"", ""count"": 3}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57270"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57933, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57446/revisions/57933""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore global data on crop yields, their changes over time, and remaining yield gaps."", ""protected"": false}, ""date_gmt"": ""2023-06-14T15:29:20"", ""modified"": ""2023-07-27T13:15:51"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2023-07-27T12:15:51"", ""comment_status"": ""closed"", ""featured_media"": 57270, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/05/Crop-Yields-150x79.png"", ""medium_large"": ""/app/uploads/2023/05/Crop-Yields-768x403.png""}}" 57249,We published a new topic page on Research & Development,rd-launch,post,publish,,"{""id"": ""wp-57249"", ""slug"": ""rd-launch"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Research & Development"", ""authors"": [""Hannah Ritchie"", ""Edouard Mathieu""], ""excerpt"": ""Explore global data on patents, researchers, R&D spending and innovation across the world."", ""dateline"": ""May 30, 2023"", ""subtitle"": ""Explore global data on patents, researchers, R&D spending and innovation across the world."", ""sidebar-toc"": false, ""featured-image"": ""Research-and-Development.png""}, ""createdAt"": ""2023-05-30T11:31:45.000Z"", ""published"": false, ""updatedAt"": ""2023-05-30T10:35:54.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-05-30T10:31:45.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-05-30 10:31:45,2024-02-16 14:22:55,,"[""Hannah Ritchie"", ""Edouard Mathieu""]","Explore global data on patents, researchers, R&D spending and innovation across the world.",2023-05-30 11:31:45,2023-05-30 10:35:54,https://ourworldindata.org/wp-content/uploads/2023/01/Research-and-Development.png,{},,"{""id"": 57249, ""date"": ""2023-05-30T11:31:45"", ""guid"": {""rendered"": ""https://owid.cloud/?p=57249""}, ""link"": ""https://owid.cloud/rd-launch"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""rd-launch"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Research & Development""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57249""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=57249"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=57249"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=57249"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=57249""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57249/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/55545"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57253, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/57249/revisions/57253""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore global data on patents, researchers, R&D spending and innovation across the world."", ""protected"": false}, ""date_gmt"": ""2023-05-30T10:31:45"", ""modified"": ""2023-05-30T11:35:54"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Edouard Mathieu""], ""modified_gmt"": ""2023-05-30T10:35:54"", ""comment_status"": ""closed"", ""featured_media"": 55545, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/01/Research-and-Development-150x79.png"", ""medium_large"": ""/app/uploads/2023/01/Research-and-Development-768x403.png""}}" 57186,How are mental illnesses defined?,how-are-mental-illnesses-defined,post,publish,"Mental illnesses are diagnosed by health professionals using manuals, which describe their symptoms, their context, and how they differ from similar conditions.
How are these manuals used? How are major mental illnesses described and diagnosed?
This article summarizes the major mental illnesses in the ICD-11 manual, the official worldwide classification.
Each section includes a summary of how the condition is medically diagnosed, a description of the symptoms, and other conditions that must be ruled out before making a diagnosis.
If you want to read the full criteria for diagnosing mental illnesses according to the ICD-11, you can find them here.
Diagnosing a mental illness begins when people see a doctor or psychiatrist.
It can be difficult to define mental illnesses because they are conditions of the mind: they are diagnosed according to people's symptoms and behavior. This means they can be subjective – they are dependent partly on people’s comfort seeing a doctor or psychiatrist, the training and experiences of healthcare professionals, and other cultural factors.
Two manuals are commonly used by doctors and psychiatrists to diagnose mental illnesses: the International Classification of Diseases (ICD) and the Diagnostic and Statistical Manual of Mental Disorders (DSM).
There are differences in how they are used.
The ICD is the official worldwide classification for healthcare and data collection. It is used by a wide range of healthcare professionals.
In contrast, the DSM was developed for psychiatrists in the United States, although its usage has expanded worldwide. It is intended to provide standardized guidelines for different psychiatrists who may have different judgments otherwise.{ref}Harrison, J. E., Weber, S., Jakob, R., & Chute, C. G. (2021). ICD-11: An international classification of diseases for the twenty-first century. BMC Medical Informatics and Decision Making, 21(S6), 206. https://doi.org/10.1186/s12911-021-01534-6
Kupfer, D. J., Regier, D. A., & Kuhl, E. A. (2008). On the road to DSM-V and ICD-11. European Archives of Psychiatry and Clinical Neuroscience, 258(S5), 2–6. https://doi.org/10.1007/s00406-008-5002-6
Tyrer, P. (2014). A comparison of DSM and ICD classifications of mental disorder. Advances in Psychiatric Treatment, 20(4), 280–285. https://doi.org/10.1192/apt.bp.113.011296 {/ref}
Each of these manuals has been revised several times: the ICD is currently on its eleventh edition (known as ICD-11) as of 2019, while the DSM is currently on its fifth edition (DSM-5) as of 2013.
Both manuals provide criteria healthcare professionals can use to decide whether people have a particular mental illness.
These criteria include a description of the symptoms, their severity, and duration. They also explain how to decide between different illnesses with similar symptoms.
Both manuals mostly group mental illnesses similarly. For common mental illnesses, their criteria are usually identical or have minor differences, but in some cases, there are major differences.{ref}This paper provides a detailed summary of the similarities and differences between the criteria for mental illnesses in the ICD-11 and DSM-5.
First, M. B., Gaebel, W., Maj, M., Stein, D. J., Kogan, C. S., Saunders, J. B., Poznyak, V. B., Gureje, O., Lewis‐Fernández, R., Maercker, A., Brewin, C. R., Cloitre, M., Claudino, A., Pike, K. M., Baird, G., Skuse, D., Krueger, R. B., Briken, P., Burke, J. D., … Reed, G. M. (2021). An organization‐ and category‐level comparison of diagnostic requirements for mental disorders in ICD ‐11 and DSM ‐5. World Psychiatry, 20(1), 34–51. https://doi.org/10.1002/wps.20825 {/ref}
In the chart, you can see the estimated prevalence of people who meet the criteria to be diagnosed with each category of mental illness, as they are classified by the ICD manual.
As you can see, anxiety disorders and depressive disorders are more common. It’s estimated that around 3–4% of people worldwide have had them in the past year.
Bipolar disorder, schizophrenia and eating disorders are less common. For example, it’s estimated that 0.3% of people have schizophrenia worldwide.
Schizophrenia is a condition that involves significant problems in perceiving reality, difficulty with memory and attention, and changes in behavior and movement.{ref}Other primary psychotic disorders include schizoaffective disorder, schizotypal disorder, acute and transient psychotic disorder, delusional disorder, symptomatic manifestations of primary psychotic disorders, and others.{/ref}
In the chart, you can see the rate of schizophrenia between genders. As the chart shows, around 0.2 to 0.5% of people are estimated to have schizophrenia across countries. Schizophrenia is relatively balanced between men and women, although in many countries it’s estimated to be slightly more common among men than women.
To diagnose schizophrenia, the patient must have at least two of the following symptoms for most of the time, lasting at least a month. They must have at least one symptom from a–d.
The symptoms shouldn’t be caused by another medical condition or the use of substances or medication.
The ICD manual describes two types of mood disorders: depressive disorders, and bipolar or related disorders.
Depressive disorders involve significant sadness or a loss of interest, along with several other symptoms.
But there are several types of depressive disorders, classified by the ICD manual. They are based on the specific symptoms, their severity, and duration.{ref}The different types of depressive disorders described by the ICD manual are: single-episode depressive disorder, recurrent depressive disorder, dysthymic disorder, mixed depressive and anxiety disorder, and others.{/ref}
In the chart, you can see the prevalence of depressive disorders in men and women. As the chart shows, it’s estimated that depressive disorders were somewhat more common in women than men in all countries. Around 2–8% of women and 1–6% of men were estimated to have depressive disorders in the past year, across countries.
Major depression
Several types of depressive disorders involve having a major depressive episode, which can be mild, moderate, or severe.
To diagnose a major depressive episode, the patient must have had at least five of the symptoms listed below for most of the day, lasting nearly every day for at least 2 weeks.
At least one of them must be from a–b. The symptoms must also significantly affect their personal life, family, social, education, work, or other areas of their life – or only have a limited effect because of significant effort.
The symptoms shouldn’t be better accounted for by bereavement (the death of someone close), another medical condition, or the use of substances or medication.
Dysthymic disorder
Unlike major depression, dysthymic disorder involves having milder symptoms for a longer minimum period.
To diagnose a dysthymic disorder, the patient must have had a persistent depressed mood lasting at least two years, for most of the day, for the majority of days. In children and adolescents, this may appear as irritability instead. The symptoms must also significantly affect their personal life, family, social, education, work, or other areas – or only be limited because of significant effort.
In addition, they typically have additional symptoms such as:
To be diagnosed, there should not have been a two-week period when they had enough symptoms to meet the criteria for a depressive episode, during the first two years of their symptoms. There should also not be any period without symptoms lasting two months or more.
They should not have a history of manic episodes, mixed episodes, or hypomanic episodes (which would meet the criteria of bipolar or related disorders; see the next section). The symptoms shouldn’t be caused by another medical condition or the use of substances or medications.
Bipolar disorder involves two different sets of symptoms. One set is depressive symptoms, while the other involves significantly increased excitement, irritability, and energy. This second set of symptoms is classified as “manic episodes”, “mixed episodes”, or “hypomanic episodes or symptoms”.{ref}In mixed episodes, people either have a mix of depressive and manic symptoms together or rapidly alternate between the different sets of symptoms. In hypomanic episodes, people have similar symptoms to those in a manic episode, but to a milder degree.{/ref}
In the chart, you can see the prevalence of bipolar disorder in different countries. Overall, around 0.2–1.8% of people are estimated to have bipolar disorder. Bipolar disorder is relatively balanced between men and women, although in many countries it’s estimated to be slightly more common among women than men.
To diagnose someone with a manic episode, the patient must have the following symptoms for most of the day, nearly every day, for at least a week{ref}Unless that period is shortened by a treatment.{/ref}.
Along with these, their behavior must include several of the following unusual symptoms.
Their symptoms should not be due to another medical condition or the use of a substance or medication, and they shouldn’t fill the requirements for a mixed episode.
To be diagnosed, their symptoms must also significantly affect their personal life, family, social, education, work, or other areas of their life; or require intensive treatment to prevent them from harming themselves or others; or are accompanied by delusions or hallucinations.
There are many illnesses related to anxiety and fear. They are classified by the ICD manual as generalized anxiety disorder, panic disorder, agoraphobia, specific phobias, social anxiety disorder, separation anxiety disorder, selective mutism, and others.
In the chart, you can see the estimated share of people who would meet the criteria for anxiety disorders in men and women. You can see that anxiety disorders are estimated to be more common in women than men in all countries. Around 2–11% of women and 2–7% of men were estimated to have anxiety disorders in the past year, across countries.
Generalized anxiety disorder involves persistent anxiety symptoms. The symptoms must also significantly affect their personal life, family, social, education, work, or other areas – or only be limited because of significant effort.
To be diagnosed with generalized anxiety disorder, patients must have either:
Along with this, they must have additional symptoms such as:
To be diagnosed, their symptoms must persist for at least several months, for the majority of days. They should not be better accounted for by another mental disorder, and should not be the result of another medical condition or the use of substances or medication.
This is a group of illnesses that involve symptoms of repetitive thoughts and behaviors, such as obsessions, intrusive thoughts, and preoccupations.
According to the ICD manual, the group includes: obsessive-compulsive disorder (OCD), body dysmorphic disorder, olfactory reference disorder, hypochondriasis, hoarding disorder, body-focused repetitive behavior disorder, and others.
To be diagnosed with obsessive-compulsive disorder, the ICD manual explains that the patient must have repetitive and persistent obsessions and/or compulsions. These can be thoughts, images, or impulses/urges that feel intrusive and unwanted and are commonly linked to anxiety. The patient usually tries to ignore them, suppress them, or comply with them by performing repetitive behaviors or rituals.
To be diagnosed, these obsessions and compulsions must be time-consuming or significantly affect their personal life, family, social, education, work, or other areas – or are only limited because of significant effort.
Their symptoms should not be better accounted for by another mental disorder and should not be the result of another medical condition or the use of substances or medication.
According to the ICD manual, eating disorders involve abnormal behaviors and preoccupations with food, along with strong concerns about body weight and shape.
In contrast, feeding disorders involve abnormal behaviors such as eating non-edible substances or regurgitating food without concerns about body weight or shape.
In the chart, you can see the estimated share of people who would meet the criteria for eating disorders, in men and women. Eating disorders are estimated to be more common among women (0.06–1.5%) than men (0.03–0.75%).
To be diagnosed with anorexia nervosa, the ICD manual explains that the patient must have significantly low body weight for their height, age, developmental stage, or weight history. Or, they must have had rapid weight loss along with other requirements. For children and adolescents, their symptoms may be the failure to gain weight as expected with age, rather than weight loss.
Their low body weight should not be better accounted for by another medical condition or a lack of available food.
They have persistent patterns of behavior aimed at reaching or maintaining an abnormally low body weight, and typically have an extreme fear of weight gain.
These behaviors can include strongly reducing energy intake (such as fasting, choosing low-calorie food, eating excessively slowly, hiding or spitting out food), purging behaviors (such as self-induced vomiting and the use of laxatives, diuretics, enemas, or not using insulin among individuals with diabetes), or increasing activity (such as excessive exercise, hyperactivity, deliberate exposure to cold, or the use of medication that increases activity).
Patients also have an excessive preoccupation with their body weight or shape, which feel central to their perception of themselves.
To be diagnosed with bulimia nervosa, the ICD manual explains that the patient must have frequent, recurrent episodes of binge eating. Binge eating is defined as a period when the person has a loss of control over their eating behavior and eats much more or differently than usual.
Along with this, they have repeated inappropriate behaviors to prevent weight gain, such as self-induced vomiting, fasting, the use of diuretics, laxatives, or enemas, or not using insulin among individuals with diabetes, or strenuous exercise.
They have an excessive preoccupation with body weight or shape and have significant distress about their behavior, or it significantly affects their personal life, family, social, education, work, or other areas.
To be diagnosed, their symptoms should not meet the diagnostic requirements for anorexia.
These conditions are related to stressful or traumatic events or a series of events or harmful experiences.{ref}They include post-traumatic stress disorder (PTSD), complex PTSD, prolonged grief disorder, adjustment disorder, reactive attachment disorder, disinhibited social engagement disorder, and others.{/ref}
To be diagnosed with PTSD, the ICD manual explains that the patient must have been exposed to an event or situation that was extremely threatening or horrific.{ref}This can include directly experiencing natural or human-made disasters, combat, serious accidents, torture, sexual violence, terrorism, assault or acute life-threatening illness (e.g., a heart attack); witnessing the threatened or actual injury or death of others in a sudden, unexpected, or violent manner; and learning about the sudden, unexpected, or violent death of a loved one.{/ref}
Following the traumatic event or situation, they must have developed a syndrome lasting at least several weeks that includes all three core elements:
To be diagnosed, their syndrome must significantly affect their personal life, family, social, education, work, or other areas.
Personality disorders were previously defined as a group of different disorders that were all related to people’s behavior, perception, and social functioning. But now, they are all classified together by the ICD manual.
To be diagnosed with personality disorders, the patient must have persistent problems about themselves (their identity, self-worth, the accuracy of their self-perception, or self-goals) or their relationships (such as maintaining close and satisfying relationships, understanding other people’s perspectives, and managing conflict in relationships). These problems are visible in patterns of thoughts, emotions, and behaviors that are rigid or poorly regulated.
These problems must have persisted over a long time – at least two years – and affect a range of personal and social situations, not specific ones.
Personality disorders can be categorized as mild, moderate, and severe. But to be diagnosed, the disorder must significantly affect the person’s personal life, family, social, education, work, or other areas.
They should not be better accounted for by another mental disorder and should not be the result of another medical condition or the use of substances or medication. They should also not be diagnosed if the patterns are typical during that period of development (such as problems establishing a self-identity during adolescence) or if they can be explained by social or cultural factors.
Our understanding of mental illnesses continues to develop over time. They are a range of conditions that significantly affect people’s lives and are reflected in their behavior, thoughts, or relationships with others.
To be diagnosed with a mental illness, people need to meet various symptom requirements. Healthcare professionals need to exclude other conditions that can cause the symptoms and the use of substances or medications.
Keep reading on Our World in Data:
Mental illnesses are diagnosed by health professionals using manuals, which describe their symptoms, their context, and how they differ from similar conditions.
\n\n\n\nHow are these manuals used? How are major mental illnesses described and diagnosed?
\n\n\n\nThis article summarizes the major mental illnesses in the ICD-11 manual, the official worldwide classification.
\n\n\n\nEach section includes a summary of how the condition is medically diagnosed, a description of the symptoms, and other conditions that must be ruled out before making a diagnosis.
\n\n\n\nIf you want to read the full criteria for diagnosing mental illnesses according to the ICD-11, you can find them here.
\n\n\n\nDiagnosing a mental illness begins when people see a doctor or psychiatrist.
\n\n\n\nIt can be difficult to define mental illnesses because they are conditions of the mind: they are diagnosed according to people’s symptoms and behavior. This means they can be subjective – they are dependent partly on people’s comfort seeing a doctor or psychiatrist, the training and experiences of healthcare professionals, and other cultural factors.
\n\n\n\nTwo manuals are commonly used by doctors and psychiatrists to diagnose mental illnesses: the International Classification of Diseases (ICD) and the Diagnostic and Statistical Manual of Mental Disorders (DSM).
\n\n\n\nThere are differences in how they are used.
\n\n\n\nThe ICD is the official worldwide classification for healthcare and data collection. It is used by a wide range of healthcare professionals.
\n\n\n\nIn contrast, the DSM was developed for psychiatrists in the United States, although its usage has expanded worldwide. It is intended to provide standardized guidelines for different psychiatrists who may have different judgments otherwise.{ref}Harrison, J. E., Weber, S., Jakob, R., & Chute, C. G. (2021). ICD-11: An international classification of diseases for the twenty-first century. BMC Medical Informatics and Decision Making, 21(S6), 206. https://doi.org/10.1186/s12911-021-01534-6
Kupfer, D. J., Regier, D. A., & Kuhl, E. A. (2008). On the road to DSM-V and ICD-11. European Archives of Psychiatry and Clinical Neuroscience, 258(S5), 2–6. https://doi.org/10.1007/s00406-008-5002-6
Tyrer, P. (2014). A comparison of DSM and ICD classifications of mental disorder. Advances in Psychiatric Treatment, 20(4), 280–285. https://doi.org/10.1192/apt.bp.113.011296 {/ref}
Each of these manuals has been revised several times: the ICD is currently on its eleventh edition (known as ICD-11) as of 2019, while the DSM is currently on its fifth edition (DSM-5) as of 2013.
\n\n\n\nBoth manuals provide criteria healthcare professionals can use to decide whether people have a particular mental illness.
\n\n\n\nThese criteria include a description of the symptoms, their severity, and duration. They also explain how to decide between different illnesses with similar symptoms.
\n\n\n\nBoth manuals mostly group mental illnesses similarly. For common mental illnesses, their criteria are usually identical or have minor differences, but in some cases, there are major differences.{ref}This paper provides a detailed summary of the similarities and differences between the criteria for mental illnesses in the ICD-11 and DSM-5.
First, M. B., Gaebel, W., Maj, M., Stein, D. J., Kogan, C. S., Saunders, J. B., Poznyak, V. B., Gureje, O., Lewis‐Fernández, R., Maercker, A., Brewin, C. R., Cloitre, M., Claudino, A., Pike, K. M., Baird, G., Skuse, D., Krueger, R. B., Briken, P., Burke, J. D., … Reed, G. M. (2021). An organization‐ and category‐level comparison of diagnostic requirements for mental disorders in ICD ‐11 and DSM ‐5. World Psychiatry, 20(1), 34–51. https://doi.org/10.1002/wps.20825 {/ref}
In the chart, you can see the estimated prevalence of people who meet the criteria to be diagnosed with each category of mental illness, as they are classified by the ICD manual.
\n\n\n\nAs you can see, anxiety disorders and depressive disorders are more common. It’s estimated that around 3–4% of people worldwide have had them in the past year.
\n\n\n\nBipolar disorder, schizophrenia and eating disorders are less common. For example, it’s estimated that 0.3% of people have schizophrenia worldwide.
\nSchizophrenia is a condition that involves significant problems in perceiving reality, difficulty with memory and attention, and changes in behavior and movement.{ref}Other primary psychotic disorders include schizoaffective disorder, schizotypal disorder, acute and transient psychotic disorder, delusional disorder, symptomatic manifestations of primary psychotic disorders, and others.{/ref}
\n\n\n\nIn the chart, you can see the rate of schizophrenia between genders. As the chart shows, around 0.2 to 0.5% of people are estimated to have schizophrenia across countries. Schizophrenia is relatively balanced between men and women, although in many countries it’s estimated to be slightly more common among men than women.
\n\n\n\nTo diagnose schizophrenia, the patient must have at least two of the following symptoms for most of the time, lasting at least a month. They must have at least one symptom from a–d.
\n\n\n\nThe symptoms shouldn’t be caused by another medical condition or the use of substances or medication.
\nThe ICD manual describes two types of mood disorders: depressive disorders, and bipolar or related disorders.
\n\n\n\nDepressive disorders involve significant sadness or a loss of interest, along with several other symptoms.
\n\n\n\nBut there are several types of depressive disorders, classified by the ICD manual. They are based on the specific symptoms, their severity, and duration.{ref}The different types of depressive disorders described by the ICD manual are: single-episode depressive disorder, recurrent depressive disorder, dysthymic disorder, mixed depressive and anxiety disorder, and others.{/ref}
\n\n\n\nIn the chart, you can see the prevalence of depressive disorders in men and women. As the chart shows, it’s estimated that depressive disorders were somewhat more common in women than men in all countries. Around 2–8% of women and 1–6% of men were estimated to have depressive disorders in the past year, across countries.
\n\n\n\nMajor depression
\n\n\n\nSeveral types of depressive disorders involve having a major depressive episode, which can be mild, moderate, or severe.
\n\n\n\nTo diagnose a major depressive episode, the patient must have had at least five of the symptoms listed below for most of the day, lasting nearly every day for at least 2 weeks.
\n\n\n\nAt least one of them must be from a–b. The symptoms must also significantly affect their personal life, family, social, education, work, or other areas of their life – or only have a limited effect because of significant effort.
\n\n\n\nThe symptoms shouldn’t be better accounted for by bereavement (the death of someone close), another medical condition, or the use of substances or medication.
\n\n\n\nDysthymic disorder
\n\n\n\nUnlike major depression, dysthymic disorder involves having milder symptoms for a longer minimum period.
\n\n\n\nTo diagnose a dysthymic disorder, the patient must have had a persistent depressed mood lasting at least two years, for most of the day, for the majority of days. In children and adolescents, this may appear as irritability instead. The symptoms must also significantly affect their personal life, family, social, education, work, or other areas – or only be limited because of significant effort.
\n\n\n\nIn addition, they typically have additional symptoms such as:
\n\n\n\nTo be diagnosed, there should not have been a two-week period when they had enough symptoms to meet the criteria for a depressive episode, during the first two years of their symptoms. There should also not be any period without symptoms lasting two months or more.
\n\n\n\nThey should not have a history of manic episodes, mixed episodes, or hypomanic episodes (which would meet the criteria of bipolar or related disorders; see the next section). The symptoms shouldn’t be caused by another medical condition or the use of substances or medications.
\nBipolar disorder involves two different sets of symptoms. One set is depressive symptoms, while the other involves significantly increased excitement, irritability, and energy. This second set of symptoms is classified as “manic episodes”, “mixed episodes”, or “hypomanic episodes or symptoms”.{ref}In mixed episodes, people either have a mix of depressive and manic symptoms together or rapidly alternate between the different sets of symptoms. In hypomanic episodes, people have similar symptoms to those in a manic episode, but to a milder degree.{/ref}
\n\n\n\nIn the chart, you can see the prevalence of bipolar disorder in different countries. Overall, around 0.2–1.8% of people are estimated to have bipolar disorder. Bipolar disorder is relatively balanced between men and women, although in many countries it’s estimated to be slightly more common among women than men.
\n\n\n\nTo diagnose someone with a manic episode, the patient must have the following symptoms for most of the day, nearly every day, for at least a week{ref}Unless that period is shortened by a treatment.{/ref}.
\n\n\n\nAlong with these, their behavior must include several of the following unusual symptoms.
\n\n\n\nTheir symptoms should not be due to another medical condition or the use of a substance or medication, and they shouldn’t fill the requirements for a mixed episode.
\n\n\n\nTo be diagnosed, their symptoms must also significantly affect their personal life, family, social, education, work, or other areas of their life; or require intensive treatment to prevent them from harming themselves or others; or are accompanied by delusions or hallucinations.
\nThere are many illnesses related to anxiety and fear. They are classified by the ICD manual as generalized anxiety disorder, panic disorder, agoraphobia, specific phobias, social anxiety disorder, separation anxiety disorder, selective mutism, and others.
\n\n\n\nIn the chart, you can see the estimated share of people who would meet the criteria for anxiety disorders in men and women. You can see that anxiety disorders are estimated to be more common in women than men in all countries. Around 2–11% of women and 2–7% of men were estimated to have anxiety disorders in the past year, across countries.
\n\n\n\nGeneralized anxiety disorder involves persistent anxiety symptoms. The symptoms must also significantly affect their personal life, family, social, education, work, or other areas – or only be limited because of significant effort.
\n\n\n\nTo be diagnosed with generalized anxiety disorder, patients must have either:
\n\n\n\nAlong with this, they must have additional symptoms such as:
\n\n\n\nTo be diagnosed, their symptoms must persist for at least several months, for the majority of days. They should not be better accounted for by another mental disorder, and should not be the result of another medical condition or the use of substances or medication.
\nThis is a group of illnesses that involve symptoms of repetitive thoughts and behaviors, such as obsessions, intrusive thoughts, and preoccupations.
\n\n\n\nAccording to the ICD manual, the group includes: obsessive-compulsive disorder (OCD), body dysmorphic disorder, olfactory reference disorder, hypochondriasis, hoarding disorder, body-focused repetitive behavior disorder, and others.
\n\n\n\nTo be diagnosed with obsessive-compulsive disorder, the ICD manual explains that the patient must have repetitive and persistent obsessions and/or compulsions. These can be thoughts, images, or impulses/urges that feel intrusive and unwanted and are commonly linked to anxiety. The patient usually tries to ignore them, suppress them, or comply with them by performing repetitive behaviors or rituals.
\n\n\n\nTo be diagnosed, these obsessions and compulsions must be time-consuming or significantly affect their personal life, family, social, education, work, or other areas – or are only limited because of significant effort.
\n\n\n\nTheir symptoms should not be better accounted for by another mental disorder and should not be the result of another medical condition or the use of substances or medication.
\nAccording to the ICD manual, eating disorders involve abnormal behaviors and preoccupations with food, along with strong concerns about body weight and shape.
\n\n\n\nIn contrast, feeding disorders involve abnormal behaviors such as eating non-edible substances or regurgitating food without concerns about body weight or shape.
\n\n\n\nIn the chart, you can see the estimated share of people who would meet the criteria for eating disorders, in men and women. Eating disorders are estimated to be more common among women (0.06–1.5%) than men (0.03–0.75%).
\n\n\n\nTo be diagnosed with anorexia nervosa, the ICD manual explains that the patient must have significantly low body weight for their height, age, developmental stage, or weight history. Or, they must have had rapid weight loss along with other requirements. For children and adolescents, their symptoms may be the failure to gain weight as expected with age, rather than weight loss.
\n\n\n\nTheir low body weight should not be better accounted for by another medical condition or a lack of available food.
\n\n\n\nThey have persistent patterns of behavior aimed at reaching or maintaining an abnormally low body weight, and typically have an extreme fear of weight gain.
\n\n\n\nThese behaviors can include strongly reducing energy intake (such as fasting, choosing low-calorie food, eating excessively slowly, hiding or spitting out food), purging behaviors (such as self-induced vomiting and the use of laxatives, diuretics, enemas, or not using insulin among individuals with diabetes), or increasing activity (such as excessive exercise, hyperactivity, deliberate exposure to cold, or the use of medication that increases activity).
\n\n\n\nPatients also have an excessive preoccupation with their body weight or shape, which feel central to their perception of themselves.
\n\n\n\nTo be diagnosed with bulimia nervosa, the ICD manual explains that the patient must have frequent, recurrent episodes of binge eating. Binge eating is defined as a period when the person has a loss of control over their eating behavior and eats much more or differently than usual.
\n\n\n\nAlong with this, they have repeated inappropriate behaviors to prevent weight gain, such as self-induced vomiting, fasting, the use of diuretics, laxatives, or enemas, or not using insulin among individuals with diabetes, or strenuous exercise.
\n\n\n\nThey have an excessive preoccupation with body weight or shape and have significant distress about their behavior, or it significantly affects their personal life, family, social, education, work, or other areas.
\n\n\n\nTo be diagnosed, their symptoms should not meet the diagnostic requirements for anorexia.
\nThese conditions are related to stressful or traumatic events or a series of events or harmful experiences.{ref}They include post-traumatic stress disorder (PTSD), complex PTSD, prolonged grief disorder, adjustment disorder, reactive attachment disorder, disinhibited social engagement disorder, and others.{/ref}
\n\n\n\nTo be diagnosed with PTSD, the ICD manual explains that the patient must have been exposed to an event or situation that was extremely threatening or horrific.{ref}This can include directly experiencing natural or human-made disasters, combat, serious accidents, torture, sexual violence, terrorism, assault or acute life-threatening illness (e.g., a heart attack); witnessing the threatened or actual injury or death of others in a sudden, unexpected, or violent manner; and learning about the sudden, unexpected, or violent death of a loved one.{/ref}
\n\n\n\nFollowing the traumatic event or situation, they must have developed a syndrome lasting at least several weeks that includes all three core elements:
\n\n\n\nTo be diagnosed, their syndrome must significantly affect their personal life, family, social, education, work, or other areas.
\nPersonality disorders were previously defined as a group of different disorders that were all related to people’s behavior, perception, and social functioning. But now, they are all classified together by the ICD manual.
\n\n\n\nTo be diagnosed with personality disorders, the patient must have persistent problems about themselves (their identity, self-worth, the accuracy of their self-perception, or self-goals) or their relationships (such as maintaining close and satisfying relationships, understanding other people’s perspectives, and managing conflict in relationships). These problems are visible in patterns of thoughts, emotions, and behaviors that are rigid or poorly regulated.
\n\n\n\nThese problems must have persisted over a long time – at least two years – and affect a range of personal and social situations, not specific ones.
\n\n\n\nPersonality disorders can be categorized as mild, moderate, and severe. But to be diagnosed, the disorder must significantly affect the person’s personal life, family, social, education, work, or other areas.
\n\n\n\nThey should not be better accounted for by another mental disorder and should not be the result of another medical condition or the use of substances or medication. They should also not be diagnosed if the patterns are typical during that period of development (such as problems establishing a self-identity during adolescence) or if they can be explained by social or cultural factors.
\nOur understanding of mental illnesses continues to develop over time. They are a range of conditions that significantly affect people’s lives and are reflected in their behavior, thoughts, or relationships with others.
\n\n\n\nTo be diagnosed with a mental illness, people need to meet various symptom requirements. Healthcare professionals need to exclude other conditions that can cause the symptoms and the use of substances or medications.
\n\n\n\nKeep reading on Our World in Data:
\n\n\nIn many countries, many people with mental illnesses go undiagnosed, meaning mental health is given less attention and support than it deserves. Even for those diagnosed, treatment can be of poor quality, if they receive it at all.{ref}Alonso, J., Liu, Z., Evans‐Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., ... & WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and anxiety, 35(3), 195-208. https://doi.org/10.1002/da.22711 {/ref}
To reduce the burden of mental illnesses, the world needs reliable data, which includes the number of people that face mental illnesses, how and when they occur, and the effectiveness of treatments.
Defining mental illnesses is complex. They are diagnosed based on people’s psychological symptoms and behavior rather than biomarkers, brain scans, or blood tests. This makes them more subjective – they are dependent on whether people share their symptoms and the way doctors diagnose them.
Mental illnesses are formally defined according to the International Classification of Diseases (ICD) and the Diagnostic and Statistical Manual of Mental Disorders (DSM). The ICD is used internationally by healthcare professionals, while the DSM is primarily used by psychiatrists in the United States.{ref}Kupfer, D. J., Regier, D. A., & Kuhl, E. A. (2008). On the road to DSM-V and ICD-11. European Archives of Psychiatry and Clinical Neuroscience, 258(S5), 2–6. https://doi.org/10.1007/s00406-008-5002-6 {/ref}
These manuals explain how to diagnose mental illnesses by observing and asking about people’s symptoms and behavior, and the context of their symptoms – for example, symptoms that appeared because of drug use or other medical conditions don’t qualify as mental illnesses.
Based on these definitions, healthcare professionals can make diagnoses, which can be used for healthcare, treatment, and national statistics.
Over time, the definitions of particular mental illnesses have changed. The DSM has been revised 5 times since it was first developed in 1952, while the ICD has been revised 11 times since 1900.{ref}American Psychiatric Association. (2022). DSM History. https://www.psychiatry.org/psychiatrists/practice/dsm/about-dsm/history-of-the-dsm
Hirsch, J. A., Nicola, G., McGinty, G., Liu, R. W., Barr, R. M., Chittle, M. D., & Manchikanti, L. (2016). ICD-10: History and Context. AJNR. American Journal of Neuroradiology, 37(4), 596–599. https://doi.org/10.3174/ajnr.A4696 {/ref} They will continue to be revised in the future, but updates have become less frequent.
Their changes are partly due to a better understanding and measurement of mental illnesses. They have also changed as a result of cultural and legal factors. There used to be larger differences in the criteria for diagnosing mental illnesses between the ICD and the DSM, but the two manuals are now more similar due to collaboration between their developers.{ref}This paper provides a detailed summary of the similarities and differences between the criteria for mental illnesses in the ICD-11 and DSM-5.
First, M. B., Gaebel, W., Maj, M., Stein, D. J., Kogan, C. S., Saunders, J. B., Poznyak, V. B., Gureje, O., Lewis‐Fernández, R., Maercker, A., Brewin, C. R., Cloitre, M., Claudino, A., Pike, K. M., Baird, G., Skuse, D., Krueger, R. B., Briken, P., Burke, J. D., … Reed, G. M. (2021). An organization‐ and category‐level comparison of diagnostic requirements for mental disorders in ICD ‐11 and DSM ‐5. World Psychiatry, 20(1), 34–51. https://doi.org/10.1002/wps.20825
Harrison, J. E., Weber, S., Jakob, R., & Chute, C. G. (2021). ICD-11: An international classification of diseases for the twenty-first century. BMC Medical Informatics and Decision Making, 21(S6), 206. https://doi.org/10.1186/s12911-021-01534-6
Tyrer, P. (2014). A comparison of DSM and ICD classifications of mental disorder. Advances in Psychiatric Treatment, 20(4), 280–285. https://doi.org/10.1192/apt.bp.113.011296 {/ref}
The process of diagnosing a mental illness usually starts with the patient consulting a healthcare professional.{ref}People can also be diagnosed through other routes. For example, in some countries, there are also screening programs to identify people who may have mental health conditions and refer them to specialists. Children and adolescents can be referred to healthcare professionals by carers.
In some countries, people can also be diagnosed and committed to mental hospitals involuntarily if they are considered to pose a danger to others. This was more common in countries like the United States before the 1960s, but since then, these laws have been reformed in many countries.
See also: Testa, M., & West, S. G. (2010). Civil commitment in the United States. Psychiatry (Edgmont (Pa.: Township)), 7(10), 30–40.
Zhang, S., Mellsop, G., Brink, J., & Wang, X. (2015). Involuntary admission and treatment of patients with mental disorder. Neuroscience Bulletin, 31(1), 99–112. https://doi.org/10.1007/s12264-014-1493-5
Appelbaum, P. S. (1997). Almost a revolution: an international perspective on the law of involuntary commitment. Journal of the American Academy of Psychiatry and the Law Online, 25(2), 135-147.{/ref} This can be a doctor in a clinic or general hospital, a psychiatrist, or another mental health specialist.
Health professionals use official medical guidance and professional judgment to decide whether to diagnose a patient with a condition.
Data on these diagnoses are collected from hospitals in many countries, but this may not include clinic visits. The data can include people’s age and sex, their reason for admission, other diagnoses, and treatments given during their visit.{ref} Otero Varela, L., Doktorchik, C., Wiebe, N., Quan, H., & Eastwood, C. (2021). Exploring the differences in ICD and hospital morbidity data collection features across countries: An international survey. BMC Health Services Research, 21(1), 308. https://doi.org/10.1186/s12913-021-06302-w
Varela, L. O., Knudsen, S., Carpendale, S., Eastwood, C., & Quan, H. (2019, October). Comparing ICD-Data Across Countries: A Case for Visualization?. In 2019 IEEE Workshop on Visual Analytics in Healthcare (VAHC) (pp. 60-61). IEEE. {/ref}
Official data from diagnoses of mental illnesses have two major strengths.
First, the diagnoses come from healthcare professionals with training and experience in recognizing mental illnesses.
They can use their knowledge to ask people more questions about their symptoms and understand their context before making a diagnosis. They can also perform additional medical tests to rule out other conditions.
Second, data on diagnoses can tell us about the number of people who seek out mental health treatment from public hospitals and clinics. This can usually be linked to data on which treatments they were prescribed and for how long.{ref}In some countries, there are also national screening programs to diagnose patients with both physical and mental illnesses.{/ref}
This can be very useful for countries to understand the resources used to treat mental illnesses.
Data on diagnoses of mental illnesses also has limitations.
One problem is that many people do not reach out to healthcare professionals about their health conditions. This might be because they lack awareness of mental illnesses or there is a lack of healthcare for these conditions in their country. They may also feel uncomfortable about sharing their symptoms with healthcare professionals.
Another problem is that the diagnoses may not be made consistently. Doctors can have different levels of training and experience in recognizing and diagnosing mental illnesses. This can be a problem when comparing diagnosis rates between countries.
A third problem is that national data on diagnosis is often collected from a limited number of sources. Data from private hospitals and clinics is usually not included. In some countries, mental health data is collected from hospitals but not clinics.{ref}Baxter, A. J., Patton, G., Scott, K. M., Degenhardt, L., & Whiteford, H. A. (2013). Global Epidemiology of Mental Disorders: What Are We Missing? PLoS ONE, 8(6), e65514. https://doi.org/10.1371/journal.pone.0065514 {/ref}
Some countries also use data from other sources. For example, the data might come from health insurance claims—which include private healthcare—or from other databases connected to many healthcare clinics across the country.
The final problem is that countries may use different definitions to diagnose patients. Some countries use modified versions of the ICD manual to collect their data, depending on their cultural context and needs.{ref}Brhlikova, P., Pollock, A. M., & Manners, R. (2011). Global Burden of Disease estimates of depression – how reliable is the epidemiological evidence? Journal of the Royal Society of Medicine, 104(1), 25–34. https://doi.org/10.1258/jrsm.2010.100080 {/ref}
These differences mean that comparing data between countries using data on diagnoses can be difficult.
Aside from the formal diagnoses, mental illnesses can be measured using surveys and screening questionnaires.
These tend to ask people about symptoms similar to those in diagnostic manuals, but they can be used more easily and widely because the data does not need to be collected by a healthcare professional.
These surveys can be conducted in different ways: over the phone, online, or in-person while anonymized.{ref}In-person surveys are usually anonymized for sensitive questions such as those on mental health. In this case, the person may be given a laptop to answer the questions while the interviewer cannot see their answers.{/ref}
During the surveys, trained professionals ask people in the general population whether they have experienced symptoms of mental illnesses. They also ask about the age when people first experienced them, how long they lasted, and how severe the symptoms were.
People may be asked about symptoms they have currently, or have had recently, or in their lifetime so far.
Some symptoms of mental illnesses may be common in the population. This chart shows data from a US survey called the National Health and Nutrition Examination Survey. It was a large-scale, in-person survey of people in the general population.{ref}Tomitaka, S., Kawasaki, Y., Ide, K., Akutagawa, M., Yamada, H., Ono, Y., & Furukawa, T. A. (2018). Distributional patterns of item responses and total scores on the PHQ-9 in the general population: Data from the National Health and Nutrition Examination Survey. BMC Psychiatry, 18(1), 108. https://doi.org/10.1186/s12888-018-1696-9 {/ref}
As you can see, around a fifth of the US population says they have had a depressed mood for several days in the past two weeks.
But having one or a few symptoms does not necessarily mean that someone can be diagnosed with depression. Instead, researchers look at the combination of symptoms that people report. They will set a threshold for the number of symptoms someone must have before they are considered to have the condition.
To diagnose someone with major depression, for example, the ICD and DSM criteria require them to have had a depressed mood or loss of interest for much of the day, nearly every day, for at least two weeks, along with several other symptoms.
Survey data on mental health has two major strengths.
First, it involves structured interviews – people are asked a consistent set of questions regardless of their interviewer. This helps to ensure the data is more comparable between interviewers and across time.{ref}Mueller, A. E., & Segal, D. L. (2014). Structured versus semistructured versus unstructured interviews. The encyclopedia of clinical psychology, 1-7. https://doi.org/10.1002/9781118625392.wbecp069 {/ref}
Second, surveys include a much wider range of people in the population, including those who would not seek treatment due to a lack of awareness, costs, or other concerns.
This can help to understand the prevalence of mental illnesses in the general population, including people never diagnosed by a healthcare professional. This can help us understand what share of people seek treatment.
One example is the World Mental Health surveys: these were large-scale surveys of mental illnesses in the general population in 21 countries of different income levels.{ref}Kessler, R. C., Green, J. G., Gruber, M. J., Sampson, N. A., Bromet, E., Cuitan, M., ... & Zaslavsky, A. M. (2010). Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. International journal of methods in psychiatric research, 19(S1), 4-22. https://doi.org/10.1002/mpr.310 {/ref}
Based on structured interviews, the authors found that around 1 in 10 people met the criteria for an anxiety disorder in the past year across countries surveyed.{ref}Alonso, J., Liu, Z., Evans-Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., Abdulmalik, J., Aguilar-Gaxiola, S., Al-Hamzawi, A., Andrade, L. H., Bruffaerts, R., Cardoso, G., Cia, A., Florescu, S., de Girolamo, G., Gureje, O., Haro, J. M., He, Y., de Jonge, P., … the WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and Anxiety, 35(3), 195–208. https://doi.org/10.1002/da.22711 {/ref}
In the chart, you can see the share of those who met the criteria who also said they had received treatment. Only around 36% with anxiety disorders received any treatment in high-income countries, while an even lower share – only 13% – received it in lower-middle-income countries.{ref}Potentially adequate treatment was defined by the authors as receiving psychotherapy, medication, or complementary alternative medicine.{/ref}
Third, surveys can tell us about milder symptoms that may be common in the population. This can help to see if a condition lies on a spectrum – where everyone has the symptoms to different degrees – or if it affects a separate share of people.
There are also several limitations of survey data to keep in mind.
One is that people may not share their symptoms in surveys, as they may not feel comfortable sharing them with researchers. Along with this, people’s comfort in sharing mental health symptoms may vary between countries and over time.{ref}Gaia, A. (2020). Social Desirability Bias and Sensitive Questions in Surveys. In SAGE Research Methods Foundations. SAGE Publications Ltd. https://doi.org/10.4135/9781526421036928979
Krosnick, J. A. (1999). Maximizing questionnaire quality. Measures of Political Attitudes, 2, 37–58.
Shoemaker, P. J. (2002). Item Nonresponse: Distinguishing between don’t Know and Refuse. International Journal of Public Opinion Research, 14(2), 193–201. https://doi.org/10.1093/ijpor/14.2.193 {/ref}
Another limitation is that people may not remember their symptoms, especially when they are asked to recall symptoms in their lifetime so far. This can be challenging for older people whose symptoms may have occurred decades ago.
This chart shows data from a study where the same people were interviewed several times about depression in their lifetimes so far.{ref}Takayanagi, Y., Spira, A. P., Roth, K. B., Gallo, J. J., Eaton, W. W., & Mojtabai, R. (2014). Accuracy of Reports of Lifetime Mental and Physical Disorders: Results From the Baltimore Epidemiological Catchment Area Study. JAMA Psychiatry, 71(3), 273. https://doi.org/10.1001/jamapsychiatry.2013.3579 {/ref} Around two-thirds of people who had described episodes of depression did not recall them in subsequent interviews.
A third limitation is that survey data does not usually exclude other diagnoses. When people visit a doctor for a diagnosis, the doctor may ask them about other existing conditions and medications they are taking, and test them for other medical conditions which could also cause their symptoms. This is not usually performed in surveys.
It can also be difficult to compare people’s responses in surveys, because people may interpret the questions differently. This can make it difficult to compare people from different backgrounds, languages, and countries.{ref}Putnick, D. L., & Bornstein, M. H. (2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review, 41, 71–90. https://doi.org/10.1016/j.dr.2016.06.004 {/ref}
Another consideration is who is included in surveys. Some include people from a wide range of backgrounds in the general population, while others only include particular groups like urban residents or university students.
This can be a big problem for mental health conditions that are less common, such as schizophrenia and bipolar disorder. If only a few people in the study had a condition, then it can be difficult to estimate the precise share of the total population with it.
It’s also important to know that surveys of the general population usually do not include people who are institutionalized in hospitals or prisons, who may have more severe physical and mental health conditions.{ref}Binswanger, I. A., Krueger, P. M., & Steiner, J. F. (2009). Prevalence of chronic medical conditions among jail and prison inmates in the USA compared with the general population. Journal of Epidemiology & Community Health, 63(11), 912–919. https://doi.org/10.1136/jech.2009.090662
Peen, J., Schoevers, R. A., Beekman, A. T., & Dekker, J. (2010). The current status of urban-rural differences in psychiatric disorders. Acta Psychiatrica Scandinavica, 121(2), 84–93. https://doi.org/10.1111/j.1600-0447.2009.01438.x
Rehm, J., Kilian, C., Rovira, P., Shield, K. D., & Manthey, J. (2021). The elusiveness of representativeness in general population surveys for alcohol. Drug and Alcohol Review, 40(2), 161–165. https://doi.org/10.1111/dar.13148 {/ref}
Data on mental health varies in two ways: the amount of data on each mental illness and the amount of data from each country.
First, there is much more data available on some mental illnesses than others.
You can see this in the chart. It shows the number of countries with primary data on the prevalence of each mental illness in the general population. The studies were used by the IHME’s Global Burden of Disease study to estimate the prevalence of mental illnesses worldwide.{ref}Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext {/ref}
As you can see, data on some mental illnesses such as attention-deficit hyperactivity disorder, cannabis use disorder, and major depressive disorder came from a large number of countries.
However, data on others, such as bipolar disorder, autism spectrum disorders, and anorexia nervosa, came from fewer than 40 countries. For personality disorders, data came from only two countries.
Second, there is much more data available from some world regions than others.
In the left-hand chart, you can see the share of the population of world regions that had any data collected on the prevalence of major depression between 1980 and 2008.
This comes from an older study published in 2013, and more data has been collected since then, which has helped improve estimates made by the Global Burden of Disease study.{ref}Baxter, A. J., Patton, G., Scott, K. M., Degenhardt, L., & Whiteford, H. A. (2013). Global Epidemiology of Mental Disorders: What Are We Missing? PLoS ONE, 8(6), e65514. https://doi.org/10.1371/journal.pone.0065514 {/ref}
In Australasia and North America, there was data relating to all of the age demographics in the population. But in regions such as Eastern Europe and Southeast Asia, less than 25% of the population was covered.
In several regions, especially in sub-Saharan Africa, Central Asia, and South America, there was almost no underlying data available.
What about other mental illnesses apart from depression?
In the chart, you can see the population that was covered in data on other common mental illnesses.
As you can see, there was more data available on anxiety disorders and major depression than on schizophrenia and bipolar disorder.
You can click ‘Change country’ to see how this varies between world regions.
Researchers can try to make comparable estimates of mental health using this underlying data and statistical methods.{ref}Mathers, C., Hogan, D., & Stevens, G. (2019). Global health estimates: modelling and predicting health outcomes. The Palgrave handbook of global health data methods for policy and practice, 403-424. https://link.springer.com/chapter/10.1057/978-1-137-54984-6_21
Whiteford, H. A., Degenhardt, L., Rehm, J., Baxter, A. J., Ferrari, A. J., Erskine, H. E., ... & Vos, T. (2013). Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. The Lancet, 382(9904), 1575-1586. https://doi.org/10.1016/s0140-6736(13)61611-6
Enders, C. K. (2022). Applied missing data analysis. Guilford Publications.
Rehm, J., & Shield, K. D. (2019). Global burden of disease and the impact of mental and addictive disorders. Current psychiatry reports, 21, 1-7. https://doi.org/10.1007/s11920-019-0997-0
Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. https://doi.org/10.1016/S0140-6736(20)30925-9 {/ref}
These methods incorporate the available data on people’s mental health, their demographics, and the level of diagnosis and testing. Then they extrapolate the results to other countries, where data has not been collected. This can be based on demographics such as age and sex, other risk factors, responses to other large-scale representative mental health surveys.{ref}For major depressive disorder, for example, the IHME uses available data on the age structure of the population; risk factors such as war mortality, intimate partner violence, and childhood sexual violence; as well as responses to the Gallup’s survey on negative experiences around the world. You can learn more in the Appendix of the Global Burden of Disease Study, which is available here.
Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext#supplementaryMaterial {/ref}
They try to adjust for the fact that the underlying data was collected from different sources (diagnoses or surveys), during different periods, and from different groups of people.
However, they come with a range of uncertainty. This is because they rely on assumptions about how the data was collected, and why some demographics and countries lacked data on the prevalence of mental illnesses.
This is especially true for some illnesses – such as eating disorders and bipolar disorder – and some world regions – including much of Asia, South America and Africa; where primary data is lacking.
Data on global mental health has two main limitations.
First, our understanding of global mental health depends on people’s willingness to share their symptoms, and contact healthcare professionals to receive a diagnosis and treatment. Because of this, many people remain undiagnosed and lack support and treatment.
Another major limitation is that data is lacking in many countries. It is often available only for some age groups, and is collected by separate one-off studies at infrequent intervals. There is much less data available for some illnesses than others.
For countries that lack data, the prevalence of mental illnesses is estimated from other similar countries with data, but this leads to large uncertainties.
Despite these limitations, the available data does give us important insights:
Importantly it shows that mental illnesses are not uncommon. For example, the World Mental Health surveys estimated that one-in-ten people met the criteria for anxiety disorders in the past year, on average, across countries.{ref}Alonso, J., Liu, Z., Evans-Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., Abdulmalik, J., Aguilar-Gaxiola, S., Al-Hamzawi, A., Andrade, L. H., Bruffaerts, R., Cardoso, G., Cia, A., Florescu, S., de Girolamo, G., Gureje, O., Haro, J. M., He, Y., de Jonge, P., … the WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and Anxiety, 35(3), 195–208. https://doi.org/10.1002/da.22711 {/ref}
Global data also tells us that mental illnesses have a large treatment gap, especially in poorer countries. For example, while around one-in-three with anxiety disorders received any treatment in high-income countries, less than one-in-eight did in lower-middle-income countries.
Mental illnesses are a major part of the global health burden and remain untreated for many people. To address this, countries need more data on these conditions for a wide range of demographics, and long-term data to understand how they develop, and how effective the treatments are.
Keep reading on Our World in Data:
Acknowledgements: I would like to thank Edouard Mathieu, Hannah Ritchie and Max Roser for their helpful suggestions to improve this article.
","{""id"": ""wp-57166"", ""slug"": ""how-do-researchers-study-the-prevalence-of-mental-illnesses"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""In many countries, many people with mental illnesses go undiagnosed, meaning mental health is given less attention and support than it deserves. Even for those diagnosed, treatment can be of poor quality, if they receive it at all.{ref}Alonso, J., Liu, Z., Evans‐Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., ... & WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and anxiety, 35(3), 195-208. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1002/da.22711"", ""children"": [{""text"": ""https://doi.org/10.1002/da.22711"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To reduce the burden of mental illnesses, the world needs reliable data, which includes the number of people that face mental illnesses, how and when they occur, and the effectiveness of treatments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How are mental illnesses defined?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Defining mental illnesses is complex. They are diagnosed based on people’s psychological symptoms and behavior rather than biomarkers, brain scans, or blood tests. This makes them more subjective – they are dependent on whether people share their symptoms and the way doctors diagnose them."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Mental illnesses are formally defined according to the International Classification of Diseases (ICD) and the Diagnostic and Statistical Manual of Mental Disorders (DSM). The ICD is used internationally by healthcare professionals, while the DSM is primarily used by psychiatrists in the United States.{ref}Kupfer, D. J., Regier, D. A., & Kuhl, E. A. (2008). On the road to DSM-V and ICD-11. European Archives of Psychiatry and Clinical Neuroscience, 258(S5), 2–6. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1007/s00406-008-5002-6"", ""children"": [{""text"": ""https://doi.org/10.1007/s00406-008-5002-6"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These manuals explain how to diagnose mental illnesses by observing and asking about people’s symptoms and behavior, and the context of their symptoms – for example, symptoms that appeared because of drug use or other medical conditions don’t qualify as mental illnesses."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Based on these definitions, healthcare professionals can make diagnoses, which can be used for healthcare, treatment, and national statistics."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Over time, the definitions of particular mental illnesses have changed. The DSM has been revised 5 times since it was first developed in 1952, while the ICD has been revised 11 times since 1900.{ref}American Psychiatric Association. (2022). DSM History. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.psychiatry.org/psychiatrists/practice/dsm/about-dsm/history-of-the-dsm"", ""children"": [{""text"": ""https://www.psychiatry.org/psychiatrists/practice/dsm/about-dsm/history-of-the-dsm"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""text"": ""Hirsch, J. A., Nicola, G., McGinty, G., Liu, R. W., Barr, R. M., Chittle, M. D., & Manchikanti, L. (2016). ICD-10: History and Context. AJNR. American Journal of Neuroradiology, 37(4), 596–599. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.3174/ajnr.A4696"", ""children"": [{""text"": ""https://doi.org/10.3174/ajnr.A4696"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref} They will continue to be revised in the future, but updates have become less frequent."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Their changes are partly due to a better understanding and measurement of mental illnesses. They have also changed as a result of cultural and legal factors. There used to be larger differences in the criteria for diagnosing mental illnesses between the ICD and the DSM, but the two manuals are now more similar due to collaboration between their developers.{ref}This paper provides a detailed summary of the similarities and differences between the criteria for mental illnesses in the ICD-11 and DSM-5."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""First, M. B., Gaebel, W., Maj, M., Stein, D. J., Kogan, C. S., Saunders, J. B., Poznyak, V. B., Gureje, O., Lewis‐Fernández, R., Maercker, A., Brewin, C. R., Cloitre, M., Claudino, A., Pike, K. M., Baird, G., Skuse, D., Krueger, R. B., Briken, P., Burke, J. D., … Reed, G. M. (2021). An organization‐ and category‐level comparison of diagnostic requirements for mental disorders in ICD ‐11 and DSM ‐5. World Psychiatry, 20(1), 34–51. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1002/wps.20825"", ""children"": [{""text"": ""https://doi.org/10.1002/wps.20825"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""text"": ""Harrison, J. E., Weber, S., Jakob, R., & Chute, C. G. (2021). ICD-11: An international classification of diseases for the twenty-first century. BMC Medical Informatics and Decision Making, 21(S6), 206. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1186/s12911-021-01534-6"", ""children"": [{""text"": ""https://doi.org/10.1186/s12911-021-01534-6"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""text"": ""Tyrer, P. (2014). A comparison of DSM and ICD classifications of mental disorder. Advances in Psychiatric Treatment, 20(4), 280–285. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1192/apt.bp.113.011296"", ""children"": [{""text"": ""https://doi.org/10.1192/apt.bp.113.011296"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Mental health data based on diagnoses"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The process of diagnosing a mental illness usually starts with the patient consulting a healthcare professional.{ref}People can also be diagnosed through other routes. For example, in some countries, there are also screening programs to identify people who may have mental health conditions and refer them to specialists. Children and adolescents can be referred to healthcare professionals by carers."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""In some countries, people can also be diagnosed and committed to mental hospitals involuntarily if they are considered to pose a danger to others. This was more common in countries like the United States before the 1960s, but since then, these laws have been reformed in many countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""See also: Testa, M., & West, S. G. (2010). Civil commitment in the United States. Psychiatry (Edgmont (Pa.: Township)), 7(10), 30–40."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""Zhang, S., Mellsop, G., Brink, J., & Wang, X. (2015). Involuntary admission and treatment of patients with mental disorder. Neuroscience Bulletin, 31(1), 99–112. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1007/s12264-014-1493-5"", ""children"": [{""text"": ""https://doi.org/10.1007/s12264-014-1493-5"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""text"": ""Appelbaum, P. S. (1997). Almost a revolution: an international perspective on the law of involuntary commitment. Journal of the American Academy of Psychiatry and the Law Online, 25(2), 135-147.{/ref} This can be a doctor in a clinic or general hospital, a psychiatrist, or another mental health specialist."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Health professionals use official medical guidance and professional judgment to decide whether to diagnose a patient with a condition."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Data on these diagnoses are collected from hospitals in many countries, but this may not include clinic visits. The data can include people’s age and sex, their reason for admission, other diagnoses, and treatments given during their visit.{ref} Otero Varela, L., Doktorchik, C., Wiebe, N., Quan, H., & Eastwood, C. (2021). Exploring the differences in ICD and hospital morbidity data collection features across countries: An international survey. BMC Health Services Research, 21(1), 308. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1186/s12913-021-06302-w"", ""children"": [{""text"": ""https://doi.org/10.1186/s12913-021-06302-w"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Varela, L. O., Knudsen, S., Carpendale, S., Eastwood, C., & Quan, H. (2019, October). Comparing ICD-Data Across Countries: A Case for Visualization?. In 2019 IEEE Workshop on Visual Analytics in Healthcare (VAHC) (pp. 60-61). IEEE. {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What are the strengths and limitations of diagnosis data?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""text"": [{""text"": ""Strengths of data based on diagnosis"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Official data from diagnoses of mental illnesses have two major strengths."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""First, the diagnoses come from healthcare professionals with training and experience in recognizing mental illnesses. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""They can use their knowledge to ask people more questions about their symptoms and understand their context before making a diagnosis. They can also perform additional medical tests to rule out other conditions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Second, data on diagnoses can tell us about the number of people who seek out mental health treatment from public hospitals and clinics. This can usually be linked to data on which treatments they were prescribed and for how long.{ref}In some countries, there are also national screening programs to diagnose patients with both physical and mental illnesses.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This can be very useful for countries to understand the resources used to treat mental illnesses."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Limitations of data based on diagnosis"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Data on diagnoses of mental illnesses also has limitations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One problem is that many people do not reach out to healthcare professionals about their health conditions. This might be because they lack awareness of mental illnesses or there is a lack of healthcare for these conditions in their country. They may also feel uncomfortable about sharing their symptoms with healthcare professionals."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Another problem is that the diagnoses may not be made consistently. Doctors can have different levels of training and experience in recognizing and diagnosing mental illnesses. This can be a problem when comparing diagnosis rates between countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A third problem is that national data on diagnosis is often collected from a limited number of sources. Data from private hospitals and clinics is usually not included. In some countries, mental health data is collected from hospitals but not clinics.{ref}Baxter, A. J., Patton, G., Scott, K. M., Degenhardt, L., & Whiteford, H. A. (2013). Global Epidemiology of Mental Disorders: What Are We Missing? PLoS ONE, 8(6), e65514. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1371/journal.pone.0065514"", ""children"": [{""text"": ""https://doi.org/10.1371/journal.pone.0065514"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Some countries also use data from other sources. For example, the data might come from health insurance claims—which include private healthcare—or from other databases connected to many healthcare clinics across the country."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The final problem is that countries may use different definitions to diagnose patients. Some countries use modified versions of the ICD manual to collect their data, depending on their cultural context and needs.{ref}Brhlikova, P., Pollock, A. M., & Manners, R. (2011). Global Burden of Disease estimates of depression – how reliable is the epidemiological evidence? Journal of the Royal Society of Medicine, 104(1), 25–34. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1258/jrsm.2010.100080"", ""children"": [{""text"": ""https://doi.org/10.1258/jrsm.2010.100080"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These differences mean that comparing data between countries using data on diagnoses can be difficult."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Mental health data based on surveys"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Aside from the formal diagnoses, mental illnesses can be measured using surveys and screening questionnaires."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These tend to ask people about symptoms similar to those in diagnostic manuals, but they can be used more easily and widely because the data does not need to be collected by a healthcare professional."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These surveys can be conducted in different ways: over the phone, online, or in-person while anonymized.{ref}In-person surveys are usually anonymized for sensitive questions such as those on mental health. In this case, the person may be given a laptop to answer the questions while the interviewer cannot see their answers.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""During the surveys, trained professionals ask people in the general population whether they have experienced symptoms of mental illnesses. They also ask about the age when people first experienced them, how long they lasted, and how severe the symptoms were."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""People may be asked about symptoms they have currently, or have had recently, or in their lifetime so far."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Some symptoms of mental illnesses may be common in the population. This chart shows data from a US survey called the National Health and Nutrition Examination Survey. It was a large-scale, in-person survey of people in the general population.{ref}Tomitaka, S., Kawasaki, Y., Ide, K., Akutagawa, M., Yamada, H., Ono, Y., & Furukawa, T. A. (2018). Distributional patterns of item responses and total scores on the PHQ-9 in the general population: Data from the National Health and Nutrition Examination Survey. BMC Psychiatry, 18(1), 108. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1186/s12888-018-1696-9"", ""children"": [{""text"": ""https://doi.org/10.1186/s12888-018-1696-9"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As you can see, around a fifth of the US population says they have had a depressed mood for several days in the past two weeks."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But having one or a few symptoms does not necessarily mean that someone can be diagnosed with depression. Instead, researchers look at the combination of symptoms that people report. They will set a threshold for the number of symptoms someone must have before they are considered to have the condition."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To diagnose someone with major depression, for example, the ICD and DSM criteria require them to have had a depressed mood or loss of interest for much of the day, nearly every day, for at least two weeks, along with several other symptoms."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/frequency-depressive-symptoms-us "", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""What are the strengths and limitations of survey data?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""text"": [{""text"": ""Strengths of survey data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Survey data on mental health has two major strengths."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""First, it involves structured interviews – people are asked a consistent set of questions regardless of their interviewer. This helps to ensure the data is more comparable between interviewers and across time.{ref}Mueller, A. E., & Segal, D. L. (2014). Structured versus semistructured versus unstructured interviews. The encyclopedia of clinical psychology, 1-7. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1002/9781118625392.wbecp069"", ""children"": [{""text"": ""https://doi.org/10.1002/9781118625392.wbecp069"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Second, surveys include a much wider range of people in the population, including those who would not seek treatment due to a lack of awareness, costs, or other concerns."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This can help to understand the prevalence of mental illnesses in the general population, including people never diagnosed by a healthcare professional. This can help us understand what share of people seek treatment. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One example is the World Mental Health surveys: these were large-scale surveys of mental illnesses in the general population in 21 countries of different income levels.{ref}Kessler, R. C., Green, J. G., Gruber, M. J., Sampson, N. A., Bromet, E., Cuitan, M., ... & Zaslavsky, A. M. (2010). Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. International journal of methods in psychiatric research, 19(S1), 4-22. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1002/mpr.310"", ""children"": [{""text"": ""https://doi.org/10.1002/mpr.310"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Based on structured interviews, the authors found that around 1 in 10 people met the criteria for an anxiety disorder in the past year across countries surveyed.{ref}Alonso, J., Liu, Z., Evans-Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., Abdulmalik, J., Aguilar-Gaxiola, S., Al-Hamzawi, A., Andrade, L. H., Bruffaerts, R., Cardoso, G., Cia, A., Florescu, S., de Girolamo, G., Gureje, O., Haro, J. M., He, Y., de Jonge, P., … the WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and Anxiety, 35(3), 195–208. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1002/da.22711"", ""children"": [{""text"": ""https://doi.org/10.1002/da.22711"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart, you can see the share of those who met the criteria who also said they had received treatment. Only around 36% with anxiety disorders received any treatment in high-income countries, while an even lower share – only 13% – received it in lower-middle-income countries.{ref}Potentially adequate treatment was defined by the authors as receiving psychotherapy, medication, or complementary alternative medicine.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Third, surveys can tell us about milder symptoms that may be common in the population. This can help to see if a condition lies on a spectrum – where everyone has the symptoms to different degrees – or if it affects a separate share of people."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/share-of-people-with-anxiety-disorders-who-received-treatment "", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Limitations of survey data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""There are also several limitations of survey data to keep in mind."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One is that people may not share their symptoms in surveys, as they may not feel comfortable sharing them with researchers. Along with this, people’s comfort in sharing mental health symptoms may vary between countries and over time.{ref}Gaia, A. (2020). Social Desirability Bias and Sensitive Questions in Surveys. In SAGE Research Methods Foundations. SAGE Publications Ltd. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.4135/9781526421036928979"", ""children"": [{""text"": ""https://doi.org/10.4135/9781526421036928979"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""text"": ""Krosnick, J. A. (1999). Maximizing questionnaire quality. Measures of Political Attitudes, 2, 37–58."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""Shoemaker, P. J. (2002). Item Nonresponse: Distinguishing between don’t Know and Refuse. International Journal of Public Opinion Research, 14(2), 193–201. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1093/ijpor/14.2.193"", ""children"": [{""text"": ""https://doi.org/10.1093/ijpor/14.2.193"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Another limitation is that people may not remember their symptoms, especially when they are asked to recall symptoms in their lifetime so far. This can be challenging for older people whose symptoms may have occurred decades ago."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This chart shows data from a study where the same people were interviewed several times about depression in their lifetimes so far.{ref}Takayanagi, Y., Spira, A. P., Roth, K. B., Gallo, J. J., Eaton, W. W., & Mojtabai, R. (2014). Accuracy of Reports of Lifetime Mental and Physical Disorders: Results From the Baltimore Epidemiological Catchment Area Study. JAMA Psychiatry, 71(3), 273. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1001/jamapsychiatry.2013.3579"", ""children"": [{""text"": ""https://doi.org/10.1001/jamapsychiatry.2013.3579"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref} Around two-thirds of people who had described episodes of depression did not recall them in subsequent interviews."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Lifetime-depression-recall-bias.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A third limitation is that survey data does not usually exclude other diagnoses. When people visit a doctor for a diagnosis, the doctor may ask them about other existing conditions and medications they are taking, and test them for other medical conditions which could also cause their symptoms. This is not usually performed in surveys."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It can also be difficult to compare people’s responses in surveys, because people may interpret the questions differently. This can make it difficult to compare people from different backgrounds, languages, and countries.{ref}Putnick, D. L., & Bornstein, M. H. (2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review, 41, 71–90. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/j.dr.2016.06.004"", ""children"": [{""text"": ""https://doi.org/10.1016/j.dr.2016.06.004"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Another consideration is who is included in surveys. Some include people from a wide range of backgrounds in the general population, while others only include particular groups like urban residents or university students."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This can be a big problem for mental health conditions that are less common, such as schizophrenia and bipolar disorder. If only a few people in the study had a condition, then it can be difficult to estimate the precise share of the total population with it."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It’s also important to know that surveys of the general population usually do not include people who are institutionalized in hospitals or prisons, who may have more severe physical and mental health conditions.{ref}Binswanger, I. A., Krueger, P. M., & Steiner, J. F. (2009). Prevalence of chronic medical conditions among jail and prison inmates in the USA compared with the general population. Journal of Epidemiology & Community Health, 63(11), 912–919. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1136/jech.2009.090662"", ""children"": [{""text"": ""https://doi.org/10.1136/jech.2009.090662"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Peen, J., Schoevers, R. A., Beekman, A. T., & Dekker, J. (2010). The current status of urban-rural differences in psychiatric disorders. Acta Psychiatrica Scandinavica, 121(2), 84–93. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1111/j.1600-0447.2009.01438.x"", ""children"": [{""text"": ""https://doi.org/10.1111/j.1600-0447.2009.01438.x"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Rehm, J., Kilian, C., Rovira, P., Shield, K. D., & Manthey, J. (2021). The elusiveness of representativeness in general population surveys for alcohol. Drug and Alcohol Review, 40(2), 161–165. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1111/dar.13148"", ""children"": [{""text"": ""https://doi.org/10.1111/dar.13148"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How much data on mental health is available around the world?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Data on mental health varies in two ways: the amount of data on each mental illness and the amount of data from each country."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""First, there is much more data available on some mental illnesses than others. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""You can see this in the chart. It shows the number of countries with primary data on the prevalence of each mental illness in the general population. The studies were used by the IHME’s Global Burden of Disease study to estimate the prevalence of mental illnesses worldwide.{ref}Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext"", ""children"": [{""text"": ""https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As you can see, data on some mental illnesses such as attention-deficit hyperactivity disorder, cannabis use disorder, and major depressive disorder came from a large number of countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""However, data on others, such as bipolar disorder, autism spectrum disorders, and anorexia nervosa, came from fewer than 40 countries. For personality disorders, data came from only two countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/number-of-countries-with-primary-data-on-prevalence-of-mental-illnesses-in-the-global-burden-of-disease-study "", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Second, there is much more data available from some world regions than others."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the left-hand chart, you can see the share of the population of world regions that had any data collected on the prevalence of major depression between 1980 and 2008. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This comes from an older study published in 2013, and more data has been collected since then, which has helped improve estimates made by the Global Burden of Disease study.{ref}Baxter, A. J., Patton, G., Scott, K. M., Degenhardt, L., & Whiteford, H. A. (2013). Global Epidemiology of Mental Disorders: What Are We Missing? PLoS ONE, 8(6), e65514. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1371/journal.pone.0065514"", ""children"": [{""text"": ""https://doi.org/10.1371/journal.pone.0065514"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In Australasia and North America, there was data relating to all of the age demographics in the population. But in regions such as Eastern Europe and Southeast Asia, less than 25% of the population was covered. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In several regions, especially in sub-Saharan Africa, Central Asia, and South America, there was almost no underlying data available."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/adult-population-covered-in-primary-data-on-the-prevalence-of-major-depression "", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""What about other mental illnesses apart from depression? "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart, you can see the population that was covered in data on other common mental illnesses. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As you can see, there was more data available on anxiety disorders and major depression than on schizophrenia and bipolar disorder. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""You can click ‘Change country’ to see how this varies between world regions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/adult-population-covered-in-primary-data-on-the-prevalence-of-mental-illnesses "", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""How do researchers extrapolate this data to make comparable estimates?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Researchers can try to make comparable estimates of mental health using this underlying data and statistical methods.{ref}Mathers, C., Hogan, D., & Stevens, G. (2019). Global health estimates: modelling and predicting health outcomes. The Palgrave handbook of global health data methods for policy and practice, 403-424. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://link.springer.com/chapter/10.1057/978-1-137-54984-6_21"", ""children"": [{""text"": ""https://link.springer.com/chapter/10.1057/978-1-137-54984-6_21"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Whiteford, H. A., Degenhardt, L., Rehm, J., Baxter, A. J., Ferrari, A. J., Erskine, H. E., ... & Vos, T. (2013). Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. The Lancet, 382(9904), 1575-1586. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/s0140-6736(13)61611-6"", ""children"": [{""text"": ""https://doi.org/10.1016/s0140-6736(13)61611-6"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Enders, C. K. (2022). Applied missing data analysis. Guilford Publications."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Rehm, J., & Shield, K. D. (2019). Global burden of disease and the impact of mental and addictive disorders. Current psychiatry reports, 21, 1-7. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1007/s11920-019-0997-0"", ""children"": [{""text"": ""https://doi.org/10.1007/s11920-019-0997-0"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/S0140-6736(20)30925-9"", ""children"": [{""text"": ""https://doi.org/10.1016/S0140-6736(20)30925-9"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These methods incorporate the available data on people’s mental health, their demographics, and the level of diagnosis and testing. Then they extrapolate the results to other countries, where data has not been collected. This can be based on demographics such as age and sex, other risk factors, responses to other large-scale representative mental health surveys.{ref}For major depressive disorder, for example, the IHME uses available data on the age structure of the population; risk factors such as war mortality, intimate partner violence, and childhood sexual violence; as well as responses to the Gallup’s survey on negative experiences around the world. You can learn more in the Appendix of the Global Burden of Disease Study, which is available here."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext#supplementaryMaterial"", ""children"": [{""text"": ""https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext#supplementaryMaterial"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""They try to adjust for the fact that the underlying data was collected from different sources (diagnoses or surveys), during different periods, and from different groups of people. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""However, they come with a range of uncertainty. This is because they rely on assumptions about how the data was collected, and why some demographics and countries lacked data on the prevalence of mental illnesses."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is especially true for some illnesses – such as eating disorders and bipolar disorder – and some world regions – including much of Asia, South America and Africa; where primary data is lacking."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Data on global mental health is limited – but nevertheless gives us important insights"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Data on global mental health has two main limitations. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""First, our understanding of global mental health depends on people’s willingness to share their symptoms, and contact healthcare professionals to receive a diagnosis and treatment. Because of this, many people remain undiagnosed and lack support and treatment."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Another major limitation is that data is lacking in many countries. It is often available only for some age groups, and is collected by separate one-off studies at infrequent intervals. There is much less data available for some illnesses than others."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""For countries that lack data, the prevalence of mental illnesses is estimated from other similar countries with data, but this leads to large uncertainties."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Despite these limitations, the available data does give us important insights:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Importantly it shows that mental illnesses are not uncommon. For example, the World Mental Health surveys estimated that one-in-ten people met the criteria for anxiety disorders in the past year, on average, across countries.{ref}Alonso, J., Liu, Z., Evans-Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., Abdulmalik, J., Aguilar-Gaxiola, S., Al-Hamzawi, A., Andrade, L. H., Bruffaerts, R., Cardoso, G., Cia, A., Florescu, S., de Girolamo, G., Gureje, O., Haro, J. M., He, Y., de Jonge, P., … the WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and Anxiety, 35(3), 195–208. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1002/da.22711"", ""children"": [{""text"": ""https://doi.org/10.1002/da.22711"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Global data also tells us that mental illnesses have a large treatment gap, especially in poorer countries. For example, while around one-in-three with anxiety disorders received any treatment in high-income countries, less than one-in-eight did in lower-middle-income countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Mental illnesses are a major part of the global health burden and remain untreated for many people. To address this, countries need more data on these conditions for a wide range of demographics, and long-term data to understand how they develop, and how effective the treatments are."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""children"": [{""text"": ""Keep reading on Our World in Data:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/mental-health"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Acknowledgements:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" I would like to thank Edouard Mathieu, Hannah Ritchie and Max Roser for their helpful suggestions to improve this article."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""How do researchers study the prevalence of mental illnesses?"", ""authors"": [""Saloni Dattani""], ""excerpt"": ""Global data on mental health is essential to understand the scale and patterns of these illnesses, and how to reduce them. How do researchers collect this data, and how reliable is it?"", ""dateline"": ""May 26, 2023"", ""subtitle"": ""Global data on mental health is essential to understand the scale and patterns of these illnesses, and how to reduce them. How do researchers collect this data, and how reliable is it?"", ""sidebar-toc"": false, ""featured-image"": ""Researchers-estimate-mental-health-thumbnail.png""}, ""createdAt"": ""2023-05-26T11:49:27.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T16:26:58.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-05-26T19:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}], ""numBlocks"": 70, ""numErrors"": 6, ""wpTagCounts"": {""html"": 5, ""image"": 1, ""column"": 26, ""columns"": 13, ""heading"": 12, ""paragraph"": 87, ""separator"": 2, ""owid/prominent-link"": 2}, ""htmlTagCounts"": {""p"": 87, ""h3"": 8, ""h4"": 4, ""hr"": 2, ""div"": 39, ""figure"": 1, ""iframe"": 5}}",2023-05-26 19:00:00,2024-02-16 14:22:55,1CKEuPmKy4VP8WTSlF6oOQN761LQyxhaoAzVm6yu5dUE,"[""Saloni Dattani""]","Global data on mental health is essential to understand the scale and patterns of these illnesses, and how to reduce them. How do researchers collect this data, and how reliable is it?",2023-05-26 11:49:27,2023-07-10 16:26:58,https://ourworldindata.org/wp-content/uploads/2023/05/Researchers-estimate-mental-health-thumbnail.png,{},"In many countries, many people with mental illnesses go undiagnosed, meaning mental health is given less attention and support than it deserves. Even for those diagnosed, treatment can be of poor quality, if they receive it at all.{ref}Alonso, J., Liu, Z., Evans‐Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., ... & WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and anxiety, 35(3), 195-208. [https://doi.org/10.1002/da.22711](https://doi.org/10.1002/da.22711) {/ref} To reduce the burden of mental illnesses, the world needs reliable data, which includes the number of people that face mental illnesses, how and when they occur, and the effectiveness of treatments. ## How are mental illnesses defined? Defining mental illnesses is complex. They are diagnosed based on people’s psychological symptoms and behavior rather than biomarkers, brain scans, or blood tests. This makes them more subjective – they are dependent on whether people share their symptoms and the way doctors diagnose them. Mental illnesses are formally defined according to the International Classification of Diseases (ICD) and the Diagnostic and Statistical Manual of Mental Disorders (DSM). The ICD is used internationally by healthcare professionals, while the DSM is primarily used by psychiatrists in the United States.{ref}Kupfer, D. J., Regier, D. A., & Kuhl, E. A. (2008). On the road to DSM-V and ICD-11. European Archives of Psychiatry and Clinical Neuroscience, 258(S5), 2–6. [https://doi.org/10.1007/s00406-008-5002-6](https://doi.org/10.1007/s00406-008-5002-6) {/ref} These manuals explain how to diagnose mental illnesses by observing and asking about people’s symptoms and behavior, and the context of their symptoms – for example, symptoms that appeared because of drug use or other medical conditions don’t qualify as mental illnesses. Based on these definitions, healthcare professionals can make diagnoses, which can be used for healthcare, treatment, and national statistics. Over time, the definitions of particular mental illnesses have changed. The DSM has been revised 5 times since it was first developed in 1952, while the ICD has been revised 11 times since 1900.{ref}American Psychiatric Association. (2022). DSM History. [https://www.psychiatry.org/psychiatrists/practice/dsm/about-dsm/history-of-the-dsm](https://www.psychiatry.org/psychiatrists/practice/dsm/about-dsm/history-of-the-dsm) Hirsch, J. A., Nicola, G., McGinty, G., Liu, R. W., Barr, R. M., Chittle, M. D., & Manchikanti, L. (2016). ICD-10: History and Context. AJNR. American Journal of Neuroradiology, 37(4), 596–599. [https://doi.org/10.3174/ajnr.A4696](https://doi.org/10.3174/ajnr.A4696) {/ref} They will continue to be revised in the future, but updates have become less frequent. Their changes are partly due to a better understanding and measurement of mental illnesses. They have also changed as a result of cultural and legal factors. There used to be larger differences in the criteria for diagnosing mental illnesses between the ICD and the DSM, but the two manuals are now more similar due to collaboration between their developers.{ref}This paper provides a detailed summary of the similarities and differences between the criteria for mental illnesses in the ICD-11 and DSM-5. First, M. B., Gaebel, W., Maj, M., Stein, D. J., Kogan, C. S., Saunders, J. B., Poznyak, V. B., Gureje, O., Lewis‐Fernández, R., Maercker, A., Brewin, C. R., Cloitre, M., Claudino, A., Pike, K. M., Baird, G., Skuse, D., Krueger, R. B., Briken, P., Burke, J. D., … Reed, G. M. (2021). An organization‐ and category‐level comparison of diagnostic requirements for mental disorders in ICD ‐11 and DSM ‐5. World Psychiatry, 20(1), 34–51. [https://doi.org/10.1002/wps.20825](https://doi.org/10.1002/wps.20825) Harrison, J. E., Weber, S., Jakob, R., & Chute, C. G. (2021). ICD-11: An international classification of diseases for the twenty-first century. BMC Medical Informatics and Decision Making, 21(S6), 206. [https://doi.org/10.1186/s12911-021-01534-6](https://doi.org/10.1186/s12911-021-01534-6) Tyrer, P. (2014). A comparison of DSM and ICD classifications of mental disorder. Advances in Psychiatric Treatment, 20(4), 280–285. [https://doi.org/10.1192/apt.bp.113.011296](https://doi.org/10.1192/apt.bp.113.011296) {/ref} ## Mental health data based on diagnoses The process of diagnosing a mental illness usually starts with the patient consulting a healthcare professional.{ref}People can also be diagnosed through other routes. For example, in some countries, there are also screening programs to identify people who may have mental health conditions and refer them to specialists. Children and adolescents can be referred to healthcare professionals by carers. In some countries, people can also be diagnosed and committed to mental hospitals involuntarily if they are considered to pose a danger to others. This was more common in countries like the United States before the 1960s, but since then, these laws have been reformed in many countries. See also: Testa, M., & West, S. G. (2010). Civil commitment in the United States. Psychiatry (Edgmont (Pa.: Township)), 7(10), 30–40. Zhang, S., Mellsop, G., Brink, J., & Wang, X. (2015). Involuntary admission and treatment of patients with mental disorder. Neuroscience Bulletin, 31(1), 99–112. [https://doi.org/10.1007/s12264-014-1493-5](https://doi.org/10.1007/s12264-014-1493-5) Appelbaum, P. S. (1997). Almost a revolution: an international perspective on the law of involuntary commitment. Journal of the American Academy of Psychiatry and the Law Online, 25(2), 135-147.{/ref} This can be a doctor in a clinic or general hospital, a psychiatrist, or another mental health specialist. Health professionals use official medical guidance and professional judgment to decide whether to diagnose a patient with a condition. Data on these diagnoses are collected from hospitals in many countries, but this may not include clinic visits. The data can include people’s age and sex, their reason for admission, other diagnoses, and treatments given during their visit.{ref} Otero Varela, L., Doktorchik, C., Wiebe, N., Quan, H., & Eastwood, C. (2021). Exploring the differences in ICD and hospital morbidity data collection features across countries: An international survey. BMC Health Services Research, 21(1), 308. [https://doi.org/10.1186/s12913-021-06302-w](https://doi.org/10.1186/s12913-021-06302-w) Varela, L. O., Knudsen, S., Carpendale, S., Eastwood, C., & Quan, H. (2019, October). Comparing ICD-Data Across Countries: A Case for Visualization?. In 2019 IEEE Workshop on Visual Analytics in Healthcare (VAHC) (pp. 60-61). IEEE. {/ref} ## What are the strengths and limitations of diagnosis data? ### Strengths of data based on diagnosis Official data from diagnoses of mental illnesses have two major strengths. First, the diagnoses come from healthcare professionals with training and experience in recognizing mental illnesses. They can use their knowledge to ask people more questions about their symptoms and understand their context before making a diagnosis. They can also perform additional medical tests to rule out other conditions. Second, data on diagnoses can tell us about the number of people who seek out mental health treatment from public hospitals and clinics. This can usually be linked to data on which treatments they were prescribed and for how long.{ref}In some countries, there are also national screening programs to diagnose patients with both physical and mental illnesses.{/ref} This can be very useful for countries to understand the resources used to treat mental illnesses. ### Limitations of data based on diagnosis Data on diagnoses of mental illnesses also has limitations. One problem is that many people do not reach out to healthcare professionals about their health conditions. This might be because they lack awareness of mental illnesses or there is a lack of healthcare for these conditions in their country. They may also feel uncomfortable about sharing their symptoms with healthcare professionals. Another problem is that the diagnoses may not be made consistently. Doctors can have different levels of training and experience in recognizing and diagnosing mental illnesses. This can be a problem when comparing diagnosis rates between countries. A third problem is that national data on diagnosis is often collected from a limited number of sources. Data from private hospitals and clinics is usually not included. In some countries, mental health data is collected from hospitals but not clinics.{ref}Baxter, A. J., Patton, G., Scott, K. M., Degenhardt, L., & Whiteford, H. A. (2013). Global Epidemiology of Mental Disorders: What Are We Missing? PLoS ONE, 8(6), e65514. [https://doi.org/10.1371/journal.pone.0065514](https://doi.org/10.1371/journal.pone.0065514) {/ref} Some countries also use data from other sources. For example, the data might come from health insurance claims—which include private healthcare—or from other databases connected to many healthcare clinics across the country. The final problem is that countries may use different definitions to diagnose patients. Some countries use modified versions of the ICD manual to collect their data, depending on their cultural context and needs.{ref}Brhlikova, P., Pollock, A. M., & Manners, R. (2011). Global Burden of Disease estimates of depression – how reliable is the epidemiological evidence? Journal of the Royal Society of Medicine, 104(1), 25–34. [https://doi.org/10.1258/jrsm.2010.100080](https://doi.org/10.1258/jrsm.2010.100080) {/ref} These differences mean that comparing data between countries using data on diagnoses can be difficult. ## Mental health data based on surveys Aside from the formal diagnoses, mental illnesses can be measured using surveys and screening questionnaires. These tend to ask people about symptoms similar to those in diagnostic manuals, but they can be used more easily and widely because the data does not need to be collected by a healthcare professional. These surveys can be conducted in different ways: over the phone, online, or in-person while anonymized.{ref}In-person surveys are usually anonymized for sensitive questions such as those on mental health. In this case, the person may be given a laptop to answer the questions while the interviewer cannot see their answers.{/ref} During the surveys, trained professionals ask people in the general population whether they have experienced symptoms of mental illnesses. They also ask about the age when people first experienced them, how long they lasted, and how severe the symptoms were. People may be asked about symptoms they have currently, or have had recently, or in their lifetime so far. Some symptoms of mental illnesses may be common in the population. This chart shows data from a US survey called the National Health and Nutrition Examination Survey. It was a large-scale, in-person survey of people in the general population.{ref}Tomitaka, S., Kawasaki, Y., Ide, K., Akutagawa, M., Yamada, H., Ono, Y., & Furukawa, T. A. (2018). Distributional patterns of item responses and total scores on the PHQ-9 in the general population: Data from the National Health and Nutrition Examination Survey. BMC Psychiatry, 18(1), 108. [https://doi.org/10.1186/s12888-018-1696-9](https://doi.org/10.1186/s12888-018-1696-9) {/ref} As you can see, around a fifth of the US population says they have had a depressed mood for several days in the past two weeks. But having one or a few symptoms does not necessarily mean that someone can be diagnosed with depression. Instead, researchers look at the combination of symptoms that people report. They will set a threshold for the number of symptoms someone must have before they are considered to have the condition. To diagnose someone with major depression, for example, the ICD and DSM criteria require them to have had a depressed mood or loss of interest for much of the day, nearly every day, for at least two weeks, along with several other symptoms.In many countries, many people with mental illnesses go undiagnosed, meaning mental health is given less attention and support than it deserves. Even for those diagnosed, treatment can be of poor quality, if they receive it at all.{ref}Alonso, J., Liu, Z., Evans‐Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., … & WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and anxiety, 35(3), 195-208. https://doi.org/10.1002/da.22711 {/ref}
\n\n\n\nTo reduce the burden of mental illnesses, the world needs reliable data, which includes the number of people that face mental illnesses, how and when they occur, and the effectiveness of treatments.
\n\n\n\nDefining mental illnesses is complex. They are diagnosed based on people’s psychological symptoms and behavior rather than biomarkers, brain scans, or blood tests. This makes them more subjective – they are dependent on whether people share their symptoms and the way doctors diagnose them.
\n\n\n\nMental illnesses are formally defined according to the International Classification of Diseases (ICD) and the Diagnostic and Statistical Manual of Mental Disorders (DSM). The ICD is used internationally by healthcare professionals, while the DSM is primarily used by psychiatrists in the United States.{ref}Kupfer, D. J., Regier, D. A., & Kuhl, E. A. (2008). On the road to DSM-V and ICD-11. European Archives of Psychiatry and Clinical Neuroscience, 258(S5), 2–6. https://doi.org/10.1007/s00406-008-5002-6 {/ref}
\n\n\n\nThese manuals explain how to diagnose mental illnesses by observing and asking about people’s symptoms and behavior, and the context of their symptoms – for example, symptoms that appeared because of drug use or other medical conditions don’t qualify as mental illnesses.
\n\n\n\nBased on these definitions, healthcare professionals can make diagnoses, which can be used for healthcare, treatment, and national statistics.
\n\n\n\nOver time, the definitions of particular mental illnesses have changed. The DSM has been revised 5 times since it was first developed in 1952, while the ICD has been revised 11 times since 1900.{ref}American Psychiatric Association. (2022). DSM History. https://www.psychiatry.org/psychiatrists/practice/dsm/about-dsm/history-of-the-dsm
Hirsch, J. A., Nicola, G., McGinty, G., Liu, R. W., Barr, R. M., Chittle, M. D., & Manchikanti, L. (2016). ICD-10: History and Context. AJNR. American Journal of Neuroradiology, 37(4), 596–599. https://doi.org/10.3174/ajnr.A4696 {/ref} They will continue to be revised in the future, but updates have become less frequent.
Their changes are partly due to a better understanding and measurement of mental illnesses. They have also changed as a result of cultural and legal factors. There used to be larger differences in the criteria for diagnosing mental illnesses between the ICD and the DSM, but the two manuals are now more similar due to collaboration between their developers.{ref}This paper provides a detailed summary of the similarities and differences between the criteria for mental illnesses in the ICD-11 and DSM-5.
First, M. B., Gaebel, W., Maj, M., Stein, D. J., Kogan, C. S., Saunders, J. B., Poznyak, V. B., Gureje, O., Lewis‐Fernández, R., Maercker, A., Brewin, C. R., Cloitre, M., Claudino, A., Pike, K. M., Baird, G., Skuse, D., Krueger, R. B., Briken, P., Burke, J. D., … Reed, G. M. (2021). An organization‐ and category‐level comparison of diagnostic requirements for mental disorders in ICD ‐11 and DSM ‐5. World Psychiatry, 20(1), 34–51. https://doi.org/10.1002/wps.20825
Harrison, J. E., Weber, S., Jakob, R., & Chute, C. G. (2021). ICD-11: An international classification of diseases for the twenty-first century. BMC Medical Informatics and Decision Making, 21(S6), 206. https://doi.org/10.1186/s12911-021-01534-6
Tyrer, P. (2014). A comparison of DSM and ICD classifications of mental disorder. Advances in Psychiatric Treatment, 20(4), 280–285. https://doi.org/10.1192/apt.bp.113.011296 {/ref}
The process of diagnosing a mental illness usually starts with the patient consulting a healthcare professional.{ref}People can also be diagnosed through other routes. For example, in some countries, there are also screening programs to identify people who may have mental health conditions and refer them to specialists. Children and adolescents can be referred to healthcare professionals by carers.
In some countries, people can also be diagnosed and committed to mental hospitals involuntarily if they are considered to pose a danger to others. This was more common in countries like the United States before the 1960s, but since then, these laws have been reformed in many countries.
See also: Testa, M., & West, S. G. (2010). Civil commitment in the United States. Psychiatry (Edgmont (Pa.: Township)), 7(10), 30–40.
Zhang, S., Mellsop, G., Brink, J., & Wang, X. (2015). Involuntary admission and treatment of patients with mental disorder. Neuroscience Bulletin, 31(1), 99–112. https://doi.org/10.1007/s12264-014-1493-5
Appelbaum, P. S. (1997). Almost a revolution: an international perspective on the law of involuntary commitment. Journal of the American Academy of Psychiatry and the Law Online, 25(2), 135-147.{/ref} This can be a doctor in a clinic or general hospital, a psychiatrist, or another mental health specialist.
Health professionals use official medical guidance and professional judgment to decide whether to diagnose a patient with a condition.
\n\n\n\nData on these diagnoses are collected from hospitals in many countries, but this may not include clinic visits. The data can include people’s age and sex, their reason for admission, other diagnoses, and treatments given during their visit.{ref} Otero Varela, L., Doktorchik, C., Wiebe, N., Quan, H., & Eastwood, C. (2021). Exploring the differences in ICD and hospital morbidity data collection features across countries: An international survey. BMC Health Services Research, 21(1), 308. https://doi.org/10.1186/s12913-021-06302-w
\n\n\n\nVarela, L. O., Knudsen, S., Carpendale, S., Eastwood, C., & Quan, H. (2019, October). Comparing ICD-Data Across Countries: A Case for Visualization?. In 2019 IEEE Workshop on Visual Analytics in Healthcare (VAHC) (pp. 60-61). IEEE. {/ref}
\nOfficial data from diagnoses of mental illnesses have two major strengths.
\n\n\n\nFirst, the diagnoses come from healthcare professionals with training and experience in recognizing mental illnesses.
\n\n\n\nThey can use their knowledge to ask people more questions about their symptoms and understand their context before making a diagnosis. They can also perform additional medical tests to rule out other conditions.
\n\n\n\nSecond, data on diagnoses can tell us about the number of people who seek out mental health treatment from public hospitals and clinics. This can usually be linked to data on which treatments they were prescribed and for how long.{ref}In some countries, there are also national screening programs to diagnose patients with both physical and mental illnesses.{/ref}
\n\n\n\nThis can be very useful for countries to understand the resources used to treat mental illnesses.
\nData on diagnoses of mental illnesses also has limitations.
\n\n\n\nOne problem is that many people do not reach out to healthcare professionals about their health conditions. This might be because they lack awareness of mental illnesses or there is a lack of healthcare for these conditions in their country. They may also feel uncomfortable about sharing their symptoms with healthcare professionals.
\n\n\n\nAnother problem is that the diagnoses may not be made consistently. Doctors can have different levels of training and experience in recognizing and diagnosing mental illnesses. This can be a problem when comparing diagnosis rates between countries.
\n\n\n\nA third problem is that national data on diagnosis is often collected from a limited number of sources. Data from private hospitals and clinics is usually not included. In some countries, mental health data is collected from hospitals but not clinics.{ref}Baxter, A. J., Patton, G., Scott, K. M., Degenhardt, L., & Whiteford, H. A. (2013). Global Epidemiology of Mental Disorders: What Are We Missing? PLoS ONE, 8(6), e65514. https://doi.org/10.1371/journal.pone.0065514 {/ref}
\n\n\n\nSome countries also use data from other sources. For example, the data might come from health insurance claims—which include private healthcare—or from other databases connected to many healthcare clinics across the country.
\n\n\n\nThe final problem is that countries may use different definitions to diagnose patients. Some countries use modified versions of the ICD manual to collect their data, depending on their cultural context and needs.{ref}Brhlikova, P., Pollock, A. M., & Manners, R. (2011). Global Burden of Disease estimates of depression – how reliable is the epidemiological evidence? Journal of the Royal Society of Medicine, 104(1), 25–34. https://doi.org/10.1258/jrsm.2010.100080 {/ref}
\n\n\n\nThese differences mean that comparing data between countries using data on diagnoses can be difficult.
\nAside from the formal diagnoses, mental illnesses can be measured using surveys and screening questionnaires.
\n\n\n\nThese tend to ask people about symptoms similar to those in diagnostic manuals, but they can be used more easily and widely because the data does not need to be collected by a healthcare professional.
\n\n\n\nThese surveys can be conducted in different ways: over the phone, online, or in-person while anonymized.{ref}In-person surveys are usually anonymized for sensitive questions such as those on mental health. In this case, the person may be given a laptop to answer the questions while the interviewer cannot see their answers.{/ref}
\n\n\n\nDuring the surveys, trained professionals ask people in the general population whether they have experienced symptoms of mental illnesses. They also ask about the age when people first experienced them, how long they lasted, and how severe the symptoms were.
\n\n\n\nPeople may be asked about symptoms they have currently, or have had recently, or in their lifetime so far.
\n\n\n\nSome symptoms of mental illnesses may be common in the population. This chart shows data from a US survey called the National Health and Nutrition Examination Survey. It was a large-scale, in-person survey of people in the general population.{ref}Tomitaka, S., Kawasaki, Y., Ide, K., Akutagawa, M., Yamada, H., Ono, Y., & Furukawa, T. A. (2018). Distributional patterns of item responses and total scores on the PHQ-9 in the general population: Data from the National Health and Nutrition Examination Survey. BMC Psychiatry, 18(1), 108. https://doi.org/10.1186/s12888-018-1696-9 {/ref}
\n\n\n\nAs you can see, around a fifth of the US population says they have had a depressed mood for several days in the past two weeks.
\n\n\n\nBut having one or a few symptoms does not necessarily mean that someone can be diagnosed with depression. Instead, researchers look at the combination of symptoms that people report. They will set a threshold for the number of symptoms someone must have before they are considered to have the condition.
\n\n\n\nTo diagnose someone with major depression, for example, the ICD and DSM criteria require them to have had a depressed mood or loss of interest for much of the day, nearly every day, for at least two weeks, along with several other symptoms.
\nSurvey data on mental health has two major strengths.
\n\n\n\nFirst, it involves structured interviews – people are asked a consistent set of questions regardless of their interviewer. This helps to ensure the data is more comparable between interviewers and across time.{ref}Mueller, A. E., & Segal, D. L. (2014). Structured versus semistructured versus unstructured interviews. The encyclopedia of clinical psychology, 1-7. https://doi.org/10.1002/9781118625392.wbecp069 {/ref}
\n\n\n\nSecond, surveys include a much wider range of people in the population, including those who would not seek treatment due to a lack of awareness, costs, or other concerns.
\n\n\n\nThis can help to understand the prevalence of mental illnesses in the general population, including people never diagnosed by a healthcare professional. This can help us understand what share of people seek treatment.
\n\n\n\nOne example is the World Mental Health surveys: these were large-scale surveys of mental illnesses in the general population in 21 countries of different income levels.{ref}Kessler, R. C., Green, J. G., Gruber, M. J., Sampson, N. A., Bromet, E., Cuitan, M., … & Zaslavsky, A. M. (2010). Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. International journal of methods in psychiatric research, 19(S1), 4-22. https://doi.org/10.1002/mpr.310 {/ref}
\n\n\n\nBased on structured interviews, the authors found that around 1 in 10 people met the criteria for an anxiety disorder in the past year across countries surveyed.{ref}Alonso, J., Liu, Z., Evans-Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., Abdulmalik, J., Aguilar-Gaxiola, S., Al-Hamzawi, A., Andrade, L. H., Bruffaerts, R., Cardoso, G., Cia, A., Florescu, S., de Girolamo, G., Gureje, O., Haro, J. M., He, Y., de Jonge, P., … the WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and Anxiety, 35(3), 195–208. https://doi.org/10.1002/da.22711 {/ref}
\n\n\n\nIn the chart, you can see the share of those who met the criteria who also said they had received treatment. Only around 36% with anxiety disorders received any treatment in high-income countries, while an even lower share – only 13% – received it in lower-middle-income countries.{ref}Potentially adequate treatment was defined by the authors as receiving psychotherapy, medication, or complementary alternative medicine.{/ref}
\n\n\n\nThird, surveys can tell us about milder symptoms that may be common in the population. This can help to see if a condition lies on a spectrum – where everyone has the symptoms to different degrees – or if it affects a separate share of people.
\nThere are also several limitations of survey data to keep in mind.
\n\n\n\nOne is that people may not share their symptoms in surveys, as they may not feel comfortable sharing them with researchers. Along with this, people’s comfort in sharing mental health symptoms may vary between countries and over time.{ref}Gaia, A. (2020). Social Desirability Bias and Sensitive Questions in Surveys. In SAGE Research Methods Foundations. SAGE Publications Ltd. https://doi.org/10.4135/9781526421036928979
Krosnick, J. A. (1999). Maximizing questionnaire quality. Measures of Political Attitudes, 2, 37–58.
Shoemaker, P. J. (2002). Item Nonresponse: Distinguishing between don’t Know and Refuse. International Journal of Public Opinion Research, 14(2), 193–201. https://doi.org/10.1093/ijpor/14.2.193 {/ref}
Another limitation is that people may not remember their symptoms, especially when they are asked to recall symptoms in their lifetime so far. This can be challenging for older people whose symptoms may have occurred decades ago.
\n\n\n\nThis chart shows data from a study where the same people were interviewed several times about depression in their lifetimes so far.{ref}Takayanagi, Y., Spira, A. P., Roth, K. B., Gallo, J. J., Eaton, W. W., & Mojtabai, R. (2014). Accuracy of Reports of Lifetime Mental and Physical Disorders: Results From the Baltimore Epidemiological Catchment Area Study. JAMA Psychiatry, 71(3), 273. https://doi.org/10.1001/jamapsychiatry.2013.3579 {/ref} Around two-thirds of people who had described episodes of depression did not recall them in subsequent interviews.
\n![\""\""](\""https://owid.cloud/app/uploads/2023/05/Lifetime-depression-recall-bias.png\"")
A third limitation is that survey data does not usually exclude other diagnoses. When people visit a doctor for a diagnosis, the doctor may ask them about other existing conditions and medications they are taking, and test them for other medical conditions which could also cause their symptoms. This is not usually performed in surveys.
\n\n\n\nIt can also be difficult to compare people’s responses in surveys, because people may interpret the questions differently. This can make it difficult to compare people from different backgrounds, languages, and countries.{ref}Putnick, D. L., & Bornstein, M. H. (2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review, 41, 71–90. https://doi.org/10.1016/j.dr.2016.06.004 {/ref}
\n\n\n\nAnother consideration is who is included in surveys. Some include people from a wide range of backgrounds in the general population, while others only include particular groups like urban residents or university students.
\n\n\n\nThis can be a big problem for mental health conditions that are less common, such as schizophrenia and bipolar disorder. If only a few people in the study had a condition, then it can be difficult to estimate the precise share of the total population with it.
\n\n\n\nIt’s also important to know that surveys of the general population usually do not include people who are institutionalized in hospitals or prisons, who may have more severe physical and mental health conditions.{ref}Binswanger, I. A., Krueger, P. M., & Steiner, J. F. (2009). Prevalence of chronic medical conditions among jail and prison inmates in the USA compared with the general population. Journal of Epidemiology & Community Health, 63(11), 912–919. https://doi.org/10.1136/jech.2009.090662
\n\n\n\nPeen, J., Schoevers, R. A., Beekman, A. T., & Dekker, J. (2010). The current status of urban-rural differences in psychiatric disorders. Acta Psychiatrica Scandinavica, 121(2), 84–93. https://doi.org/10.1111/j.1600-0447.2009.01438.x
\n\n\n\nRehm, J., Kilian, C., Rovira, P., Shield, K. D., & Manthey, J. (2021). The elusiveness of representativeness in general population surveys for alcohol. Drug and Alcohol Review, 40(2), 161–165. https://doi.org/10.1111/dar.13148 {/ref}
\nData on mental health varies in two ways: the amount of data on each mental illness and the amount of data from each country.
\n\n\n\nFirst, there is much more data available on some mental illnesses than others.
\n\n\n\nYou can see this in the chart. It shows the number of countries with primary data on the prevalence of each mental illness in the general population. The studies were used by the IHME’s Global Burden of Disease study to estimate the prevalence of mental illnesses worldwide.{ref}Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext {/ref}
\n\n\n\nAs you can see, data on some mental illnesses such as attention-deficit hyperactivity disorder, cannabis use disorder, and major depressive disorder came from a large number of countries.
\n\n\n\nHowever, data on others, such as bipolar disorder, autism spectrum disorders, and anorexia nervosa, came from fewer than 40 countries. For personality disorders, data came from only two countries.
\nSecond, there is much more data available from some world regions than others.
\n\n\n\nIn the left-hand chart, you can see the share of the population of world regions that had any data collected on the prevalence of major depression between 1980 and 2008.
\n\n\n\nThis comes from an older study published in 2013, and more data has been collected since then, which has helped improve estimates made by the Global Burden of Disease study.{ref}Baxter, A. J., Patton, G., Scott, K. M., Degenhardt, L., & Whiteford, H. A. (2013). Global Epidemiology of Mental Disorders: What Are We Missing? PLoS ONE, 8(6), e65514. https://doi.org/10.1371/journal.pone.0065514 {/ref}
\n\n\n\nIn Australasia and North America, there was data relating to all of the age demographics in the population. But in regions such as Eastern Europe and Southeast Asia, less than 25% of the population was covered.
\n\n\n\nIn several regions, especially in sub-Saharan Africa, Central Asia, and South America, there was almost no underlying data available.
\nWhat about other mental illnesses apart from depression?
\n\n\n\nIn the chart, you can see the population that was covered in data on other common mental illnesses.
\n\n\n\nAs you can see, there was more data available on anxiety disorders and major depression than on schizophrenia and bipolar disorder.
\n\n\n\nYou can click ‘Change country’ to see how this varies between world regions.
\nResearchers can try to make comparable estimates of mental health using this underlying data and statistical methods.{ref}Mathers, C., Hogan, D., & Stevens, G. (2019). Global health estimates: modelling and predicting health outcomes. The Palgrave handbook of global health data methods for policy and practice, 403-424. https://link.springer.com/chapter/10.1057/978-1-137-54984-6_21
\n\n\n\nWhiteford, H. A., Degenhardt, L., Rehm, J., Baxter, A. J., Ferrari, A. J., Erskine, H. E., … & Vos, T. (2013). Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. The Lancet, 382(9904), 1575-1586. https://doi.org/10.1016/s0140-6736(13)61611-6
\n\n\n\nEnders, C. K. (2022). Applied missing data analysis. Guilford Publications.
\n\n\n\nRehm, J., & Shield, K. D. (2019). Global burden of disease and the impact of mental and addictive disorders. Current psychiatry reports, 21, 1-7. https://doi.org/10.1007/s11920-019-0997-0
\n\n\n\nVos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. https://doi.org/10.1016/S0140-6736(20)30925-9 {/ref}
\n\n\n\nThese methods incorporate the available data on people’s mental health, their demographics, and the level of diagnosis and testing. Then they extrapolate the results to other countries, where data has not been collected. This can be based on demographics such as age and sex, other risk factors, responses to other large-scale representative mental health surveys.{ref}For major depressive disorder, for example, the IHME uses available data on the age structure of the population; risk factors such as war mortality, intimate partner violence, and childhood sexual violence; as well as responses to the Gallup’s survey on negative experiences around the world. You can learn more in the Appendix of the Global Burden of Disease Study, which is available here.
Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari, M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I., Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., Abu-Raddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30925-9/fulltext#supplementaryMaterial {/ref}
They try to adjust for the fact that the underlying data was collected from different sources (diagnoses or surveys), during different periods, and from different groups of people.
\n\n\n\nHowever, they come with a range of uncertainty. This is because they rely on assumptions about how the data was collected, and why some demographics and countries lacked data on the prevalence of mental illnesses.
\n\n\n\nThis is especially true for some illnesses – such as eating disorders and bipolar disorder – and some world regions – including much of Asia, South America and Africa; where primary data is lacking.
\nData on global mental health has two main limitations.
\n\n\n\nFirst, our understanding of global mental health depends on people’s willingness to share their symptoms, and contact healthcare professionals to receive a diagnosis and treatment. Because of this, many people remain undiagnosed and lack support and treatment.
\n\n\n\nAnother major limitation is that data is lacking in many countries. It is often available only for some age groups, and is collected by separate one-off studies at infrequent intervals. There is much less data available for some illnesses than others.
\n\n\n\nFor countries that lack data, the prevalence of mental illnesses is estimated from other similar countries with data, but this leads to large uncertainties.
\n\n\n\nDespite these limitations, the available data does give us important insights:
\n\n\n\nImportantly it shows that mental illnesses are not uncommon. For example, the World Mental Health surveys estimated that one-in-ten people met the criteria for anxiety disorders in the past year, on average, across countries.{ref}Alonso, J., Liu, Z., Evans-Lacko, S., Sadikova, E., Sampson, N., Chatterji, S., Abdulmalik, J., Aguilar-Gaxiola, S., Al-Hamzawi, A., Andrade, L. H., Bruffaerts, R., Cardoso, G., Cia, A., Florescu, S., de Girolamo, G., Gureje, O., Haro, J. M., He, Y., de Jonge, P., … the WHO World Mental Health Survey Collaborators. (2018). Treatment gap for anxiety disorders is global: Results of the World Mental Health Surveys in 21 countries. Depression and Anxiety, 35(3), 195–208. https://doi.org/10.1002/da.22711 {/ref}
\n\n\n\nGlobal data also tells us that mental illnesses have a large treatment gap, especially in poorer countries. For example, while around one-in-three with anxiety disorders received any treatment in high-income countries, less than one-in-eight did in lower-middle-income countries.
\n\n\n\nMental illnesses are a major part of the global health burden and remain untreated for many people. To address this, countries need more data on these conditions for a wide range of demographics, and long-term data to understand how they develop, and how effective the treatments are.
\nKeep reading on Our World in Data:
\n\n\nAcknowledgements: I would like to thank Edouard Mathieu, Hannah Ritchie and Max Roser for their helpful suggestions to improve this article.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Global data on mental health is essential to understand the scale and patterns of these illnesses, and how to reduce them. How do researchers collect this data, and how reliable is it?"", ""protected"": false}, ""date_gmt"": ""2023-05-26T19:00:00"", ""modified"": ""2023-07-10T17:26:58"", ""template"": """", ""categories"": [46], ""ping_status"": ""closed"", ""authors_name"": [""Saloni Dattani""], ""modified_gmt"": ""2023-07-10T16:26:58"", ""comment_status"": ""closed"", ""featured_media"": 57185, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/05/Researchers-estimate-mental-health-thumbnail-150x59.png"", ""medium_large"": ""/app/uploads/2023/05/Researchers-estimate-mental-health-thumbnail-768x303.png""}}" 56984,We published a new topic page on Influenza,influenza-launch,post,publish,,"{""id"": ""wp-56984"", ""slug"": ""influenza-launch"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a new topic page on Influenza"", ""authors"": [""Saloni Dattani"", ""Fiona Spooner""], ""excerpt"": ""Explore global data on influenza cases, deaths, and vaccinations."", ""dateline"": ""May 18, 2023"", ""subtitle"": ""Explore global data on influenza cases, deaths, and vaccinations."", ""sidebar-toc"": false, ""featured-image"": ""flu-topic-page-thumbnail.png""}, ""createdAt"": ""2023-05-17T09:16:57.000Z"", ""published"": false, ""updatedAt"": ""2023-05-18T09:25:21.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-05-18T08:16:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-05-18 08:16:00,2024-02-16 14:22:55,,"[""Saloni Dattani"", ""Fiona Spooner""]","Explore global data on influenza cases, deaths, and vaccinations.",2023-05-17 09:16:57,2023-05-18 09:25:21,https://ourworldindata.org/wp-content/uploads/2023/05/flu-topic-page-thumbnail.png,{},,"{""id"": 56984, ""date"": ""2023-05-18T09:16:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=56984""}, ""link"": ""https://owid.cloud/influenza-launch"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""influenza-launch"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a new topic page on Influenza""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56984""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/47"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=56984"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=56984"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=56984"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=56984""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56984/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/56858"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 56988, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56984/revisions/56988""}]}, ""author"": 47, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore global data on influenza cases, deaths, and vaccinations."", ""protected"": false}, ""date_gmt"": ""2023-05-18T08:16:00"", ""modified"": ""2023-05-18T10:25:21"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Saloni Dattani"", ""Fiona Spooner""], ""modified_gmt"": ""2023-05-18T09:25:21"", ""comment_status"": ""closed"", ""featured_media"": 56858, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/05/flu-topic-page-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2023/05/flu-topic-page-thumbnail-768x403.png""}}" 56935,LGBT+ Rights,lgbt-rights,page,publish,"This page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview
LGBT+ rights are much more protected than in the past, but are often a lot less protected than those of straight and cisgender people.
\n""}}, ""slug"": ""lgbt-rights"", ""tags"": [251], ""type"": ""page"", ""title"": {""rendered"": ""LGBT+ Rights""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56935""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/page""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=56935"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=56935"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=56935"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=56935""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56935/revisions"", ""count"": 13}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57050"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57900, ""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56935/revisions/57900""}]}, ""author"": 49, ""parent"": 0, ""status"": ""publish"", ""content"": {""rendered"": ""\n\nThis page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/1RvMWl3OpP6zrygM49jcvI6SotoWWq5er14gLBnPLxUg/preview
\nThis page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/195wtinOB2zsiGevmvUfAGPZv4NrG4jY_vpib49e5gt4/preview
","{""id"": ""wp-56926"", ""slug"": ""women-rights"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""This page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/195wtinOB2zsiGevmvUfAGPZv4NrG4jY_vpib49e5gt4/preview"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Women's Rights"", ""authors"": [""Bastian Herre"", ""Pablo Arriagada"", ""Esteban Ortiz-Ospina"", ""Hannah Ritchie"", ""Joe Hasell"", ""Max Roser""], ""excerpt"": ""How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data and research on women’s rights."", ""dateline"": ""May 12, 2023"", ""subtitle"": ""How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data and research on women’s rights."", ""sidebar-toc"": false, ""featured-image"": ""Womens-Rights.png""}, ""createdAt"": ""2023-05-12T10:03:09.000Z"", ""published"": false, ""updatedAt"": ""2023-07-18T21:12:37.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-05-12T09:03:25.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 1, ""numErrors"": 0, ""wpTagCounts"": {""paragraph"": 1}, ""htmlTagCounts"": {""p"": 1}}",2023-05-12 09:03:25,2024-02-16 14:22:42,,"[""Bastian Herre"", ""Pablo Arriagada"", ""Esteban Ortiz-Ospina"", ""Hannah Ritchie"", ""Joe Hasell""]",How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data and research on women’s rights.,2023-05-12 10:03:09,2023-07-18 21:12:37,https://ourworldindata.org/wp-content/uploads/2023/05/Womens-Rights.png,{},This page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/195wtinOB2zsiGevmvUfAGPZv4NrG4jY_vpib49e5gt4/preview,"{""id"": 56926, ""date"": ""2023-05-12T10:03:25"", ""guid"": {""rendered"": ""https://owid.cloud/?page_id=56926""}, ""link"": ""https://owid.cloud/women-rights"", ""meta"": {""owid_publication_context_meta_field"": [], ""owid_key_performance_indicators_meta_field"": {""raw"": ""Women's rights are much more protected than in the past, but are often a lot less protected than those of men."", ""rendered"": ""Women's rights are much more protected than in the past, but are often a lot less protected than those of men.
\n""}}, ""slug"": ""women-rights"", ""tags"": [], ""type"": ""page"", ""title"": {""rendered"": ""Women’s Rights""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56926""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/page""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=56926"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=56926"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=56926"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=56926""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56926/revisions"", ""count"": 22}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/57051"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57907, ""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56926/revisions/57907""}]}, ""author"": 49, ""parent"": 0, ""status"": ""publish"", ""content"": {""rendered"": ""\nThis page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/195wtinOB2zsiGevmvUfAGPZv4NrG4jY_vpib49e5gt4/preview
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""How has the protection of women’s rights changed over time? How does it differ across countries? Explore global data and research on women’s rights."", ""protected"": false}, ""date_gmt"": ""2023-05-12T09:03:25"", ""modified"": ""2023-07-18T22:12:37"", ""template"": """", ""categories"": [44, 56, 239], ""menu_order"": 16, ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre"", ""Pablo Arriagada"", ""Esteban Ortiz-Ospina"", ""Hannah Ritchie"", ""Joe Hasell""], ""modified_gmt"": ""2023-07-18T21:12:37"", ""comment_status"": ""closed"", ""featured_media"": 57051, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/05/Womens-Rights-150x79.png"", ""medium_large"": ""/app/uploads/2023/05/Womens-Rights-768x403.png""}}" 56832,Influenza,influenza,page,publish,"Seasonal flu is a contagious illness caused by the influenza virus. It kills around 400,000 people from respiratory disease on average each year. In large pandemics, when new strains have evolved, the death toll has been much higher.
Yet, data on the flu is limited. With better testing, countries could improve their response to flu epidemics. It could help to rapidly identify new strains, detect epidemics early, and design better-matched vaccines to target flu strains circulating in the population.
This page therefore shows estimates of deaths during seasonal flu epidemics, historical flu pandemics, patterns of flu seasons in different countries, and confirmed cases of flu and flu-like symptoms across the world.
It also includes our Flu Explorer, a resource for epidemiologists, infectious disease researchers, and public health experts to monitor the global spread of the influenza virus.
Related topics
The flu is estimated to cause 400,000 respiratory deaths each year on average across the world. These deaths come from pneumonia and other respiratory symptoms caused by the flu.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. https://jogh.org/documents/issue201902/jogh-09-020421.pdf
This shows the mean estimate of annual influenza mortality between 2002–2011, excluding the 2009 “Swine flu” pandemic influenza season. You can find estimated numbers for world regions in Table 2 of the paper. Rates for other age groups can be found here: https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf {/ref}
People also die from other complications of the flu – such as a stroke or heart attack – but global estimates have not been made of their death toll.{ref}The global number of people who die from other complications of the flu is unclear.
Paget et al. (the authors of the Global Pandemic Mortality project, i.e. GLaMOR) state in their paper that their estimate “does not cover cardiovascular deaths, something that could at least double the estimate of influenza-associated deaths.”
In recent meta-analyses, Behrouzi et al. found that influenza vaccination reduces the chances of major cardiovascular events (such as heart attacks and strokes) by around 34%, in clinical trials of the elderly.
This suggests the death toll from other complications could be large. However, global estimates have not been made of these types of deaths from flu.
Paget, J., Danielle Iuliano, A., Taylor, R. J., Simonsen, L., Viboud, C., & Spreeuwenberg, P. (2022). Estimates of mortality associated with seasonal influenza for the European Union from the GLaMOR project. Vaccine, 40(9), 1361–1369. https://doi.org/10.1016/j.vaccine.2021.11.080
Behrouzi, B., Bhatt, D. L., Cannon, C. P., Vardeny, O., Lee, D. S., Solomon, S. D., & Udell, J. A. (2022). Association of Influenza Vaccination With Cardiovascular Risk: A Meta-analysis. JAMA Network Open, 5(4), e228873. https://doi.org/10.1001/jamanetworkopen.2022.8873{/ref}
The flu is most severe in infants and the elderly.{ref}Metcalf, C. J. E., Paireau, J., O’Driscoll, M., Pivette, M., Hubert, B., Pontais, I., Nickbakhsh, S., Cummings, D. A. T., Cauchemez, S., & Salje, H. (2022). Comparing the age and sex trajectories of SARS-CoV-2 morbidity and mortality with other respiratory pathogens. Royal Society Open Science, 9(6), 211498. https://doi.org/10.1098/rsos.211498 {/ref}
Among those over 65, the flu kills around 31 people per 100,000 each year from respiratory disease in Europe. You can see this on the map.
But it’s not only age that matters, as the map shows. Death rates from the flu are higher in countries in South America, Africa, and South Asia, than in Europe and North America, due to poverty, poorer underlying health, lower access to healthcare, and lower vaccination rates.
In this article, we provide more detail:
In many countries, flu became much rarer during the COVID-19 pandemic, due to the impact of social distancing.
You can see this in the chart. It shows the share of flu tests that were positive. In 2020 and 2021, there was a large decline in flu and the rates of positive tests were low.
Because the influenza virus is spread between people, through respiratory droplets and human contact{ref}Kutter, J. S., Spronken, M. I., Fraaij, P. L., Fouchier, R. A., & Herfst, S. (2018). Transmission routes of respiratory viruses among humans. Current Opinion in Virology, 28, 142–151. https://doi.org/10.1016/j.coviro.2018.01.001{/ref}, social distancing led to a large reduction in contact between people and limited the virus from spreading.{ref}Farboodi, M., Jarosch, G., & Shimer, R. (2021). Internal and external effects of social distancing in a pandemic. Journal of Economic Theory, 196, 105293. https://doi.org/10.1016/j.jet.2021.105293
Woskie, L. R., Hennessy, J., Espinosa, V., Tsai, T. C., Vispute, S., Jacobson, B. H., Cattuto, C., Gauvin, L., Tizzoni, M., Fabrikant, A., Gadepalli, K., Boulanger, A., Pearce, A., Kamath, C., Schlosberg, A., Stanton, C., Bavadekar, S., Abueg, M., Hogue, M., … Gabrilovich, E. (2021). Early social distancing policies in Europe, changes in mobility & COVID-19 case trajectories: Insights from Spring 2020. PLOS ONE, 16(6), e0253071. https://doi.org/10.1371/journal.pone.0253071 {/ref}
This decline was very large because of the mathematics of epidemics.
The reproductive number (also called the R number) can help to understand why. This refers to the average number of people who will be infected by someone with the virus. When the R number is greater than 1, the average person who is infected will spread the virus to more than one person, who spread it to more and more people; the number of cases rises exponentially and leads to an epidemic.{ref}Rothman, K. J., Lash, T. L., VanderWeele, T. J., & Haneuse, S. (2021). Modern epidemiology (Fourth edition). Wolters Kluwer.{/ref}
However, when the R number is lower than 1, the virus does not lead to an epidemic, and the number of cases falls exponentially.
Seasonal flu viruses tend to have an R number that is slightly above 1 at the start of an epidemic.{ref}Biggerstaff, M., Cauchemez, S., Reed, C., Gambhir, M., & Finelli, L. (2014). Estimates of the reproduction number for seasonal, pandemic, and zoonotic influenza: A systematic review of the literature. BMC Infectious Diseases, 14(1), 480. https://doi.org/10.1186/1471-2334-14-480{/ref} Social distancing cut the number of contacts between people, and led to the R number of flu to dip much below 1 for a long time. This is why the spread of flu dwindled worldwide and was only seen in limited circumstances.{ref}This effect was so large that it may have led to the extinction of a lineage of flu called influenza B Yamagata.
Dhanasekaran, V., Sullivan, S., Edwards, K. M., Xie, R., Khvorov, A., Valkenburg, S. A., Cowling, B. J., & Barr, I. G. (2022). Human seasonal influenza under COVID-19 and the potential consequences of influenza lineage elimination. Nature Communications, 13(1), 1721. https://doi.org/10.1038/s41467-022-29402-5
Paget, J., Caini, S., Del Riccio, M., van Waarden, W., & Meijer, A. (2022). Has influenza B/Yamagata become extinct and what implications might this have for quadrivalent influenza vaccines? Eurosurveillance, 27(39). https://doi.org/10.2807/1560-7917.ES.2022.27.39.2200753 {/ref}
Over time, the severity of the flu has declined among people of the same age, as the chart shows.{ref}Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y {/ref}
This is because of flu vaccination, which began in the 1930s and 1940s, as well as improvements in sanitation, neonatal healthcare, and childhood vaccination for other diseases.{ref}Barberis, I., Myles, P., Ault, S. K., Bragazzi, N. L., & Martini, M. (2016). History and evolution of influenza control through vaccination: From the first monovalent vaccine to universal vaccines. Journal of Preventive Medicine and Hygiene, 57(3), E115–E120. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5139605/
Centers for Disease Control and Prevention, & National Center for Immunization and Respiratory Diseases. (2021). Historical Reference of Seasonal Influenza Vaccine Doses Distributed. https://www.cdc.gov/flu/prevent/vaccine-supply-historical.htm {/ref} These benefits carried forward as people aged: they protected people from being vulnerable to diseases including influenza.
But the flu still causes a large burden today, especially in countries that have poor sanitation, healthcare, and low vaccination rates.
Another challenge is that populations have been aging rapidly.{ref}Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y {/ref} In lower-income countries, the flu could become a larger burden as their populations continue to age.
In this article, we provide more detail:
Several respiratory infections, including the flu, are more common in the winter.
This is because they transmit more efficiently at lower temperatures and humidity, and when there is more social contact between people indoors.{ref}Petrova, V. N., & Russell, C. A. (2018). The evolution of seasonal influenza viruses. Nature Reviews Microbiology, 16(1), 47–60. https://doi.org/10.1038/nrmicro.2017.118 {/ref}
In the chart, you can see the share of flu tests that were positive in different countries.
Although the precise start and end of a flu season vary between years, flu epidemics tend to occur between November and May in the Northern Hemisphere. Meanwhile, in the Southern Hemisphere, they generally occur between June and October.
But in countries closer to the equator, there tend to be multiple peaks each year, or flu is present throughout the year. This may be because of rainy seasons, when people have more indoor contact.{ref}Chen, C., Jiang, D., Yan, D., Pi, L., Zhang, X., Du, Y., Liu, X., Yang, M., Zhou, Y., Ding, C., Lan, L., & Yang, S. (2023). The global region-specific epidemiologic characteristics of influenza: World Health Organization FluNet data from 1996 to 2021. International Journal of Infectious Diseases, 129, 118–124. https://pubmed.ncbi.nlm.nih.gov/36773717/
Petrova, V. N., & Russell, C. A. (2018). The evolution of seasonal influenza viruses. Nature Reviews Microbiology, 16(1), 47–60. https://doi.org/10.1038/nrmicro.2017.118
Paynter, S. (2015). Humidity and respiratory virus transmission in tropical and temperate settings. Epidemiology & Infection, 143(6), 1110–1118. https://doi.org/10.1017/S0950268814002702
Igboh, L. S., Roguski, K., Marcenac, P., Emukule, G. O., Charles, M. D., Tempia, S., Herring, B., Vandemaele, K., Moen, A., Olsen, S. J., Wentworth, D. E., Kondor, R., Mott, J. A., Hirve, S., Bresee, J. S., Mangtani, P., Nguipdop-Djomo, P., & Azziz-Baumgartner, E. (2023). Timing of seasonal influenza epidemics for 25 countries in Africa during 2010–19: A retrospective analysis. The Lancet Global Health, 11(5), e729–e739. https://doi.org/10.1016/S2214-109X(23)00109-2
Newman, L. P., Bhat, N., Fleming, J. A., & Neuzil, K. M. (2018). Global influenza seasonality to inform country-level vaccine programs: An analysis of WHO FluNet influenza surveillance data between 2011 and 2016. PLOS ONE, 13(2), e0193263. https://doi.org/10.1371/journal.pone.0193263 {/ref}
You can see this in the chart for Singapore and Thailand, for example.
Direct testing for the presence of the influenza virus is limited in many countries. For this reason, flu cases recorded in public databases – and the global data shown in our data explorer – greatly underestimate the true number of cases.
Because of the lack of direct testing, it is useful to track flu-like symptoms.
It is important to note however that these symptoms are not specific to the flu: people with other diseases – such as rhinovirus, COVID-19, common colds, malaria, and others – can also have these symptoms and meet the following criteria.
Acute respiratory infections (ARIs) are the broadest type of metric. They can include anyone with sudden onset of at least one of the following symptoms: cough, sore throat, shortness of breath or rhinitis (inflammation of the mucous lining of the nose), but only if they were judged by a doctor to be caused by an infection.
Influenza-like illnesses (ILIs) are narrower – they include only people with a sudden respiratory infection with a fever above 38ºC and a cough within the last 10 days.{ref}This definition has been used since 2011, after the Swine flu pandemic. Since then, most countries, but not all, have adopted it.
Fitzner, J., Qasmieh, S., Mounts, A. W., Alexander, B., Besselaar, T., Briand, S., Brown, C., Clark, S., Dueger, E., Gross, D., Hauge, S., Hirve, S., Jorgensen, P., Katz, M. A., Mafi, A., Malik, M., McCarron, M., Meerhoff, T., Mori, Y., … Vandemaele, K. (2018). Revision of clinical case definitions: Influenza-like illness and severe acute respiratory infection. Bulletin of the World Health Organization, 96(2), 122–128.
https://doi.org/10.2471/BLT.17.194514
World Health Organization. (2022). Respiratory Viruses Surveillance Country, Territory and Area Profiles, 2021. https://apps.who.int/iris/bitstream/handle/10665/352183/WHO-EURO-2022-4760-44523-63025-eng.pdf?sequence=1&isAllowed=y
Simpson, R. B., Gottlieb, J., Zhou, B., Hartwick, M. A., & Naumova, E. N. (2021). Completeness of open access FluNet influenza surveillance data for Pan-America in 2005–2019. Scientific Reports, 11(1), 795. https://doi.org/10.1038/s41598-020-80842-9 {/ref}
Severe acute respiratory infections (SARIs) are severe cases of ILIs: they include only people with a sudden respiratory infection who had a fever above 38ºC, a cough, and required hospitalization.
In many countries, only a fraction of clinics in the country report flu-like metrics to their national system. This means that the number of reported cases does not tell us about the total number of people with these infections.
Since some countries provide data from a larger number of healthcare clinics than others, this needs to be kept in mind when comparing different countries.{ref}World Health Organization. (2022). Respiratory Viruses Surveillance Country, Territory and Area Profiles, 2021. https://apps.who.int/iris/bitstream/handle/10665/352183/WHO-EURO-2022-4760-44523-63025-eng.pdf?sequence=1&isAllowed=y {/ref}
Each year, flu vaccines need to be updated because different viruses circulate in the population.
This happens for two reasons. One is that flu viruses gradually evolve, and can evade people’s immunity and cause reinfections.{ref}Krammer, F., Smith, G. J. D., Fouchier, R. A. M., Peiris, M., Kedzierska, K., Doherty, P. C., Palese, P., Shaw, M. L., Treanor, J., Webster, R. G., & García-Sastre, A. (2018). Influenza. Nature Reviews Disease Primers, 4(1), 3. https://doi.org/10.1038/s41572-018-0002-y {/ref}
Another reason is that there are different types of flu that circulate each season.
You can see this in the chart. It shows cases of two types of influenza: A and B, which commonly infect humans.{ref}There are four types of influenza viruses: A, B, C, and D. Influenza A and B tend to spread between people around the world each year. In contrast, influenza C and D mainly infect birds and other animals.{/ref}
There are many subtypes of influenza A, including H1N1 and H3N2 among others.{ref}The subtypes are named according to two proteins the virus has on its surface: hemagglutinin (H) and neuraminidase (N). There are many subtypes of each of these two proteins (18 hemagglutinin subtypes and 11 neuraminidase subtypes), but only some of the combinations have been observed. For example, H3N2 is a type of influenza A virus which has hemagglutinin subtype 3 and neuraminidase subtype 2.
Long, J.S., Mistry, B., Haslam, S.M. et al. Host and viral determinants of influenza A virus species specificity. Nat Rev Microbiol 17, 67–81 (2019). https://doi.org/10.1038/s41579-018-0115-z {/ref} In contrast, there are two lineages of influenza B, which are called Victoria and Yamagata.{ref}Caini, S., Kusznierz, G., Garate, V. V., Wangchuk, S., Thapa, B., de Paula Júnior, F. J., Ferreira de Almeida, W. A., Njouom, R., Fasce, R. A., Bustos, P., Feng, L., Peng, Z., Araya, J. L., Bruno, A., de Mora, D., Barahona de Gámez, M. J., Pebody, R., Zambon, M., Higueros, R., … the Global Influenza B Study team. (2019). The epidemiological signature of influenza B virus and its B/Victoria and B/Yamagata lineages in the 21st century. PLOS ONE, 14(9), e0222381. https://doi.org/10.1371/journal.pone.0222381 {/ref}
Unfortunately, testing for the flu is limited, and many countries lack testing to identify specific flu strains. This is why the number of confirmed cases shown in the chart greatly underestimates the actual number of infections, and why some cases are shown as ‘unknown subtype/lineage’.{ref}The subtype or lineage of a flu virus is not always determined during testing. This tends to be because some clinics do not test for all subtypes of influenza, due to a lack of testing resources. These are listed as unknown subtypes/lineages of influenza. For example, with influenza A, labs tend to test only whether they are the H1 or H3 strain.
However, unknown subtypes/lineages can also include novel influenza strains, which have gone through significant evolution.{/ref}
This lack of testing is a problem. When there is a mismatch between the strains in the vaccine and the viruses circulating in the population, vaccines tend to have lower efficacy and the flu season tends to be more severe.{ref}Tricco, A. C., Chit, A., Soobiah, C., Hallett, D., Meier, G., Chen, M. H., Tashkandi, M., Bauch, C. T., & Loeb, M. (2013). Comparing influenza vaccine efficacy against mismatched and matched strains: A systematic review and meta-analysis. BMC Medicine, 11(1), 153. https://doi.org/10.1186/1741-7015-11-153 {/ref} Additionally, limited testing also means the world is less able to detect new strains that may cause pandemics.{ref}Jernigan, D. B., Lindstrom, S. . L., Johnson, J. . R., Miller, J. D., Hoelscher, M., Humes, R., Shively, R., Brammer, L., Burke, S. A., Villanueva, J. M., Balish, A., Uyeki, T., Mustaquim, D., Bishop, A., Handsfield, J. H., Astles, R., Xu, X., Klimov, A. I., Cox, N. J., & Shaw, M. W. (2011). Detecting 2009 Pandemic Influenza A (H1N1) Virus Infection: Availability of Diagnostic Testing Led to Rapid Pandemic Response. Clinical Infectious Diseases, 52(suppl_1), S36–S43. https://doi.org/10.1093/cid/ciq020 {/ref}
To address this, the world needs more routine testing for the flu.
Some flu seasons are far more severe than usual seasonal influenza.
This happens when influenza viruses combine with each other to make new strains which are more infectious and lethal, and lead to deadly pandemics.{ref}Krammer, F., Smith, G. J. D., Fouchier, R. A. M., Peiris, M., Kedzierska, K., Doherty, P. C., Palese, P., Shaw, M. L., Treanor, J., Webster, R. G., & García-Sastre, A. (2018). Influenza. Nature Reviews Disease Primers, 4(1), 3. https://doi.org/10.1038/s41572-018-0002-y {/ref}
For example, the Spanish flu pandemic was caused by a combination of human influenza and another animal influenza. Together, they formed the new H1N1 virus.{ref}Worobey, M., Han, G.-Z., & Rambaut, A. (2014). Genesis and pathogenesis of the 1918 pandemic H1N1 influenza A virus. Proceedings of the National Academy of Sciences, 111(22), 8107–8112. https://doi.org/10.1073/pnas.1324197111{/ref}
As you can see in the chart, it led to the most devastating influenza pandemic in recorded history. Estimates of the death toll vary: some studies estimate that 17.4 million people died globally from the Spanish flu between 1918 and 1920, while others estimate a much higher death toll of 50 to 100 million deaths.{ref}P. Spreeuwenberg; et al. (1 December 2018). “Reassessing the Global Mortality Burden of the 1918 Influenza Pandemic”. American Journal of Epidemiology. 187 (12): 2561–2567. doi:10.1093/aje/kwy191. PMID 30202996. Online here.{/ref}
The Spanish flu pandemic was most severe among children and young adults. Life expectancy at birth and at young ages declined by more than ten years.
But surprisingly, it did not have a significant impact on older people. Research suggests that this is because older generations had been exposed to similar H1 influenza viruses decades before the pandemic began, which gave them some protection against the Spanish flu strain.{ref}Gagnon, A., Miller, M. S., Hallman, S. A., Bourbeau, R., Herring, D. A., Earn, D. J. D., & Madrenas, J. (2013). Age-specific mortality during the 1918 influenza pandemic: Unravelling the mystery of high young adult mortality. PloS One, 8(8), e69586. https://doi.org/10.1371/journal.pone.0069586
Luk, J., Gross, P., & Thompson, W. W. (2001). Observations on Mortality during the 1918 Influenza Pandemic. Clinical Infectious Diseases, 33(8), 1375–1378. https://doi.org/10.1086/322662
Ma, J., Dushoff, J., & Earn, D. J. D. (2011). Age-specific mortality risk from pandemic influenza. Journal of Theoretical Biology, 288, 29–34. https://doi.org/10.1016/j.jtbi.2011.08.003 {/ref}
In this article, we provide more detail:
With this Flu Explorer, we aim to provide a helpful resource for epidemiologists, infectious disease researchers, and public health experts to understand the global spread of the influenza virus.
It differs from our widely-used infectious diseases projects, such as the COVID-19 Explorer and the Mpox Explorer. These tools are designed for a broad audience. Unfortunately, flu data is incomplete in many ways, making it harder to communicate. This tool is therefore designed for users with pre-existing knowledge to navigate effectively the complex data published by the World Health Organization.
The explorer also highlights the significant gaps in influenza data. It is an important reminder of the need to improve data collection and reporting.
Saloni Dattani
Max Roser
Seasonal flu is a contagious illness caused by the influenza virus. It kills around 400,000 people from respiratory disease on average each year. In large pandemics, when new strains have evolved, the death toll has been much higher.
\n\n\n\nYet, data on the flu is limited. With better testing, countries could improve their response to flu epidemics. It could help to rapidly identify new strains, detect epidemics early, and design better-matched vaccines to target flu strains circulating in the population.
\n\n\n\nThis page therefore shows estimates of deaths during seasonal flu epidemics, historical flu pandemics, patterns of flu seasons in different countries, and confirmed cases of flu and flu-like symptoms across the world.
\n\n\n\nIt also includes our Flu Explorer, a resource for epidemiologists, infectious disease researchers, and public health experts to monitor the global spread of the influenza virus.
\nRelated topics
\n\n\n\n\nThe flu is estimated to cause 400,000 respiratory deaths each year on average across the world. These deaths come from pneumonia and other respiratory symptoms caused by the flu.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. https://jogh.org/documents/issue201902/jogh-09-020421.pdf
\n\n\n\nThis shows the mean estimate of annual influenza mortality between 2002–2011, excluding the 2009 “Swine flu” pandemic influenza season. You can find estimated numbers for world regions in Table 2 of the paper. Rates for other age groups can be found here: https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf {/ref}
\n\n\n\nPeople also die from other complications of the flu – such as a stroke or heart attack – but global estimates have not been made of their death toll.{ref}The global number of people who die from other complications of the flu is unclear.
\n\n\n\nPaget et al. (the authors of the Global Pandemic Mortality project, i.e. GLaMOR) state in their paper that their estimate “does not cover cardiovascular deaths, something that could at least double the estimate of influenza-associated deaths.”
\n\n\n\nIn recent meta-analyses, Behrouzi et al. found that influenza vaccination reduces the chances of major cardiovascular events (such as heart attacks and strokes) by around 34%, in clinical trials of the elderly.
\n\n\n\nThis suggests the death toll from other complications could be large. However, global estimates have not been made of these types of deaths from flu.
\n\n\n\nPaget, J., Danielle Iuliano, A., Taylor, R. J., Simonsen, L., Viboud, C., & Spreeuwenberg, P. (2022). Estimates of mortality associated with seasonal influenza for the European Union from the GLaMOR project. Vaccine, 40(9), 1361–1369. https://doi.org/10.1016/j.vaccine.2021.11.080
\n\n\n\nBehrouzi, B., Bhatt, D. L., Cannon, C. P., Vardeny, O., Lee, D. S., Solomon, S. D., & Udell, J. A. (2022). Association of Influenza Vaccination With Cardiovascular Risk: A Meta-analysis. JAMA Network Open, 5(4), e228873. https://doi.org/10.1001/jamanetworkopen.2022.8873{/ref}
\n\n\n\nThe flu is most severe in infants and the elderly.{ref}Metcalf, C. J. E., Paireau, J., O’Driscoll, M., Pivette, M., Hubert, B., Pontais, I., Nickbakhsh, S., Cummings, D. A. T., Cauchemez, S., & Salje, H. (2022). Comparing the age and sex trajectories of SARS-CoV-2 morbidity and mortality with other respiratory pathogens. Royal Society Open Science, 9(6), 211498. https://doi.org/10.1098/rsos.211498 {/ref}
\n\n\n\nAmong those over 65, the flu kills around 31 people per 100,000 each year from respiratory disease in Europe. You can see this on the map.
\n\n\n\nBut it’s not only age that matters, as the map shows. Death rates from the flu are higher in countries in South America, Africa, and South Asia, than in Europe and North America, due to poverty, poorer underlying health, lower access to healthcare, and lower vaccination rates.
\n\n\n\nIn this article, we provide more detail:
\n\n\nIn many countries, flu became much rarer during the COVID-19 pandemic, due to the impact of social distancing.
\n\n\n\nYou can see this in the chart. It shows the share of flu tests that were positive. In 2020 and 2021, there was a large decline in flu and the rates of positive tests were low.
\n\n\n\nBecause the influenza virus is spread between people, through respiratory droplets and human contact{ref}Kutter, J. S., Spronken, M. I., Fraaij, P. L., Fouchier, R. A., & Herfst, S. (2018). Transmission routes of respiratory viruses among humans. Current Opinion in Virology, 28, 142–151. https://doi.org/10.1016/j.coviro.2018.01.001{/ref}, social distancing led to a large reduction in contact between people and limited the virus from spreading.{ref}Farboodi, M., Jarosch, G., & Shimer, R. (2021). Internal and external effects of social distancing in a pandemic. Journal of Economic Theory, 196, 105293. https://doi.org/10.1016/j.jet.2021.105293
\n\n\n\nWoskie, L. R., Hennessy, J., Espinosa, V., Tsai, T. C., Vispute, S., Jacobson, B. H., Cattuto, C., Gauvin, L., Tizzoni, M., Fabrikant, A., Gadepalli, K., Boulanger, A., Pearce, A., Kamath, C., Schlosberg, A., Stanton, C., Bavadekar, S., Abueg, M., Hogue, M., … Gabrilovich, E. (2021). Early social distancing policies in Europe, changes in mobility & COVID-19 case trajectories: Insights from Spring 2020. PLOS ONE, 16(6), e0253071. https://doi.org/10.1371/journal.pone.0253071 {/ref}
\n\n\n\nThis decline was very large because of the mathematics of epidemics.
\n\n\n\nThe reproductive number (also called the R number) can help to understand why. This refers to the average number of people who will be infected by someone with the virus. When the R number is greater than 1, the average person who is infected will spread the virus to more than one person, who spread it to more and more people; the number of cases rises exponentially and leads to an epidemic.{ref}Rothman, K. J., Lash, T. L., VanderWeele, T. J., & Haneuse, S. (2021). Modern epidemiology (Fourth edition). Wolters Kluwer.{/ref}
\n\n\n\nHowever, when the R number is lower than 1, the virus does not lead to an epidemic, and the number of cases falls exponentially.
\n\n\n\nSeasonal flu viruses tend to have an R number that is slightly above 1 at the start of an epidemic.{ref}Biggerstaff, M., Cauchemez, S., Reed, C., Gambhir, M., & Finelli, L. (2014). Estimates of the reproduction number for seasonal, pandemic, and zoonotic influenza: A systematic review of the literature. BMC Infectious Diseases, 14(1), 480. https://doi.org/10.1186/1471-2334-14-480{/ref} Social distancing cut the number of contacts between people, and led to the R number of flu to dip much below 1 for a long time. This is why the spread of flu dwindled worldwide and was only seen in limited circumstances.{ref}This effect was so large that it may have led to the extinction of a lineage of flu called influenza B Yamagata.
\n\n\n\nDhanasekaran, V., Sullivan, S., Edwards, K. M., Xie, R., Khvorov, A., Valkenburg, S. A., Cowling, B. J., & Barr, I. G. (2022). Human seasonal influenza under COVID-19 and the potential consequences of influenza lineage elimination. Nature Communications, 13(1), 1721. https://doi.org/10.1038/s41467-022-29402-5
Paget, J., Caini, S., Del Riccio, M., van Waarden, W., & Meijer, A. (2022). Has influenza B/Yamagata become extinct and what implications might this have for quadrivalent influenza vaccines? Eurosurveillance, 27(39). https://doi.org/10.2807/1560-7917.ES.2022.27.39.2200753 {/ref}
Over time, the severity of the flu has declined among people of the same age, as the chart shows.{ref}Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y {/ref}
\n\n\n\nThis is because of flu vaccination, which began in the 1930s and 1940s, as well as improvements in sanitation, neonatal healthcare, and childhood vaccination for other diseases.{ref}Barberis, I., Myles, P., Ault, S. K., Bragazzi, N. L., & Martini, M. (2016). History and evolution of influenza control through vaccination: From the first monovalent vaccine to universal vaccines. Journal of Preventive Medicine and Hygiene, 57(3), E115–E120. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5139605/
\n\n\n\nCenters for Disease Control and Prevention, & National Center for Immunization and Respiratory Diseases. (2021). Historical Reference of Seasonal Influenza Vaccine Doses Distributed. https://www.cdc.gov/flu/prevent/vaccine-supply-historical.htm {/ref} These benefits carried forward as people aged: they protected people from being vulnerable to diseases including influenza.
\n\n\n\nBut the flu still causes a large burden today, especially in countries that have poor sanitation, healthcare, and low vaccination rates.
\n\n\n\nAnother challenge is that populations have been aging rapidly.{ref}Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y {/ref} In lower-income countries, the flu could become a larger burden as their populations continue to age.
\n\n\n\nIn this article, we provide more detail:
\n\n\n![\""People](\""https://owid.cloud/app/uploads/2023/05/Historical-decline-in-influenza-deaths-by-birth-cohort-635x550.png\"")
Several respiratory infections, including the flu, are more common in the winter.
\n\n\n\nThis is because they transmit more efficiently at lower temperatures and humidity, and when there is more social contact between people indoors.{ref}Petrova, V. N., & Russell, C. A. (2018). The evolution of seasonal influenza viruses. Nature Reviews Microbiology, 16(1), 47–60. https://doi.org/10.1038/nrmicro.2017.118 {/ref}
\n\n\n\nIn the chart, you can see the share of flu tests that were positive in different countries.
\n\n\n\nAlthough the precise start and end of a flu season vary between years, flu epidemics tend to occur between November and May in the Northern Hemisphere. Meanwhile, in the Southern Hemisphere, they generally occur between June and October.
\n\n\n\nBut in countries closer to the equator, there tend to be multiple peaks each year, or flu is present throughout the year. This may be because of rainy seasons, when people have more indoor contact.{ref}Chen, C., Jiang, D., Yan, D., Pi, L., Zhang, X., Du, Y., Liu, X., Yang, M., Zhou, Y., Ding, C., Lan, L., & Yang, S. (2023). The global region-specific epidemiologic characteristics of influenza: World Health Organization FluNet data from 1996 to 2021. International Journal of Infectious Diseases, 129, 118–124. https://pubmed.ncbi.nlm.nih.gov/36773717/
Petrova, V. N., & Russell, C. A. (2018). The evolution of seasonal influenza viruses. Nature Reviews Microbiology, 16(1), 47–60. https://doi.org/10.1038/nrmicro.2017.118
Paynter, S. (2015). Humidity and respiratory virus transmission in tropical and temperate settings. Epidemiology & Infection, 143(6), 1110–1118. https://doi.org/10.1017/S0950268814002702
Igboh, L. S., Roguski, K., Marcenac, P., Emukule, G. O., Charles, M. D., Tempia, S., Herring, B., Vandemaele, K., Moen, A., Olsen, S. J., Wentworth, D. E., Kondor, R., Mott, J. A., Hirve, S., Bresee, J. S., Mangtani, P., Nguipdop-Djomo, P., & Azziz-Baumgartner, E. (2023). Timing of seasonal influenza epidemics for 25 countries in Africa during 2010–19: A retrospective analysis. The Lancet Global Health, 11(5), e729–e739. https://doi.org/10.1016/S2214-109X(23)00109-2
Newman, L. P., Bhat, N., Fleming, J. A., & Neuzil, K. M. (2018). Global influenza seasonality to inform country-level vaccine programs: An analysis of WHO FluNet influenza surveillance data between 2011 and 2016. PLOS ONE, 13(2), e0193263. https://doi.org/10.1371/journal.pone.0193263 {/ref}
\n\n\n\nYou can see this in the chart for Singapore and Thailand, for example.
\n\n\n\tDirect testing for the presence of the influenza virus is limited in many countries. For this reason, flu cases recorded in public databases – and the global data shown in our data explorer – greatly underestimate the true number of cases.
\n\n\n\nBecause of the lack of direct testing, it is useful to track flu-like symptoms.
\n\n\n\nIt is important to note however that these symptoms are not specific to the flu: people with other diseases – such as rhinovirus, COVID-19, common colds, malaria, and others – can also have these symptoms and meet the following criteria.
\n\n\n\nAcute respiratory infections (ARIs) are the broadest type of metric. They can include anyone with sudden onset of at least one of the following symptoms: cough, sore throat, shortness of breath or rhinitis (inflammation of the mucous lining of the nose), but only if they were judged by a doctor to be caused by an infection.
\n\n\n\nInfluenza-like illnesses (ILIs) are narrower – they include only people with a sudden respiratory infection with a fever above 38ºC and a cough within the last 10 days.{ref}This definition has been used since 2011, after the Swine flu pandemic. Since then, most countries, but not all, have adopted it.
Fitzner, J., Qasmieh, S., Mounts, A. W., Alexander, B., Besselaar, T., Briand, S., Brown, C., Clark, S., Dueger, E., Gross, D., Hauge, S., Hirve, S., Jorgensen, P., Katz, M. A., Mafi, A., Malik, M., McCarron, M., Meerhoff, T., Mori, Y., … Vandemaele, K. (2018). Revision of clinical case definitions: Influenza-like illness and severe acute respiratory infection. Bulletin of the World Health Organization, 96(2), 122–128.
https://doi.org/10.2471/BLT.17.194514
World Health Organization. (2022). Respiratory Viruses Surveillance Country, Territory and Area Profiles, 2021. https://apps.who.int/iris/bitstream/handle/10665/352183/WHO-EURO-2022-4760-44523-63025-eng.pdf?sequence=1&isAllowed=y
Simpson, R. B., Gottlieb, J., Zhou, B., Hartwick, M. A., & Naumova, E. N. (2021). Completeness of open access FluNet influenza surveillance data for Pan-America in 2005–2019. Scientific Reports, 11(1), 795. https://doi.org/10.1038/s41598-020-80842-9 {/ref}
Severe acute respiratory infections (SARIs) are severe cases of ILIs: they include only people with a sudden respiratory infection who had a fever above 38ºC, a cough, and required hospitalization.
\n\n\n\nIn many countries, only a fraction of clinics in the country report flu-like metrics to their national system. This means that the number of reported cases does not tell us about the total number of people with these infections.
\n\n\n\nSince some countries provide data from a larger number of healthcare clinics than others, this needs to be kept in mind when comparing different countries.{ref}World Health Organization. (2022). Respiratory Viruses Surveillance Country, Territory and Area Profiles, 2021. https://apps.who.int/iris/bitstream/handle/10665/352183/WHO-EURO-2022-4760-44523-63025-eng.pdf?sequence=1&isAllowed=y {/ref}
\n\n\n\tEach year, flu vaccines need to be updated because different viruses circulate in the population.
\n\n\n\nThis happens for two reasons. One is that flu viruses gradually evolve, and can evade people’s immunity and cause reinfections.{ref}Krammer, F., Smith, G. J. D., Fouchier, R. A. M., Peiris, M., Kedzierska, K., Doherty, P. C., Palese, P., Shaw, M. L., Treanor, J., Webster, R. G., & García-Sastre, A. (2018). Influenza. Nature Reviews Disease Primers, 4(1), 3. https://doi.org/10.1038/s41572-018-0002-y {/ref}
\n\n\n\nAnother reason is that there are different types of flu that circulate each season.
\n\n\n\nYou can see this in the chart. It shows cases of two types of influenza: A and B, which commonly infect humans.{ref}There are four types of influenza viruses: A, B, C, and D. Influenza A and B tend to spread between people around the world each year. In contrast, influenza C and D mainly infect birds and other animals.{/ref}
\n\n\n\nThere are many subtypes of influenza A, including H1N1 and H3N2 among others.{ref}The subtypes are named according to two proteins the virus has on its surface: hemagglutinin (H) and neuraminidase (N). There are many subtypes of each of these two proteins (18 hemagglutinin subtypes and 11 neuraminidase subtypes), but only some of the combinations have been observed. For example, H3N2 is a type of influenza A virus which has hemagglutinin subtype 3 and neuraminidase subtype 2.
Long, J.S., Mistry, B., Haslam, S.M. et al. Host and viral determinants of influenza A virus species specificity. Nat Rev Microbiol 17, 67–81 (2019). https://doi.org/10.1038/s41579-018-0115-z {/ref} In contrast, there are two lineages of influenza B, which are called Victoria and Yamagata.{ref}Caini, S., Kusznierz, G., Garate, V. V., Wangchuk, S., Thapa, B., de Paula Júnior, F. J., Ferreira de Almeida, W. A., Njouom, R., Fasce, R. A., Bustos, P., Feng, L., Peng, Z., Araya, J. L., Bruno, A., de Mora, D., Barahona de Gámez, M. J., Pebody, R., Zambon, M., Higueros, R., … the Global Influenza B Study team. (2019). The epidemiological signature of influenza B virus and its B/Victoria and B/Yamagata lineages in the 21st century. PLOS ONE, 14(9), e0222381. https://doi.org/10.1371/journal.pone.0222381 {/ref}
Unfortunately, testing for the flu is limited, and many countries lack testing to identify specific flu strains. This is why the number of confirmed cases shown in the chart greatly underestimates the actual number of infections, and why some cases are shown as ‘unknown subtype/lineage’.{ref}The subtype or lineage of a flu virus is not always determined during testing. This tends to be because some clinics do not test for all subtypes of influenza, due to a lack of testing resources. These are listed as unknown subtypes/lineages of influenza. For example, with influenza A, labs tend to test only whether they are the H1 or H3 strain.
\n\n\n\nHowever, unknown subtypes/lineages can also include novel influenza strains, which have gone through significant evolution.{/ref}
\n\n\n\nThis lack of testing is a problem. When there is a mismatch between the strains in the vaccine and the viruses circulating in the population, vaccines tend to have lower efficacy and the flu season tends to be more severe.{ref}Tricco, A. C., Chit, A., Soobiah, C., Hallett, D., Meier, G., Chen, M. H., Tashkandi, M., Bauch, C. T., & Loeb, M. (2013). Comparing influenza vaccine efficacy against mismatched and matched strains: A systematic review and meta-analysis. BMC Medicine, 11(1), 153. https://doi.org/10.1186/1741-7015-11-153 {/ref} Additionally, limited testing also means the world is less able to detect new strains that may cause pandemics.{ref}Jernigan, D. B., Lindstrom, S. . L., Johnson, J. . R., Miller, J. D., Hoelscher, M., Humes, R., Shively, R., Brammer, L., Burke, S. A., Villanueva, J. M., Balish, A., Uyeki, T., Mustaquim, D., Bishop, A., Handsfield, J. H., Astles, R., Xu, X., Klimov, A. I., Cox, N. J., & Shaw, M. W. (2011). Detecting 2009 Pandemic Influenza A (H1N1) Virus Infection: Availability of Diagnostic Testing Led to Rapid Pandemic Response. Clinical Infectious Diseases, 52(suppl_1), S36–S43. https://doi.org/10.1093/cid/ciq020 {/ref}
\n\n\n\nTo address this, the world needs more routine testing for the flu.
\n\n\n\tSome flu seasons are far more severe than usual seasonal influenza.
\n\n\n\nThis happens when influenza viruses combine with each other to make new strains which are more infectious and lethal, and lead to deadly pandemics.{ref}Krammer, F., Smith, G. J. D., Fouchier, R. A. M., Peiris, M., Kedzierska, K., Doherty, P. C., Palese, P., Shaw, M. L., Treanor, J., Webster, R. G., & García-Sastre, A. (2018). Influenza. Nature Reviews Disease Primers, 4(1), 3. https://doi.org/10.1038/s41572-018-0002-y {/ref}
\n\n\n\nFor example, the Spanish flu pandemic was caused by a combination of human influenza and another animal influenza. Together, they formed the new H1N1 virus.{ref}Worobey, M., Han, G.-Z., & Rambaut, A. (2014). Genesis and pathogenesis of the 1918 pandemic H1N1 influenza A virus. Proceedings of the National Academy of Sciences, 111(22), 8107–8112. https://doi.org/10.1073/pnas.1324197111{/ref}
\n\n\n\nAs you can see in the chart, it led to the most devastating influenza pandemic in recorded history. Estimates of the death toll vary: some studies estimate that 17.4 million people died globally from the Spanish flu between 1918 and 1920, while others estimate a much higher death toll of 50 to 100 million deaths.{ref}P. Spreeuwenberg; et al. (1 December 2018). “Reassessing the Global Mortality Burden of the 1918 Influenza Pandemic”. American Journal of Epidemiology. 187 (12): 2561–2567. doi:10.1093/aje/kwy191. PMID 30202996. Online here.{/ref}
\n\n\n\nThe Spanish flu pandemic was most severe among children and young adults. Life expectancy at birth and at young ages declined by more than ten years.
\n\n\n\nBut surprisingly, it did not have a significant impact on older people. Research suggests that this is because older generations had been exposed to similar H1 influenza viruses decades before the pandemic began, which gave them some protection against the Spanish flu strain.{ref}Gagnon, A., Miller, M. S., Hallman, S. A., Bourbeau, R., Herring, D. A., Earn, D. J. D., & Madrenas, J. (2013). Age-specific mortality during the 1918 influenza pandemic: Unravelling the mystery of high young adult mortality. PloS One, 8(8), e69586. https://doi.org/10.1371/journal.pone.0069586
\n\n\n\nLuk, J., Gross, P., & Thompson, W. W. (2001). Observations on Mortality during the 1918 Influenza Pandemic. Clinical Infectious Diseases, 33(8), 1375–1378. https://doi.org/10.1086/322662
\n\n\n\nMa, J., Dushoff, J., & Earn, D. J. D. (2011). Age-specific mortality risk from pandemic influenza. Journal of Theoretical Biology, 288, 29–34. https://doi.org/10.1016/j.jtbi.2011.08.003 {/ref}
\n\n\n\nIn this article, we provide more detail:
\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2023/05/Influenza-pandemics-in-comparison-1-750x550.png\"")
With this Flu Explorer, we aim to provide a helpful resource for epidemiologists, infectious disease researchers, and public health experts to understand the global spread of the influenza virus.
\n\n\n\nIt differs from our widely-used infectious diseases projects, such as the COVID-19 Explorer and the Mpox Explorer. These tools are designed for a broad audience. Unfortunately, flu data is incomplete in many ways, making it harder to communicate. This tool is therefore designed for users with pre-existing knowledge to navigate effectively the complex data published by the World Health Organization.
\n\n\n\nThe explorer also highlights the significant gaps in influenza data. It is an important reminder of the need to improve data collection and reporting.
\nFood insecurity is defined by the Food and Agriculture Organization (FAO) of the United Nations as the “situation when people lack secure access to sufficient amounts of safe and nutritious food for normal growth and development and an active and healthy life.”
The chart below shows the share of the population that is ‘severely food insecure'. Let’s look at what this means, and how this is measured.
Food insecurity is one of the major causes of poor nutrition.
Food insecurity can be caused by several factors: food might be physically unavailable in a particular country or region; it can be unaffordable even if it is available to buy; or there might be an unequal distribution of food between household members.
Food insecurity is measured by the FAO using its Food Insecurity Experience Scale (FIES) global reference scale.
Food insecurity can be based on how having enough food (the quantity) or having inadequate quality and diversity of food. Someone might get enough food to meet their energy requirements, but they might rely on only a few basic foods (such as cereals) and have a diet with very little diversity.
The FIES measures the share of the population that has experienced food insecurity at moderate or severe levels during the period of measurement. Data is collected at the household level using a food security questionnaire. This survey asks households about a number of conditions that someone with food insecurity would typically experience.
It’s based on eight questions (the full list is given at the end of this article), such as:
“During the last 12 months, was there a time when you (or any other adult in the household) were worried you would not have enough food to eat because of a lack of money or other resources?”.
And:
“Was there a time when you (or any other adult in the household) had to skip a meal because there was not enough money or other resources to get food?”
These eight questions increase in severity: the first question is about worrying that you might not have enough food to eat at some point over the year. This evolves to actually eating less than is sufficient, then finally to going a whole day or more without any food.
The scale shows the definitions of food insecurity, ranging from mild to severe.
Moderate food insecurity is generally associated with the inability to regularly eat healthy, nutritious diets. It’s an important indicator of poor dietary quality and a high risk of micronutrient deficiencies.
The share of people that are either moderately or severely food insecure is shown in the chart.
Severe food insecurity is more strongly related to insufficient quantity of food (energy) and therefore strongly related to undernourishment or hunger.
In principle, the share of people that are severely food insecure should be similar to the share of the people that are defined as undernourished. Both metrics are used to estimate how many do not get enough food (in terms of energy) to eat. They just try to measure this in different ways. Food insecurity is based on household survey responses, which are more qualitative measure. Undernourishment is estimated based on actual food availability across the population; its a more quantitative measure.
In the chart below we see the correlation between these two metrics: on the y-axis we have the prevalence of severe food insecurity, and on the x-axis we have the prevalence of undernourishment. If these estimates were identical, they’d lie along the diagonal gray line.
We see that these metrics are very strongly correlated, although not perfectly. This may be for several reasons.
Peoples’ subjective experience of food availability might vary from their actual availability.
Undernourishment measurements are based on average calorie availability across the year. People might suffer from severe food insecurity on very short time scales i.e. they might have experienced a week of very little food, but have enough for the rest of the year. That means estimates of food insecurity from surveys would give a higher prevalence than undernourishment metrics.
As explained above, the UN FAO uses a survey of eight questions to evaluate whether someone is food insecure or not. These eight questions are:
Food insecurity is defined by the Food and Agriculture Organization (FAO) of the United Nations as the “situation when people lack secure access to sufficient amounts of safe and nutritious food for normal growth and development and an active and healthy life.”
\n\n\n\nThe chart below shows the share of the population that is ‘severely food insecure’. Let’s look at what this means, and how this is measured.
\nFood insecurity is one of the major causes of poor nutrition.
\n\n\n\nFood insecurity can be caused by several factors: food might be physically unavailable in a particular country or region; it can be unaffordable even if it is available to buy; or there might be an unequal distribution of food between household members.
\n\n\n\nFood insecurity is measured by the FAO using its Food Insecurity Experience Scale (FIES) global reference scale.
\n\n\n\nFood insecurity can be based on how having enough food (the quantity) or having inadequate quality and diversity of food. Someone might get enough food to meet their energy requirements, but they might rely on only a few basic foods (such as cereals) and have a diet with very little diversity.
\n\n\n\nThe FIES measures the share of the population that has experienced food insecurity at moderate or severe levels during the period of measurement. Data is collected at the household level using a food security questionnaire. This survey asks households about a number of conditions that someone with food insecurity would typically experience.
\n\n\n\nIt’s based on eight questions (the full list is given at the end of this article), such as:
\n\n\n\n“During the last 12 months, was there a time when you (or any other adult in the household) were worried you would not have enough food to eat because of a lack of money or other resources?”.
\n\n\n\nAnd:
\n\n\n\n“Was there a time when you (or any other adult in the household) had to skip a meal because there was not enough money or other resources to get food?”
\n\n\n\nThese eight questions increase in severity: the first question is about worrying that you might not have enough food to eat at some point over the year. This evolves to actually eating less than is sufficient, then finally to going a whole day or more without any food.
\n\n\n\nThe scale shows the definitions of food insecurity, ranging from mild to severe.
\n![\""\""](\""https://owid.cloud/app/uploads/2023/04/Food-Insecurity-Scale.png\"")
Moderate food insecurity is generally associated with the inability to regularly eat healthy, nutritious diets. It’s an important indicator of poor dietary quality and a high risk of micronutrient deficiencies.
\n\n\n\nThe share of people that are either moderately or severely food insecure is shown in the chart.
\nSevere food insecurity is more strongly related to insufficient quantity of food (energy) and therefore strongly related to undernourishment or hunger.
\n\n\n\nIn principle, the share of people that are severely food insecure should be similar to the share of the people that are defined as undernourished. Both metrics are used to estimate how many do not get enough food (in terms of energy) to eat. They just try to measure this in different ways. Food insecurity is based on household survey responses, which are more qualitative measure. Undernourishment is estimated based on actual food availability across the population; its a more quantitative measure.
\n\n\n\nIn the chart below we see the correlation between these two metrics: on the y-axis we have the prevalence of severe food insecurity, and on the x-axis we have the prevalence of undernourishment. If these estimates were identical, they’d lie along the diagonal gray line.
\n\n\n\nWe see that these metrics are very strongly correlated, although not perfectly. This may be for several reasons.
\n\n\n\nPeoples’ subjective experience of food availability might vary from their actual availability.
\n\n\n\nUndernourishment measurements are based on average calorie availability across the year. People might suffer from severe food insecurity on very short time scales i.e. they might have experienced a week of very little food, but have enough for the rest of the year. That means estimates of food insecurity from surveys would give a higher prevalence than undernourishment metrics.
\nAs explained above, the UN FAO uses a survey of eight questions to evaluate whether someone is food insecure or not. These eight questions are:
\n\n\n\nOur World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Technological Change.
Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases.
The best way to explain what this means is to look at a concrete example.
The time series in the chart shows the deployment of solar panels on the horizontal axis and the price of solar panels on the vertical axis. The orange line that describes the relationship between these two metrics over time is called the learning curve of that technology.
As the cumulative installed capacity increased, the price of solar declined exponentially. Solar technology is a prime example. For more than four decades, the price of solar panels declined by 20% with each doubling of global cumulative capacity.
The fact that both metrics changed exponentially can be nicely seen in this chart because both axes are logarithmic. On a logarithmic axis, a measure that declines exponentially follows a straight line.
That more production leads to falling prices is not surprising – such ‘economies of scale’ are found in the production of many goods. If you are already making one pizza, making a second one isn’t that much extra work.
What is exceptional about technologies that follow a learning curve is that this effect persists, and the rate at which the price declines stays roughly constant. This is what it means for a technology to follow Wright’s Law.
Solar power is not the only technology where we see trends of exponential change. The most famous case of exponential technological change is Moore’s Law – the observation of Intel’s co-founder Gordon Moore who noticed that the number of transistors on microprocessors doubled every two years.
We have another article on Moore’s Law on Our World in Data: What is Moore’s Law?
Integrated circuits are the fundamental technology of computers, and Moore’s Law has driven a range of changes in computer technology in recent decades – computers became rapidly cheaper, more energy efficient, and faster.
Moore’s Law, however, is not given in the same way that we just looked at for solar prices. Moore’s Law describes technological change as a function of time. In the example of solar technology we looked at price changes not as a function of time, but of experience – measured as the cumulative amount of solar panels that were ever installed.
This relationship that each doubling in experience leads to the same relative price decline was discovered earlier than Moore’s Law by aerospace engineer Theodore Paul Wright in 1936.{ref}Theodore Paul Wright (1936) – Factors affecting the cost of airplanes. J. Aeronaut. Sci., 3 (4) (1936), pp. 122-128 {/ref} It’s called Wright’s Law, after him.
Moore’s observation of the progress in computing technology can be seen as a special case of Wright’s Law.{ref}Plausibly, it isn’t just the passing of time that drives the progress in computer chips, but there too it is the learning that comes with continuously expanding the production of these chips. Lafond et al (2018) explain that the two laws produce the same forecasts when cumulative production grows exponentially, which is the case when production grows exponentially. More precisely, if production grows exponentially with some noise/fluctuations, then cumulative production grows exponentially with very little noise/fluctuations. As a result, the log of cumulative production is a linear trend and therefore predicting cost by the linear trend of time or the linear trend of log cumulative production give the same results.
Fracois Lafond, Aimee G. Bailey, Jan D. Bakker, Dylan Rebois, Rubina Zadourian, Patrick McSharry, and J. Doyne Farmer (2018) – How well do experience curves predict technological progress? A method for making distributional forecasts In Technological Forecasting and Social Change 128, pp 104-117, 2018. arXiv, Publisher, Data, Code.
See also Nagy B, Farmer JD, Bui QM, Trancik JE (2013) Statistical Basis for Predicting Technological Progress. PLoS ONE 8(2): e52669. https://doi.org/10.1371/journal.pone.0052669
Wright’s Law for solar PV modules has also been given its own name; some call it Swanson’s Law (Wiki).{/ref}
Solar panels are not the only technologies that follow this law. Look at our visualization of the price declines of 66 different technologies and the research referenced in the footnote{ref}Nagy B, Farmer JD, Bui QM, Trancik JE (2013) Statistical Basis for Predicting Technological Progress. PLoS ONE 8(2): e52669. https://doi.org/10.1371/journal.pone.0052669
Many more references can be found in Doyne Farmer and Fracois Lafond (2016) – How predictable is technological progress? Research Policy. Volume 45, Issue 3, April 2016, Pages 647-665. https://doi.org/10.1016/j.respol.2015.11.001
The price of Ford’s Model T followed Wright’s Law: each doubling of cumulative production led to the same relative decline in prices. What’s fascinating is that this decline hasn’t stopped until today. An 8hp car, as the Model T, costs what you’d expect: See Sam Korus (2019) – Wright’s Law Predicted 109 Years of Auto Production Costs, and Now Tesla’s {/ref}
The relative price decline associated with each doubling of cumulative experience is the learning rate of a technology.
The learning rate of solar panels is 20%. This means that with each doubling of the installed cumulative capacity, the price of solar panels declined by 20%.
In the footnote, you can find more information about the scientific literature on the learning rate in solar technology, and an example of how the learning rate is calculated.{ref}As one would expect, the exact learning rate for a given technology differs slightly across studies, mostly due to differences in the chosen data source, the chosen proxy measure for ‘experience’, the geographic location or the considered time-span.
To give the fairest estimate and avoid relying on one unusual data point I am therefore reporting an average across several experience curve studies for PV that was conducted by de La Tour et al. 2013. The authors find an average learning rate over many studies of 20.2% (see Table 1 of their publication).
de La Tour, A., Glachant, M. & Ménière, Y. (2013) – Predicting the costs of photovoltaic solar modules in 2020 using experience curve models. In Energy 62, 341–348.
The learning rate implied by the data that I’m presenting here is very similar (19.3%) and can be calculated as follows:
Cumulative capacity
– in 1976 0.3 MW
– in 2019 578,553 MW
Module Cost ($ per W)
– in 1976 106.09
– in 2019 0.377
The number of doublings of the capacity is: log2(578,553 / 0.3)=20.879
The rate of change of the price at each doubling is: (106.09 / 0.37725) ^ (1/(20.879)) - 1=0.31=31%
So the learning rate is 1-2^(-0.31)=0.193399911=19.3%{/ref}
The high learning rate meant that the price of solar declined dramatically. As the chart above showed, the price declined from $106 to $0.38 per watt in these four decades. A decline of 99.6%.
That the price of technology declines when more of that technology is produced is a classic case of learning by doing. Increasing production gives the engineers the chance to learn how to improve the process.
This effect creates a virtuous cycle of increasing demand and falling prices. More of that technology gets deployed to satisfy increasing demand, leading to falling prices. At those lower prices, the technology becomes cost-effective in new applications, which in turn means that demand increases. In this positive feedback loop, these technologies power themselves forward to lower and lower prices.
The specifics, of course, differ between the different technologies. For more information on what is behind the price reduction of solar panels, see the footnote.{ref}According to the research cited below it involved: larger, more efficient factories are producing the modules; R&D efforts increase; technological advances increase the efficiency of the panels; engineering advances improve the production processes of the silicon ingots and wafers; the mining and processing of the raw materials increases in scale and becomes cheaper; operational experience accumulates; the modules are more durable and live longer; market competition ensures that profits are low; and capital costs for the production decline. A myriad of small improvements across a large collective process drives this continuous price decline.
Kavlak, Goksin and McNerney, James and Trancik, Jessika E. (2017) – Evaluating the Causes of Cost Reduction in Photovoltaic Modules (August 9, 2017). In Energy Policy, 123:700-710, 2018, http://dx.doi.org/10.2139/ssrn.2891516{/ref}
How do we know that increasing experience is causing lower prices? After all, it could be the other way around: production only increases after costs have fallen.
In most settings, this is difficult to disentangle empirically, but researchers François Lafond, Diana Greenwald, and Doyne Farmer found an instance where this question can be answered. In their paper “Can Stimulating Demand Drive Costs Down?”, they study the price changes at a time when reverse causality can be ruled out: the demand for military technology in the Second World War. In that case it is clear that demand was driven by the circumstances of the war, and not by lower prices.{ref}Lafond, Francois and Greenwald, Diana Seave and Farmer, J. Doyne, Can Stimulating Demand Drive Costs Down? World War II as a Natural Experiment (June 1, 2020). http://dx.doi.org/10.2139/ssrn.3519913{/ref}
They found that as demand for weapons grew, production experience increased sharply, and prices declined. When the war was over and demand shrank, the price decline reverted back to a slower rate. It was the cumulative experience that drove a decline in prices, not the other way around.
If you want to know what the future looks like, one of the most useful questions to ask is which technologies follow a learning curve.
Most technologies do not follow Wright’s Law – the prices of bicycles, fridges, or coal power plants do not decline exponentially as we produce more of them. But those which do follow Wright’s Law – like computers, solar panels, and batteries – are the ones to look out for. In their infancy, they might only be found in very niche applications, but a few decades later they are everywhere.
This means that if you are unaware that a technology follows Wright’s Law, you can get your predictions very wrong. At the dawn of the computer age in 1943, IBM president Thomas Watson famously said, ""I think there is a world market for maybe five computers.""{ref}The first reference to Watson saying this is in an article from Der Spiegel from 26th May 1965 – Sieg der Mikrosekunde{/ref} At the price point of computers at the time, that was perhaps perfectly true, but what he didn’t foresee was how rapidly the price of computers would come down. From their initial niche computers expanded to more and more applications, and the virtuous cycle meant that the price of computers declined continuously. The exponential progress of computer technology expanded their use from a tiny niche to the defining technology of our time.
Solar panels are on the same trajectory as we’ve seen before. At the price of solar panels in the 1950s, it would have sounded quite reasonable to say, “I think there is a world market for maybe five solar panels.” But as a prediction for the future, this statement too, would have been ridiculously wrong.
To get our expectations about the future right, we are well-advised to take the exponential change of Wright’s Law seriously. Doyne Farmer, François Lafond, Penny Mealy, Rupert Way, Cameron Hepburn, and others have done important pioneering work in this field. A central paper of their work is Farmer’s and Lafond’s “How predictable is technological progress?”.{ref}Doyne Farmer and Fracois Lafond (2016) – How predictable is technological progress? Research Policy. Volume 45, Issue 3, April 2016, Pages 647-665. https://doi.org/10.1016/j.respol.2015.11.001
See also: de La Tour, A., Glachant, M. & Ménière, Y. (2013) – Predicting the costs of photovoltaic solar modules in 2020 using experience curve models. In Energy 62, 341–348.{/ref} The focus of this research paper is the price of solar, so that we avoid repeating Watson’s mistake with renewable energy. They lay out in detail what I discussed here: how solar panels decline in price, how demand drives this change, and how we can learn about the future by relying on these insights.
To get our expectations for the future right, we need to pay particular attention to the technologies that follow learning curves. Initially, we might only find them in a few high-tech applications, but the future belongs to them.
","{""id"": ""wp-56742"", ""slug"": ""learning-curve"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Our World in Data presents the data and research to make progress against the world’s largest problems."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""This article draws on data and research discussed in our entry on "", ""spanType"": ""span-simple-text""}, {""children"": [{""url"": ""https://ourworldindata.org/technological-change"", ""children"": [{""text"": ""Technological Change"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""spanType"": ""span-bold""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The best way to explain what this means is to look at a concrete example."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Solar technology: an example of a technology that follows Wright’s Law"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The time series in the chart shows the deployment of solar panels on the horizontal axis and the price of solar panels on the vertical axis. The orange line that describes the relationship between these two metrics over time is called the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""learning curve"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of that technology. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As the cumulative installed capacity increased, the price of solar "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""declined"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""children"": [{""text"": ""exponentially"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". Solar technology is a prime example. For more than four decades, the price of solar panels declined by 20% with each doubling of global cumulative capacity."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The fact that both metrics changed exponentially can be nicely seen in this chart because both axes are logarithmic. On a logarithmic axis, a measure that declines exponentially "", ""spanType"": ""span-simple-text""}, {""url"": ""https://blog.datawrapper.de/weeklychart-logscale/"", ""children"": [{""text"": ""follows a straight line"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That more production leads to falling prices is not surprising – such ‘economies of scale’ are found in the production of many goods. If you are already making one pizza, making a second one isn’t that much extra work."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""What is exceptional about technologies that follow a learning curve is that this effect persists, and "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""the rate at which the price declines stays roughly constant"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". This is what it means for a technology to follow Wright’s Law."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""solar-pv-prices-vs-cumulative-capacity.png"", ""parseErrors"": []}, {""text"": [{""text"": ""The relationship between the laws of Gordon Moore and Theodore Paul Wright"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Solar power is not the only technology where we see trends of exponential change. The most famous case of exponential technological change is Moore’s Law – the observation of Intel’s co-founder Gordon Moore who noticed that the number of transistors on microprocessors doubled every two years. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We have another article on Moore’s Law on Our World in Data: "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/moores-law"", ""children"": [{""text"": ""What is Moore’s Law?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Integrated circuits are the fundamental technology of computers, and Moore’s Law has driven a range of changes in computer technology in recent decades – computers became rapidly cheaper, more energy efficient, and faster."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Moore’s Law, however, is not given in the same way that we just looked at for solar prices. Moore’s Law describes technological change as a function of "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""time. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""In the example of solar technology we looked at price changes not as a function of time, but of "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""experience"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" – measured as the cumulative amount of solar panels that were ever installed. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This relationship that each doubling in experience leads to the same relative price decline was discovered earlier than Moore’s Law by aerospace engineer Theodore Paul Wright in 1936.{ref}Theodore Paul Wright (1936) – Factors affecting the cost of airplanes. J. Aeronaut. Sci., 3 (4) (1936), pp. 122-128 {/ref} It’s called "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Wright’s Law"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", after him. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Moore’s observation of the progress in computing technology can be seen as a special case of Wright’s Law.{ref}Plausibly, it isn’t just the passing of time that drives the progress in computer chips, but there too it is the learning that comes with continuously expanding the production of these chips. Lafond et al (2018) explain that the two laws produce the same forecasts when cumulative production grows exponentially, which is the case when production grows exponentially. More precisely, if production grows exponentially with some noise/fluctuations, then cumulative production grows exponentially with very little noise/fluctuations. As a result, the log of cumulative production is a linear trend and therefore predicting cost by the linear trend of time or the linear trend of log cumulative production give the same results. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Fracois Lafond, Aimee G. Bailey, Jan D. Bakker, Dylan Rebois, Rubina Zadourian, Patrick McSharry, and "", ""spanType"": ""span-simple-text""}, {""url"": ""http://www.doynefarmer.com/"", ""children"": [{""text"": ""J. Doyne Farmer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (2018) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://francoislafond.files.wordpress.com/2015/11/wrightslawpaper20.pdf"", ""children"": [{""text"": ""How well do experience curves predict technological progress? A method for making distributional forecasts"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" In Technological Forecasting and Social Change 128, pp 104-117, 2018. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.arxiv.org/abs/1703.05979"", ""children"": [{""text"": ""arXiv"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""url"": ""http://www.sciencedirect.com/science/article/pii/S0040162517303736"", ""children"": [{""text"": ""Publisher"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.dropbox.com/sh/w7jvzijblb4nkex/AAC2R-ml3JvIjFfBZtUTPlkta?dl=0"", ""children"": [{""text"": ""Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""url"": ""https://francoislafond.files.wordpress.com/2019/12/forecast_tech_progress-1.zip"", ""children"": [{""text"": ""Code"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""See also Nagy B, Farmer JD, Bui QM, Trancik JE (2013) Statistical Basis for Predicting Technological Progress. PLoS ONE 8(2): e52669. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1371/journal.pone.0052669"", ""children"": [{""text"": ""https://doi.org/10.1371/journal.pone.0052669"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""spanType"": ""span-newline""}, {""text"": ""Wright’s Law for solar PV modules has also been given its own name; some call it "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Swanson%27s_law"", ""children"": [{""text"": ""Swanson’s Law (Wiki)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Solar panels are not the only technologies that follow this law. Look at "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/costs-of-66-different-technologies-over-time"", ""children"": [{""text"": ""our visualization"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of the price declines of 66 different technologies and the research referenced in the footnote{ref}Nagy B, Farmer JD, Bui QM, Trancik JE (2013) Statistical Basis for Predicting Technological Progress. PLoS ONE 8(2): e52669. https://doi.org/10.1371/journal.pone.0052669"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""Many more references can be found in Doyne Farmer and Fracois Lafond (2016) – How predictable is technological progress? Research Policy. Volume 45, Issue 3, April 2016, Pages 647-665. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/j.respol.2015.11.001"", ""children"": [{""text"": ""https://doi.org/10.1016/j.respol.2015.11.001"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}], ""spanType"": ""span-link""}, {""text"": ""The price of Ford’s Model T followed Wright’s Law: each doubling of cumulative production led to the same relative decline in prices. What’s fascinating is that this decline hasn’t stopped until today. An 8hp car, as the Model T, costs what you’d expect: See Sam Korus (2019) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ark-invest.com/analyst-research/wrights-law-predicts-teslas-gross-margin/"", ""children"": [{""text"": ""Wright’s Law Predicted 109 Years of Auto Production Costs, and Now Tesla’s"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""The learning rate"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The relative price decline associated with each doubling of cumulative experience is the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""learning rate"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of a technology. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The learning rate of solar panels is 20%. This means that with each doubling of the installed cumulative capacity, the price of solar panels declined by 20%."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the footnote, you can find more information about the scientific literature on the learning rate in solar technology, and an example of how the learning rate is calculated.{ref}As one would expect, the exact learning rate for a given technology differs slightly across studies, mostly due to differences in the chosen data source, the chosen proxy measure for ‘experience’, the geographic location or the considered time-span."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To give the fairest estimate and avoid relying on one unusual data point I am therefore reporting an average across several experience curve studies for PV that was conducted by de La Tour et al. 2013. The authors find an average learning rate over many studies of 20.2% (see Table 1 of their publication)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""de La Tour, A., Glachant, M. & Ménière, Y. (2013) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.sciencedirect.com/science/article/abs/pii/S0360544213007883"", ""children"": [{""text"": ""Predicting the costs of photovoltaic solar modules in 2020 using experience curve models"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". In "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Energy"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" 62, 341–348."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The learning rate implied by the data that I’m presenting here is very similar (19.3%) and can be calculated as follows:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Cumulative capacity"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""– in 1976 0.3 MW"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""– in 2019 578,553 MW"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Module Cost ($ per W)"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""– in 1976 106.09"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""– in 2019 0.377"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The number of doublings of the capacity is: log2(578,553 / 0.3)=20.879"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""The rate of change of the price at each doubling is: (106.09 / 0.37725) ^ (1/(20.879)) - 1=0.31="", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""31%"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""children"": [{""spanType"": ""span-newline""}], ""spanType"": ""span-bold""}, {""text"": ""So the learning rate is 1-2^(-0.31)=0.193399911="", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""19.3%"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The high learning rate meant that the price of solar declined dramatically. As the chart above showed, the price declined from $106 to $0.38 per watt in these four decades. A decline of 99.6%."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How is this possible? And is the relationship between experience and price causal?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That the price of technology declines when more of that technology is produced is a classic case of learning by doing. Increasing production gives the engineers the chance to learn how to improve the process. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This effect creates a virtuous cycle of increasing demand and falling prices. More of that technology gets deployed to satisfy increasing demand, leading to falling prices. At those lower prices, the technology becomes cost-effective in new applications, which in turn means that demand increases. In this positive feedback loop, these technologies power themselves forward to lower and lower prices."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The specifics, of course, differ between the different technologies. For more information on what is behind the price reduction of solar panels, see the footnote.{ref}According to the research cited below it involved: larger, more efficient factories are producing the modules; R&D efforts increase; technological advances increase the efficiency of the panels; engineering advances improve the production processes of the silicon ingots and wafers; the mining and processing of the raw materials increases in scale and becomes cheaper; operational experience accumulates; the modules are more durable and live longer; market competition ensures that profits are low; and capital costs for the production decline. A myriad of small improvements across a large collective process drives this continuous price decline."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Kavlak, Goksin and McNerney, James and Trancik, Jessika E. (2017) – Evaluating the Causes of Cost Reduction in Photovoltaic Modules (August 9, 2017). In Energy Policy, 123:700-710, 2018, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://dx.doi.org/10.2139/ssrn.2891516"", ""children"": [{""text"": ""http://dx.doi.org/10.2139/ssrn.2891516"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How do we know that increasing experience is causing lower prices? After all, it could be the other way around: production only increases after costs have fallen."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In most settings, this is difficult to disentangle empirically, but researchers François Lafond, Diana Greenwald, and Doyne Farmer found an instance where this question can be answered. In their paper “Can Stimulating Demand Drive Costs Down?”, they study the price changes at a time when reverse causality can be ruled out: the demand for military technology in the Second World War. In that case it is clear that demand was driven by the circumstances of the war, and not by lower prices.{ref}Lafond, Francois and Greenwald, Diana Seave and Farmer, J. Doyne, Can Stimulating Demand Drive Costs Down? World War II as a Natural Experiment (June 1, 2020). "", ""spanType"": ""span-simple-text""}, {""url"": ""http://dx.doi.org/10.2139/ssrn.3519913"", ""children"": [{""text"": ""http://dx.doi.org/10.2139/ssrn.3519913"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""They found that as demand for weapons grew, production experience increased sharply, and prices declined. When the war was over and demand shrank, the price decline reverted back to a slower rate. It was the cumulative experience that drove a decline in prices, not the other way around."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What can we learn from the learning curve of a technology?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If you want to know what the future looks like, one of the most useful questions to ask is which technologies follow a learning curve."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Most technologies do not follow Wright’s Law – the prices of bicycles, fridges, or coal power plants do not decline exponentially as we produce more of them. But those which do follow Wright’s Law – like computers, solar panels, and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/battery-price-decline"", ""children"": [{""text"": ""batteries"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" – are the ones to look out for. In their infancy, they might only be found in very niche applications, but a few decades later they are everywhere."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This means that if you are unaware that a technology follows Wright’s Law, you can get your predictions very wrong. At the dawn of the computer age in 1943, IBM president Thomas Watson famously said, \""I think there is a world market for maybe five computers.\""{ref}The first reference to Watson saying this is in an article from Der Spiegel from 26th May 1965 – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.spiegel.de/spiegel/print/d-46272769.html"", ""children"": [{""text"": ""Sieg der Mikrosekunde"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} At the price point of computers at the time, that was perhaps perfectly true, but what he didn’t foresee was how rapidly the price of computers would come down. From their initial niche computers expanded to more and more applications, and the virtuous cycle meant that the price of computers declined continuously. The exponential progress of computer technology expanded their use from a tiny niche to the defining technology of our time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Solar panels are on the same trajectory as we’ve seen before. At the price of solar panels in the 1950s, it would have sounded quite reasonable to say, “I think there is a world market for maybe five solar panels.” But as a prediction for the future, this statement too, would have been "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/energy?tab=chart&facet=none&uniformYAxis=0&country=~OWID_WRL&Total+or+Breakdown=Select+a+source&Energy+or+Electricity=Electricity+only&Metric=Annual+generation&Select+a+source=Solar"", ""children"": [{""text"": ""ridiculously wrong"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To get our expectations about the future right, we are well-advised to take the exponential change of Wright’s Law seriously. Doyne Farmer, François Lafond, Penny Mealy, Rupert Way, Cameron Hepburn, and others have done important pioneering work in this field. A central paper of their work is Farmer’s and Lafond’s “How predictable is technological progress?”.{ref}Doyne Farmer and Fracois Lafond (2016) – How predictable is technological progress? Research Policy. Volume 45, Issue 3, April 2016, Pages 647-665. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/j.respol.2015.11.001"", ""children"": [{""text"": ""https://doi.org/10.1016/j.respol.2015.11.001"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}], ""spanType"": ""span-link""}, {""text"": ""See also: de La Tour, A., Glachant, M. & Ménière, Y. (2013) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.sciencedirect.com/science/article/abs/pii/S0360544213007883"", ""children"": [{""text"": ""Predicting the costs of photovoltaic solar modules in 2020 using experience curve models"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". In "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Energy"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" 62, 341–348.{/ref} The focus of this research paper is the price of solar, so that we avoid repeating Watson’s mistake with renewable energy. They lay out in detail what I discussed here: how solar panels decline in price, how demand drives this change, and how we can learn about the future by relying on these insights."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To get our expectations for the future right, we need to pay particular attention to the technologies that follow learning curves. Initially, we might only find them in a few high-tech applications, but the future belongs to them."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Learning curves: What does it mean for a technology to follow Wright’s Law?"", ""authors"": [""Max Roser""], ""excerpt"": ""Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases."", ""dateline"": ""April 18, 2023"", ""subtitle"": ""Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases."", ""sidebar-toc"": false, ""featured-image"": ""Screenshot-2023-04-17-at-20.57.11.png""}, ""createdAt"": ""2023-04-18T01:57:26.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T16:24:50.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-04-18T00:57:56.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}], ""numBlocks"": 47, ""numErrors"": 1, ""wpTagCounts"": {""html"": 1, ""image"": 1, ""heading"": 5, ""paragraph"": 41}, ""htmlTagCounts"": {""p"": 42, ""h4"": 5, ""div"": 1, ""figure"": 1}}",2023-04-18 00:57:56,2024-02-16 14:22:55,1THmmJVzdsf_Bl2xyf9xlYG4YZBamhvhD7SrlEmzOKlU,"[""Max Roser""]","Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases.",2023-04-18 01:57:26,2023-07-10 16:24:50,https://ourworldindata.org/wp-content/uploads/2023/04/Screenshot-2023-04-17-at-20.57.11.png,{},"Our World in Data presents the data and research to make progress against the world’s largest problems. This article draws on data and research discussed in our entry on **[Technological Change](https://ourworldindata.org/technological-change)**. Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases. The best way to explain what this means is to look at a concrete example. ## Solar technology: an example of a technology that follows Wright’s Law The time series in the chart shows the deployment of solar panels on the horizontal axis and the price of solar panels on the vertical axis. The orange line that describes the relationship between these two metrics over time is called the _learning curve_ of that technology. As the cumulative installed capacity increased, the price of solar _declined__exponentially_. Solar technology is a prime example. For more than four decades, the price of solar panels declined by 20% with each doubling of global cumulative capacity. The fact that both metrics changed exponentially can be nicely seen in this chart because both axes are logarithmic. On a logarithmic axis, a measure that declines exponentially [follows a straight line](https://blog.datawrapper.de/weeklychart-logscale/). That more production leads to falling prices is not surprising – such ‘economies of scale’ are found in the production of many goods. If you are already making one pizza, making a second one isn’t that much extra work. What is exceptional about technologies that follow a learning curve is that this effect persists, and _the rate at which the price declines stays roughly constant_. This is what it means for a technology to follow Wright’s Law.Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Technological Change.
Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases.
\n\n\n\nThe best way to explain what this means is to look at a concrete example.
\n\n\n\nThe time series in the chart shows the deployment of solar panels on the horizontal axis and the price of solar panels on the vertical axis. The orange line that describes the relationship between these two metrics over time is called the learning curve of that technology.
\n\n\n\nAs the cumulative installed capacity increased, the price of solar declined exponentially. Solar technology is a prime example. For more than four decades, the price of solar panels declined by 20% with each doubling of global cumulative capacity.
\n\n\n\nThe fact that both metrics changed exponentially can be nicely seen in this chart because both axes are logarithmic. On a logarithmic axis, a measure that declines exponentially follows a straight line.
\n\n\n\nThat more production leads to falling prices is not surprising – such ‘economies of scale’ are found in the production of many goods. If you are already making one pizza, making a second one isn’t that much extra work.
\n\n\n\nWhat is exceptional about technologies that follow a learning curve is that this effect persists, and the rate at which the price declines stays roughly constant. This is what it means for a technology to follow Wright’s Law.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2023/04/solar-pv-prices-vs-cumulative-capacity.png\"")
Solar power is not the only technology where we see trends of exponential change. The most famous case of exponential technological change is Moore’s Law – the observation of Intel’s co-founder Gordon Moore who noticed that the number of transistors on microprocessors doubled every two years.
\n\n\n\nWe have another article on Moore’s Law on Our World in Data: What is Moore’s Law?
\n\n\n\nIntegrated circuits are the fundamental technology of computers, and Moore’s Law has driven a range of changes in computer technology in recent decades – computers became rapidly cheaper, more energy efficient, and faster.
\n\n\n\nMoore’s Law, however, is not given in the same way that we just looked at for solar prices. Moore’s Law describes technological change as a function of time. In the example of solar technology we looked at price changes not as a function of time, but of experience – measured as the cumulative amount of solar panels that were ever installed.
\n\n\n\nThis relationship that each doubling in experience leads to the same relative price decline was discovered earlier than Moore’s Law by aerospace engineer Theodore Paul Wright in 1936.{ref}Theodore Paul Wright (1936) – Factors affecting the cost of airplanes. J. Aeronaut. Sci., 3 (4) (1936), pp. 122-128 {/ref} It’s called Wright’s Law, after him.
\n\n\n\nMoore’s observation of the progress in computing technology can be seen as a special case of Wright’s Law.{ref}Plausibly, it isn’t just the passing of time that drives the progress in computer chips, but there too it is the learning that comes with continuously expanding the production of these chips. Lafond et al (2018) explain that the two laws produce the same forecasts when cumulative production grows exponentially, which is the case when production grows exponentially. More precisely, if production grows exponentially with some noise/fluctuations, then cumulative production grows exponentially with very little noise/fluctuations. As a result, the log of cumulative production is a linear trend and therefore predicting cost by the linear trend of time or the linear trend of log cumulative production give the same results.
\n\n\n\nFracois Lafond, Aimee G. Bailey, Jan D. Bakker, Dylan Rebois, Rubina Zadourian, Patrick McSharry, and J. Doyne Farmer (2018) – How well do experience curves predict technological progress? A method for making distributional forecasts In Technological Forecasting and Social Change 128, pp 104-117, 2018. arXiv, Publisher, Data, Code.
\n\n\n\nSee also Nagy B, Farmer JD, Bui QM, Trancik JE (2013) Statistical Basis for Predicting Technological Progress. PLoS ONE 8(2): e52669. https://doi.org/10.1371/journal.pone.0052669
\n\n\n\n
Wright’s Law for solar PV modules has also been given its own name; some call it Swanson’s Law (Wiki).{/ref}
Solar panels are not the only technologies that follow this law. Look at our visualization of the price declines of 66 different technologies and the research referenced in the footnote{ref}Nagy B, Farmer JD, Bui QM, Trancik JE (2013) Statistical Basis for Predicting Technological Progress. PLoS ONE 8(2): e52669. https://doi.org/10.1371/journal.pone.0052669
Many more references can be found in Doyne Farmer and Fracois Lafond (2016) – How predictable is technological progress? Research Policy. Volume 45, Issue 3, April 2016, Pages 647-665. https://doi.org/10.1016/j.respol.2015.11.001
The price of Ford’s Model T followed Wright’s Law: each doubling of cumulative production led to the same relative decline in prices. What’s fascinating is that this decline hasn’t stopped until today. An 8hp car, as the Model T, costs what you’d expect: See Sam Korus (2019) – Wright’s Law Predicted 109 Years of Auto Production Costs, and Now Tesla’s {/ref}
The relative price decline associated with each doubling of cumulative experience is the learning rate of a technology.
\n\n\n\nThe learning rate of solar panels is 20%. This means that with each doubling of the installed cumulative capacity, the price of solar panels declined by 20%.
\n\n\n\nIn the footnote, you can find more information about the scientific literature on the learning rate in solar technology, and an example of how the learning rate is calculated.{ref}As one would expect, the exact learning rate for a given technology differs slightly across studies, mostly due to differences in the chosen data source, the chosen proxy measure for ‘experience’, the geographic location or the considered time-span.
\n\n\n\nTo give the fairest estimate and avoid relying on one unusual data point I am therefore reporting an average across several experience curve studies for PV that was conducted by de La Tour et al. 2013. The authors find an average learning rate over many studies of 20.2% (see Table 1 of their publication).
\n\n\n\nde La Tour, A., Glachant, M. & Ménière, Y. (2013) – Predicting the costs of photovoltaic solar modules in 2020 using experience curve models. In Energy 62, 341–348.
\n\n\n\nThe learning rate implied by the data that I’m presenting here is very similar (19.3%) and can be calculated as follows:
\n\n\n\nCumulative capacity
– in 1976 0.3 MW
– in 2019 578,553 MW
Module Cost ($ per W)
– in 1976 106.09
– in 2019 0.377
The number of doublings of the capacity is: log2(578,553 / 0.3)=20.879
The rate of change of the price at each doubling is: (106.09 / 0.37725) ^ (1/(20.879)) – 1=0.31=31%
So the learning rate is 1-2^(-0.31)=0.193399911=19.3%{/ref}
The high learning rate meant that the price of solar declined dramatically. As the chart above showed, the price declined from $106 to $0.38 per watt in these four decades. A decline of 99.6%.
\n\n\n\nThat the price of technology declines when more of that technology is produced is a classic case of learning by doing. Increasing production gives the engineers the chance to learn how to improve the process.
\n\n\n\nThis effect creates a virtuous cycle of increasing demand and falling prices. More of that technology gets deployed to satisfy increasing demand, leading to falling prices. At those lower prices, the technology becomes cost-effective in new applications, which in turn means that demand increases. In this positive feedback loop, these technologies power themselves forward to lower and lower prices.
\n\n\n\nThe specifics, of course, differ between the different technologies. For more information on what is behind the price reduction of solar panels, see the footnote.{ref}According to the research cited below it involved: larger, more efficient factories are producing the modules; R&D efforts increase; technological advances increase the efficiency of the panels; engineering advances improve the production processes of the silicon ingots and wafers; the mining and processing of the raw materials increases in scale and becomes cheaper; operational experience accumulates; the modules are more durable and live longer; market competition ensures that profits are low; and capital costs for the production decline. A myriad of small improvements across a large collective process drives this continuous price decline.
\n\n\n\nKavlak, Goksin and McNerney, James and Trancik, Jessika E. (2017) – Evaluating the Causes of Cost Reduction in Photovoltaic Modules (August 9, 2017). In Energy Policy, 123:700-710, 2018, http://dx.doi.org/10.2139/ssrn.2891516{/ref}
\n\n\n\nHow do we know that increasing experience is causing lower prices? After all, it could be the other way around: production only increases after costs have fallen.
\n\n\n\nIn most settings, this is difficult to disentangle empirically, but researchers François Lafond, Diana Greenwald, and Doyne Farmer found an instance where this question can be answered. In their paper “Can Stimulating Demand Drive Costs Down?”, they study the price changes at a time when reverse causality can be ruled out: the demand for military technology in the Second World War. In that case it is clear that demand was driven by the circumstances of the war, and not by lower prices.{ref}Lafond, Francois and Greenwald, Diana Seave and Farmer, J. Doyne, Can Stimulating Demand Drive Costs Down? World War II as a Natural Experiment (June 1, 2020). http://dx.doi.org/10.2139/ssrn.3519913{/ref}
\n\n\n\nThey found that as demand for weapons grew, production experience increased sharply, and prices declined. When the war was over and demand shrank, the price decline reverted back to a slower rate. It was the cumulative experience that drove a decline in prices, not the other way around.
\n\n\n\nIf you want to know what the future looks like, one of the most useful questions to ask is which technologies follow a learning curve.
\n\n\n\nMost technologies do not follow Wright’s Law – the prices of bicycles, fridges, or coal power plants do not decline exponentially as we produce more of them. But those which do follow Wright’s Law – like computers, solar panels, and batteries – are the ones to look out for. In their infancy, they might only be found in very niche applications, but a few decades later they are everywhere.
\n\n\n\nThis means that if you are unaware that a technology follows Wright’s Law, you can get your predictions very wrong. At the dawn of the computer age in 1943, IBM president Thomas Watson famously said, “I think there is a world market for maybe five computers.”{ref}The first reference to Watson saying this is in an article from Der Spiegel from 26th May 1965 – Sieg der Mikrosekunde{/ref} At the price point of computers at the time, that was perhaps perfectly true, but what he didn’t foresee was how rapidly the price of computers would come down. From their initial niche computers expanded to more and more applications, and the virtuous cycle meant that the price of computers declined continuously. The exponential progress of computer technology expanded their use from a tiny niche to the defining technology of our time.
\n\n\n\nSolar panels are on the same trajectory as we’ve seen before. At the price of solar panels in the 1950s, it would have sounded quite reasonable to say, “I think there is a world market for maybe five solar panels.” But as a prediction for the future, this statement too, would have been ridiculously wrong.
\n\n\n\nTo get our expectations about the future right, we are well-advised to take the exponential change of Wright’s Law seriously. Doyne Farmer, François Lafond, Penny Mealy, Rupert Way, Cameron Hepburn, and others have done important pioneering work in this field. A central paper of their work is Farmer’s and Lafond’s “How predictable is technological progress?”.{ref}Doyne Farmer and Fracois Lafond (2016) – How predictable is technological progress? Research Policy. Volume 45, Issue 3, April 2016, Pages 647-665. https://doi.org/10.1016/j.respol.2015.11.001
See also: de La Tour, A., Glachant, M. & Ménière, Y. (2013) – Predicting the costs of photovoltaic solar modules in 2020 using experience curve models. In Energy 62, 341–348.{/ref} The focus of this research paper is the price of solar, so that we avoid repeating Watson’s mistake with renewable energy. They lay out in detail what I discussed here: how solar panels decline in price, how demand drives this change, and how we can learn about the future by relying on these insights.
To get our expectations for the future right, we need to pay particular attention to the technologies that follow learning curves. Initially, we might only find them in a few high-tech applications, but the future belongs to them.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Technologies that follow Wright’s Law get cheaper at a consistent rate, as the cumulative production of that technology increases."", ""protected"": false}, ""date_gmt"": ""2023-04-18T00:57:56"", ""modified"": ""2023-07-10T17:24:50"", ""template"": """", ""categories"": [234], ""ping_status"": ""closed"", ""authors_name"": [""Max Roser""], ""modified_gmt"": ""2023-07-10T16:24:50"", ""comment_status"": ""closed"", ""featured_media"": 56744, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/04/Screenshot-2023-04-17-at-20.57.11-150x70.png"", ""medium_large"": ""/app/uploads/2023/04/Screenshot-2023-04-17-at-20.57.11-768x360.png""}}" 56732,Flu Explorer - Introductory text,flu-explorer-introductory-text,wp_block,publish,"With this Flu Explorer, we aim to provide a helpful resource for epidemiologists, infectious disease researchers, and public health experts to understand the global spread of the influenza virus.
While influenza is not always at the forefront of the public's mind, it is a significant cause of death globally — it causes 400,000 respiratory deaths each year on average.
Additionally, it is considered one of the leading candidates for a future pandemic. To detect these risks early, the world must closely monitor the situation. Accessible data must be available for those who can help understand and manage the potential threats.
This Flu Data Explorer differs from our widely-used infectious diseases projects, such as the COVID-19 Explorer and the Mpox Explorer. These tools are designed for a broad audience. Unfortunately, flu data is incomplete in many ways, making it harder to communicate. This tool is therefore designed for users with pre-existing knowledge to navigate effectively the complex data published by the World Health Organization.
The explorer also highlights the significant gaps in influenza data. It is an important reminder of the need to improve data collection and reporting.
Before navigating this Explorer, we recommend reading the key insights on our Influenza page; they provide more detail on how to interpret these metrics meaningfully.
","{""id"": ""wp-56732"", ""slug"": ""flu-explorer-introductory-text"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""Why we provide this Influenza Data Explorer"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""With this Flu Explorer, we aim to provide a helpful resource for epidemiologists, infectious disease researchers, and public health experts to understand the global spread of the influenza virus."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""While influenza is not always at the forefront of the public's mind, it is a significant cause of death globally — it causes "", ""spanType"": ""span-simple-text""}, {""url"": ""https://pubmed.ncbi.nlm.nih.gov/31673337/"", ""children"": [{""text"": ""400,000 respiratory deaths"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" each year on average."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Additionally, it is considered "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1038/s12276-021-00603-0"", ""children"": [{""text"": ""one of the leading candidates"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" for a future pandemic. To detect these risks early, the world must closely monitor the situation. Accessible data must be available for those who can help understand and manage the potential threats."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This Flu Data Explorer differs from our widely-used infectious diseases projects, such as the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/coronavirus-data-explorer"", ""children"": [{""text"": ""COVID-19 Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/monkeypox"", ""children"": [{""text"": ""Mpox Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". These tools are designed for a broad audience. Unfortunately, flu data is incomplete in many ways, making it harder to communicate. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""This tool is therefore designed for users with pre-existing knowledge to navigate effectively the complex data published by the World Health Organization."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The explorer also highlights the significant gaps in influenza data. It is an important reminder of the need to improve data collection and reporting."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Before navigating this Explorer, we recommend reading the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/influenza#key-insights-on-influenza"", ""children"": [{""text"": ""key insights"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on our Influenza page; they provide more detail on how to interpret these metrics meaningfully."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Flu Explorer - Introductory text"", ""authors"": [null], ""dateline"": ""April 13, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-04-13T14:59:19.000Z"", ""published"": false, ""updatedAt"": ""2023-04-26T12:58:26.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-04-13T13:59:19.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 7, ""numErrors"": 0, ""wpTagCounts"": {""heading"": 1, ""paragraph"": 6}, ""htmlTagCounts"": {""p"": 6, ""h2"": 1}}",2023-04-13 13:59:19,2024-02-16 14:23:03,,[null],,2023-04-13 14:59:19,2023-04-26 12:58:26,,{},"## Why we provide this Influenza Data Explorer **With this Flu Explorer, we aim to provide a helpful resource for epidemiologists, infectious disease researchers, and public health experts to understand the global spread of the influenza virus.** While influenza is not always at the forefront of the public's mind, it is a significant cause of death globally — it causes [400,000 respiratory deaths](https://pubmed.ncbi.nlm.nih.gov/31673337/) each year on average. Additionally, it is considered [one of the leading candidates](https://doi.org/10.1038/s12276-021-00603-0) for a future pandemic. To detect these risks early, the world must closely monitor the situation. Accessible data must be available for those who can help understand and manage the potential threats. This Flu Data Explorer differs from our widely-used infectious diseases projects, such as the [COVID-19 Explorer](https://ourworldindata.org/explorers/coronavirus-data-explorer) and the [Mpox Explorer](https://ourworldindata.org/explorers/monkeypox). These tools are designed for a broad audience. Unfortunately, flu data is incomplete in many ways, making it harder to communicate. **This tool is therefore designed for users with pre-existing knowledge to navigate effectively the complex data published by the World Health Organization.** The explorer also highlights the significant gaps in influenza data. It is an important reminder of the need to improve data collection and reporting. Before navigating this Explorer, we recommend reading the [key insights](https://ourworldindata.org/influenza#key-insights-on-influenza) on our Influenza page; they provide more detail on how to interpret these metrics meaningfully.","{""data"": {""wpBlock"": {""content"": ""\nWith this Flu Explorer, we aim to provide a helpful resource for epidemiologists, infectious disease researchers, and public health experts to understand the global spread of the influenza virus.
\n\n\n\nWhile influenza is not always at the forefront of the public’s mind, it is a significant cause of death globally — it causes 400,000 respiratory deaths each year on average.
\n\n\n\nAdditionally, it is considered one of the leading candidates for a future pandemic. To detect these risks early, the world must closely monitor the situation. Accessible data must be available for those who can help understand and manage the potential threats.
\n\n\n\nThis Flu Data Explorer differs from our widely-used infectious diseases projects, such as the COVID-19 Explorer and the Mpox Explorer. These tools are designed for a broad audience. Unfortunately, flu data is incomplete in many ways, making it harder to communicate. This tool is therefore designed for users with pre-existing knowledge to navigate effectively the complex data published by the World Health Organization.
\n\n\n\nThe explorer also highlights the significant gaps in influenza data. It is an important reminder of the need to improve data collection and reporting.
\n\n\n\nBefore navigating this Explorer, we recommend reading the key insights on our Influenza page; they provide more detail on how to interpret these metrics meaningfully.
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 56724,We published a redesign of our work on the Internet,overhaul-internet,post,publish,,"{""id"": ""wp-56724"", ""slug"": ""overhaul-internet"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a redesign of our work on the Internet"", ""authors"": [""Hannah Ritchie"", ""Edouard Mathieu""], ""excerpt"": ""Explore our updated design, charts, and writing on global data on the Internet."", ""dateline"": ""April 13, 2023"", ""subtitle"": ""Explore our updated design, charts, and writing on global data on the Internet."", ""sidebar-toc"": false, ""featured-image"": ""Internet.png""}, ""createdAt"": ""2023-04-13T09:13:24.000Z"", ""published"": false, ""updatedAt"": ""2023-04-26T09:31:55.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-04-13T08:13:24.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-04-13 08:13:24,2024-02-16 14:22:55,,"[""Hannah Ritchie"", ""Edouard Mathieu""]","Explore our updated design, charts, and writing on global data on the Internet.",2023-04-13 09:13:24,2023-04-26 09:31:55,https://ourworldindata.org/wp-content/uploads/2023/04/Internet.png,{},,"{""id"": 56724, ""date"": ""2023-04-13T09:13:24"", ""guid"": {""rendered"": ""https://owid.cloud/?p=56724""}, ""link"": ""https://owid.cloud/overhaul-internet"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""overhaul-internet"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a redesign of our work on the Internet""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56724""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=56724"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=56724"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=56724"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=56724""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56724/revisions"", ""count"": 4}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/56553"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 56775, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56724/revisions/56775""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore our updated design, charts, and writing on global data on the Internet."", ""protected"": false}, ""date_gmt"": ""2023-04-13T08:13:24"", ""modified"": ""2023-04-26T10:31:55"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Edouard Mathieu""], ""modified_gmt"": ""2023-04-26T09:31:55"", ""comment_status"": ""closed"", ""featured_media"": 56553, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/04/Internet-150x79.png"", ""medium_large"": ""/app/uploads/2023/04/Internet-768x403.png""}}" 56663,Population Growth,population-growth,page,publish,"Population growth is one of the most important topics we cover at Our World in Data.
For most of human history, the global population was a tiny fraction of what it is today. Over the last few centuries, the human population has gone through an extraordinary change. In 1800, there were one billion people. Today there are more than 8 billion of us.
But after a period of very fast population growth, demographers expect the world population to peak by the end of this century.
On this page, you will find all of our data, charts, and writing on changes in population growth. This includes how populations are distributed worldwide, how this has changed, and what demographers expect for the future.
Related topics
Geographical maps show us where the world's landmasses are; not where people are. That means they don't always give us an accurate picture of how global living standards are changing.
One way to understand the distribution of people worldwide is to redraw the world map – not based on the area but according to population.
This is shown here as a population cartogram: a geographical presentation of the world where the size of countries is not drawn according to the distribution of land but by the distribution of people. It’s shown for the year 2018.
As the population size rather than the territory is shown in this map, you can see some significant differences when you compare it to the standard geographical map we’re most familiar with.
Small countries with a high population density increase in size in this cartogram relative to the world maps we are used to – look at Bangladesh, Taiwan, or the Netherlands. Large countries with a small population shrink in size – look for Canada, Mongolia, Australia, or Russia.
You can find more details on this cartogram in our article about it:
The speed of global population growth over the last few centuries has been staggering. For most of human history, the world population was well under one million.{ref}See, for example, Kremer (1993) – Population growth and technological change: one million BC to 1990. In the Quarterly Journal of Economics, Vol. 108, No. 3, 681-716.{/ref}
As recently as 12,000 years ago, there were only 4 million people worldwide.
The chart shows the rapid increase in the global population since 1700.
The one-billion mark wasn’t broken until the early 1800s. It was only a century ago that there were 2 billion people.
Since then, the global population has quadrupled to eight billion.
Around 108 billion people have ever lived on our planet. This means that today’s population size makes up 6.5% of the total number of people ever born.{ref}As per 2011 estimates from Carl Haub (2011), “How Many People Have Ever Lived on Earth?” Population Reference Bureau.{/ref}
This increase has been the result of advances in living conditions and health that reduced death rates – especially in children – and increases in life expectancy.
There’s a popular misconception that the global population is growing exponentially. But it’s not.
While the global population is still increasing in absolute numbers, population growth peaked decades ago.
In the chart, we see the global population growth rate per year. This is based on historical UN estimates and its medium projection to 2100.
Global population growth peaked in the 1960s at over 2% per year. Since then, rates have more than halved, falling to less than 1%.
The UN expects rates to continue to fall until the end of the century. In fact, towards the end of the century, it projects negative growth, meaning the global population will shrink instead of grow.
Global population growth, in absolute terms – which is the number of births minus the number of deaths – has also peaked. You can see this in our interactive chart:
Hans Rosling famously coined the term ""peak child"" for the moment in global demographic history when the number of children stopped increasing.
According to the UN data, the world has passed ""peak child"", which is defined as the number of children under the age of five.
The chart shows the UN’s historical estimates and projections of the number of children under five.
It estimates that the number of children in the world peaked in 2017. For the coming decades, demographers expect a decades-long plateau before the number will decline more rapidly in the second half of the century.
When will population growth come to an end?
The UN’s historical estimates and latest projections for the global population are shown in the chart.
The UN projects that the global population will peak before the end of the century – in 2086, at just over 10.4 billion people.
What would the work look like if each country's area was in proportion to its population?
Max Roser
The world population has increased rapidly in recent centuries. But this is slowing.
Max Roser and Hannah Ritchie
Hannah Ritchie and Edouard Mathieu
Hannah Ritchie, Edouard Mathieu and Lucas Rodés-Guirao
Hannah Ritchie and Edouard Mathieu
Max Roser and Hannah Ritchie
Hannah Ritchie
Hannah Ritchie
Max Roser
Max Roser
Max Roser
Edouard Mathieu and Lucas Rodés-Guirao
Hannah Ritchie
Joe Hasell
Joe Hasell
Population growth is one of the most important topics we cover at Our World in Data.
\n\n\n\nFor most of human history, the global population was a tiny fraction of what it is today. Over the last few centuries, the human population has gone through an extraordinary change. In 1800, there were one billion people. Today there are more than 8 billion of us.
\n\n\n\nBut after a period of very fast population growth, demographers expect the world population to peak by the end of this century.
\n\n\n\nOn this page, you will find all of our data, charts, and writing on changes in population growth. This includes how populations are distributed worldwide, how this has changed, and what demographers expect for the future.
\nRelated topics
\n\n\n\n\nGeographical maps show us where the world’s landmasses are; not where people are. That means they don’t always give us an accurate picture of how global living standards are changing.
\n\n\n\nOne way to understand the distribution of people worldwide is to redraw the world map – not based on the area but according to population.
\n\n\n\nThis is shown here as a population cartogram: a geographical presentation of the world where the size of countries is not drawn according to the distribution of land but by the distribution of people. It’s shown for the year 2018.
\n\n\n\nAs the population size rather than the territory is shown in this map, you can see some significant differences when you compare it to the standard geographical map we’re most familiar with.
\n\n\n\nSmall countries with a high population density increase in size in this cartogram relative to the world maps we are used to – look at Bangladesh, Taiwan, or the Netherlands. Large countries with a small population shrink in size – look for Canada, Mongolia, Australia, or Russia.
\n\n\n\nYou can find more details on this cartogram in our article about it:
\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2021/12/Population-cartogram_World-800x306.png\"")
The speed of global population growth over the last few centuries has been staggering. For most of human history, the world population was well under one million.{ref}See, for example, Kremer (1993) – Population growth and technological change: one million BC to 1990. In the Quarterly Journal of Economics, Vol. 108, No. 3, 681-716.{/ref}
\n\n\n\nAs recently as 12,000 years ago, there were only 4 million people worldwide.
\n\n\n\nThe chart shows the rapid increase in the global population since 1700.
\n\n\n\nThe one-billion mark wasn’t broken until the early 1800s. It was only a century ago that there were 2 billion people.
\n\n\n\nSince then, the global population has quadrupled to eight billion.
\n\n\n\nAround 108 billion people have ever lived on our planet. This means that today’s population size makes up 6.5% of the total number of people ever born.{ref}As per 2011 estimates from Carl Haub (2011), “How Many People Have Ever Lived on Earth?” Population Reference Bureau.{/ref}
\n\n\n\nThis increase has been the result of advances in living conditions and health that reduced death rates – especially in children – and increases in life expectancy.
\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2023/07/Annual-World-Population-since-10-thousand-BCE-1-584x550.png\"")
There’s a popular misconception that the global population is growing exponentially. But it’s not.
\n\n\n\nWhile the global population is still increasing in absolute numbers, population growth peaked decades ago.
\n\n\n\nIn the chart, we see the global population growth rate per year. This is based on historical UN estimates and its medium projection to 2100.
\n\n\n\nGlobal population growth peaked in the 1960s at over 2% per year. Since then, rates have more than halved, falling to less than 1%.
\n\n\n\nThe UN expects rates to continue to fall until the end of the century. In fact, towards the end of the century, it projects negative growth, meaning the global population will shrink instead of grow.
\n\n\n\nGlobal population growth, in absolute terms – which is the number of births minus the number of deaths – has also peaked. You can see this in our interactive chart:
\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2019/06/2019-Revision-–-World-Population-Growth-1700-2100.png\"")
Hans Rosling famously coined the term “peak child” for the moment in global demographic history when the number of children stopped increasing.
\n\n\n\nAccording to the UN data, the world has passed “peak child”, which is defined as the number of children under the age of five.
\n\n\n\nThe chart shows the UN’s historical estimates and projections of the number of children under five.
\n\n\n\nIt estimates that the number of children in the world peaked in 2017. For the coming decades, demographers expect a decades-long plateau before the number will decline more rapidly in the second half of the century.
\n\n\n\n\n\n\n\tWhen will population growth come to an end?
\n\n\n\nThe UN’s historical estimates and latest projections for the global population are shown in the chart.
\n\n\n\nThe UN projects that the global population will peak before the end of the century – in 2086, at just over 10.4 billion people.
\n\n\n\tHannah Ritchie and Edouard Mathieu
\nHannah Ritchie, Edouard Mathieu and Lucas Rodés-Guirao
\nHannah Ritchie and Edouard Mathieu
\nAge standardization can be helpful when comparing rates of health outcomes (for example, deaths from cancer) between populations, because age can be a significant risk factor for many diseases.
Since age structure varies between countries and in the same country over time, this adjustment allows us to see how mortality and morbidity vary without age differences.
Age standardization involves adjusting the observed rates of a particular outcome to a “standard population” with a specific age structure.
The proportion of people in different age groups in a country is called its ""age structure."" Countries across the world have significant differences in age structure. For example, only 2% of the population of Uganda is older than 65, against 30% in Japan.
But many diseases or health outcomes occur more frequently as people get older: for example, cancers, heart disease, or dementia.
This difference in age structure means that a condition that affects older people will tend to occur more frequently in the Japanese population than in the Ugandan population, simply because of the higher proportion of older people in the country.
Standardizing the data for age can tell us how a health condition compares between Japan and Uganda if there were no differences in the age structure of the two countries.
A crude rate is the unadjusted rate of a health outcome before any age standardization is applied.
Suppose a researcher wants to compare the death rate from cancer between Japan and Uganda. They would first calculate the death rate of people from cancer in each country.
To calculate this, for each year, the researcher would take the number of people who died of cancer and divide it by the country's population. This is called the crude rate.
This chart shows the crude death rate from cancer in Uganda and Japan, based on Institute for Health Metrics and Evaluation (IHME) data. It is expressed per 100,000 people in the population.
A crude rate is not inaccurate but can give us a biased view of the situation. By comparing the rates of cancer deaths between Uganda and Japan, we may conclude that something in Japan (such as nutrition, lifestyle, or healthcare) is causing many more people to die from cancer than in Uganda.
But we cannot make such a statement based on the crude rate, because the difference in the death rates may be due to Japan’s higher proportion of older people compared to Uganda.
If we want to accurately compare the rates of a health outcome that becomes more frequent with age, we need to apply age standardization by adjusting these rates to a standard population.
We need a “standard population” to apply age standardization. A standard population is a specific age structure used as a reference for these adjustments.
Many different standard populations can be used. Some of the most commonly used ones are the World Standard Population (published by the WHO) and the European Standard Population (published by Eurostat).
There are also national standard populations that can be used to compare health outcomes between different locations within a country. For example, the US National Cancer Institute maintains a standard population for the United States.
In a standard population, age groups are often broken down into 5-year intervals. Next to each age group, the proportion that the age group represents in the standard population is given.
For example, in the WHO’s World Standard Population, the under-5 age group represents 8.86% of the standard population, the 5 to 9 years age group represents 8.69% of the standard population, etc.
Age standardization is applied in three steps.
(1) Researchers start with the observed rates of the relevant health outcome for each age group. In our example, these would be the cancer death rates for each age group in Japan and Uganda.
For example, if we have 5-year intervals in our standard population, we would start from the cancer death rates of people in Japan aged 0–4, 5–9, 10–14, etc. These rates are called age-specific rates.
(2) These rates are then multiplied by the proportion of the corresponding age group in the standard population. For example, as the group aged 60–64 represents 3.72% of the World Standard Population, we would multiply the age-specific rate for this group by 3.72%.
(3) After multiplying all age-specific rates by the corresponding proportion of each age group in the standard population, we sum the rates to obtain one number. This number is the age-standardized rate.{ref}The national statistical office of Canada provides more details and a full example of how to calculate age-standardized rates.{/ref}
These age-standardized rates allow us to compare cancer death rates in Uganda and Japan as if they applied to two countries with identical age structures.
In this example, we see that Uganda's age-standardized cancer death rate has surpassed Japan's since 2000.
Age-standardized rates provide a better comparison between populations with different age structures but do not represent the actual rate in each population. Using the age-standardized rate to say that 114 per 100,000 people in the Japanese population die from cancer each year would be inaccurate. We should refer back to the crude rate (346 per 100,000) for this.
Age standardization is often used to compare health metrics between countries.
But we can also use it to measure health outcomes in the same location over time. Again, if we have a disease that primarily affects older people, we expect rates of that disease to rise over time as populations age.
Age standardization then allows us to see how the rate would have changed if the population's age structure hadn’t changed.
This can have a significant impact on our interpretation of health outcomes. One example of this comes from global death rates from cancer. On this chart, you can see that if we look at crude death rates (without age standardization), it looks like they have increased a lot in recent decades.
But a big part of this effect is explained by the fact that the world is getting older. When we adjust for age, we see that death rates have fallen.
In this related post, we discuss this chart – and the question of whether or not the world is making progress against cancer – in more detail.
","{""id"": ""wp-56533"", ""slug"": ""age-standardization"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Age standardization can be helpful when comparing rates of health outcomes (for example, deaths from cancer) between populations, because age can be a significant risk factor for many diseases."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Since age structure varies between countries and in the same country over time, this adjustment allows us to see how mortality and morbidity vary without age differences. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Age standardization involves adjusting the observed rates of a particular outcome to a “standard population” with a specific age structure."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Why is it helpful to standardize health metrics for age?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The proportion of people in different age groups in a country is called its \""age structure.\"" Countries across the world have significant differences in age structure. For example, only "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/population-by-broad-age-group?stackMode=relative&country=~UGA"", ""children"": [{""text"": ""2% of the population of Uganda"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" is older than 65, against "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/population-by-broad-age-group?stackMode=relative&country=~JPN"", ""children"": [{""text"": ""30% in Japan"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But many diseases or health outcomes occur more frequently as people get older: for example, cancers, heart disease, or dementia."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This difference in age structure means that a condition that affects older people will tend to occur more frequently in the Japanese population than in the Ugandan population, simply because of the higher proportion of older people in the country."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Standardizing the data for age can tell us how a health condition compares between Japan and Uganda if there were no differences in the age structure of the two countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What is a crude rate?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A crude rate is the unadjusted rate of a health outcome before any age standardization is applied."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Suppose a researcher wants to compare the death rate from cancer between Japan and Uganda. They would first calculate the death rate of people from cancer in each country."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To calculate this, for each year, the researcher would take the number of people who died of cancer and divide it by the country's population. This is called the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""crude rate"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This chart shows the crude death rate from cancer in Uganda and Japan, based on Institute for Health Metrics and Evaluation (IHME) data. It is expressed per 100,000 people in the population."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""death-rate-from-cancer-gbd.png"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A crude rate is not inaccurate but can give us a biased view of the situation. By comparing the rates of cancer deaths between Uganda and Japan, we may conclude that something in Japan (such as nutrition, lifestyle, or healthcare) is causing many more people to die from cancer than in Uganda."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But we cannot make such a statement based on the crude rate, because the difference in the death rates may be due to Japan’s higher proportion of older people compared to Uganda."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If we want to accurately compare the rates of a health outcome that becomes more frequent with age, we need to apply age standardization by adjusting these rates to a standard population."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What is a standard population?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We need a “standard population” to apply age standardization. A standard population is a specific age structure used as a reference for these adjustments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many different standard populations can be used. Some of the most commonly used ones are the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.researchgate.net/publication/284696312_Age_Standardization_of_Rates_A_New_WHO_Standard"", ""children"": [{""text"": ""World Standard Population"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (published by the WHO) and the "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ec.europa.eu/eurostat/documents/3859598/5926869/KS-RA-13-028-EN.PDF/e713fa79-1add-44e8-b23d-5e8fa09b3f8f"", ""children"": [{""text"": ""European Standard Population"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (published by Eurostat)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are also national standard populations that can be used to compare health outcomes between different locations within a country. For example, the US National Cancer Institute maintains "", ""spanType"": ""span-simple-text""}, {""url"": ""https://seer.cancer.gov/stdpopulations/"", ""children"": [{""text"": ""a standard population"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" for the United States."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In a standard population, age groups are often broken down into 5-year intervals. Next to each age group, the proportion that the age group represents in the standard population is given."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""For example, in the WHO’s World Standard Population, the under-5 age group represents 8.86% of the standard population, the 5 to 9 years age group represents 8.69% of the standard population, etc. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""the-world-health-organizations-world-standard-population1.png"", ""parseErrors"": []}, {""text"": [{""text"": ""How does age standardization work?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Age standardization is applied in three steps."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""(1) Researchers start with the observed rates of the relevant health outcome for each age group. In our example, these would be the cancer death rates for each age group in Japan and Uganda."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""For example, if we have 5-year intervals in our standard population, we would start from the cancer death rates of people in Japan aged 0–4, 5–9, 10–14, etc. These rates are called "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""age-specific rates"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""(2) These rates are then multiplied by the proportion of the corresponding age group in the standard population. For example, as the group aged 60–64 represents 3.72% of the World Standard Population, we would multiply the age-specific rate for this group by 3.72%."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""(3) After multiplying all age-specific rates by the corresponding proportion of each age group in the standard population, we sum the rates to obtain one number. This number is the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""age-standardized rate"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "".{ref}The national statistical office of Canada provides "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20230402125741/https://www.statcan.gc.ca/en/dai/btd/asr"", ""children"": [{""text"": ""more details and a full example"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of how to calculate age-standardized rates.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These age-standardized rates allow us to compare cancer death rates in Uganda and Japan as if they applied to two countries with identical age structures."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this example, we see that Uganda's age-standardized cancer death rate has surpassed Japan's since 2000."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""cancer-death-rates.png"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Age-standardized rates provide a better comparison between populations with different age structures but do not represent the actual rate in each population. Using the age-standardized rate to say that 114 per 100,000 people in the Japanese population die from cancer each year would be inaccurate. We should refer back to the crude rate (346 per 100,000) for this."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Age standardization can also be used across time or within a country"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Age standardization is often used to compare health metrics between countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But we can also use it to measure health outcomes in the same location over time. Again, if we have a disease that primarily affects older people, we expect rates of that disease to rise over time as populations age."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Age standardization then allows us to see how the rate would have changed if the population's age structure hadn’t changed."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This can have a significant impact on our interpretation of health outcomes. One example of this comes from global death rates from cancer. On this chart, you can see that if we look at crude death rates (without age standardization), it looks like they have increased a lot in recent decades."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/cancer-deaths-rate-and-age-standardized-rate-index"", ""type"": ""chart"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But a big part of this effect is explained by the fact that the world is getting older. When we adjust for age, we see that death rates have "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""fallen"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/progress-against-cancer"", ""children"": [{""text"": ""In this related post"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", we discuss this chart – and the question of whether or not the world is making progress against cancer – in more detail."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""How does age standardization make health metrics comparable?"", ""authors"": [""Edouard Mathieu""], ""excerpt"": ""Age standardization is a statistical method used to compare disease rates, or other health indicators, between populations while accounting for differences in their age structure."", ""dateline"": ""April 4, 2023"", ""subtitle"": ""Age standardization is a statistical method used to compare disease rates, or other health indicators, between populations while accounting for differences in their age structure."", ""sidebar-toc"": false, ""featured-image"": ""Living-planet-index-region-thumbnail.png""}, ""createdAt"": ""2023-04-04T10:42:16.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T16:23:55.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-04-04T09:58:49.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}], ""numBlocks"": 42, ""numErrors"": 3, ""wpTagCounts"": {""html"": 1, ""image"": 3, ""heading"": 5, ""paragraph"": 33}, ""htmlTagCounts"": {""p"": 33, ""h4"": 5, ""figure"": 3, ""iframe"": 1}}",2023-04-04 09:58:49,2024-03-12 20:21:39,1abYtLfS2RLS4Ph2pyHXcuTg61vhxFMeUybfUHtKqMnE,"[""Edouard Mathieu""]","Age standardization is a statistical method used to compare disease rates, or other health indicators, between populations while accounting for differences in their age structure.",2023-04-04 10:42:16,2023-07-10 16:23:55,https://ourworldindata.org/wp-content/uploads/2022/10/Living-planet-index-region-thumbnail.png,{},"Age standardization can be helpful when comparing rates of health outcomes (for example, deaths from cancer) between populations, because age can be a significant risk factor for many diseases. Since age structure varies between countries and in the same country over time, this adjustment allows us to see how mortality and morbidity vary without age differences. Age standardization involves adjusting the observed rates of a particular outcome to a “standard population” with a specific age structure. ## Why is it helpful to standardize health metrics for age? The proportion of people in different age groups in a country is called its ""age structure."" Countries across the world have significant differences in age structure. For example, only [2% of the population of Uganda](https://ourworldindata.org/grapher/population-by-broad-age-group?stackMode=relative&country=~UGA) is older than 65, against [30% in Japan](https://ourworldindata.org/grapher/population-by-broad-age-group?stackMode=relative&country=~JPN). But many diseases or health outcomes occur more frequently as people get older: for example, cancers, heart disease, or dementia. This difference in age structure means that a condition that affects older people will tend to occur more frequently in the Japanese population than in the Ugandan population, simply because of the higher proportion of older people in the country. Standardizing the data for age can tell us how a health condition compares between Japan and Uganda if there were no differences in the age structure of the two countries. ## What is a crude rate? A crude rate is the unadjusted rate of a health outcome before any age standardization is applied. Suppose a researcher wants to compare the death rate from cancer between Japan and Uganda. They would first calculate the death rate of people from cancer in each country. To calculate this, for each year, the researcher would take the number of people who died of cancer and divide it by the country's population. This is called the **crude rate**. This chart shows the crude death rate from cancer in Uganda and Japan, based on Institute for Health Metrics and Evaluation (IHME) data. It is expressed per 100,000 people in the population.Age standardization can be helpful when comparing rates of health outcomes (for example, deaths from cancer) between populations, because age can be a significant risk factor for many diseases.
\n\n\n\nSince age structure varies between countries and in the same country over time, this adjustment allows us to see how mortality and morbidity vary without age differences.
\n\n\n\nAge standardization involves adjusting the observed rates of a particular outcome to a “standard population” with a specific age structure.
\n\n\n\nThe proportion of people in different age groups in a country is called its “age structure.” Countries across the world have significant differences in age structure. For example, only 2% of the population of Uganda is older than 65, against 30% in Japan.
\n\n\n\nBut many diseases or health outcomes occur more frequently as people get older: for example, cancers, heart disease, or dementia.
\n\n\n\nThis difference in age structure means that a condition that affects older people will tend to occur more frequently in the Japanese population than in the Ugandan population, simply because of the higher proportion of older people in the country.
\n\n\n\nStandardizing the data for age can tell us how a health condition compares between Japan and Uganda if there were no differences in the age structure of the two countries.
\n\n\n\nA crude rate is the unadjusted rate of a health outcome before any age standardization is applied.
\n\n\n\nSuppose a researcher wants to compare the death rate from cancer between Japan and Uganda. They would first calculate the death rate of people from cancer in each country.
\n\n\n\nTo calculate this, for each year, the researcher would take the number of people who died of cancer and divide it by the country’s population. This is called the crude rate.
\n\n\n\nThis chart shows the crude death rate from cancer in Uganda and Japan, based on Institute for Health Metrics and Evaluation (IHME) data. It is expressed per 100,000 people in the population.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2023/04/death-rate-from-cancer-gbd-779x550.png\"")
A crude rate is not inaccurate but can give us a biased view of the situation. By comparing the rates of cancer deaths between Uganda and Japan, we may conclude that something in Japan (such as nutrition, lifestyle, or healthcare) is causing many more people to die from cancer than in Uganda.
\n\n\n\nBut we cannot make such a statement based on the crude rate, because the difference in the death rates may be due to Japan’s higher proportion of older people compared to Uganda.
\n\n\n\nIf we want to accurately compare the rates of a health outcome that becomes more frequent with age, we need to apply age standardization by adjusting these rates to a standard population.
\n\n\n\nWe need a “standard population” to apply age standardization. A standard population is a specific age structure used as a reference for these adjustments.
\n\n\n\nMany different standard populations can be used. Some of the most commonly used ones are the World Standard Population (published by the WHO) and the European Standard Population (published by Eurostat).
\n\n\n\nThere are also national standard populations that can be used to compare health outcomes between different locations within a country. For example, the US National Cancer Institute maintains a standard population for the United States.
\n\n\n\nIn a standard population, age groups are often broken down into 5-year intervals. Next to each age group, the proportion that the age group represents in the standard population is given.
\n\n\n\nFor example, in the WHO’s World Standard Population, the under-5 age group represents 8.86% of the standard population, the 5 to 9 years age group represents 8.69% of the standard population, etc.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2023/04/the-world-health-organizations-world-standard-population1-779x550.png\"")
Age standardization is applied in three steps.
\n\n\n\n(1) Researchers start with the observed rates of the relevant health outcome for each age group. In our example, these would be the cancer death rates for each age group in Japan and Uganda.
\n\n\n\nFor example, if we have 5-year intervals in our standard population, we would start from the cancer death rates of people in Japan aged 0–4, 5–9, 10–14, etc. These rates are called age-specific rates.
\n\n\n\n(2) These rates are then multiplied by the proportion of the corresponding age group in the standard population. For example, as the group aged 60–64 represents 3.72% of the World Standard Population, we would multiply the age-specific rate for this group by 3.72%.
\n\n\n\n(3) After multiplying all age-specific rates by the corresponding proportion of each age group in the standard population, we sum the rates to obtain one number. This number is the age-standardized rate.{ref}The national statistical office of Canada provides more details and a full example of how to calculate age-standardized rates.{/ref}
\n\n\n\nThese age-standardized rates allow us to compare cancer death rates in Uganda and Japan as if they applied to two countries with identical age structures.
\n\n\n\nIn this example, we see that Uganda’s age-standardized cancer death rate has surpassed Japan’s since 2000.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2023/04/cancer-death-rates-779x550.png\"")
Age-standardized rates provide a better comparison between populations with different age structures but do not represent the actual rate in each population. Using the age-standardized rate to say that 114 per 100,000 people in the Japanese population die from cancer each year would be inaccurate. We should refer back to the crude rate (346 per 100,000) for this.
\n\n\n\nAge standardization is often used to compare health metrics between countries.
\n\n\n\nBut we can also use it to measure health outcomes in the same location over time. Again, if we have a disease that primarily affects older people, we expect rates of that disease to rise over time as populations age.
\n\n\n\nAge standardization then allows us to see how the rate would have changed if the population’s age structure hadn’t changed.
\n\n\n\nThis can have a significant impact on our interpretation of health outcomes. One example of this comes from global death rates from cancer. On this chart, you can see that if we look at crude death rates (without age standardization), it looks like they have increased a lot in recent decades.
\n\n\n\n\n\n\n\nBut a big part of this effect is explained by the fact that the world is getting older. When we adjust for age, we see that death rates have fallen.
\n\n\n\nIn this related post, we discuss this chart – and the question of whether or not the world is making progress against cancer – in more detail.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Age standardization is a statistical method used to compare disease rates, or other health indicators, between populations while accounting for differences in their age structure."", ""protected"": false}, ""date_gmt"": ""2023-04-04T09:58:49"", ""modified"": ""2023-07-10T17:23:55"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Edouard Mathieu""], ""modified_gmt"": ""2023-07-10T16:23:55"", ""comment_status"": ""closed"", ""featured_media"": 54364, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/10/Living-planet-index-region-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2022/10/Living-planet-index-region-thumbnail-768x402.png""}}" 56226,"What is the ozone layer, and why is it important?",ozone-layer-context,post,publish,"One of the most pressing environmental problems over the last century has been the depletion of the ozone layer. But what is the ozone layer, and why does it matter?
Ozone is a gas present naturally within Earth’s atmosphere. It is formed of three oxygen atoms (giving it the chemical formula O3). Its structure makes it unstable: it can be easily formed and broken down through interaction with other compounds.
Ozone is most highly concentrated at two very different altitudes in the atmosphere: near the surface, and high in the atmosphere (in the stratosphere). Its function is very different in these two zones.
Ozone close to the surface is called tropospheric ozone, and it is often referred to as ‘bad ozone’. Ozone concentrations are lower in the troposphere than in the stratosphere.
We can see this in the diagram.
However, ozone concentrations close to the Earth’s surface can be temporarily and locally higher, because of emissions from motor vehicle exhausts, industrial processes, electric utilities, and chemical solvents. Ground-level ozone is a local air pollutant, and can negatively impact human health. Breathing ozone is particularly harmful to the young, elderly, and people with underlying respiratory problems.
This is very different from ozone high in the atmosphere: stratospheric ozone. It’s referred to as ‘good ozone’.
As shown in the diagram, ozone concentrations are higher in the stratosphere than in the troposphere.
The stratosphere includes the zone commonly called the ‘ozone layer’. It plays a crucial role in keeping the planet habitable by absorbing potentially dangerous ultraviolet (UV-B) radiation from the sun. Before its depletion, the ozone layer typically absorbed 97 to 99% of incoming UV-B radiation.
This means we need high ozone concentrations in the stratosphere to ensure that life — including human life — is not exposed to harmful concentrations of UV-B radiation.
In our work on the ozone layer, we focus on this ozone high in the atmosphere (the ‘good ozone’). The impact of ozone near the surface (‘bad ozone’) is covered in our work on air pollution.
The ozone layer absorbs 97% to 99% of the sun’s incoming ultraviolet radiation (UV-B).
This is fundamental to protecting life on Earth’s surface from exposure to harmful levels of this radiation, which can damage and disrupt DNA.
In the 1970s and ‘80s, humans emitted large amounts of gases that depleted this ozone in the upper atmosphere. As ozone concentrations in the stratosphere fell, and a hole in the ozone layer opened up, there have been measurable increases in the amount of UV-B radiation reaching the surface.
The chart shows the measured change in annual quantities of UV irradiance reaching Earth’s surface, in 2008 compared to 1979.{ref}This is given for UV at a wavelength of 305 nanometers (nm), which is well within the range where it has maximum damage to DNA.
Herman, J. R. (2010). Global increase in UV irradiance during the past 30 years (1979–2008) estimated from satellite data. Journal of Geophysical Research: Atmospheres, 115(D4).{/ref}
What’s noticeable is that ozone depletion and UV irradiance have increased much more in the Southern Hemisphere. This is because ozone depletion is also impacted by temperature and sunlight. Temperatures are colder at high latitudes in the Southern Hemisphere, so polar stratospheric clouds can form. These clouds can accelerate the reactions that break ozone down.
You will also notice that ozone depletion is worse at higher latitudes. It’s non-existent at the equator, and rises steeply towards the poles. Again, this is influenced by temperature and sunlight. That’s why ozone holes form at the poles, rather than the equator.
This increase in UV-B irradiation reaching the surface matters for life on Earth. One of the biggest concerns has been an increased risk of skin cancer (as well as skin damage and aging).{ref}Pitcher, H. M., & Longstreth, J. D. (1991). Melanoma mortality and exposure to ultraviolet radiation: an empirical relationship. Environment International, 17(1), 7-21.
Clydesdale, G. J., Dandie, G. W., & Muller, H. K. (2001). Ultraviolet light induced injury: immunological and inflammatory effects. Immunology and Cell Biology, 79(6), 547.{/ref} This is because UV-B irradiation can damage skin DNA.
Since the 1980s, the world has achieved rapid progress: the near-elimination of ozone-depleting substances and the trend toward recovering the ozone layer are among the most successful international environmental achievements to date.
Several studies have estimated that millions of excess skin cancer cases have been avoided due to the Montreal Protocol and its follow-up treaties.{ref}Dijk, A., Slaper, H., den Outer, P. N., Morgenstern, O., Braesicke, P., Pyle, J. A., & Tourpali, K. (2013). Skin Cancer Risks Avoided by the Montreal Protocol—Worldwide Modeling Integrating Coupled Climate: Chemistry Models with a Risk Model for UV. Photochemistry and Photobiology, 89(1), 234-246.
Slaper, H., G. J. M. Velders, J. S. Daniel, F. R. de Gruijl and J. C. van der Leun (1996) Estimates of ozone depletion and skin cancer incidence to examine the Vienna convention achievements. Nature 384(6606), 256–258.{/ref}
One of the most pressing environmental problems over the last century has been the depletion of the ozone layer. But what is the ozone layer, and why does it matter?
\n\n\n\nOzone is a gas present naturally within Earth’s atmosphere. It is formed of three oxygen atoms (giving it the chemical formula O3). Its structure makes it unstable: it can be easily formed and broken down through interaction with other compounds.
\n\n\n\nOzone is most highly concentrated at two very different altitudes in the atmosphere: near the surface, and high in the atmosphere (in the stratosphere). Its function is very different in these two zones.
\n\n\n\nOzone close to the surface is called tropospheric ozone, and it is often referred to as ‘bad ozone’. Ozone concentrations are lower in the troposphere than in the stratosphere.
\n\n\n\nWe can see this in the diagram.
\n\n\n\nHowever, ozone concentrations close to the Earth’s surface can be temporarily and locally higher, because of emissions from motor vehicle exhausts, industrial processes, electric utilities, and chemical solvents. Ground-level ozone is a local air pollutant, and can negatively impact human health. Breathing ozone is particularly harmful to the young, elderly, and people with underlying respiratory problems.
\n\n\n\nThis is very different from ozone high in the atmosphere: stratospheric ozone. It’s referred to as ‘good ozone’.
\n\n\n\nAs shown in the diagram, ozone concentrations are higher in the stratosphere than in the troposphere.
\n\n\n\nThe stratosphere includes the zone commonly called the ‘ozone layer’. It plays a crucial role in keeping the planet habitable by absorbing potentially dangerous ultraviolet (UV-B) radiation from the sun. Before its depletion, the ozone layer typically absorbed 97 to 99% of incoming UV-B radiation.
\n\n\n\nThis means we need high ozone concentrations in the stratosphere to ensure that life — including human life — is not exposed to harmful concentrations of UV-B radiation.
\n\n\n\nIn our work on the ozone layer, we focus on this ozone high in the atmosphere (the ‘good ozone’). The impact of ozone near the surface (‘bad ozone’) is covered in our work on air pollution.
\n![\""\""](\""https://owid.cloud/app/uploads/2018/06/Ozone-zones-in-atmosphere-01.png\"")
The ozone layer absorbs 97% to 99% of the sun’s incoming ultraviolet radiation (UV-B).
\n\n\n\nThis is fundamental to protecting life on Earth’s surface from exposure to harmful levels of this radiation, which can damage and disrupt DNA.
\n\n\n\nIn the 1970s and ‘80s, humans emitted large amounts of gases that depleted this ozone in the upper atmosphere. As ozone concentrations in the stratosphere fell, and a hole in the ozone layer opened up, there have been measurable increases in the amount of UV-B radiation reaching the surface.
\n\n\n\nThe chart shows the measured change in annual quantities of UV irradiance reaching Earth’s surface, in 2008 compared to 1979.{ref}This is given for UV at a wavelength of 305 nanometers (nm), which is well within the range where it has maximum damage to DNA.
\n\n\n\nHerman, J. R. (2010). Global increase in UV irradiance during the past 30 years (1979–2008) estimated from satellite data. Journal of Geophysical Research: Atmospheres, 115(D4).{/ref}
\n\n\n\nWhat’s noticeable is that ozone depletion and UV irradiance have increased much more in the Southern Hemisphere. This is because ozone depletion is also impacted by temperature and sunlight. Temperatures are colder at high latitudes in the Southern Hemisphere, so polar stratospheric clouds can form. These clouds can accelerate the reactions that break ozone down.
\n\n\n\nYou will also notice that ozone depletion is worse at higher latitudes. It’s non-existent at the equator, and rises steeply towards the poles. Again, this is influenced by temperature and sunlight. That’s why ozone holes form at the poles, rather than the equator.
\n\n\n\nThis increase in UV-B irradiation reaching the surface matters for life on Earth. One of the biggest concerns has been an increased risk of skin cancer (as well as skin damage and aging).{ref}Pitcher, H. M., & Longstreth, J. D. (1991). Melanoma mortality and exposure to ultraviolet radiation: an empirical relationship. Environment International, 17(1), 7-21.
\n\n\n\nClydesdale, G. J., Dandie, G. W., & Muller, H. K. (2001). Ultraviolet light induced injury: immunological and inflammatory effects. Immunology and Cell Biology, 79(6), 547.{/ref} This is because UV-B irradiation can damage skin DNA.
\n\n\n\nSince the 1980s, the world has achieved rapid progress: the near-elimination of ozone-depleting substances and the trend toward recovering the ozone layer are among the most successful international environmental achievements to date.
\n\n\n\nSeveral studies have estimated that millions of excess skin cancer cases have been avoided due to the Montreal Protocol and its follow-up treaties.{ref}Dijk, A., Slaper, H., den Outer, P. N., Morgenstern, O., Braesicke, P., Pyle, J. A., & Tourpali, K. (2013). Skin Cancer Risks Avoided by the Montreal Protocol—Worldwide Modeling Integrating Coupled Climate: Chemistry Models with a Risk Model for UV. Photochemistry and Photobiology, 89(1), 234-246.
\n\n\n\nSlaper, H., G. J. M. Velders, J. S. Daniel, F. R. de Gruijl and J. C. van der Leun (1996) Estimates of ozone depletion and skin cancer incidence to examine the Vienna convention achievements. Nature 384(6606), 256–258.{/ref}
\n![\""\""](\""https://owid.cloud/app/uploads/2018/04/Changes-in-UV-irradiation-reaching-surface-01.png\"")
The observation that the number of transistors on computer chips doubles approximately every two years is known as Moore’s Law.
Moore’s Law is not a law of nature, but an observation of a long-term trend in how technology is changing.
The law was first described by Gordon E. Moore, the co-founder of Intel, in 1965.{ref}The original paper is Gordon E. Moore (1965) – Cramming more components onto integrated circuits. In Electronics, Volume 38, Number 8, April 19, 1965.{/ref}
The chart shows Moore’s original graph that he drew in 1965 to describe this regularity. At the time, he had only a handful of data points. Note that he drew it on a logarithmic scale, and remember that a straight line on a log-axis means that the growth rate is constant and it is therefore showing the exponential growth of the number of transistors.
However, he hypothesized that this relationship would continue at a similar rate: “There is no reason to believe it will not remain constant for at least 10 years”.{ref}Quoted from Gordon E. Moore (1965) – Cramming more components onto integrated circuits. In Electronics, Volume 38, Number 8, April 19, 1965.{/ref}
In 1965, Gordon Moore predicted that this growth would continue for another 10 years, at least. Was he right?
In the chart, we’ve visualized the growth in transistor density – the number of transistors on integrated circuits – from 1970 onwards.
It looks strikingly similar to Moore’s simple plot from 1965. Note again that the transistor count is on a logarithmic axis, so the linear relationship over time means that the growth rate has been constant.
This means that the growth of the transistor count has, in fact, been exponential. You can also see this on our interactive chart, which shows the average transistor count over time and where you can switch between a linear and a log axis
Transistor counts have doubled approximately every two years, just as Moore predicted.
This has held true for more than 50 years now.
Moore’s Law describes the increasing number of transistors on integrated circuits, which in itself doesn’t matter for us as users of computer equipment. But it matters for those aspects that we do care about, like the speed and cost of computing.
Many related metrics show a similar pattern of exponential growth. The computational capacity of computers has increased exponentially, doubling every 1.5 years, from 1975 to 2009.{ref}Koomey, Berard, Sanchez, and Wong (2011) – Implications of Historical Trends in the Electrical Efficiency of Computing. In IEEE Annals of the History of Computing, 33, 3, 46–54.{/ref}
More recent data is shown in the interactive chart. It shows the increase in supercomputer power, measured as the largest supercomputer in any given year. The unit of measurement is FLOPS: the number of computations the machine can carry out per second.
Computing efficiency – measuring the energy use of computers – has halved every 1.5 years over the last 60 years.{ref}A short ungated article on this research is in the MIT Technology Review.{/ref}
Exponential progress is also found in the cost of computer memory and storage. In the chart, we see the cost of computer storage across different mediums – disks, flash drives, and internal memory – since the 1950s. This is measured as the price per terabyte.
Moore’s observation that the transistor count on integrated circuits grows exponentially is at the heart of many of the most consequential changes of our time. In our work on artificial intelligence we explore how the exponential growth translates in computing technology translated into more and more powerful AI systems.
The observation that the number of transistors on computer chips doubles approximately every two years is known as Moore’s Law.
\n\n\n\nMoore’s Law is not a law of nature, but an observation of a long-term trend in how technology is changing.
\n\n\n\nThe law was first described by Gordon E. Moore, the co-founder of Intel, in 1965.{ref}The original paper is Gordon E. Moore (1965) – Cramming more components onto integrated circuits. In Electronics, Volume 38, Number 8, April 19, 1965.{/ref}
\n\n\n\nThe chart shows Moore’s original graph that he drew in 1965 to describe this regularity. At the time, he had only a handful of data points. Note that he drew it on a logarithmic scale, and remember that a straight line on a log-axis means that the growth rate is constant and it is therefore showing the exponential growth of the number of transistors.
\n\n\n\nHowever, he hypothesized that this relationship would continue at a similar rate: “There is no reason to believe it will not remain constant for at least 10 years”.{ref}Quoted from Gordon E. Moore (1965) – Cramming more components onto integrated circuits. In Electronics, Volume 38, Number 8, April 19, 1965.{/ref}
\n\n\n\n\n![\""The](\""http://ourworldindata.org/app/uploads/2013/05/The-Number-of-Components-per-Integrated-Function-Moores-Original-Graph-Moore0.png\"")
In 1965, Gordon Moore predicted that this growth would continue for another 10 years, at least. Was he right?
\n\n\n\nIn the chart, we’ve visualized the growth in transistor density – the number of transistors on integrated circuits – from 1970 onwards.
\n\n\n\nIt looks strikingly similar to Moore’s simple plot from 1965. Note again that the transistor count is on a logarithmic axis, so the linear relationship over time means that the growth rate has been constant.
\n\n\n\nThis means that the growth of the transistor count has, in fact, been exponential. You can also see this on our interactive chart, which shows the average transistor count over time and where you can switch between a linear and a log axis
\n\n\n\nTransistor counts have doubled approximately every two years, just as Moore predicted.
\n\n\n\nThis has held true for more than 50 years now.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/03/Transistor-Count-over-time.png\"")
Moore’s Law describes the increasing number of transistors on integrated circuits, which in itself doesn’t matter for us as users of computer equipment. But it matters for those aspects that we do care about, like the speed and cost of computing.
\n\n\n\nMany related metrics show a similar pattern of exponential growth. The computational capacity of computers has increased exponentially, doubling every 1.5 years, from 1975 to 2009.{ref}Koomey, Berard, Sanchez, and Wong (2011) – Implications of Historical Trends in the Electrical Efficiency of Computing. In IEEE Annals of the History of Computing, 33, 3, 46–54.{/ref}
\n\n\n\nMore recent data is shown in the interactive chart. It shows the increase in supercomputer power, measured as the largest supercomputer in any given year. The unit of measurement is FLOPS: the number of computations the machine can carry out per second.
\nComputing efficiency – measuring the energy use of computers – has halved every 1.5 years over the last 60 years.{ref}A short ungated article on this research is in the MIT Technology Review.{/ref}
\n\n\n\nExponential progress is also found in the cost of computer memory and storage. In the chart, we see the cost of computer storage across different mediums – disks, flash drives, and internal memory – since the 1950s. This is measured as the price per terabyte.
\nMoore’s observation that the transistor count on integrated circuits grows exponentially is at the heart of many of the most consequential changes of our time. In our work on artificial intelligence we explore how the exponential growth translates in computing technology translated into more and more powerful AI systems.
\nJohns Hopkins University has been at the forefront of collecting and reporting data on confirmed COVID-19 cases and deaths since the outbreak began in 2020. However, it recently announced that it would stop updating its data on 10 March 2023.
Our team at Our World in Data recognizes the importance of up-to-date data on the pandemic. We will switch our primary source to the World Health Organization (WHO), which updates its dataset weekly.
To keep this data consistent over time, we will replace the entire time series with WHO data on 8 March 2023.
Our goal is to continue providing the most complete and timely data on the pandemic, and relying on the WHO data is the best way to achieve this.
This change will be seamless for our users. The URL of all our charts and the variable names in our dataset will remain the same.
All other COVID-19 datasets we maintain will remain unchanged.
An archive of the last version (8 March 2023) of our dataset based on data from Johns Hopkins University is available for download here (CSV file, 69 Mo).
We want to take this opportunity to thank the team at Johns Hopkins for their crucial work since the beginning of the pandemic. Their efforts have been invaluable in helping us understand the impact of this virus across the world.
","{""id"": ""wp-56089"", ""slug"": ""covid-jhu-who"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Johns Hopkins University has been at the forefront of collecting and reporting data on confirmed COVID-19 cases and deaths since the outbreak began in 2020. However, it "", ""spanType"": ""span-simple-text""}, {""url"": ""https://github.com/CSSEGISandData/COVID-19/issues/6577"", ""children"": [{""text"": ""recently announced"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" that it would stop updating its data on 10 March 2023."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Our team at Our World in Data recognizes the importance of up-to-date data on the pandemic. We will switch our primary source to "", ""spanType"": ""span-simple-text""}, {""url"": ""https://covid19.who.int/data"", ""children"": [{""text"": ""the World Health Organization (WHO)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which updates its dataset weekly."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To keep this data consistent over time, we will replace the entire time series with WHO data on 8 March 2023."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Our goal is to continue providing the most complete and timely data on the pandemic, and relying on the WHO data is the best way to achieve this."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This change will be seamless for our users. The URL of all our charts and the variable names in our dataset will remain the same."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://github.com/owid/covid-19-data/tree/master/public/data"", ""children"": [{""text"": ""All other COVID-19 datasets"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" we maintain will remain unchanged."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""An archive of the last version (8 March 2023) of our dataset based on data from Johns Hopkins University is available for download "", ""spanType"": ""span-simple-text""}, {""url"": ""https://covid.ourworldindata.org/data/owid-covid-data-old.csv"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (CSV file, 69 Mo)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We want to take this opportunity to thank the team at Johns Hopkins for their crucial work since the beginning of the pandemic. Their efforts have been invaluable in helping us understand the impact of this virus across the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Our World in Data will rely on data from the WHO to track confirmed COVID-19 cases and deaths"", ""authors"": [""Edouard Mathieu"", ""Lucas Rodés-Guirao""], ""excerpt"": ""Johns Hopkins University will stop publishing data on confirmed COVID-19 cases and deaths. Our team will replace our entire time series with WHO's weekly-updated data on 8 March. This change will not affect users of our charts and dataset."", ""dateline"": ""February 28, 2023"", ""subtitle"": ""Johns Hopkins University will stop publishing data on confirmed COVID-19 cases and deaths. Our team will replace our entire time series with WHO's weekly-updated data on 8 March. This change will not affect users of our charts and dataset."", ""sidebar-toc"": false, ""featured-image"": ""coronavirus.png""}, ""createdAt"": ""2023-02-28T12:45:15.000Z"", ""published"": false, ""updatedAt"": ""2023-03-13T11:40:47.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-28T12:49:05.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 8, ""numErrors"": 0, ""wpTagCounts"": {""paragraph"": 8}, ""htmlTagCounts"": {""p"": 8}}",2023-02-28 12:49:05,2024-03-10 13:18:58,1pB2dwxPSw-i1TWhuEJqx6n3jBTwxB9bMJQNyinB6NOg,"[""Edouard Mathieu"", ""Lucas Rodés-Guirao""]",Johns Hopkins University will stop publishing data on confirmed COVID-19 cases and deaths. Our team will replace our entire time series with WHO's weekly-updated data on 8 March. This change will not affect users of our charts and dataset.,2023-02-28 12:45:15,2023-03-13 11:40:47,https://ourworldindata.org/wp-content/uploads/2021/02/coronavirus.png,{},"Johns Hopkins University has been at the forefront of collecting and reporting data on confirmed COVID-19 cases and deaths since the outbreak began in 2020. However, it [recently announced](https://github.com/CSSEGISandData/COVID-19/issues/6577) that it would stop updating its data on 10 March 2023. Our team at Our World in Data recognizes the importance of up-to-date data on the pandemic. We will switch our primary source to [the World Health Organization (WHO)](https://covid19.who.int/data), which updates its dataset weekly. To keep this data consistent over time, we will replace the entire time series with WHO data on 8 March 2023. Our goal is to continue providing the most complete and timely data on the pandemic, and relying on the WHO data is the best way to achieve this. This change will be seamless for our users. The URL of all our charts and the variable names in our dataset will remain the same. [All other COVID-19 datasets](https://github.com/owid/covid-19-data/tree/master/public/data) we maintain will remain unchanged. An archive of the last version (8 March 2023) of our dataset based on data from Johns Hopkins University is available for download [here](https://covid.ourworldindata.org/data/owid-covid-data-old.csv) (CSV file, 69 Mo). We want to take this opportunity to thank the team at Johns Hopkins for their crucial work since the beginning of the pandemic. Their efforts have been invaluable in helping us understand the impact of this virus across the world.","{""id"": 56089, ""date"": ""2023-02-28T12:49:05"", ""guid"": {""rendered"": ""https://owid.cloud/?p=56089""}, ""link"": ""https://owid.cloud/covid-jhu-who"", ""meta"": {""owid_publication_context_meta_field"": []}, ""slug"": ""covid-jhu-who"", ""tags"": [199], ""type"": ""post"", ""title"": {""rendered"": ""Our World in Data will rely on data from the WHO to track confirmed COVID-19 cases and deaths""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56089""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/41"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=56089"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=56089"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=56089"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=56089""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56089/revisions"", ""count"": 4}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/39688"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 56227, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56089/revisions/56227""}]}, ""author"": 41, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": ""\nJohns Hopkins University has been at the forefront of collecting and reporting data on confirmed COVID-19 cases and deaths since the outbreak began in 2020. However, it recently announced that it would stop updating its data on 10 March 2023.
\n\n\n\nOur team at Our World in Data recognizes the importance of up-to-date data on the pandemic. We will switch our primary source to the World Health Organization (WHO), which updates its dataset weekly.
\n\n\n\nTo keep this data consistent over time, we will replace the entire time series with WHO data on 8 March 2023.
\n\n\n\nOur goal is to continue providing the most complete and timely data on the pandemic, and relying on the WHO data is the best way to achieve this.
\n\n\n\nThis change will be seamless for our users. The URL of all our charts and the variable names in our dataset will remain the same.
\n\n\n\nAll other COVID-19 datasets we maintain will remain unchanged.
\n\n\n\nAn archive of the last version (8 March 2023) of our dataset based on data from Johns Hopkins University is available for download here (CSV file, 69 Mo).
\n\n\n\nWe want to take this opportunity to thank the team at Johns Hopkins for their crucial work since the beginning of the pandemic. Their efforts have been invaluable in helping us understand the impact of this virus across the world.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Johns Hopkins University will stop publishing data on confirmed COVID-19 cases and deaths. Our team will replace our entire time series with WHO’s weekly-updated data on 8 March. This change will not affect users of our charts and dataset."", ""protected"": false}, ""date_gmt"": ""2023-02-28T12:49:05"", ""modified"": ""2023-03-13T11:40:47"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Edouard Mathieu"", ""Lucas Rodés-Guirao""], ""modified_gmt"": ""2023-03-13T11:40:47"", ""comment_status"": ""closed"", ""featured_media"": 39688, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2021/02/coronavirus-150x78.png"", ""medium_large"": ""/app/uploads/2021/02/coronavirus-768x401.png""}}" 56065,how do we choose metrics?,how-do-we-choose-metrics,wp_block,publish,"For each topic, we work to provide the best metrics to understand it. What metrics are 'best' will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria:
The topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at info@ourworldindata.org.
","{""id"": ""wp-56065"", ""slug"": ""how-do-we-choose-metrics"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""For each topic, we work to provide the best metrics to understand it. What metrics are 'best' will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It covers large parts of the world. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""True to our name, we seek metrics which cover as much of the world as possible. Only then can they help us understand global differences and changes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It covers a lot of time. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""This means both that the measure goes as far back in time as possible, and that it is as recent as possible. It then can help us understand both historical and very recent developments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is comparable across time and space"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "". This means that we prefer metrics that can be compared across years and countries. This allows us to evaluate whether countries are making progress or falling behind, and how countries are doing relative to another."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It captures what we are trying to measure"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "". This means that the metric does not give an incomplete or misleading answer to the question we have. For example, an inadequate measure for whether a country is a democracy is the share of the population that voted. Looking only at voter turnout ignores whether citizens had more than one choice at the ballot box. And at the same time, it inadvertently considers citizens that were coerced to vote."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is reliable."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that the metric is consistent, i.e. it captures the phenomenon similarly when measured repeatedly, and therefore is precise, and captures the phenomenon with little error. A consistent and precise metric makes us more confident in what it tells us about the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Its construction is transparent."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that we prefer metrics that come with a detailed description of how it was constructed, why it was constructed in this way, and with the underlying code and raw data. We, and you as our reader, then can evaluate its strengths and weaknesses in detail."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is easy to understand"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "". This means that the metric captures something that people are broadly familiar with, and they can broadly make sense of its construction. It then can provide answers that people beyond experts can learn from."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is maintained well."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that the data source updates the metric frequently, and provides reasonably up-to-date data. We often favor data from international institutions (such as the World Bank and the UN) and research institutions (such as the Global Carbon Project and the Varieties of Democracy project) over data from individual academic publications, because the former have the mandate and resources to keep this data up-to-date."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Its values differ a lot from the same measure by another trusted source."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means a metric captures disagreement across sources. It then helps us to be appropriately uncertain of our answers in light of disagreeing sources."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is accessible."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that the data is published in a publicly accessible document and is licensed to be reused by us and preferably others. Only then can it help people answer their questions, on and beyond our site."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""We have the tools to visualize it. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""This means a metric is structured such that our in-house visualization tool — the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/owid-grapher"", ""children"": [{""text"": ""Our World in Data Grapher"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" — can display its information well. For example, our maps are set up to visualize national data, and currently cannot display metrics at the sub-national level or gridded data."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at "", ""spanType"": ""span-simple-text""}, {""url"": ""mailto:info@ourworldindata.org"", ""children"": [{""text"": ""info@ourworldindata.org"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""how do we choose metrics?"", ""authors"": [null], ""dateline"": ""February 27, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-02-27T09:05:29.000Z"", ""published"": false, ""updatedAt"": ""2023-02-27T09:05:29.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-27T09:05:29.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 13, ""numErrors"": 11, ""wpTagCounts"": {""list"": 11, ""paragraph"": 13}, ""htmlTagCounts"": {""p"": 13, ""ul"": 11}}",2023-02-27 09:05:29,2024-02-16 14:23:03,,[null],,,2023-02-27 09:05:29,,{},"For each topic, we work to provide the best metrics to understand it. What metrics are 'best' will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria: * **It covers large parts of the world. **True to our name, we seek metrics which cover as much of the world as possible. Only then can they help us understand global differences and changes. * **It covers a lot of time. **This means both that the measure goes as far back in time as possible, and that it is as recent as possible. It then can help us understand both historical and very recent developments. * **It is comparable across time and space**. This means that we prefer metrics that can be compared across years and countries. This allows us to evaluate whether countries are making progress or falling behind, and how countries are doing relative to another. * **It captures what we are trying to measure**. This means that the metric does not give an incomplete or misleading answer to the question we have. For example, an inadequate measure for whether a country is a democracy is the share of the population that voted. Looking only at voter turnout ignores whether citizens had more than one choice at the ballot box. And at the same time, it inadvertently considers citizens that were coerced to vote. * **It is reliable.** This means that the metric is consistent, i.e. it captures the phenomenon similarly when measured repeatedly, and therefore is precise, and captures the phenomenon with little error. A consistent and precise metric makes us more confident in what it tells us about the world. * **Its construction is transparent.** This means that we prefer metrics that come with a detailed description of how it was constructed, why it was constructed in this way, and with the underlying code and raw data. We, and you as our reader, then can evaluate its strengths and weaknesses in detail. * **It is easy to understand**. This means that the metric captures something that people are broadly familiar with, and they can broadly make sense of its construction. It then can provide answers that people beyond experts can learn from. * **It is maintained well.** This means that the data source updates the metric frequently, and provides reasonably up-to-date data. We often favor data from international institutions (such as the World Bank and the UN) and research institutions (such as the Global Carbon Project and the Varieties of Democracy project) over data from individual academic publications, because the former have the mandate and resources to keep this data up-to-date. * **Its values differ a lot from the same measure by another trusted source.** This means a metric captures disagreement across sources. It then helps us to be appropriately uncertain of our answers in light of disagreeing sources. * **It is accessible.** This means that the data is published in a publicly accessible document and is licensed to be reused by us and preferably others. Only then can it help people answer their questions, on and beyond our site. * **We have the tools to visualize it. **This means a metric is structured such that our in-house visualization tool — the [Our World in Data Grapher](https://ourworldindata.org/owid-grapher) — can display its information well. For example, our maps are set up to visualize national data, and currently cannot display metrics at the sub-national level or gridded data. The topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at [info@ourworldindata.org](mailto:info@ourworldindata.org).","{""data"": {""wpBlock"": {""content"": ""\nFor each topic, we work to provide the best metrics to understand it. What metrics are ‘best’ will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria:
\n\n\n\nThe topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at info@ourworldindata.org.
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 56064,how do we choose topics?,how-do-we-choose-topics,wp_block,publish,"We cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria:
We evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed.
","{""id"": ""wp-56064"", ""slug"": ""how-do-we-choose-topics"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""We cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It affects many countries and people."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This can mean that it concerns every person, such as "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/health-meta"", ""children"": [{""text"": ""health"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". It can mean that it affects many people in all countries, such as "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/poverty"", ""children"": [{""text"": ""poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Or it can mean that it affects many people in fewer countries, such as "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/malaria"", ""children"": [{""text"": ""malaria"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It comes with great costs or benefits. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""The costs or benefits can be direct, and shorten people’s lives or mean they lead happier lives. An example is the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/coronavirus"", ""children"": [{""text"": ""COVID-19 pandemic"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which has immediately affected people’s well-being. But the costs or benefits can also be indirect, and worsen or alleviate other problems. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/agricultural-production"", ""children"": [{""text"": ""agricultural production"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which affects many people’s access to nutrition."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It poses significant risks."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that it may not impose great costs at the moment, but may do so in the future. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/nuclear-weapons"", ""children"": [{""text"": ""nuclear weapons"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which have not been used in decades but whose use would be devastating."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It will remain important, or become more important in the future."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""url"": ""https://ourworldindata.org/poverty"", ""children"": [{""text"": ""Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" is an example of a topic that will remain important, as many people remain impoverished even if fewer people live in extreme poverty than in the past. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""Artificial intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" is an example of a topic that will become more important, as technological advances continuously expand its effects on people’s lives."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is helpful to understand other topics."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" Many of the topics we focus on are problems in themselves. But we also provide data and research on major changes that help us understand and address these problems. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/world-population-growth"", ""children"": [{""text"": ""population changes"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which are crucial to better understand energy and education needs."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is poorly understood."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means the public knows little about a problem or frequently misunderstands it, such as because the data on it is not described well. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/plastic-pollution"", ""children"": [{""text"": ""plastic pollution"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", where data and research were often missing from the public conversation."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is neglected elsewhere."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This can mean that other organizations do not cover it, or do so in a limited fashion. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/biodiversity"", ""children"": [{""text"": ""biodiversity"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", where data on global changes are hard to find elsewhere. This also means that if others cover a topic well, we are less likely to cover it ourselves."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""We have expertise on it."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" If we have someone on our team with deep knowledge of the area, we are more likely to cover the topic. Ideally, we would have both a researcher and a data scientist with this expertise. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracy"", ""children"": [{""text"": ""democracy"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", where we expanded our work as our team grew."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""We have funding available for it."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" While most of our funding comes from unrestricted resources, including "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/donate"", ""children"": [{""text"": ""reader donations"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", we partially fund our work through restricted grants. Importantly, we only apply for them if they are on topics we want to cover in depth anyway, and the grant does not come with any requirements on "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""how"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" to cover the topic."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""how do we choose topics?"", ""authors"": [null], ""dateline"": ""February 27, 2023"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2023-02-27T09:05:08.000Z"", ""published"": false, ""updatedAt"": ""2023-02-27T09:05:08.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-27T09:05:08.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 11, ""numErrors"": 9, ""wpTagCounts"": {""list"": 9, ""paragraph"": 12}, ""htmlTagCounts"": {""p"": 12, ""ul"": 9}}",2023-02-27 09:05:08,2024-02-16 14:23:03,,[null],,,2023-02-27 09:05:08,,{},"We cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria: * **It affects many countries and people.** This can mean that it concerns every person, such as [health](https://ourworldindata.org/health-meta). It can mean that it affects many people in all countries, such as [poverty](https://ourworldindata.org/poverty). Or it can mean that it affects many people in fewer countries, such as [malaria](https://ourworldindata.org/malaria). * **It comes with great costs or benefits. **The costs or benefits can be direct, and shorten people’s lives or mean they lead happier lives. An example is the [COVID-19 pandemic](https://ourworldindata.org/coronavirus), which has immediately affected people’s well-being. But the costs or benefits can also be indirect, and worsen or alleviate other problems. An example is [agricultural production](https://ourworldindata.org/agricultural-production), which affects many people’s access to nutrition. * **It poses significant risks.** This means that it may not impose great costs at the moment, but may do so in the future. An example is [nuclear weapons](https://ourworldindata.org/nuclear-weapons), which have not been used in decades but whose use would be devastating. * **It will remain important, or become more important in the future.**[Poverty](https://ourworldindata.org/poverty) is an example of a topic that will remain important, as many people remain impoverished even if fewer people live in extreme poverty than in the past. [Artificial intelligence](https://ourworldindata.org/artificial-intelligence) is an example of a topic that will become more important, as technological advances continuously expand its effects on people’s lives. * **It is helpful to understand other topics.** Many of the topics we focus on are problems in themselves. But we also provide data and research on major changes that help us understand and address these problems. An example is [population changes](https://ourworldindata.org/world-population-growth), which are crucial to better understand energy and education needs. * **It is poorly understood.** This means the public knows little about a problem or frequently misunderstands it, such as because the data on it is not described well. An example is [plastic pollution](http://ourworldindata.org/plastic-pollution), where data and research were often missing from the public conversation. * **It is neglected elsewhere.** This can mean that other organizations do not cover it, or do so in a limited fashion. An example is [biodiversity](http://ourworldindata.org/biodiversity), where data on global changes are hard to find elsewhere. This also means that if others cover a topic well, we are less likely to cover it ourselves. * **We have expertise on it.** If we have someone on our team with deep knowledge of the area, we are more likely to cover the topic. Ideally, we would have both a researcher and a data scientist with this expertise. An example is [democracy](https://ourworldindata.org/democracy), where we expanded our work as our team grew. * **We have funding available for it.** While most of our funding comes from unrestricted resources, including [reader donations](https://ourworldindata.org/donate), we partially fund our work through restricted grants. Importantly, we only apply for them if they are on topics we want to cover in depth anyway, and the grant does not come with any requirements on _how_ to cover the topic. We evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed.","{""data"": {""wpBlock"": {""content"": ""\nWe cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria:
\n\n\n\n\n\n\n\nWe evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed.
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 56043,Optimism and Pessimism,optimism-and-pessimism,post,publish,"This page is dedicated to the research why people are optimistic or pessimistic about certain things and how this is influenced by human nature, the media, and social circumstances.
We are interested in this topic also because it is closely linked to our motivation for publishing Our World in Data. We face big global problems, but living conditions around the world have improved in important ways; fewer people are dying of disease, conflict and famine; more of us are receiving a basic education; the world is becoming more democratic; we live longer and lead healthier lives.
Why is that we – especially those in the developed world – often have a negative view on how the world has changed over the last decades and centuries? Why we are so pessimistic about our collective future?
It is a peculiar empirical phenomenon that while people tend to be optimistic about their own future, they can at the same time be deeply pessimistic about the future of their nation or the world. Tali Sharot, associate professor of psychology at UCL, has popularised the idea of an innate optimism bias built into the human brain.{ref}Sharot, Tali. The Optimism Bias: A Tour of the Irrationally Positive Brain. New York: Pantheon Books, 2011.
Sharot, Tali. The Science of Optimism Why We're Hard-wired for Hope. New York: Ted Conferences, 2012.{/ref}
That is, we tend to be optimistic rather than realistic when considering our individual future. If you were to ask newlywed couples to estimate the probability they will divorce in the future, they would likely reject the possibility outright. Yet today roughly 40% of marriages in the UK end in divorce. Another example is asking smokers to estimate their chances of getting cancer and again, most would underestimate their risk. This optimism persists even when people are presented with the relevant statistics.
Consider the following graphs from the European Union's Eurobarometer surveys; they report people's expectations about their own personal job situation and of the economic situation in their home country. From the end of 1995 to the middle of 2015, around 60% of people predict that their job situation will remain the same, while 20% expect their situation to improve. Compare that with the response of the same group of individuals considering the future of the economic situation in their home country. Although far less stable, the results show that most people expect the economic situation in their home country to get worse or stay the same. The expectation that things are going to worsen nationally is correlated with recessions, yet there is remarkable stability in the results for individual expectations. Does the response to the question about national economic well being better correspond to an individual's true job prospects?
This pattern is also observed on a larger scale. This chart shows how individuals in the UK respond to the question: ""Thinking about ..., how much of a problem do you think each of the following are in your local area and in the whole of the UK?"" Individuals tend to believe problems are more pronounced nationally than in their local area.
This chart shows how many individuals rate the environment in their local area as fairly or very bad, compared with the environment nationally and globally. Again, we observe a similar pattern for most countries. No matter where you ask people are much more negative about places that are far away – places which they know less from their own experience and more through the media.
How can we reconcile this individual optimism with social pessimism? Paul Dolan, professor of behavioural science at LSE, believes people respond pessimistically to questions about national or international performance for three reasons:
This explanation suggests there is a problem of information. If we do not pay attention to human development, then our judgement may suffer from a bias related to transient events or framing. The Gapminder Ignorance Project – which studied how wrong or right people are informed about global development – suggests the reason for all this ignorance is:
""Statistical facts don’t come to people naturally. Quite the opposite. Most people understand the world by generalizing personal experiences which are very biased. In the media the “news-worthy” events exaggerate the unusual and put the focus on swift changes. Slow and steady changes in major trends don’t get much attention. Unintentionally, people end-up carrying around a sack of outdated facts that you got in school (including knowledge that often was outdated when acquired in school).""
Another explanation put forward by Martin Seligman, professor of psychology at the University of Pennsylvania, suggests a link between control and optimism. If we feel more in control of our lives, we tend to be happier, healthier and more optimistic about the future. This could also help to explain the gap between individual and societal optimism: since we are in direct control of our own lives but not the destiny of the nation we feel more optimistic about ourselves.
At Our World in Data we aim to bring together the empirical data and research to show how living conditions around the world are changing. Is that necessary?
The opinion research organization Ipsos MORI conducted a detailed survey of 26,489 people across 28 countries that gives us an answer.{ref}The full reference of the survey is ‘Chris Jackson (2017) – Global Perceptions of Development Progress: ‘Perils of Perceptions’ Research’, published by Ipsos MORI, 18 September 2017.{/ref}
The first chart shows how the surveyed people answered the following question: “In the last 20 years, the proportion of the world population living in extreme poverty has decreased, increased, or remained the same?”
The majority of people – 52% – believe that the share of people in extreme poverty is rising. The opposite is true. In fact, the share of people living in extreme poverty across the world has been declining for two centuries and in the last 20 years this positive development has been faster than ever before (see our work on Poverty). For the recent era it doesn't even matter what poverty line you choose, the share of people below any poverty line has fallen (see here).
There are some people who answered the question correctly: every fifth person knows that poverty is falling. But it’s interesting that the share of correct answers differs substantially across countries. The countries I marked with a star are those that were a low-income or lower-middle-income countries a generation ago (in 1990). In these poorer countries more people understand how global poverty has changed. People in richer countries on the other hand – in which the majority of the population escaped extreme poverty some generations ago – have a very wrong perception about what is happening to global poverty.
We are not just wrong about global poverty. In the same survey people were asked: “In the last 20 years, has the child mortality rate in developing regions increased, decreased or stayed about the same?”
Here again the data is very clear. The child mortality rate in both the less- and least-developed countries has halved in the last 20 years.{ref}In fact not only the average child mortality rate has fallen, but the child mortality rate has fallen in all countries (except for two very small ones).{/ref}
The survey once more shows that most people are not aware of this. On average only 39% know that the mortality of children is falling. And what greater achievement has humanity ever achieved than making it more and more likely that children survive the first, vulnerable years of their lives and sparing parents the sadness of losing their babies? This has to be one of humanity’s greatest achievements.
And just as with knowledge about extreme poverty, the share of uninformed people is much higher in the rich countries of the world. So is our work at Our World in Data needed? This survey shows that few Senegalese or Kenyans will learn something new; but if you have some friends in the US or Japan you will probably help them if you share our work.
The widespread ignorance about these truly important changes in the world feeds into a general discontent about how the world is changing. When YouGov asked in a separate survey the more general question: “All things considered, do you think the world is getting better or worse?” there were very few who gave a positive answer. In France and Australia only 3%(!) think the world is getting better.
And again we see that in poorer countries the share of people who answer positively is higher.{ref}This chart shows the income level difference between the very negative rich countries and the somewhat less negative poorer countries.{/ref}
What should we make of the fact that many perceive the world to be stagnating or even declining in global health or poverty while we are in fact achieving the most rapid improvements in our history in these very same aspects?
First, this is simply sad. It means that we think worse of the world than we should. We think more poorly than we should about the time we are living in, and we think more poorly than we should about what people around the world are achieving right now.
Second it makes clear that we are doing a terrible job at understanding and communicating what is happening in the world. Particularly in rich countries the education systems and media are failing to convey an accurate perspective on how the world is changing – arguably one of the main expectations we should have of them.{ref}This finding is the starting point for the recently published book ’Factfulness’ coauthored by Anna Rosling Rönnlund, and Hans and Ola Rosling. The authors go on to explain that our perception is so very wrong because our minds are paying attention to extremes – the very richest, very poorest, most violent and most corrupt aspects of our world – so that we end up with what they call the ‘overdramatic worldview’, which is pieced together by all the most dramatic aspects of our world, but has a massive blind spot for the world that is the reality of most people in the world. The overdramatic worldview leaves us with a picture of the world that includes all the stories that are in fact rare (the fact that they are extraordinary is why they are reported in the media), but which has no understanding of what is actually common.{/ref}
Our perception of how the world is changing matters for what we believe is possible in the future.
If we ask people about what is possible for the world, then most of us answer ‘not much’. This chart documents the survey answers to the question “over the next 15 years, do you think living conditions for people around the world get better or worse?”. More than half of the people expect stagnation or that things will be getting worse. Fortunately, the places in which people currently have the worst living conditions are more optimistic about what is possible in the coming years.
On the whole, the findings from the surveys are clear: we do not only believe that the world is stagnating or declining, we also expect that this perceived stagnation or decline will continue into the future.
This pessimism about what is possible for the world matters politically. Those who don’t expect that things get better in the first place will be less likely to demand actions that can bring positive developments about. The few optimists on the other hand will want to see the necessary changes for the improvements they are expecting.
Finally the survey suggests that there is a connection between our perception of the past and our hope for the future. This chart shows that the degree of optimism about the future differs hugely by the level of people’s knowledge about global development.
Those that were most pessimistic about the future tended to have the least basic knowledge on how the world has changed. Of those who could not give a single correct answer to the survey questions, only 17% expect the world to be better off in the future. At the other end of the spectrum, those who had very good knowledge about how the world has changed were the most optimistic about the changes that we can achieve in the next 15 years.
This is a correlation and as we know, correlation does not imply causation. To understand whether there is a causal link we would need to know whether getting a more accurate picture of how the world is changing makes one change one’s belief about what will happen in the future. Unfortunately I am not aware of a study that looked into this question.{ref}Please do get in contact if you are aware of a study that investigates this question and I will update this section of the post – and I’d be really grateful as I would very much like to understand this link.{/ref}
Of course no one can know how the future turns out and there is nothing that would make the progress we have seen in recent decades continue inevitably and not every global development pessimist is ill-informed. But what we do know from these surveys is that these two views go together: Those who are pessimistic are much more likely to have little understanding about what is happening in the world.
Obviously the question then is, why is it that better informed people are more optimistic about the future?
As we have seen, being wrong about global development mostly means being too negative about how the world is changing. Being wrong in these questions means having a cynical worldview. Cynicism suggests that nothing can be done to improve our situation and every effort to do so is bound to fail. Our history, the cynics say, is a history of failures and what we can expect for the future is more of the same.
In contrast to this, answering the questions correctly means that you understand that things can change. An accurate understanding of how global health and poverty are improving leaves no space for cynicism. Those who are optimistic about the future can base their view on the knowledge that it is possible to change the world for the better, because they know that we did.
Declinism refers to the belief that a country or some other institution is in decline. Declinism was a prevalent feature of British political and economic history, whereby the decline of Britain as a world power was seen as the result of internal failures rather than international forces or global convergence. David Edgerton writes: ""Declinism is beginning to appear as one of the last vestiges of imperial grandeur: for declinism holds, implicitly but clearly, that if Britain had done better it would have remained a much larger player on the world stage.""{ref}David Edgerton. ""The Decline of Declinism."" The Business History Review, Vol. 71, No. 2 (Summer, 1997), pp. 201-206. Available online at http://www.jstor.org/stable/3116157{/ref}
Today declinism in the United States is fashionable with many politicians. Donald Trump's campaign slogan for the 2016 Republican nomination election was ""Make America Great Again!""
The major flaw in much of the declinist narrative is the failure to distinguish between absolute and relative changes. Between 2010-14, US real GDP growth rates have fluctuated between 1.5-2.5% and yet, the US economy was recently overtaken by the Chinese economy measured in PPP-adjusted terms.{ref}For more information on PPP-adjustments, please visit the economic growth page.{/ref}
In many ways this may capture the reason why the most developed nations tend to believe that their economy is in decline: relative decline is interpreted as absolute decline. Unsurprisingly, new EU member states tend to be much more optimistic about the future. The four largest economies -- the UK, France, Germany and Italy -- are the most pessimistic. This pattern persists when considering economies at different stages of development: developing countries are more optimistic about the future, while developed ones tend to be pessimistic.
One interesting explanation for declinism is that it is the result of the way we encode memories and what we remember. Firstly, researchers have long established a robust pattern in the age at which we retain the most memories. In old age, memories from our lives are not evenly distributed but instead concentrated in two regions. These regions are (1) memories formed in adolescence and early adulthood, between the ages of 10-30, and (2) recent memory of events. The following figure is a useful representation of this distribution.
Secondly, research finds that as we get older we tend to have – on average – fewer negative experiences and that we are more likely to remember the positive ones over the negative ones.{ref}Mara Mather, Laura L. Carstensen, Aging and motivated cognition: the positivity effect in attention and memory, Trends in Cognitive Sciences, Volume 9, Issue 10, October 2005, Pages 496-502, ISSN 1364-6613. Available online here.{/ref}
This effect combined with the reminiscence bump could explain why declinism exists among older generations, and why your parents could never stand the music you listened to! The universality of this effect is illustrated by Harvey Daniels with the use of these quotes about the decline of the English language{ref}Daniels, Harvey. Famous last words: The American language crisis reconsidered. Southern Illinois University Press, 1983.{/ref}:
In the light of this research on human nature it is then not surprising that one of the earliest Sumerian tablets discovered and deciphered by modern scholars was a complaint by a teacher about his students' writing ability.
There are three main reasons we should try to combat social pessimism and declinism. The first reason is simple; indicators of living standards are significantly improving around the world. By monitoring and researching these changes we can identify ways in which progress can be achieved. Over the long-run, say 50-100 years, human progress has been staggering with the benefits not confined to the richest or most powerful. The second reason is that if our perceptions of the reality are wrong, we can end up prioritising the wrong things and making ineffectual change. Finally, being optimistic can be good for your health, while having a pessimistic outlook can be detrimental to your health.
The public perception of these indicators matters because it directly influences the priorities of voters in democratic countries and politicians. If, as in the example above, the public believes crime is increasing, it is likely that it demands more policing not for a reason grounded in reality, but for an imagined worsening of the society they live it. This is one reason why incorrect public perceptions can be a problem.
The following figures underline just how sizable these effects can be. The first shows how spending on crime has moved with the public's confidence in the government's ability to crack down on crime. As the public's confidence fell, spending on crime increased and recorded crime fell; without any uptick in the public's confidence.
One contributing factor to some of the widespread misinformation seems to be the content consumed through media channels.
Of those who believed crime was increasing, more than half suggested that information on TV was a reason they believed there was more crime. In addition to this, almost half suggested that what they read in newspapers was a factor.
Research conducted by Stefano DellaVigna and Ethan Kaplan highlights the degree to which the media can influence voting behaviour.{ref}DellaVigna, Stefano, and Ethan Daniel Kaplan. ""The Fox News Effect: Media Bias and Voting."" The Quarterly Journal of Economics 122, no. 3 (2007): 1187-1234. Available online here.{/ref}
DellaVigna and Kaplan looked at how the introduction of Fox News between 1996 and 2000 in different towns affected voting patterns and turnout in the Presidential election of 2000. They find ""a significant effect of the introduction of Fox News on the vote share in Presidential elections between 1996 and 2000. Republicans gained 0.4 to 0.7 percentage points in the towns that broadcast Fox News. Fox News also affected voter turnout and the Republican vote share in the Senate. Our estimates imply that Fox News convinced 3 to 28 percent of its viewers to vote Republican, depending on the audience measure. The Fox News effect could be a temporary learning effect for rational voters, or a permanent effect for nonrational voters subject to persuasion.""
Another dimension to this debate is the extent to which the perceived terrorism threat affects the willingness of individuals to trade-off civil liberties.{ref}Davis, Darren W., and Brian D. Silver. ""Civil liberties vs. security: Public opinion in the context of the terrorist attacks on America."" American Journal of Political Science 48, no. 1 (2004): 28-46. Available online here.{/ref}
Darren Davis and Brian Silver find that ""the greater people’s sense of threat, the lower their support for civil liberties."" This effect is attenuated by the people's trust in government but fairly consistent across nearly all political affiliations and demographics.
With all the negative news stories and sensationalism that exists in the media it may be hard to believe things are improving. These events can be contextualized as short-term fluctuations in an otherwise positive global trend.
Quantifying this progress and identifying its causes will help researchers develop successful strategies to combat the world's problems. We discuss many important improvements in our history of global living conditions.
There is a large literature that links an optimistic outlook on life to positive health outcomes. While it is interesting to read and think about this, one should be prudent not to over-interpret these findings and consider carefully if it is possible to think of these relationships as causal:
Studies have found a link between an individual's optimism/pessimism (measured by surveys) and their health outcomes. Julia Boehm and Laura Kubzansky reviewed over 200 published studies to investigate the link between a positive psychological outlook (optimism, life satisfaction and happiness) and cardiovascular health.{ref}Boehm, Julia K., and Laura D. Kubzansky. ""The heart's content: the association between positive psychological well-being and cardiovascular health."" Psychological bulletin 138, no. 4 (2012): 655. Available online here.{/ref}
They found that a positive psychological outlook was strongly associated with a reduced risk of cardiovascular disease: “For example, the most optimistic individuals had an approximately 50% reduced risk of experiencing an initial cardiovascular event compared to their less optimistic peers.""{ref}Positive feelings may help protect cardiovascular health. Harvard School of Public Health (2012). Available online here.{/ref}
Boehm et al. (2011) also find a link between optimism and the composition of cholesterol in the blood. Optimistic individuals had higher levels of good cholesterol and lower levels of triglycerides.{ref}Boehm, Julia K., Christopher Peterson, Mika Kivimaki, and Laura Kubzansky. ""A prospective study of positive psychological well-being and coronary heart disease."" Health Psychology 30, no. 3 (2011): 259. Available online here.{/ref}
Further research using data from the Women's Health Initiative found that over an eight year period, the most optimistic women had a 9% lower risk of developing coronary heart disease and a 14% lower risk of dying from any cause.{ref}Tindle, Hilary A., Yue-Fang Chang, Lewis H. Kuller, JoAnn E. Manson, Jennifer G. Robinson, Milagros C. Rosal, Greg J. Siegle, and Karen A. Matthews. ""Optimism, cynical hostility, and incident coronary heart disease and mortality in the Women’s Health Initiative."" Circulation 120, no. 8 (2009): 656-662. Available online here.{/ref} Similar results were also found by researchers writing in the Archives of General Psychiatry; using data from the Netherlands, they found that the most optimistic individuals had a 55% reduced risk of all-cause mortality and a 23% reduced risk of cardiovascular death.
Dire predictions for the future are nothing new. Indeed we can go back centuries or even millennia and find plenty of examples of pessimistic accounts of the future of the world.
This infographic shows a series of predictions for the year in which the world will end – from religious figures to scientists like John Napier and Isaac Newton.
Predictions of dire futures are also common in fictional literature.
This beautiful visualization presents a time of predictions of the future as foretold in novels.
This page is dedicated to the research why people are optimistic or pessimistic about certain things and how this is influenced by human nature, the media, and social circumstances.
\n\n\n\nWe are interested in this topic also because it is closely linked to our motivation for publishing Our World in Data. We face big global problems, but living conditions around the world have improved in important ways; fewer people are dying of disease, conflict and famine; more of us are receiving a basic education; the world is becoming more democratic; we live longer and lead healthier lives.
\n\n\n\nWhy is that we – especially those in the developed world – often have a negative view on how the world has changed over the last decades and centuries? Why we are so pessimistic about our collective future?
\n\n\n\nIt is a peculiar empirical phenomenon that while people tend to be optimistic about their own future, they can at the same time be deeply pessimistic about the future of their nation or the world. Tali Sharot, associate professor of psychology at UCL, has popularised the idea of an innate optimism bias built into the human brain.{ref}Sharot, Tali. The Optimism Bias: A Tour of the Irrationally Positive Brain. New York: Pantheon Books, 2011.
Sharot, Tali. The Science of Optimism Why We’re Hard-wired for Hope. New York: Ted Conferences, 2012.{/ref}
That is, we tend to be optimistic rather than realistic when considering our individual future. If you were to ask newlywed couples to estimate the probability they will divorce in the future, they would likely reject the possibility outright. Yet today roughly 40% of marriages in the UK end in divorce. Another example is asking smokers to estimate their chances of getting cancer and again, most would underestimate their risk. This optimism persists even when people are presented with the relevant statistics.
\n\n\n\nConsider the following graphs from the European Union’s Eurobarometer surveys; they report people’s expectations about their own personal job situation and of the economic situation in their home country. From the end of 1995 to the middle of 2015, around 60% of people predict that their job situation will remain the same, while 20% expect their situation to improve. Compare that with the response of the same group of individuals considering the future of the economic situation in their home country. Although far less stable, the results show that most people expect the economic situation in their home country to get worse or stay the same. The expectation that things are going to worsen nationally is correlated with recessions, yet there is remarkable stability in the results for individual expectations. Does the response to the question about national economic well being better correspond to an individual’s true job prospects?
\n\n\n\n![\""EU](\""http://ourworldindata.org/app/uploads/2015/08/ourworldindata_eu-survey-responses-on-optimism.png\"")
This pattern is also observed on a larger scale. This chart shows how individuals in the UK respond to the question: “Thinking about …, how much of a problem do you think each of the following are in your local area and in the whole of the UK?” Individuals tend to believe problems are more pronounced nationally than in their local area.
\n\n\n\n![\""LocalOptimismGlobalPessimism_IPSOSMori\""](\""http://ourworldindata.org/app/uploads/2014/11/ourworldindata_localoptimismglobalpessimism_ipsosmori.png\"")
This chart shows how many individuals rate the environment in their local area as fairly or very bad, compared with the environment nationally and globally. Again, we observe a similar pattern for most countries. No matter where you ask people are much more negative about places that are far away – places which they know less from their own experience and more through the media.
\n\n\n\n![\""Percentage](\""http://ourworldindata.org/app/uploads/2014/11/ourworldindata_percentage-of-respondents-who-evaluate-the-environmental-quality.png\"")
How can we reconcile this individual optimism with social pessimism? Paul Dolan, professor of behavioural science at LSE, believes people respond pessimistically to questions about national or international performance for three reasons:
\n\n\n\nThis explanation suggests there is a problem of information. If we do not pay attention to human development, then our judgement may suffer from a bias related to transient events or framing. The Gapminder Ignorance Project – which studied how wrong or right people are informed about global development – suggests the reason for all this ignorance is:
\n\n\n\n\n\n\n\n“Statistical facts don’t come to people naturally. Quite the opposite. Most people understand the world by generalizing personal experiences which are very biased. In the media the “news-worthy” events exaggerate the unusual and put the focus on swift changes. Slow and steady changes in major trends don’t get much attention. Unintentionally, people end-up carrying around a sack of outdated facts that you got in school (including knowledge that often was outdated when acquired in school).”
Another explanation put forward by Martin Seligman, professor of psychology at the University of Pennsylvania, suggests a link between control and optimism. If we feel more in control of our lives, we tend to be happier, healthier and more optimistic about the future. This could also help to explain the gap between individual and societal optimism: since we are in direct control of our own lives but not the destiny of the nation we feel more optimistic about ourselves.
\n\n\n\nAt Our World in Data we aim to bring together the empirical data and research to show how living conditions around the world are changing. Is that necessary?
\n\n\n\nThe opinion research organization Ipsos MORI conducted a detailed survey of 26,489 people across 28 countries that gives us an answer.{ref}The full reference of the survey is ‘Chris Jackson (2017) – Global Perceptions of Development Progress: ‘Perils of Perceptions’ Research’, published by Ipsos MORI, 18 September 2017.{/ref}
\n\n\n\nThe first chart shows how the surveyed people answered the following question: “In the last 20 years, the proportion of the world population living in extreme poverty has decreased, increased, or remained the same?”
\n\n\n\nThe majority of people – 52% – believe that the share of people in extreme poverty is rising. The opposite is true. In fact, the share of people living in extreme poverty across the world has been declining for two centuries and in the last 20 years this positive development has been faster than ever before (see our work on Poverty). For the recent era it doesn’t even matter what poverty line you choose, the share of people below any poverty line has fallen (see here).
\n\n\n\nThere are some people who answered the question correctly: every fifth person knows that poverty is falling. But it’s interesting that the share of correct answers differs substantially across countries. The countries I marked with a star are those that were a low-income or lower-middle-income countries a generation ago (in 1990). In these poorer countries more people understand how global poverty has changed. People in richer countries on the other hand – in which the majority of the population escaped extreme poverty some generations ago – have a very wrong perception about what is happening to global poverty.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2018/07/Public-perception-of-Change-in-poverty.png\"")
We are not just wrong about global poverty. In the same survey people were asked: “In the last 20 years, has the child mortality rate in developing regions increased, decreased or stayed about the same?”
\n\n\n\nHere again the data is very clear. The child mortality rate in both the less- and least-developed countries has halved in the last 20 years.{ref}In fact not only the average child mortality rate has fallen, but the child mortality rate has fallen in all countries (except for two very small ones).{/ref}
The survey once more shows that most people are not aware of this. On average only 39% know that the mortality of children is falling. And what greater achievement has humanity ever achieved than making it more and more likely that children survive the first, vulnerable years of their lives and sparing parents the sadness of losing their babies? This has to be one of humanity’s greatest achievements.
And just as with knowledge about extreme poverty, the share of uninformed people is much higher in the rich countries of the world. So is our work at Our World in Data needed? This survey shows that few Senegalese or Kenyans will learn something new; but if you have some friends in the US or Japan you will probably help them if you share our work.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2018/07/Public-perception-of-Change-in-u5mr.png\"")
The widespread ignorance about these truly important changes in the world feeds into a general discontent about how the world is changing. When YouGov asked in a separate survey the more general question: “All things considered, do you think the world is getting better or worse?” there were very few who gave a positive answer. In France and Australia only 3%(!) think the world is getting better.
\n\n\n\nAnd again we see that in poorer countries the share of people who answer positively is higher.{ref}This chart shows the income level difference between the very negative rich countries and the somewhat less negative poorer countries.{/ref}
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2018/07/Optimistic-about-the-future.png\"")
What should we make of the fact that many perceive the world to be stagnating or even declining in global health or poverty while we are in fact achieving the most rapid improvements in our history in these very same aspects?
\n\n\n\nFirst, this is simply sad. It means that we think worse of the world than we should. We think more poorly than we should about the time we are living in, and we think more poorly than we should about what people around the world are achieving right now.
\n\n\n\nSecond it makes clear that we are doing a terrible job at understanding and communicating what is happening in the world. Particularly in rich countries the education systems and media are failing to convey an accurate perspective on how the world is changing – arguably one of the main expectations we should have of them.{ref}This finding is the starting point for the recently published book ’Factfulness’ coauthored by Anna Rosling Rönnlund, and Hans and Ola Rosling. The authors go on to explain that our perception is so very wrong because our minds are paying attention to extremes – the very richest, very poorest, most violent and most corrupt aspects of our world – so that we end up with what they call the ‘overdramatic worldview’, which is pieced together by all the most dramatic aspects of our world, but has a massive blind spot for the world that is the reality of most people in the world. The overdramatic worldview leaves us with a picture of the world that includes all the stories that are in fact rare (the fact that they are extraordinary is why they are reported in the media), but which has no understanding of what is actually common.{/ref}
\n\n\n\nOur perception of how the world is changing matters for what we believe is possible in the future.
\n\n\n\nIf we ask people about what is possible for the world, then most of us answer ‘not much’. This chart documents the survey answers to the question “over the next 15 years, do you think living conditions for people around the world get better or worse?”. More than half of the people expect stagnation or that things will be getting worse. Fortunately, the places in which people currently have the worst living conditions are more optimistic about what is possible in the coming years.
\n\n\n\nOn the whole, the findings from the surveys are clear: we do not only believe that the world is stagnating or declining, we also expect that this perceived stagnation or decline will continue into the future.
\n\n\n\nThis pessimism about what is possible for the world matters politically. Those who don’t expect that things get better in the first place will be less likely to demand actions that can bring positive developments about. The few optimists on the other hand will want to see the necessary changes for the improvements they are expecting.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2018/07/Optimistic-about-the-future-of-the-world-.png\"")
Finally the survey suggests that there is a connection between our perception of the past and our hope for the future. This chart shows that the degree of optimism about the future differs hugely by the level of people’s knowledge about global development.
\n\n\n\nThose that were most pessimistic about the future tended to have the least basic knowledge on how the world has changed. Of those who could not give a single correct answer to the survey questions, only 17% expect the world to be better off in the future. At the other end of the spectrum, those who had very good knowledge about how the world has changed were the most optimistic about the changes that we can achieve in the next 15 years.
\n\n\n\nThis is a correlation and as we know, correlation does not imply causation. To understand whether there is a causal link we would need to know whether getting a more accurate picture of how the world is changing makes one change one’s belief about what will happen in the future. Unfortunately I am not aware of a study that looked into this question.{ref}Please do get in contact if you are aware of a study that investigates this question and I will update this section of the post – and I’d be really grateful as I would very much like to understand this link.{/ref}
\n\n\n\nOf course no one can know how the future turns out and there is nothing that would make the progress we have seen in recent decades continue inevitably and not every global development pessimist is ill-informed. But what we do know from these surveys is that these two views go together: Those who are pessimistic are much more likely to have little understanding about what is happening in the world.
\n\n\n\nObviously the question then is, why is it that better informed people are more optimistic about the future?
\n\n\n\nAs we have seen, being wrong about global development mostly means being too negative about how the world is changing. Being wrong in these questions means having a cynical worldview. Cynicism suggests that nothing can be done to improve our situation and every effort to do so is bound to fail. Our history, the cynics say, is a history of failures and what we can expect for the future is more of the same.
\n\n\n\nIn contrast to this, answering the questions correctly means that you understand that things can change. An accurate understanding of how global health and poverty are improving leaves no space for cynicism. Those who are optimistic about the future can base their view on the knowledge that it is possible to change the world for the better, because they know that we did.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2018/07/Share-who-believe-that-the-world-will-be-better_by-knowledge-about-the-world.png\"")
Declinism refers to the belief that a country or some other institution is in decline. Declinism was a prevalent feature of British political and economic history, whereby the decline of Britain as a world power was seen as the result of internal failures rather than international forces or global convergence. David Edgerton writes: “Declinism is beginning to appear as one of the last vestiges of imperial grandeur: for declinism holds, implicitly but clearly, that if Britain had done better it would have remained a much larger player on the world stage.”{ref}David Edgerton. “The Decline of Declinism.” The Business History Review, Vol. 71, No. 2 (Summer, 1997), pp. 201-206. Available online at http://www.jstor.org/stable/3116157{/ref}
\n\n\n\nToday declinism in the United States is fashionable with many politicians. Donald Trump’s campaign slogan for the 2016 Republican nomination election was “Make America Great Again!”
\n\n\n\nThe major flaw in much of the declinist narrative is the failure to distinguish between absolute and relative changes. Between 2010-14, US real GDP growth rates have fluctuated between 1.5-2.5% and yet, the US economy was recently overtaken by the Chinese economy measured in PPP-adjusted terms.{ref}For more information on PPP-adjustments, please visit the economic growth page.{/ref}
\n\n\n\nIn many ways this may capture the reason why the most developed nations tend to believe that their economy is in decline: relative decline is interpreted as absolute decline. Unsurprisingly, new EU member states tend to be much more optimistic about the future. The four largest economies — the UK, France, Germany and Italy — are the most pessimistic. This pattern persists when considering economies at different stages of development: developing countries are more optimistic about the future, while developed ones tend to be pessimistic.
\n\n\n\n![\""Optimism](\""http://ourworldindata.org/app/uploads/2015/08/ourworldindata_optimism-surveys-on-the-next-generation-by-country.png\"")
One interesting explanation for declinism is that it is the result of the way we encode memories and what we remember. Firstly, researchers have long established a robust pattern in the age at which we retain the most memories. In old age, memories from our lives are not evenly distributed but instead concentrated in two regions. These regions are (1) memories formed in adolescence and early adulthood, between the ages of 10-30, and (2) recent memory of events. The following figure is a useful representation of this distribution.
\n\n\n\n![\""Lifespan](\""http://ourworldindata.org/app/uploads/2015/08/ourworldindata_lifespan-retrieval-curve.jpg\"")
Secondly, research finds that as we get older we tend to have – on average – fewer negative experiences and that we are more likely to remember the positive ones over the negative ones.{ref}Mara Mather, Laura L. Carstensen, Aging and motivated cognition: the positivity effect in attention and memory, Trends in Cognitive Sciences, Volume 9, Issue 10, October 2005, Pages 496-502, ISSN 1364-6613. Available online here.{/ref}
\n\n\n\nThis effect combined with the reminiscence bump could explain why declinism exists among older generations, and why your parents could never stand the music you listened to! The universality of this effect is illustrated by Harvey Daniels with the use of these quotes about the decline of the English language{ref}Daniels, Harvey. Famous last words: The American language crisis reconsidered. Southern Illinois University Press, 1983.{/ref}:
\n\n\n\nIn the light of this research on human nature it is then not surprising that one of the earliest Sumerian tablets discovered and deciphered by modern scholars was a complaint by a teacher about his students’ writing ability.
\n\n\n\nThere are three main reasons we should try to combat social pessimism and declinism. The first reason is simple; indicators of living standards are significantly improving around the world. By monitoring and researching these changes we can identify ways in which progress can be achieved. Over the long-run, say 50-100 years, human progress has been staggering with the benefits not confined to the richest or most powerful. The second reason is that if our perceptions of the reality are wrong, we can end up prioritising the wrong things and making ineffectual change. Finally, being optimistic can be good for your health, while having a pessimistic outlook can be detrimental to your health.
\n\n\n\nThe public perception of these indicators matters because it directly influences the priorities of voters in democratic countries and politicians. If, as in the example above, the public believes crime is increasing, it is likely that it demands more policing not for a reason grounded in reality, but for an imagined worsening of the society they live it. This is one reason why incorrect public perceptions can be a problem.
\n\n\n\nThe following figures underline just how sizable these effects can be. The first shows how spending on crime has moved with the public’s confidence in the government’s ability to crack down on crime. As the public’s confidence fell, spending on crime increased and recorded crime fell; without any uptick in the public’s confidence.
\n\n\n\n![\""Spending](\""http://ourworldindata.org/app/uploads/2015/08/ourworldindata_spending-on-crime-crime-levels-and-public-confidence-ipsos-mori.png\"")
One contributing factor to some of the widespread misinformation seems to be the content consumed through media channels.
\n\n\n\nOf those who believed crime was increasing, more than half suggested that information on TV was a reason they believed there was more crime. In addition to this, almost half suggested that what they read in newspapers was a factor.
\n\n\n\n![\""Why](\""http://ourworldindata.org/app/uploads/2015/08/ourworldindata_why-do-people-think-there-is-more-crime-ipsos-mori.png\"")
Research conducted by Stefano DellaVigna and Ethan Kaplan highlights the degree to which the media can influence voting behaviour.{ref}DellaVigna, Stefano, and Ethan Daniel Kaplan. “The Fox News Effect: Media Bias and Voting.” The Quarterly Journal of Economics 122, no. 3 (2007): 1187-1234. Available online here.{/ref}
\n\n\n\nDellaVigna and Kaplan looked at how the introduction of Fox News between 1996 and 2000 in different towns affected voting patterns and turnout in the Presidential election of 2000. They find “a significant effect of the introduction of Fox News on the vote share in Presidential elections between 1996 and 2000. Republicans gained 0.4 to 0.7 percentage points in the towns that broadcast Fox News. Fox News also affected voter turnout and the Republican vote share in the Senate. Our estimates imply that Fox News convinced 3 to 28 percent of its viewers to vote Republican, depending on the audience measure. The Fox News effect could be a temporary learning effect for rational voters, or a permanent effect for nonrational voters subject to persuasion.”
\n\n\n\nAnother dimension to this debate is the extent to which the perceived terrorism threat affects the willingness of individuals to trade-off civil liberties.{ref}Davis, Darren W., and Brian D. Silver. “Civil liberties vs. security: Public opinion in the context of the terrorist attacks on America.” American Journal of Political Science 48, no. 1 (2004): 28-46. Available online here.{/ref}
\n\n\n\nDarren Davis and Brian Silver find that “the greater people’s sense of threat, the lower their support for civil liberties.” This effect is attenuated by the people’s trust in government but fairly consistent across nearly all political affiliations and demographics.
\n\n\n\nWith all the negative news stories and sensationalism that exists in the media it may be hard to believe things are improving. These events can be contextualized as short-term fluctuations in an otherwise positive global trend.
\n\n\n\nQuantifying this progress and identifying its causes will help researchers develop successful strategies to combat the world’s problems. We discuss many important improvements in our history of global living conditions.
\n\n\n\nThere is a large literature that links an optimistic outlook on life to positive health outcomes. While it is interesting to read and think about this, one should be prudent not to over-interpret these findings and consider carefully if it is possible to think of these relationships as causal:
\n\n\n\nStudies have found a link between an individual’s optimism/pessimism (measured by surveys) and their health outcomes. Julia Boehm and Laura Kubzansky reviewed over 200 published studies to investigate the link between a positive psychological outlook (optimism, life satisfaction and happiness) and cardiovascular health.{ref}Boehm, Julia K., and Laura D. Kubzansky. “The heart’s content: the association between positive psychological well-being and cardiovascular health.” Psychological bulletin 138, no. 4 (2012): 655. Available online here.{/ref}
\n\n\n\nThey found that a positive psychological outlook was strongly associated with a reduced risk of cardiovascular disease: “For example, the most optimistic individuals had an approximately 50% reduced risk of experiencing an initial cardiovascular event compared to their less optimistic peers.”{ref}Positive feelings may help protect cardiovascular health. Harvard School of Public Health (2012). Available online here.{/ref}
\n\n\n\nBoehm et al. (2011) also find a link between optimism and the composition of cholesterol in the blood. Optimistic individuals had higher levels of good cholesterol and lower levels of triglycerides.{ref}Boehm, Julia K., Christopher Peterson, Mika Kivimaki, and Laura Kubzansky. “A prospective study of positive psychological well-being and coronary heart disease.” Health Psychology 30, no. 3 (2011): 259. Available online here.{/ref}
\n\n\n\nFurther research using data from the Women’s Health Initiative found that over an eight year period, the most optimistic women had a 9% lower risk of developing coronary heart disease and a 14% lower risk of dying from any cause.{ref}Tindle, Hilary A., Yue-Fang Chang, Lewis H. Kuller, JoAnn E. Manson, Jennifer G. Robinson, Milagros C. Rosal, Greg J. Siegle, and Karen A. Matthews. “Optimism, cynical hostility, and incident coronary heart disease and mortality in the Women’s Health Initiative.” Circulation 120, no. 8 (2009): 656-662. Available online here.{/ref} Similar results were also found by researchers writing in the Archives of General Psychiatry; using data from the Netherlands, they found that the most optimistic individuals had a 55% reduced risk of all-cause mortality and a 23% reduced risk of cardiovascular death.
\n\n\n\nDire predictions for the future are nothing new. Indeed we can go back centuries or even millennia and find plenty of examples of pessimistic accounts of the future of the world.
\n\n\n\nThis infographic shows a series of predictions for the year in which the world will end – from religious figures to scientists like John Napier and Isaac Newton.
\n\n\n\n![\""20121222_woc6530\""](\""http://ourworldindata.org/app/uploads/2014/11/ourworldindata_20121222_woc6530.png\"")
Predictions of dire futures are also common in fictional literature.
\n\n\n\nThis beautiful visualization presents a time of predictions of the future as foretold in novels.
\n\n\n\n![\""futureevents_giorgialupi_large1\""](\""http://ourworldindata.org/app/uploads/2014/11/ourworldindata_futureevents_giorgialupi_large1.jpg\"")
Now in GDoc: https://owid.cloud/admin/gdocs/1Ame1LHwda2O45NasOLFDDbWY3HWaTAToB0FbmYdLVwI/preview
","{""id"": ""wp-56037"", ""slug"": ""waste-management"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Now in GDoc: https://owid.cloud/admin/gdocs/1Ame1LHwda2O45NasOLFDDbWY3HWaTAToB0FbmYdLVwI/preview"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Waste Management"", ""authors"": [""Hannah Ritchie"", ""Edouard Mathieu""], ""dateline"": ""February 23, 2023"", ""sidebar-toc"": false, ""featured-image"": ""Waste-Management.png""}, ""createdAt"": ""2023-02-23T14:46:08.000Z"", ""published"": false, ""updatedAt"": ""2023-08-11T14:20:14.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-23T14:46:28.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 1, ""numErrors"": 0, ""wpTagCounts"": {""paragraph"": 1}, ""htmlTagCounts"": {""p"": 1}}",2023-02-23 14:46:28,2024-02-16 14:22:42,,"[""Hannah Ritchie"", ""Edouard Mathieu""]",,2023-02-23 14:46:08,2023-08-11 14:20:14,https://ourworldindata.org/wp-content/uploads/2023/02/Waste-Management.png,{},Now in GDoc: https://owid.cloud/admin/gdocs/1Ame1LHwda2O45NasOLFDDbWY3HWaTAToB0FbmYdLVwI/preview,"{""id"": 56037, ""date"": ""2023-02-23T14:46:28"", ""guid"": {""rendered"": ""https://owid.cloud/?page_id=56037""}, ""link"": ""https://owid.cloud/waste-management"", ""meta"": {""owid_publication_context_meta_field"": [], ""owid_key_performance_indicators_meta_field"": []}, ""slug"": ""waste-management"", ""tags"": [], ""type"": ""page"", ""title"": {""rendered"": ""Waste Management""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56037""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/page""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=56037"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=56037"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=56037"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=56037""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56037/revisions"", ""count"": 3}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/56040"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57975, ""href"": ""https://owid.cloud/wp-json/wp/v2/pages/56037/revisions/57975""}]}, ""author"": 17, ""parent"": 0, ""status"": ""publish"", ""content"": {""rendered"": ""\nNow in GDoc: https://owid.cloud/admin/gdocs/1Ame1LHwda2O45NasOLFDDbWY3HWaTAToB0FbmYdLVwI/preview
\n"", ""protected"": false}, ""excerpt"": {""rendered"": """", ""protected"": false}, ""date_gmt"": ""2023-02-23T14:46:28"", ""modified"": ""2023-08-11T15:20:14"", ""template"": """", ""categories"": [], ""menu_order"": 20, ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Edouard Mathieu""], ""modified_gmt"": ""2023-08-11T14:20:14"", ""comment_status"": ""closed"", ""featured_media"": 56040, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/02/Waste-Management-150x79.png"", ""medium_large"": ""/app/uploads/2023/02/Waste-Management-768x403.png""}}" 56007,"How we choose which topics to work on, and which metrics to provide",choosing-our-topics-and-metrics,post,publish,"On Our World in Data, we present thousands of metrics on hundreds of topics.
But there are many more topics that we could work on, and metrics we could present. How do we choose them?
We cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria:
We evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed.
For each topic, we work to provide the best metrics to understand it. What metrics are 'best' will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria:
The topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at info@ourworldindata.org.
I thank Edouard Mathieu, Esteban Ortiz-Ospina, Hannah Ritchie, and Max Roser for their very helpful comments and ideas about how to improve this article.
","{""id"": ""wp-56007"", ""slug"": ""choosing-our-topics-and-metrics"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""On "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Our World in Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", we present thousands of metrics on hundreds of topics."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But there are many more topics that we could work on, and metrics we could present. How do we choose them?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How do we choose which topics to cover?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It affects many countries and people."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This can mean that it concerns every person, such as "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/health-meta"", ""children"": [{""text"": ""health"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". It can mean that it affects many people in all countries, such as "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/poverty"", ""children"": [{""text"": ""poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Or it can mean that it affects many people in fewer countries, such as "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/malaria"", ""children"": [{""text"": ""malaria"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It comes with great costs or benefits. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""The costs or benefits can be direct, and shorten people’s lives or mean they lead happier lives. An example is the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/coronavirus"", ""children"": [{""text"": ""COVID-19 pandemic"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which has immediately affected people’s well-being. But the costs or benefits can also be indirect, and worsen or alleviate other problems. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/agricultural-production"", ""children"": [{""text"": ""agricultural production"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which affects many people’s access to nutrition."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It poses significant risks."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that it may not impose great costs at the moment, but may do so in the future. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/nuclear-weapons"", ""children"": [{""text"": ""nuclear weapons"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which have not been used in decades but whose use would be devastating."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It will remain important, or become more important in the future."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""url"": ""https://ourworldindata.org/poverty"", ""children"": [{""text"": ""Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" is an example of a topic that will remain important, as many people remain impoverished even if fewer people live in extreme poverty than in the past. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""Artificial intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" is an example of a topic that will become more important, as technological advances continuously expand its effects on people’s lives."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is helpful to understand other topics."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" Many of the topics we focus on are problems in themselves. But we also provide data and research on major changes that help us understand and address these problems. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/world-population-growth"", ""children"": [{""text"": ""population changes"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which are crucial to better understand energy and education needs."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is poorly understood."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means the public knows little about a problem or frequently misunderstands it, such as because the data on it is not described well. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/plastic-pollution"", ""children"": [{""text"": ""plastic pollution"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", where data and research were often missing from the public conversation."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is neglected elsewhere."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This can mean that other organizations do not cover it, or do so in a limited fashion. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/biodiversity"", ""children"": [{""text"": ""biodiversity"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", where data on global changes are hard to find elsewhere. This also means that if others cover a topic well, we are less likely to cover it ourselves."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""We have expertise on it."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" If we have someone on our team with deep knowledge of the area, we are more likely to cover the topic. Ideally, we would have both a researcher and a data scientist with this expertise. An example is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracy"", ""children"": [{""text"": ""democracy"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", where we expanded our work as our team grew."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""We have funding available for it. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""While most of our funding comes from unrestricted resources, including "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/donate"", ""children"": [{""text"": ""reader donations"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", we partially fund our work through grants that cover work on specific areas. Importantly, we only apply for these grants if we have editorial independence: that they are on topics we want to cover in depth anyway, and there are no requirements on "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""how"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" to cover the topic."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How do we choose which metrics to provide?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""For each topic, we work to provide the best metrics to understand it. What metrics are 'best' will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It covers large parts of the world. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""True to our name, we seek metrics which cover as much of the world as possible. Only then can they help us understand global differences and changes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It covers a lot of time. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""This means both that the measure goes as far back in time as possible, and that it is as recent as possible. It then can help us understand both historical and very recent developments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is comparable across time and space"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "". This means that we prefer metrics that can be compared across years and countries. This allows us to evaluate whether countries are making progress or falling behind, and how countries are doing relative to another."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It captures what we are trying to measure"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "". This means that the metric does not give an incomplete or misleading answer to the question we have. For example, an inadequate measure for whether a country is a democracy is the share of the population that voted. Looking only at voter turnout ignores whether citizens had more than one choice at the ballot box. And at the same time, it inadvertently considers citizens that were coerced to vote."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is reliable."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that the metric is consistent, i.e. it captures the phenomenon similarly when measured repeatedly, and therefore is precise, and captures the phenomenon with little error. A consistent and precise metric makes us more confident in what it tells us about the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Its construction is transparent."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that we prefer metrics that come with a detailed description of how it was constructed, why it was constructed in this way, and with the underlying code and raw data. We, and you as our reader, then can evaluate its strengths and weaknesses in detail."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is easy to understand"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "". This means that the metric captures something that people are broadly familiar with, and they can broadly make sense of its construction. It then can provide answers that people beyond experts can learn from."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is maintained well."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that the data source updates the metric frequently, and provides reasonably up-to-date data. We often favor data from international institutions (such as the World Bank and the UN) and research institutions (such as the Global Carbon Project and the Varieties of Democracy project) over data from individual academic publications, because the former have the mandate and resources to keep this data up-to-date."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Its values differ a lot from the same measure by another trusted source."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means a metric captures disagreement across sources. It then helps us to be appropriately uncertain of our answers in light of disagreeing sources."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""It is accessible."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" This means that the data is published in a publicly accessible document and is licensed to be reused by us and preferably others. Only then can it help people answer their questions, on and beyond our site."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""We have the tools to visualize it. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""This means a metric is structured such that our in-house visualization tool — the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/owid-grapher"", ""children"": [{""text"": ""Our World in Data Grapher"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" — can display its information well. For example, our maps are set up to visualize national data, and currently cannot display metrics at the sub-national level or gridded data."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at "", ""spanType"": ""span-simple-text""}, {""url"": ""mailto:info@ourworldindata.org"", ""children"": [{""text"": ""info@ourworldindata.org"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""Acknowledgements"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I thank Edouard Mathieu, Esteban Ortiz-Ospina, Hannah Ritchie, and Max Roser for their very helpful comments and ideas about how to improve this article."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""How we choose which topics to work on, and which metrics to provide"", ""authors"": [""Bastian Herre""], ""excerpt"": ""On Our World in Data, we present thousands of metrics on hundreds of topics. How do we choose them?"", ""dateline"": ""February 27, 2023"", ""subtitle"": ""On Our World in Data, we present thousands of metrics on hundreds of topics. How do we choose them?"", ""sidebar-toc"": false, ""featured-image"": ""OurWorldInData.png""}, ""createdAt"": ""2023-02-22T13:16:34.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T16:46:10.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-27T08:30:01.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 30, ""numErrors"": 20, ""wpTagCounts"": {""list"": 20, ""heading"": 3, ""paragraph"": 31}, ""htmlTagCounts"": {""p"": 31, ""h4"": 3, ""ul"": 20}}",2023-02-27 08:30:01,2024-02-16 14:22:54,1KKAs2S3PBqQLeDtaC-QrSGT0IzewyNF7DyGToEIIb1M,"[""Bastian Herre""]","On Our World in Data, we present thousands of metrics on hundreds of topics. How do we choose them?",2023-02-22 13:16:34,2023-07-10 16:46:10,https://ourworldindata.org/wp-content/uploads/2016/06/OurWorldInData.png,{},"On _Our World in Data_, we present thousands of metrics on hundreds of topics. But there are many more topics that we could work on, and metrics we could present. How do we choose them? ## How do we choose which topics to cover? We cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria: * **It affects many countries and people.** This can mean that it concerns every person, such as [health](https://ourworldindata.org/health-meta). It can mean that it affects many people in all countries, such as [poverty](https://ourworldindata.org/poverty). Or it can mean that it affects many people in fewer countries, such as [malaria](https://ourworldindata.org/malaria). * **It comes with great costs or benefits. **The costs or benefits can be direct, and shorten people’s lives or mean they lead happier lives. An example is the [COVID-19 pandemic](https://ourworldindata.org/coronavirus), which has immediately affected people’s well-being. But the costs or benefits can also be indirect, and worsen or alleviate other problems. An example is [agricultural production](https://ourworldindata.org/agricultural-production), which affects many people’s access to nutrition. * **It poses significant risks.** This means that it may not impose great costs at the moment, but may do so in the future. An example is [nuclear weapons](https://ourworldindata.org/nuclear-weapons), which have not been used in decades but whose use would be devastating. * **It will remain important, or become more important in the future.**[Poverty](https://ourworldindata.org/poverty) is an example of a topic that will remain important, as many people remain impoverished even if fewer people live in extreme poverty than in the past. [Artificial intelligence](https://ourworldindata.org/artificial-intelligence) is an example of a topic that will become more important, as technological advances continuously expand its effects on people’s lives. * **It is helpful to understand other topics.** Many of the topics we focus on are problems in themselves. But we also provide data and research on major changes that help us understand and address these problems. An example is [population changes](https://ourworldindata.org/world-population-growth), which are crucial to better understand energy and education needs. * **It is poorly understood.** This means the public knows little about a problem or frequently misunderstands it, such as because the data on it is not described well. An example is [plastic pollution](http://ourworldindata.org/plastic-pollution), where data and research were often missing from the public conversation. * **It is neglected elsewhere.** This can mean that other organizations do not cover it, or do so in a limited fashion. An example is [biodiversity](http://ourworldindata.org/biodiversity), where data on global changes are hard to find elsewhere. This also means that if others cover a topic well, we are less likely to cover it ourselves. * **We have expertise on it.** If we have someone on our team with deep knowledge of the area, we are more likely to cover the topic. Ideally, we would have both a researcher and a data scientist with this expertise. An example is [democracy](https://ourworldindata.org/democracy), where we expanded our work as our team grew. * **We have funding available for it. **While most of our funding comes from unrestricted resources, including [reader donations](https://ourworldindata.org/donate), we partially fund our work through grants that cover work on specific areas. Importantly, we only apply for these grants if we have editorial independence: that they are on topics we want to cover in depth anyway, and there are no requirements on _how_ to cover the topic. We evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed. ## How do we choose which metrics to provide? For each topic, we work to provide the best metrics to understand it. What metrics are 'best' will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria: * **It covers large parts of the world. **True to our name, we seek metrics which cover as much of the world as possible. Only then can they help us understand global differences and changes. * **It covers a lot of time. **This means both that the measure goes as far back in time as possible, and that it is as recent as possible. It then can help us understand both historical and very recent developments. * **It is comparable across time and space**. This means that we prefer metrics that can be compared across years and countries. This allows us to evaluate whether countries are making progress or falling behind, and how countries are doing relative to another. * **It captures what we are trying to measure**. This means that the metric does not give an incomplete or misleading answer to the question we have. For example, an inadequate measure for whether a country is a democracy is the share of the population that voted. Looking only at voter turnout ignores whether citizens had more than one choice at the ballot box. And at the same time, it inadvertently considers citizens that were coerced to vote. * **It is reliable.** This means that the metric is consistent, i.e. it captures the phenomenon similarly when measured repeatedly, and therefore is precise, and captures the phenomenon with little error. A consistent and precise metric makes us more confident in what it tells us about the world. * **Its construction is transparent.** This means that we prefer metrics that come with a detailed description of how it was constructed, why it was constructed in this way, and with the underlying code and raw data. We, and you as our reader, then can evaluate its strengths and weaknesses in detail. * **It is easy to understand**. This means that the metric captures something that people are broadly familiar with, and they can broadly make sense of its construction. It then can provide answers that people beyond experts can learn from. * **It is maintained well.** This means that the data source updates the metric frequently, and provides reasonably up-to-date data. We often favor data from international institutions (such as the World Bank and the UN) and research institutions (such as the Global Carbon Project and the Varieties of Democracy project) over data from individual academic publications, because the former have the mandate and resources to keep this data up-to-date. * **Its values differ a lot from the same measure by another trusted source.** This means a metric captures disagreement across sources. It then helps us to be appropriately uncertain of our answers in light of disagreeing sources. * **It is accessible.** This means that the data is published in a publicly accessible document and is licensed to be reused by us and preferably others. Only then can it help people answer their questions, on and beyond our site. * **We have the tools to visualize it. **This means a metric is structured such that our in-house visualization tool — the [Our World in Data Grapher](https://ourworldindata.org/owid-grapher) — can display its information well. For example, our maps are set up to visualize national data, and currently cannot display metrics at the sub-national level or gridded data. The topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at [info@ourworldindata.org](mailto:info@ourworldindata.org). ## **Acknowledgements** I thank Edouard Mathieu, Esteban Ortiz-Ospina, Hannah Ritchie, and Max Roser for their very helpful comments and ideas about how to improve this article.","{""id"": 56007, ""date"": ""2023-02-27T08:30:01"", ""guid"": {""rendered"": ""https://owid.cloud/?p=56007""}, ""link"": ""https://owid.cloud/choosing-our-topics-and-metrics"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""choosing-our-topics-and-metrics"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""How we choose which topics to work on, and which metrics to provide""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56007""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=56007"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=56007"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=56007"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=56007""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56007/revisions"", ""count"": 24}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/8771"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 56078, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/56007/revisions/56078""}]}, ""author"": 49, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": ""\nOn Our World in Data, we present thousands of metrics on hundreds of topics.
\n\n\n\nBut there are many more topics that we could work on, and metrics we could present. How do we choose them?
\n\n\n\n\n\n\n\nWe cover a topic if we believe it helps our readers understand one or several of the world’s largest problems. More specifically, this means that a topic will fit many of the following criteria:
\n\n\n\n\n\n\n\nWe evaluate ourselves how a topic fits these criteria. But we rely heavily on related research, especially research that is peer-reviewed.
\n\n\n\n\n\n\n\nFor each topic, we work to provide the best metrics to understand it. What metrics are ‘best’ will often depend on our specific questions. Overall, a metric we provide will fit many of the following criteria:
\n\n\n\nThe topics and metrics we present are not set in stone, and we keep thinking about which ones to add. So if you think a topic or metric fits the criteria outlined here, please reach out to us at info@ourworldindata.org.
\n\n\n\n\n\n\n\nI thank Edouard Mathieu, Esteban Ortiz-Ospina, Hannah Ritchie, and Max Roser for their very helpful comments and ideas about how to improve this article.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""On Our World in Data, we present thousands of metrics on hundreds of topics. How do we choose them?"", ""protected"": false}, ""date_gmt"": ""2023-02-27T08:30:01"", ""modified"": ""2023-07-10T17:46:10"", ""template"": """", ""categories"": [69, 1], ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre""], ""modified_gmt"": ""2023-07-10T16:46:10"", ""comment_status"": ""closed"", ""featured_media"": 8771, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2016/06/OurWorldInData-150x77.png"", ""medium_large"": ""/app/uploads/2016/06/OurWorldInData-768x393.png""}}" 55967,What does it mean for a species to be at risk of extinction?,extinction-risk-definition,post,publish,"The International Union for Conservation of Nature (IUCN) reports that tens of thousands of species are threatened with extinction.
But what does it mean for a species to be ‘threatened with extinction’? How do researchers evaluate extinction risk?
The IUCN Red List of Threatened Species is regarded as the definitive source of extinction risk. Every year, the IUCN publishes its latest assessment of the status of each evaluated species.
In this article, we look at how species are categorized and assessed for their extinction risk.
Each species is assessed against a standardized list of criteria.{ref}Here we use the term ‘species’ to talk about different groups of organisms. This is the main level that wildlife is evaluated for extinction risk. But, the IUCN also uses the broader term ‘taxon’ which applies at the species level or sub-species level. This means that the framework can be applied to evaluate extinction risk at multiple levels.{/ref} Based on these answers, a species is assigned to one of nine categories:
How these categories relate to each other is shown in the visualization.
The first two categories are about species that we can’t say much about.
Not Evaluated (NE). The first question researchers ask is whether we know enough about a species to evaluate their status. If a species doesn’t meet this criterion, they are not evaluated and removed from the assessment process. As part of the IUCN’s aims, each species is re-evaluated every five to ten years; if our understanding has developed by this point, then a species can be evaluated later.
Data Deficient (DD). For some species, we do not have sufficient data to assess their risk of extinction. We might know a lot about their biology, but without adequate data on their populations and occurrence, a species’ extinction risk can’t be evaluated.
That leaves us with seven categories where we have enough data to make an assessment.
Researchers draw on a large number of extensive studies of a species’ population size, range, and habitat, and how these are changing over time. Later we’ll look at the specific criteria that they’re evaluated against.
These are the seven categories, from extinct through to species that are abundant and of little concern:
An Extinct (EX) species is one for which there is no reasonable doubt that the last individual has died. Researchers declare this when no individuals are found in exhaustive studies of its known or expected habitats for a long period of time. This includes the golden toad and the Japanese sea lion.
An Extinct in the Wild (EW) species fits the criteria for ‘Extinct’ – exhaustive surveys have found no individuals in its expected habitats – but individuals still exist in captivity or as naturalized populations outside of its normal, historical range.
That then leaves us with five categories by which species are assessed for their risk of extinction.
A Critically Endangered (CR) species faces an extremely high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid decline in population, or a large restriction in the range of a species. One of the metrics that categorizes a critically endangered species is if quantitative analysis shows that there is a greater than 50% chance that it goes extinct in the wild within 10 years. This does not necessarily apply to species classified as critically endangered under other criteria.
An Endangered (EN) species faces a very high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid decline in population, or a large restriction in the range of a species. One of the metrics that categorizes an endangered species is if quantitative analysis shows that there is a greater than 20% chance that it goes extinct in the wild within 20 years. Again, this does not necessarily apply to species classified as endangered under other criteria.
A Vulnerable (VU) species faces a high risk of extinction. One of the metrics that categorizes a vulnerable species is if quantitative analysis shows that there is a greater than 10% chance that it goes extinct in the wild within 100 years.
A Near Threatened (NT) species does not qualify as Critically Endangered, Endangered, or Vulnerable now, but according to the researchers’ assessment, it is close to meeting this definition soon, based on recent trends.
A Least-Concern (LC) species is widespread, abundant, and not threatened with extinction.
The IUCN will often discuss species as being threatened with extinction. That’s the most commonly reported metric. On Our World in Data, we show both the number or share of studied species that are ‘threatened with extinction’ according to the IUCN.
Threatened species are those classified as either critically endangered, endangered, or vulnerable. It is the sum of species in each of these three categories.
We’ve covered all the categories that species are classified into. But how do researchers arrive at the categorization? How do researchers decide how vulnerable a species is to extinction?
Species are evaluated across four key metrics. It’s important to note that this doesn’t include every possible way to assess the future health of a species: aspects such as food supplies or future hunting threats aren’t defined as measure risks.
Any of these would indicate that a species is at risk:
Within each of these measurements, researchers set thresholds that detail whether a species is vulnerable, endangered, or critically endangered.
Take metric (1): the size of the population. A species is considered ‘Critically Endangered’ if there are fewer than 50 mature individuals globally. It’s ‘Endangered’ if there are fewer than 250 individuals, and ‘Vulnerable’ if there are fewer than 1,000.
Or metric (4): the geographic range of the species. A species is considered ‘Critically Endangered’ if its area of occupancy is less than 10 square kilometers (km2). It’s ‘Endangered’ if this area is less than 500 km2, and ‘Vulnerable’ if it’s less than 2,000 km2.{ref}To be put into any of these categories, a species also has to meet 2 of these 3 conditions: its habitats are very fragmented or small in number; have a continued decline in its geographical range; or have extreme fluctuations in its range.{/ref}
The final metric that researchers use to classify species is the probabilistic assessment on the likelihood of extinction within a given time period. For example, a critically endangered species is expected to have a greater than 50% chance of extinction within 10 years if threats are not reduced, and protections put in place.
You can find a complete list of the criteria for each extinction risk category in the IUCN’s definition guide.These assessments are conducted and prepared by the IUCN’s network of specialist groups: this includes 10,500 volunteer experts across more than 160 Specialist Groups, Conservation Committees, and Task Forces. There is, for example, an IUCN African Elephant Specialist Group that focuses on the monitoring of African elephant populations. These volunteers and expert groups do not necessarily do all of the original research themselves: they will often form their assessments based on peer-reviewed academic literature conducted by other researchers.
The number of described species in the world is now over 2 million (with more than 1 million being insects, which the IUCN does not assess).{ref}The total number of species is estimated to be much higher – widely-cited figures are in the range of 8 to 9 million.{/ref} The IUCN has assessed more than 150,000 species for their extinction risk. A large number, but a small fraction of the total.
They aim to re-evaluate every species in a peer-reviewed process every five years – every ten years at most. Many at-risk species are monitored and evaluated more frequently. That’s an incredible achievement. But it’s still just a fraction of the species that exist. We don’t know the extinction risk for most of the world’s species.
Additionally, the uncertainty can be high for the species that the IUCN researchers have evaluated. The dynamics of wildlife populations can change quickly; can vary a lot from one location to another; and measurements are prone to error.
In this case, a species might be classified as ‘data deficient’. But often the data is deemed adequate, just with a higher level of uncertainty.
A species moving from one category to another does not always mean a dramatic improvement or regression in its health. It can simply be a re-evaluation with better data.
Despite these challenges, the IUCN Red List provides an essential resource to understand the extinction risk of our world’s species. It highlights which species are most at-risk, so we can protect them.
","{""id"": ""wp-55967"", ""slug"": ""extinction-risk-definition"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""The International Union for Conservation of Nature (IUCN) reports that tens of thousands of species are "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/number-species-threatened"", ""children"": [{""text"": ""threatened with extinction"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But what does it mean for a species to be ‘threatened with extinction’? How do researchers evaluate extinction risk?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.iucnredlist.org/"", ""children"": [{""text"": ""IUCN "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Red List of Threatened Species"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""spanType"": ""span-link""}, {""text"": "" is regarded as the definitive source of extinction risk. Every year, the IUCN publishes its latest assessment of the status of each evaluated species."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this article, we look at how species are categorized and assessed for their extinction risk."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""The 9 categories of extinction risk"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Each species is assessed against a standardized list of criteria.{ref}Here we use the term ‘species’ to talk about different groups of organisms. This is the main level that wildlife is evaluated for extinction risk. But, the IUCN also uses the broader term ‘taxon’ which applies at the species level or sub-species level. This means that the framework can be applied to evaluate extinction risk at multiple levels.{/ref} Based on these answers, a species is assigned to one of "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""nine categories"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "":"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""text"": ""Not Evaluated (NE)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Data Deficient (DD)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Least Concern (LC)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Near Threatened (NT)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Vulnerable (VU)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Endangered (EN)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Critically Endangered (CR)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Extinct in the Wild (EW)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Extinct (EX)"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How these categories relate to each other is shown in the visualization."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The first two categories are about species that we can’t say much about."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Not Evaluated (NE). "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""The first question researchers ask is whether we know enough about a species to evaluate their status. If a species doesn’t meet this criterion, they are not evaluated and removed from the assessment process. As part of the IUCN’s aims, each species is re-evaluated every five to ten years; if our understanding has developed by this point, then a species can be evaluated later."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Data Deficient (DD)."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" For some species, we do not have sufficient data to assess their risk of extinction. We might know a lot about their biology, but without adequate data on their populations and occurrence, a species’ extinction risk can’t be evaluated."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That leaves us with seven categories where we have enough data to make an assessment."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Researchers draw on a large number of extensive studies of a species’ population size, range, and habitat, and how these are changing over time. Later we’ll look at the specific criteria that they’re evaluated against."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These are the seven categories, from extinct through to species that are abundant and of little concern:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""An "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Extinct (EX)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" species is one for which there is no reasonable doubt that the last individual has died. Researchers declare this when no individuals are found in exhaustive studies of its known or expected habitats for a long period of time. This includes the golden toad and the Japanese sea lion."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""An "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Extinct in the Wild (EW) "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""species fits the criteria for ‘Extinct’ – exhaustive surveys have found no individuals in its expected habitats – but individuals still exist in captivity or as naturalized populations outside of its normal, historical range."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That then leaves us with five categories by which species are assessed for their risk of extinction."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Critically Endangered (CR) "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""species faces an extremely high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""decline"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" in population, or a large restriction in the range of a species. One of the metrics that categorizes a critically endangered species is if quantitative analysis shows that there is a greater than 50% chance that it goes extinct in the wild within 10 years. This does not necessarily apply to species classified as critically endangered under other criteria."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""An "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Endangered (EN)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" species faces a very high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""decline"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" in population, or a large restriction in the range of a species. One of the metrics that categorizes an endangered species is if quantitative analysis shows that there is a greater than 20% chance that it goes extinct in the wild within 20 years. Again, this does not necessarily apply to species classified as endangered under other criteria."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Vulnerable (VU)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" species faces a high risk of extinction. One of the metrics that categorizes a vulnerable species is if quantitative analysis shows that there is a greater than 10% chance that it goes extinct in the wild within 100 years."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Near Threatened (NT)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" species does not qualify as Critically Endangered, Endangered, or Vulnerable now, but according to the researchers’ assessment, it is close to meeting this definition soon, based on recent trends."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Least-Concern (LC)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" species is widespread, abundant, and not threatened with extinction."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The IUCN will often discuss species as being "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""threatened with extinction"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". That’s the most commonly reported metric. On "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Our World in Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", we show both the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/number-species-threatened"", ""children"": [{""text"": ""number"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" or "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/share-threatened-species"", ""children"": [{""text"": ""share"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of studied species that are ‘threatened with extinction’ according to the IUCN."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Threatened species"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" are those classified as either "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""critically endangered"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""endangered"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "", or "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""vulnerable"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "". It is the sum of species in each of these three categories."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""IUCN-Red-List-Extinction-Categories-01.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""How do researchers assess the risk of extinction?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We’ve covered all the categories that species are classified into. But how do researchers arrive at the categorization? How do researchers decide how vulnerable a species is to extinction?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Species are evaluated across four key metrics. It’s important to note that this doesn’t include "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""every"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" possible way to assess the future health of a species: aspects such as food supplies or future hunting threats aren’t defined as measure risks."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Any of these would indicate that a species is at risk:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""numbered-list"", ""items"": [{""type"": ""text"", ""value"": [{""text"": ""If the population size is very small."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If the population size is small "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""and"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" it’s declining."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If there has been a large "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""decline"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" in population size (regardless of its total size)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The geographic range of the species – the area or region it lives in – is small and declining."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Within each of these measurements, researchers set thresholds that detail whether a species is vulnerable, endangered, or critically endangered. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Take metric (1): the size of the population. A species is considered ‘Critically Endangered’ if there are fewer than 50 mature individuals globally. It’s ‘Endangered’ if there are fewer than 250 individuals, and ‘Vulnerable’ if there are fewer than 1,000."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Or metric (4): the geographic range of the species. A species is considered ‘Critically Endangered’ if its area of occupancy is less than 10 square kilometers (km"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-superscript""}, {""text"": ""). It’s ‘Endangered’ if this area is less than 500 km"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-superscript""}, {""text"": "", and ‘Vulnerable’ if it’s less than 2,000 km"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-superscript""}, {""text"": "".{ref}To be put into any of these categories, a species also has to meet 2 of these 3 conditions: its habitats are very fragmented or small in number; have a continued decline in its geographical range; or have extreme fluctuations in its range.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The final metric that researchers use to classify species is the probabilistic assessment on the likelihood of extinction within a given time period. For example, a critically endangered species is expected to have a greater than 50% chance of extinction within 10 years if threats are not reduced, and protections put in place."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""You can find a complete list of the criteria for each extinction risk category in the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://portals.iucn.org/library/sites/library/files/documents/RL-2001-001-2nd.pdf"", ""children"": [{""text"": ""IUCN’s definition guide"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".These assessments are conducted and prepared by the IUCN’s network of specialist groups: "", ""spanType"": ""span-simple-text""}, {""url"": ""https://iucn.org/our-union/commissions/group/1445"", ""children"": [{""text"": ""this includes "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""10,500 volunteer experts across more than 160 Specialist Groups, Conservation Committees, and Task Forces. There is, for example, an "", ""spanType"": ""span-simple-text""}, {""url"": ""https://iucn.org/our-union/commissions/group/iucn-ssc-african-elephant-specialist-group"", ""children"": [{""text"": ""IUCN African Elephant Specialist Group"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" that focuses on the monitoring of African elephant populations. These volunteers and expert groups do not necessarily do all of the original research themselves: they will often form their assessments based on peer-reviewed academic literature conducted by other researchers."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Assessments of wildlife populations can be uncertain"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The number of "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/how-many-species-are-there"", ""children"": [{""text"": ""described species"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in the world is now over 2 million (with more than 1 million being insects, which the IUCN does not assess).{ref}The total "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/how-many-species-are-there"", ""children"": [{""text"": ""number of species"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" is estimated to be much higher – widely-cited figures are in the range of 8 to 9 million.{/ref} The IUCN has assessed more than 150,000 species for their extinction risk. A large number, but a small fraction of the total."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""They aim to "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ipbes.net/policy-support/tools-instruments/iucn-red-list-threatened-species"", ""children"": [{""text"": ""re-evaluate every species"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in a peer-reviewed process every five years – every ten years at most. Many at-risk species are monitored and evaluated more frequently. That’s an incredible achievement. But it’s still just a fraction of the species that exist. We don’t know the extinction risk for most of the world’s species."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Additionally, the uncertainty can be high for the species that the IUCN researchers have evaluated. The dynamics of wildlife populations can change quickly; can vary a lot from one location to another; and measurements are prone to error."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this case, a species might be classified as ‘data deficient’. But often the data is deemed adequate, just with a higher level of uncertainty."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A species moving from one category to another does not always mean a dramatic improvement or regression in its health. It can simply be a re-evaluation with better data."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Despite these challenges, the IUCN Red List provides an essential resource to understand the extinction risk of our world’s species. It highlights which species are most at-risk, so we can protect them."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""What does it mean for a species to be at risk of extinction?"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""We need to focus on the most threatened species to protect them from extinction. But what does it mean for a species to be at risk, and how is it measured?"", ""dateline"": ""February 20, 2023"", ""subtitle"": ""We need to focus on the most threatened species to protect them from extinction. But what does it mean for a species to be at risk, and how is it measured?"", ""sidebar-toc"": false, ""featured-image"": ""IUCN-Thumbnail-01.png""}, ""createdAt"": ""2023-02-20T10:56:33.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T16:16:20.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-20T10:56:33.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 23, ""numErrors"": 7, ""wpTagCounts"": {""list"": 2, ""image"": 1, ""column"": 2, ""columns"": 1, ""heading"": 3, ""paragraph"": 36, ""separator"": 2}, ""htmlTagCounts"": {""p"": 36, ""h3"": 3, ""hr"": 2, ""ol"": 1, ""ul"": 1, ""div"": 3, ""figure"": 1}}",2023-02-20 10:56:33,2024-03-10 13:14:25,14SGOgwH7D0N51LGCXHhHMnRR1FbZqEfTNI98ipiRmrk,"[""Hannah Ritchie""]","We need to focus on the most threatened species to protect them from extinction. But what does it mean for a species to be at risk, and how is it measured?",2023-02-20 10:56:33,2023-07-10 16:16:20,https://ourworldindata.org/wp-content/uploads/2023/02/IUCN-Thumbnail-01.png,{},"The International Union for Conservation of Nature (IUCN) reports that tens of thousands of species are [threatened with extinction](https://ourworldindata.org/grapher/number-species-threatened). But what does it mean for a species to be ‘threatened with extinction’? How do researchers evaluate extinction risk? The [IUCN _Red List of Threatened Species_](https://www.iucnredlist.org/) is regarded as the definitive source of extinction risk. Every year, the IUCN publishes its latest assessment of the status of each evaluated species. In this article, we look at how species are categorized and assessed for their extinction risk. ## The 9 categories of extinction risk Each species is assessed against a standardized list of criteria.{ref}Here we use the term ‘species’ to talk about different groups of organisms. This is the main level that wildlife is evaluated for extinction risk. But, the IUCN also uses the broader term ‘taxon’ which applies at the species level or sub-species level. This means that the framework can be applied to evaluate extinction risk at multiple levels.{/ref} Based on these answers, a species is assigned to one of **nine categories**: * Not Evaluated (NE) * Data Deficient (DD) * Least Concern (LC) * Near Threatened (NT) * Vulnerable (VU) * Endangered (EN) * Critically Endangered (CR) * Extinct in the Wild (EW) * Extinct (EX) How these categories relate to each other is shown in the visualization. The first two categories are about species that we can’t say much about. **Not Evaluated (NE). **The first question researchers ask is whether we know enough about a species to evaluate their status. If a species doesn’t meet this criterion, they are not evaluated and removed from the assessment process. As part of the IUCN’s aims, each species is re-evaluated every five to ten years; if our understanding has developed by this point, then a species can be evaluated later. **Data Deficient (DD).** For some species, we do not have sufficient data to assess their risk of extinction. We might know a lot about their biology, but without adequate data on their populations and occurrence, a species’ extinction risk can’t be evaluated. That leaves us with seven categories where we have enough data to make an assessment. Researchers draw on a large number of extensive studies of a species’ population size, range, and habitat, and how these are changing over time. Later we’ll look at the specific criteria that they’re evaluated against. These are the seven categories, from extinct through to species that are abundant and of little concern: An **Extinct (EX)** species is one for which there is no reasonable doubt that the last individual has died. Researchers declare this when no individuals are found in exhaustive studies of its known or expected habitats for a long period of time. This includes the golden toad and the Japanese sea lion. An **Extinct in the Wild (EW) **species fits the criteria for ‘Extinct’ – exhaustive surveys have found no individuals in its expected habitats – but individuals still exist in captivity or as naturalized populations outside of its normal, historical range. That then leaves us with five categories by which species are assessed for their risk of extinction. A **Critically Endangered (CR) **species faces an extremely high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid _decline_ in population, or a large restriction in the range of a species. One of the metrics that categorizes a critically endangered species is if quantitative analysis shows that there is a greater than 50% chance that it goes extinct in the wild within 10 years. This does not necessarily apply to species classified as critically endangered under other criteria. An **Endangered (EN)** species faces a very high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid _decline_ in population, or a large restriction in the range of a species. One of the metrics that categorizes an endangered species is if quantitative analysis shows that there is a greater than 20% chance that it goes extinct in the wild within 20 years. Again, this does not necessarily apply to species classified as endangered under other criteria. A **Vulnerable (VU)** species faces a high risk of extinction. One of the metrics that categorizes a vulnerable species is if quantitative analysis shows that there is a greater than 10% chance that it goes extinct in the wild within 100 years. A **Near Threatened (NT)** species does not qualify as Critically Endangered, Endangered, or Vulnerable now, but according to the researchers’ assessment, it is close to meeting this definition soon, based on recent trends. A **Least-Concern (LC)** species is widespread, abundant, and not threatened with extinction. The IUCN will often discuss species as being _threatened with extinction_. That’s the most commonly reported metric. On _Our World in Data_, we show both the [number](https://ourworldindata.org/grapher/number-species-threatened) or [share](https://ourworldindata.org/grapher/share-threatened-species) of studied species that are ‘threatened with extinction’ according to the IUCN. **Threatened species** are those classified as either **critically endangered**, **endangered**, or **vulnerable**. It is the sum of species in each of these three categories.The International Union for Conservation of Nature (IUCN) reports that tens of thousands of species are threatened with extinction.
\n\n\n\nBut what does it mean for a species to be ‘threatened with extinction’? How do researchers evaluate extinction risk?
\n\n\n\nThe IUCN Red List of Threatened Species is regarded as the definitive source of extinction risk. Every year, the IUCN publishes its latest assessment of the status of each evaluated species.
\n\n\n\nIn this article, we look at how species are categorized and assessed for their extinction risk.
\n\n\n\nEach species is assessed against a standardized list of criteria.{ref}Here we use the term ‘species’ to talk about different groups of organisms. This is the main level that wildlife is evaluated for extinction risk. But, the IUCN also uses the broader term ‘taxon’ which applies at the species level or sub-species level. This means that the framework can be applied to evaluate extinction risk at multiple levels.{/ref} Based on these answers, a species is assigned to one of nine categories:
\n\n\n\nHow these categories relate to each other is shown in the visualization.
\n\n\n\nThe first two categories are about species that we can’t say much about.
\n\n\n\nNot Evaluated (NE). The first question researchers ask is whether we know enough about a species to evaluate their status. If a species doesn’t meet this criterion, they are not evaluated and removed from the assessment process. As part of the IUCN’s aims, each species is re-evaluated every five to ten years; if our understanding has developed by this point, then a species can be evaluated later.
\n\n\n\nData Deficient (DD). For some species, we do not have sufficient data to assess their risk of extinction. We might know a lot about their biology, but without adequate data on their populations and occurrence, a species’ extinction risk can’t be evaluated.
\n\n\n\nThat leaves us with seven categories where we have enough data to make an assessment.
\n\n\n\nResearchers draw on a large number of extensive studies of a species’ population size, range, and habitat, and how these are changing over time. Later we’ll look at the specific criteria that they’re evaluated against.
\n\n\n\nThese are the seven categories, from extinct through to species that are abundant and of little concern:
\n\n\n\nAn Extinct (EX) species is one for which there is no reasonable doubt that the last individual has died. Researchers declare this when no individuals are found in exhaustive studies of its known or expected habitats for a long period of time. This includes the golden toad and the Japanese sea lion.
\n\n\n\nAn Extinct in the Wild (EW) species fits the criteria for ‘Extinct’ – exhaustive surveys have found no individuals in its expected habitats – but individuals still exist in captivity or as naturalized populations outside of its normal, historical range.
\n\n\n\nThat then leaves us with five categories by which species are assessed for their risk of extinction.
\n\n\n\nA Critically Endangered (CR) species faces an extremely high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid decline in population, or a large restriction in the range of a species. One of the metrics that categorizes a critically endangered species is if quantitative analysis shows that there is a greater than 50% chance that it goes extinct in the wild within 10 years. This does not necessarily apply to species classified as critically endangered under other criteria.
\n\n\n\nAn Endangered (EN) species faces a very high risk of extinction. As explained in the next section, this is based on the criteria of having a very small population size, a very rapid decline in population, or a large restriction in the range of a species. One of the metrics that categorizes an endangered species is if quantitative analysis shows that there is a greater than 20% chance that it goes extinct in the wild within 20 years. Again, this does not necessarily apply to species classified as endangered under other criteria.
\n\n\n\nA Vulnerable (VU) species faces a high risk of extinction. One of the metrics that categorizes a vulnerable species is if quantitative analysis shows that there is a greater than 10% chance that it goes extinct in the wild within 100 years.
\n\n\n\nA Near Threatened (NT) species does not qualify as Critically Endangered, Endangered, or Vulnerable now, but according to the researchers’ assessment, it is close to meeting this definition soon, based on recent trends.
\n\n\n\nA Least-Concern (LC) species is widespread, abundant, and not threatened with extinction.
\n\n\n\nThe IUCN will often discuss species as being threatened with extinction. That’s the most commonly reported metric. On Our World in Data, we show both the number or share of studied species that are ‘threatened with extinction’ according to the IUCN.
\n\n\n\nThreatened species are those classified as either critically endangered, endangered, or vulnerable. It is the sum of species in each of these three categories.
\n![\""\""](\""https://owid.cloud/app/uploads/2023/02/IUCN-Red-List-Extinction-Categories-01.png\"")
We’ve covered all the categories that species are classified into. But how do researchers arrive at the categorization? How do researchers decide how vulnerable a species is to extinction?
\n\n\n\nSpecies are evaluated across four key metrics. It’s important to note that this doesn’t include every possible way to assess the future health of a species: aspects such as food supplies or future hunting threats aren’t defined as measure risks.
\n\n\n\nAny of these would indicate that a species is at risk:
\n\n\n\nWithin each of these measurements, researchers set thresholds that detail whether a species is vulnerable, endangered, or critically endangered.
\n\n\n\nTake metric (1): the size of the population. A species is considered ‘Critically Endangered’ if there are fewer than 50 mature individuals globally. It’s ‘Endangered’ if there are fewer than 250 individuals, and ‘Vulnerable’ if there are fewer than 1,000.
\n\n\n\nOr metric (4): the geographic range of the species. A species is considered ‘Critically Endangered’ if its area of occupancy is less than 10 square kilometers (km2). It’s ‘Endangered’ if this area is less than 500 km2, and ‘Vulnerable’ if it’s less than 2,000 km2.{ref}To be put into any of these categories, a species also has to meet 2 of these 3 conditions: its habitats are very fragmented or small in number; have a continued decline in its geographical range; or have extreme fluctuations in its range.{/ref}
\n\n\n\nThe final metric that researchers use to classify species is the probabilistic assessment on the likelihood of extinction within a given time period. For example, a critically endangered species is expected to have a greater than 50% chance of extinction within 10 years if threats are not reduced, and protections put in place.
\n\n\n\nYou can find a complete list of the criteria for each extinction risk category in the IUCN’s definition guide.These assessments are conducted and prepared by the IUCN’s network of specialist groups: this includes 10,500 volunteer experts across more than 160 Specialist Groups, Conservation Committees, and Task Forces. There is, for example, an IUCN African Elephant Specialist Group that focuses on the monitoring of African elephant populations. These volunteers and expert groups do not necessarily do all of the original research themselves: they will often form their assessments based on peer-reviewed academic literature conducted by other researchers.
\n\n\n\nThe number of described species in the world is now over 2 million (with more than 1 million being insects, which the IUCN does not assess).{ref}The total number of species is estimated to be much higher – widely-cited figures are in the range of 8 to 9 million.{/ref} The IUCN has assessed more than 150,000 species for their extinction risk. A large number, but a small fraction of the total.
\n\n\n\nThey aim to re-evaluate every species in a peer-reviewed process every five years – every ten years at most. Many at-risk species are monitored and evaluated more frequently. That’s an incredible achievement. But it’s still just a fraction of the species that exist. We don’t know the extinction risk for most of the world’s species.
\n\n\n\nAdditionally, the uncertainty can be high for the species that the IUCN researchers have evaluated. The dynamics of wildlife populations can change quickly; can vary a lot from one location to another; and measurements are prone to error.
\n\n\n\nIn this case, a species might be classified as ‘data deficient’. But often the data is deemed adequate, just with a higher level of uncertainty.
\n\n\n\nA species moving from one category to another does not always mean a dramatic improvement or regression in its health. It can simply be a re-evaluation with better data.
\n\n\n\nDespite these challenges, the IUCN Red List provides an essential resource to understand the extinction risk of our world’s species. It highlights which species are most at-risk, so we can protect them.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""We need to focus on the most threatened species to protect them from extinction. But what does it mean for a species to be at risk, and how is it measured?"", ""protected"": false}, ""date_gmt"": ""2023-02-20T10:56:33"", ""modified"": ""2023-07-10T17:16:20"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2023-07-10T16:16:20"", ""comment_status"": ""closed"", ""featured_media"": 55972, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2023/02/IUCN-Thumbnail-01-150x79.png"", ""medium_large"": ""/app/uploads/2023/02/IUCN-Thumbnail-01-768x402.png""}}" 55521,"Engel's Law: Richer people spend more money on food, but it makes up a smaller share of their income",engels-law-food-spending,post,publish,"Richer people tend to spend more money on food.
We see this relationship when we look at data on food expenditure from across the world.
In the chart you see food expenditure plotted against total consumer expenditure per person. Total consumer expenditure is any personal expenditure on goods and services.
On the y-axis, we have the average amount of money spent on food per person, in countries across the world.{ref}This data comes from the United States Department of Agriculture (USDA) Economic Research Service, which estimates this for countries where available market data is sufficient.{/ref} This measures money spent on food consumed at home – from stores and supermarkets. The source does not include food eaten out-of-home, from restaurants and cafes (however, that would be useful data to have). Tobacco and alcoholic beverages are also excluded.
On the x-axis, we have a measure of prosperity, total expenditure. Expenditure tells us how much disposable income people have to spend on goods and services such as food, housing, education, health, and leisure activities. It’s important to ensure that we can afford all of the essentials that give us a high and comfortable standard of living.
Countries which are poor – and where therefore people’s expenditure is very low – can be found, richer countries are further towards the right. This is shown on a log axis by default, but you can change it to a linear scale on the interactive chart.
Both metrics are measured in US dollars per person per year.
What we see is a strong positive relationship: where expenditures are high people tend to spend more money on food
At expenditures below $5,000, the average person tends to spend less than $1,000 on food. At expenditures above $10,000, it's common for the average to be $2,000 or more per person per year.
The amount of money spent on food might increase in absolute terms — as we saw in the previous chart — it falls as a share of people’s expenditure. This is what the following chart shows.
In the chart, we’ve plotted food spending against total expenditure again, but this time food spending is measured as a percentage of someone’s total expenditure.{ref}Again, this data comes from the United States Department of Agriculture (USDA) Economic Research Service, which estimates this for countries where available market data is sufficient.{/ref}
In countries with expenditures below $10,000, it’s common for people to spend a quarter or more on food.
As total spending rises, a smaller and smaller share goes toward food. At high expenditures, it’s common for people to spend 10% or less. This is despite the fact that they spend more on food in absolute terms.
Both of these points are positive signs for consumers. Spending more on food in dollar terms often means that people are eating more diverse diets. At low incomes, people tend to rely on cheaper staple crops such as cereals and tubers for most of their calories. As people get richer they can afford other foods such as fruits, vegetables, legumes, meats, and dairy products which have a wider range of micronutrients, protein, and fats.
Spending less of our money on food means we have more disposable income for other essentials such as education, healthcare, and housing.
That gives us two important insights into food spending.
First, richer people tend to spend more money on food in absolute terms. Second, food tends to account for a smaller percentage of their total expenditure.
The chart illustrates this relationship: in absolute terms spending on food increases with income (shown in purple), in percentage terms it decreases (shown in green).
Combine them and we get a relationship called ‘Engel’s Law’: as household income (and expenditure) increases, the percentage that is spent on food decreases, but the absolute amount spent on food increases.
Research has shown that this relationship holds true within countries: studies looking at household expenditures in South Africa and China, for example, found the same pattern.{ref}Mulamba, K. C. (2022). Relationship between households’ share of food expenditure and income across South African districts: a multilevel regression analysis. Humanities and Social Sciences Communications, 9(1), 1-11.
Chen, M. (2022). Engel’s law in China: Some new evidence. Review of Development Economics.{/ref}
In a related article, I look at what happens at the bottom of the global income distribution: it’s estimated that around three billion people can’t afford a healthy, diverse diet even when they spend most (or all) of their income on food. We want them to be able to spend more on food so they can have a diversified diet, and we don’t want them to have to spend all of their money to achieve this.
","{""id"": ""wp-55521"", ""slug"": ""engels-law-food-spending"", ""content"": {""toc"": [], ""body"": [{""left"": [{""type"": ""text"", ""value"": [{""text"": ""Richer people tend to spend more money on food."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We see this relationship when we look at data on food expenditure from across the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart you see food expenditure plotted against total consumer expenditure per person. Total consumer expenditure is any personal expenditure on goods and services."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""On the y-axis, we have the average amount of money spent on food per person, in countries across the world.{ref}This data comes from the United States Department of Agriculture (USDA) Economic Research Service, which estimates this for countries where available market data is sufficient.{/ref} This measures money spent on food consumed at home – from stores and supermarkets. The source "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""does not "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""include food eaten out-of-home, from restaurants and cafes (however, that would be useful data to have). Tobacco and alcoholic beverages are also excluded."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""On the x-axis, we have a measure of prosperity, total expenditure. Expenditure tells us how much disposable income people have to spend on goods and services such as food, housing, education, health, and leisure activities. It’s important to ensure that we can afford all of the essentials that give us a high and comfortable standard of living."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Countries which are poor – and where therefore people’s expenditure is very low – can be found, richer countries are further towards the right. This is shown on a log axis by default, but you can change it to a linear scale on the interactive chart."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Both metrics are measured in US dollars per person per year."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""What we see is a strong positive relationship: where expenditures are high people tend to spend more money on food"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""At expenditures below $5,000, the average person tends to spend less than $1,000 on food. At expenditures above $10,000, it's common for the average to be $2,000 or more per person per year."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/food-expenditure-vs-gdp"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""For richer people, food makes up a smaller share of their expenditure"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The amount of money spent on food might increase in absolute terms — as we saw in the previous chart — it falls as a share of people’s expenditure. This is what the following chart shows."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart, we’ve plotted food spending against total expenditure again, but this time food spending is measured as a percentage of someone’s total expenditure.{ref}Again, this data comes from the United States Department of Agriculture (USDA) Economic Research Service, which estimates this for countries where available market data is sufficient.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In countries with expenditures below $10,000, it’s common for people to spend a quarter or more on food."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As total spending rises, a smaller and smaller share goes toward food. At high expenditures, it’s common for people to spend 10% or less. This is despite the fact that they spend "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""more"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" on food in absolute terms."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Both of these points are positive signs for consumers. Spending more on food in dollar terms often means that people are eating more diverse diets. At low incomes, people tend to "", ""spanType"": ""span-simple-text""}, {""url"": ""https://owid.cloud/grapher/share-of-energy-from-cereals-roots-and-tubers-vs-gdp-per-capita"", ""children"": [{""text"": ""rely on cheaper staple crops"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" such as cereals and tubers for most of their calories. As people get richer they can afford other foods such as fruits, vegetables, legumes, meats, and dairy products which have a wider range of micronutrients, protein, and fats."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Spending less of our money on food means we have more disposable income for other essentials such as education, healthcare, and housing."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/food-expenditure-share-gdp"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Engel’s Law: how food spending changes with income"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""That gives us two important insights into food spending."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""First, richer people tend to spend more money on food in absolute terms. Second, food tends to account for a smaller percentage of their total expenditure."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The chart illustrates this relationship: in absolute terms spending on food increases with income (shown in purple), in percentage terms it decreases (shown in green)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Combine them and we get a relationship called ‘"", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Engel%27s_law"", ""children"": [{""text"": ""Engel’s Law"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""’: as household income (and expenditure) increases, the percentage that is spent on food decreases, but the absolute amount spent on food increases. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Research has shown that this relationship holds true "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""within"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" countries: studies looking at household expenditures in South Africa and China, for example, found the same pattern.{ref}Mulamba, K. C. (2022). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/s41599-022-01454-4"", ""children"": [{""text"": ""Relationship between households’ share of food expenditure and income across South African districts: a multilevel regression analysis"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Humanities and Social Sciences Communications, 9(1), 1-11."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Chen, M. (2022). Engel’s law in China: Some new evidence. Review of Development Economics.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Engelsl-Law-Schematic.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Keep reading at "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Our World in Data..."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""type"": ""heading"", ""level"": 4, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In a "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/diet-affordability"", ""children"": [{""children"": [{""text"": ""related article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""spanType"": ""span-link""}, {""text"": "", I look at what happens at the bottom of the global income distribution: it’s estimated that around three billion people can’t afford a healthy, diverse diet even when they spend most (or all) of their income on food. We want them to be able to spend more on food so they can have a diversified diet, and we don’t want them to have to spend all of their money to achieve this."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/food-prices"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/diet-affordability"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Engel's Law: Richer people spend more money on food, but it makes up a smaller share of their income"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""How does spending on food change as incomes rise?"", ""dateline"": ""January 19, 2023"", ""subtitle"": ""How does spending on food change as incomes rise?"", ""sidebar-toc"": false, ""featured-image"": ""Engels-Law-Featured-Image-01.png""}, ""createdAt"": ""2023-01-19T11:43:15.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T15:14:46.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-01-19T11:43:15.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}], ""numBlocks"": 9, ""numErrors"": 3, ""wpTagCounts"": {""html"": 2, ""image"": 1, ""column"": 6, ""columns"": 3, ""heading"": 3, ""paragraph"": 22, ""separator"": 1, ""owid/prominent-link"": 2}, ""htmlTagCounts"": {""p"": 22, ""h3"": 2, ""h5"": 1, ""hr"": 1, ""div"": 9, ""figure"": 1, ""iframe"": 2}}",2023-01-19 11:43:15,2024-02-16 14:22:54,1cA7lv3ROTMuTWV48zIkIx21_C3MD-qbH7UGntiCDK6g,"[""Hannah Ritchie""]",How does spending on food change as incomes rise?,2023-01-19 11:43:15,2023-07-10 15:14:46,https://ourworldindata.org/wp-content/uploads/2023/01/Engels-Law-Featured-Image-01.png,{},"Richer people tend to spend more money on food. We see this relationship when we look at data on food expenditure from across the world. In the chart you see food expenditure plotted against total consumer expenditure per person. Total consumer expenditure is any personal expenditure on goods and services. On the y-axis, we have the average amount of money spent on food per person, in countries across the world.{ref}This data comes from the United States Department of Agriculture (USDA) Economic Research Service, which estimates this for countries where available market data is sufficient.{/ref} This measures money spent on food consumed at home – from stores and supermarkets. The source _does not _include food eaten out-of-home, from restaurants and cafes (however, that would be useful data to have). Tobacco and alcoholic beverages are also excluded. On the x-axis, we have a measure of prosperity, total expenditure. Expenditure tells us how much disposable income people have to spend on goods and services such as food, housing, education, health, and leisure activities. It’s important to ensure that we can afford all of the essentials that give us a high and comfortable standard of living. Countries which are poor – and where therefore people’s expenditure is very low – can be found, richer countries are further towards the right. This is shown on a log axis by default, but you can change it to a linear scale on the interactive chart. Both metrics are measured in US dollars per person per year. What we see is a strong positive relationship: where expenditures are high people tend to spend more money on food At expenditures below $5,000, the average person tends to spend less than $1,000 on food. At expenditures above $10,000, it's common for the average to be $2,000 or more per person per year.Richer people tend to spend more money on food.
\n\n\n\nWe see this relationship when we look at data on food expenditure from across the world.
\n\n\n\nIn the chart you see food expenditure plotted against total consumer expenditure per person. Total consumer expenditure is any personal expenditure on goods and services.
\n\n\n\nOn the y-axis, we have the average amount of money spent on food per person, in countries across the world.{ref}This data comes from the United States Department of Agriculture (USDA) Economic Research Service, which estimates this for countries where available market data is sufficient.{/ref} This measures money spent on food consumed at home – from stores and supermarkets. The source does not include food eaten out-of-home, from restaurants and cafes (however, that would be useful data to have). Tobacco and alcoholic beverages are also excluded.
\n\n\n\nOn the x-axis, we have a measure of prosperity, total expenditure. Expenditure tells us how much disposable income people have to spend on goods and services such as food, housing, education, health, and leisure activities. It’s important to ensure that we can afford all of the essentials that give us a high and comfortable standard of living.
\n\n\n\nCountries which are poor – and where therefore people’s expenditure is very low – can be found, richer countries are further towards the right. This is shown on a log axis by default, but you can change it to a linear scale on the interactive chart.
\n\n\n\nBoth metrics are measured in US dollars per person per year.
\n\n\n\nWhat we see is a strong positive relationship: where expenditures are high people tend to spend more money on food
\n\n\n\nAt expenditures below $5,000, the average person tends to spend less than $1,000 on food. At expenditures above $10,000, it’s common for the average to be $2,000 or more per person per year.
\nThe amount of money spent on food might increase in absolute terms — as we saw in the previous chart — it falls as a share of people’s expenditure. This is what the following chart shows.
\n\n\n\nIn the chart, we’ve plotted food spending against total expenditure again, but this time food spending is measured as a percentage of someone’s total expenditure.{ref}Again, this data comes from the United States Department of Agriculture (USDA) Economic Research Service, which estimates this for countries where available market data is sufficient.{/ref}
\n\n\n\nIn countries with expenditures below $10,000, it’s common for people to spend a quarter or more on food.
\n\n\n\nAs total spending rises, a smaller and smaller share goes toward food. At high expenditures, it’s common for people to spend 10% or less. This is despite the fact that they spend more on food in absolute terms.
\n\n\n\nBoth of these points are positive signs for consumers. Spending more on food in dollar terms often means that people are eating more diverse diets. At low incomes, people tend to rely on cheaper staple crops such as cereals and tubers for most of their calories. As people get richer they can afford other foods such as fruits, vegetables, legumes, meats, and dairy products which have a wider range of micronutrients, protein, and fats.
\n\n\n\nSpending less of our money on food means we have more disposable income for other essentials such as education, healthcare, and housing.
\nThat gives us two important insights into food spending.
\n\n\n\nFirst, richer people tend to spend more money on food in absolute terms. Second, food tends to account for a smaller percentage of their total expenditure.
\n\n\n\nThe chart illustrates this relationship: in absolute terms spending on food increases with income (shown in purple), in percentage terms it decreases (shown in green).
\n\n\n\nCombine them and we get a relationship called ‘Engel’s Law’: as household income (and expenditure) increases, the percentage that is spent on food decreases, but the absolute amount spent on food increases.
\n\n\n\nResearch has shown that this relationship holds true within countries: studies looking at household expenditures in South Africa and China, for example, found the same pattern.{ref}Mulamba, K. C. (2022). Relationship between households’ share of food expenditure and income across South African districts: a multilevel regression analysis. Humanities and Social Sciences Communications, 9(1), 1-11.
\n\n\n\nChen, M. (2022). Engel’s law in China: Some new evidence. Review of Development Economics.{/ref}
\n![\""\""](\""https://owid.cloud/app/uploads/2023/01/Engelsl-Law-Schematic.png\"")
In a related article, I look at what happens at the bottom of the global income distribution: it’s estimated that around three billion people can’t afford a healthy, diverse diet even when they spend most (or all) of their income on food. We want them to be able to spend more on food so they can have a diversified diet, and we don’t want them to have to spend all of their money to achieve this.
\n\n\nMeasuring the state of human rights across the world helps us understand the extent to which people have fundamental personal and civil rights and freedoms.
But it can be challenging to measure how well-protected these human rights are.. People do not always agree on what rights are fundamental. These rights — such as whether people are free to voice their opinions — are difficult to define and assess. The judgement of experts is to some degree subjective. They may disagree about a specific characteristic or how several characteristics can be reduced into a single measure.
How do researchers address these challenges and measure how much people’s human rights are protected?
In some of our work on democracy, we rely on data published by the Varieties of Democracy (V-Dem) project.{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Nazifa Alizada, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Sandra Grahn, Allen Hicken, Garry Hindle, Nina Ilchenko, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Ryden, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2022. VDem [Country–Year/Country–Date] Dataset v12. Varieties of Democracy (V-Dem) Project.{/ref}
The project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is run by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators.
V-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation.
Our team at Our World in Data uses V-Dem’s Civil Liberties Index to measure human rights.
V-Dem characterizes civil liberties as three types of freedoms: physical integrity rights, private civil liberties, and political civil liberties. More specifically, this means:
V-Dem’s Civil Liberties Index scores each country on a spectrum, with some countries protecting human rights more than others.
The spectrum ranges from 0 (‘fewest rights’) to 1 (‘most rights’).
As of version 12 of the dataset, V-Dem covers 202 countries, going back in time as far as 1789. Many countries have been covered since 1900, including before they became independent from their colonial powers.
To actually measure what it wants to capture, V-Dem assesses the characteristics of human rights mostly through evaluations by experts.{ref} For more details, see: Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
These anonymous experts are primarily academics and members of the media and civil society. They are also often nationals or residents of the country they assess, and therefore know its political system well and can evaluate aspects that are difficult to observe.
V-Dem’s own team of researchers supplements the expert evaluations. They code some easier-to-observe rules and laws of the political system, such as whether the legislature has a lower and upper house.
V-Dem uses several experts per country, year, and topic, to make its assessments less subjective. In total, around 3,500 country-experts fill surveys for V-Dem every year.
While there are fewer experts for small countries and for the time before 1900, they rely typically on 25 experts per country and 5 experts per topic.
V-Dem also works to make their coders’ assessments comparable across countries and time.
The surveys ask the experts to answer very specific questions on completely explained scales about sub-characteristics of human rights — such as whether women can freely move in their own country — instead of making them rely on their broad impressions.
The surveys are available in English, Arabic, French, Portuguese, Russian, and Spanish to reduce misunderstandings.
Experts further evaluate hypothetical countries, many coded several countries, and they denote their own uncertainty and personal demographic information.
V-Dem then uses this information to investigate expert biases, which they have found to be limited: they only find that experts from a country tend to be stricter in their assessments. {ref}“We have run extensive tests on how well such individual-level factors predict country-ratings but have found that the only factor consistently associated with country-ratings is country of origin (with “domestic” experts being harsher in their judgments).”
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
V-Dem uses a statistical model to address any remaining differences between coders.{ref}Specifically, it uses a Bayesian Item-Response Theory estimation strategy.
Marquardt, Kyle, and Daniel Pemstein. 2018. IRT Models for Expert-Coded Panel Data. Political Analysis 26(4): 431-456.{/ref}
The model combines the experts’ ratings of actual countries and hypothetical countries, as well as the experts’ stated uncertainties and personal demographics to produce best, upper-, and lower-bound estimates of many characteristics.{ref}Expressed precisely, V-Dem’s measurement model produces a probability distribution over the country-year scores. The best estimate is the distribution’s median, while the upper and lower bound estimates demarcate the interval in which the model places 68 percent of the probability mass.{/ref}
V-Dem provides these different estimates for all of its main and supplementary indices, including the Civil Liberties Index and the subindices for physical integrity rights, private civil liberties, and political civil liberties.
With the different estimates, V-Dem explicitly acknowledges that its coders can be uncertain or make errors in their measurement.
The overall Civil Liberties Index score is the result of averaging the three subindices.
V-Dem releases its data publicly, and makes it straightforward to download and use.
It publishes the overall scores, the underlying subindices, and several hundred specific questions by country-year, country-date, and coder.
V-Dem also releases detailed descriptions of the questions and coding procedures that guide the experts and researchers.
In our work, we expand the years covered by V-Dem further.
To expand the time coverage of today’s countries and include more of the period when they were still non-sovereign territories, we identified the historical entity they were a part of and used that regime’s data whenever available.{ref}For example, V-Dem only provides data since Bangladesh’s independence in 1971. There is, however, data for Pakistan and the colony of India, both of which the current territory of Bangladesh was a part. We, therefore, use the data of Pakistan for Bangladesh from 1947 to 1970, and the data of India from 1789 to 1946. We did so for all countries with a past or current population of more than one million.{/ref}
We also calculated regional and global averages of the Civil Liberties Index and its sub-indices, weighted and unweighted by population.
Our code and data are available on GitHub and record our revisions in detail.
V-Dem releases a new version of the data each year in March.
We at Our World in Data aim to update our own data within a few weeks of the release.
There are shortcomings in the way that V-Dem’s Civil Liberties Index characterizes and measures human rights.{ref}This and the following section draw on an article summarizing and reviewing some of the leading human rights datasets:
Cope, Kevin, Charles Crabtree, and Christopher Fariss. 2020. Patterns of disagreement in indicators of state repression. Political Science Research and Methods 8(1): 178-187.{/ref}
The index focuses on human rights as civil liberties and does not account for other characterizations, such as rights to food, health, or education. This means that countries with good health and education outcomes but restricted civil liberties, such as Iran and Singapore in recent years, still receive relatively low scores.
The index also does not tell us anything about how human rights differ across parts of the population, such as between men and women, or between different ethnic groups.{ref}Though some of the index’s specific indicators distinguish between rights for men and women.{/ref}
V-Dem also does not cover some countries with very small populations.
Furthermore, the index is more difficult to interpret than other measures. The Civil Liberties Index does not identify whether a country grants or protects human rights or not, but only allows us to say whether a country is protecting human rights by comparing it to the range of the index, to other countries, or to the same country at another point in time. And when doing so, it is still difficult to say how large these differences are.{ref}This can be made easier by comparing how a score relates to the index’s overall distribution or its distribution for a specific year.{/ref}
The assessment of the Civil Liberties Index remains to some extent subjective. Its index is built on difficult evaluations by experts and relies less on easier-to-observe characteristics, such as whether forming a civil society organization independent of the state is legal, or the number of allegations made in human rights reports against the government.
Finally, the index’s aggregation remains to some extent arbitrary. V-Dem does not say why these specific subindices were chosen, and why the subindices are given the same weight.
Despite these shortcomings, the index tells us a lot about how protected human rights are around the world, in the past and today.
Its characterization of human rights as people enjoying physical integrity, as well as private and political civil liberties, is commonly recognized to be at the core of human rights.
Because it treats human rights as a spectrum, the index is able to capture both big and small differences in their protection across countries, and to record small changes within countries over time. This allows us to observe whether one country protects human rights more than another, or whether a country has protected human rights more or less over time.
The index also covers many countries and years. With the exception of microstates, it covers all countries in the world. Many countries are covered since 1900 — even while they were colonized by another country — and some of them as far back as 1789.
Finally, V-Dem takes many steps to make its assessments valid, precise, comparable across countries and time, and transparent. It relies on many country and subject experts answering detailed surveys to measure aspects of political systems that are often difficult to observe and acknowledges the remaining uncertainty in their assessments.
Whether V-Dem’s Civil Liberties Index is a useful measure of human rights will depend on the questions we want to answer.
The index will not give us a satisfying answer if we are interested in an understanding of human rights as also including rights to health or education; in differences in the protection of human rights by gender and ethnicity; and if we are also interested in the political systems of microstates.
In these cases, we may have to rely on other measures.
But if we value a sophisticated measure based on the knowledge of many country experts and are interested in big and small differences in civil liberties, within and across countries, and far into the past, we can learn a lot from this data.
It is for these latter purposes we use the measure in some of our reporting on human rights.
I thank Edouard Mathieu for his very helpful comments and ideas about how to improve this article.
","{""id"": ""wp-55153"", ""slug"": ""vdem-human-rights-data"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Measuring the state of human rights across the world helps us understand the extent to which people have fundamental personal and civil rights and freedoms."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But it can be challenging to measure how well-protected these human rights are.. People do not always agree on what rights are fundamental. These rights — such as whether people are free to voice their opinions — are difficult to define and assess. The judgement of experts is to some degree subjective. They may disagree about a specific characteristic or how several characteristics can be reduced into a single measure."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How do researchers address these challenges and measure how much people’s human rights are protected?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What are the Varieties of Democracy project?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In some of our work on democracy, we rely on data published by the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.v-dem.net/vdemds.html"", ""children"": [{""text"": ""Varieties of Democracy (V-Dem) project"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Nazifa Alizada, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Sandra Grahn, Allen Hicken, Garry Hindle, Nina Ilchenko, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Ryden, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2022."", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Carl+Henrik+Knutsen%2C+Staffan+I.+Lindberg%2C+Jan+Teorell%2C+Nazifa+Alizada%2C+David+Altman%2C+Michael+Bernhard%2C+Agnes+Cornell%2C+M.+Steven+Fish%2C+Lisa+Gastaldi%2C+Haakon+Gjerl%C3%B8w%2C+Adam+Glynn%2C+Sandra+Grahn%2C+Allen+Hicken%2C+Garry+Hindle%2C+Nina+Ilchenko%2C+Katrin+Kinzelbach%2C+Joshua+Krusell%2C+Kyle+L.+Marquardt%2C+Kelly+McMann%2C+Valeriya+Mechkova%2C+Juraj+Medzihorsky%2C+Pamela+Paxton%2C+Daniel+Pemstein%2C+Josefine+Pernes%2C+Oskar+Ryd%E2%80%80en%2C+Johannes+von+R%C3%B6mer%2C+Brigitte+Seim%2C+Rachel+Sigman%2C+Svend-Erik+Skaaning%2C+Jeffrey+Staton%2C+Aksel+Sundstr%C3%B6m%2C+Eitan+Tzelgov%2C+Yi-ting+Wang%2C+Tore+Wig%2C+Steven+Wilson+and+Daniel+Ziblatt.+2022.+VDem+%5BCountry%E2%80%93Year%2FCountry%E2%80%93Date%5D+Dataset+v12.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG="", ""children"": [{""text"": "" VDem [Country–Year/Country–Date] Dataset v12"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Varieties of Democracy (V-Dem) Project.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.v-dem.net/about/v-dem-project/"", ""children"": [{""text"": ""run"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem characterize human rights?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Our team at Our World in Data uses V-Dem’s Civil Liberties Index to measure human rights."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem characterizes civil liberties as three types of freedoms: physical integrity rights, private civil liberties, and political civil liberties. More specifically, this means:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Physical integrity rights"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": people are free and protected from government torture and political killings"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Private civil liberties"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": people are free from forced labor, have property rights, and enjoy freedoms of movement (move unrestricted within, to, and from the country) and religion (choose and practice their faith)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Political civil liberties"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": people enjoy freedoms of association (parties and civil society organizations can form and operate freely) and expression (they can voice their views, and the media can present different political perspectives)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How are human rights scored?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem’s Civil Liberties Index scores each country on a spectrum, with some countries protecting human rights more than others."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The spectrum ranges from 0 (‘fewest rights’) to 1 (‘most rights’)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/human-rights-vdem"", ""type"": ""chart"", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-human-rights-vdem"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What years and countries are covered?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As of version 12 of the dataset, V-Dem covers 202 countries, going back in time as far as 1789. Many countries have been covered since 1900, including before they became independent from their colonial powers."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How is democracy measured?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem work to make its assessments valid?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To actually measure what it wants to capture, V-Dem assesses the characteristics of human rights mostly through evaluations by experts.{ref} For more details, see: Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These anonymous experts are primarily academics and members of the media and civil society. They are also often nationals or residents of the country they assess, and therefore know its political system well and can evaluate aspects that are difficult to observe."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem’s own team of researchers supplements the expert evaluations. They code some easier-to-observe rules and laws of the political system, such as whether the legislature has a lower and upper house."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem work to make its assessments precise and reliable?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem uses several experts per country, year, and topic, to make its assessments less subjective. In total, around 3,500 country-experts fill surveys for V-Dem every year."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""While there are fewer experts for small countries and for the time before 1900, they rely typically on 25 experts per country and 5 experts per topic."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem work to make its assessments comparable?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem also works to make their coders’ assessments comparable across countries and time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The surveys ask the experts to answer very specific questions on completely explained scales about sub-characteristics of human rights — such as whether women can freely move in their own country — instead of making them rely on their broad impressions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The surveys are available in English, Arabic, French, Portuguese, Russian, and Spanish to reduce misunderstandings."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Experts further evaluate hypothetical countries, many coded several countries, and they denote their own uncertainty and personal demographic information."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem then uses this information to investigate expert biases, which they have found to be limited: they only find that experts from a country tend to be stricter in their assessments. {ref}“We have run extensive tests on how well such individual-level factors predict country-ratings but have found that the only factor consistently associated with country-ratings is country of origin (with “domestic” experts being harsher in their judgments).”"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How are the remaining differences in the data dealt with?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem uses a statistical model to address any remaining differences between coders.{ref}Specifically, it uses a Bayesian Item-Response Theory estimation strategy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Marquardt, Kyle, and Daniel Pemstein. 2018. IRT Models for Expert-Coded Panel Data. Political Analysis 26(4): 431-456.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The model combines the experts’ ratings of actual countries and hypothetical countries, as well as the experts’ stated uncertainties and personal demographics to produce best, upper-, and lower-bound estimates of many characteristics.{ref}Expressed precisely, V-Dem’s measurement model produces a probability distribution over the country-year scores. The best estimate is the distribution’s median, while the upper and lower bound estimates demarcate the interval in which the model places 68 percent of the probability mass.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem provides these different estimates for all of its main and supplementary indices, including the Civil Liberties Index and the subindices for physical integrity rights, private civil liberties, and political civil liberties."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""With the different estimates, V-Dem explicitly acknowledges that its coders can be uncertain or make errors in their measurement."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The overall Civil Liberties Index score is the result of averaging the three subindices."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How is the data made accessible and transparent?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem releases "", ""spanType"": ""span-simple-text""}, {""url"": ""https://v-dem.net/data/the-v-dem-dataset/"", ""children"": [{""text"": ""its data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" publicly, and makes it straightforward to download and use."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It publishes the overall scores, the underlying subindices, and several hundred specific questions by country-year, country-date, and coder."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem also releases detailed descriptions of the questions and coding procedures that guide the experts and researchers."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How do we change the data?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In our work, we expand the years covered by V-Dem further."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To expand the time coverage of today’s countries and include more of the period when they were still non-sovereign territories, we identified the historical entity they were a part of and used that regime’s data whenever available.{ref}For example, V-Dem only provides data since Bangladesh’s independence in 1971. There is, however, data for Pakistan and the colony of India, both of which the current territory of Bangladesh was a part. We, therefore, use the data of Pakistan for Bangladesh from 1947 to 1970, and the data of India from 1789 to 1946. We did so for all countries with a past or current population of more than one million.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We also calculated regional and global averages of the Civil Liberties Index and its sub-indices, weighted and unweighted by population."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Our code and data are available "", ""spanType"": ""span-simple-text""}, {""url"": ""https://github.com/owid/notebooks/tree/main/BastianHerre/democracy"", ""children"": [{""text"": ""on GitHub"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and record our revisions in detail."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How often and when is the data updated?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem releases a new version of the data each year in March."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We at Our World in Data aim to update our own data within a few weeks of the release."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What are the data’s shortcomings?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are shortcomings in the way that V-Dem’s Civil Liberties Index characterizes and measures human rights.{ref}This and the following section draw on an article summarizing and reviewing some of the leading human rights datasets:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Cope, Kevin, Charles Crabtree, and Christopher Fariss. 2020. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.cambridge.org/core/journals/political-science-research-and-methods/article/abs/patterns-of-disagreement-in-indicators-of-state-repression/FBCA92BDEFD630ED364DEFE73600172F"", ""children"": [{""text"": ""Patterns of disagreement in indicators of state repression"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Political Science Research and Methods 8(1): 178-187.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The index focuses on human rights as civil liberties and does not account for other characterizations, such as rights to food, health, or education. This means that countries with good health and education outcomes but restricted civil liberties, such as Iran and Singapore in recent years, still receive relatively low scores."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The index also does not tell us anything about how human rights differ across parts of the population, such as between men and women, or between different ethnic groups.{ref}Though some of the index’s specific indicators distinguish between rights for men and women.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem also does not cover some countries with very small populations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Furthermore, the index is more difficult to interpret than other measures. The Civil Liberties Index does not identify whether a country grants or protects human rights or not, but only allows us to say whether a country is protecting human rights by comparing it to the range of the index, to other countries, or to the same country at another point in time. And when doing so, it is still difficult to say how large these differences are.{ref}This can be made easier by comparing how a score relates to the index’s overall distribution or its distribution for a specific year.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The assessment of the Civil Liberties Index remains to some extent subjective. Its index is built on difficult evaluations by experts and relies less on easier-to-observe characteristics, such as whether forming a civil society organization independent of the state is legal, or the number of allegations made in human rights reports against the government."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Finally, the index’s aggregation remains to some extent arbitrary. V-Dem does not say why these specific subindices were chosen, and why the subindices are given the same weight."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What are the data’s strengths?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Despite these shortcomings, the index tells us a lot about how protected human rights are around the world, in the past and today."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Its characterization of human rights as people enjoying physical integrity, as well as private and political civil liberties, is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.cambridge.org/core/journals/american-political-science-review/article/human-rights-are-increasingly-plural-learning-the-changing-taxonomy-of-human-rights-from-largescale-text-reveals-information-effects/F202F327EA8F4CF52D2E65EB48D409D3"", ""children"": [{""text"": ""commonly recognized"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to be at the core of human rights."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Because it treats human rights as a spectrum, the index is able to capture both big and small differences in their protection across countries, and to record small changes within countries over time. This allows us to observe whether one country protects human rights more than another, or whether a country has protected human rights more or less over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The index also covers many countries and years. With the exception of microstates, it covers all countries in the world. Many countries are covered since 1900 — even while they were colonized by another country — and some of them as far back as 1789."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Finally, V-Dem takes many steps to make its assessments valid, precise, comparable across countries and time, and transparent. It relies on many country and subject experts answering detailed surveys to measure aspects of political systems that are often difficult to observe and acknowledges the remaining uncertainty in their assessments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What is our summary assessment?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Whether V-Dem’s Civil Liberties Index is a useful measure of human rights will depend on the questions we want to answer."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The index will not give us a satisfying answer if we are interested in an understanding of human rights as also including rights to health or education; in differences in the protection of human rights by gender and ethnicity; and if we are also interested in the political systems of microstates."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In these cases, we may have to rely on "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/physical-integrity-rights-fkr"", ""children"": [{""text"": ""other measures"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But if we value a sophisticated measure based on the knowledge of many country experts and are interested in big and small differences in civil liberties, within and across countries, and far into the past, we can learn a lot from this data."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It is for these latter purposes we use the measure in some of our reporting on human rights."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""Keep reading on "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""children"": [{""children"": [{""text"": ""Our World in Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/human-rights"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""Acknowledgments"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I thank Edouard Mathieu for his very helpful comments and ideas about how to improve this article."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""The 'Varieties of Democracy' data: how do researchers measure human rights?"", ""authors"": [""Bastian Herre""], ""excerpt"": ""There are several ways to measure human rights. Here is how the Varieties of Democracy project does it, one of the leading sources of global human rights data."", ""dateline"": ""December 16, 2022"", ""subtitle"": ""There are several ways to measure human rights. Here is how the Varieties of Democracy project does it, one of the leading sources of global human rights data."", ""sidebar-toc"": false, ""featured-image"": ""distribution-human-rights-vdem.png""}, ""createdAt"": ""2022-12-16T17:12:36.000Z"", ""published"": false, ""updatedAt"": ""2023-06-08T09:57:14.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-12-16T17:33:07.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag meta""}], ""numBlocks"": 77, ""numErrors"": 2, ""wpTagCounts"": {""html"": 2, ""list"": 1, ""heading"": 17, ""paragraph"": 56, ""owid/prominent-link"": 1}, ""htmlTagCounts"": {""p"": 56, ""h4"": 13, ""h5"": 4, ""ul"": 1, ""meta"": 1, ""iframe"": 2}}",2022-12-16 17:33:07,2024-02-16 14:22:54,1Hy_mYebgaQCiZLU2AM3aRUBMebz6zGrj7uFLjdnDJkU,"[""Bastian Herre""]","There are several ways to measure human rights. Here is how the Varieties of Democracy project does it, one of the leading sources of global human rights data.",2022-12-16 17:12:36,2023-06-08 09:57:14,https://ourworldindata.org/wp-content/uploads/2022/12/distribution-human-rights-vdem.png,{},"Measuring the state of human rights across the world helps us understand the extent to which people have fundamental personal and civil rights and freedoms. But it can be challenging to measure how well-protected these human rights are.. People do not always agree on what rights are fundamental. These rights — such as whether people are free to voice their opinions — are difficult to define and assess. The judgement of experts is to some degree subjective. They may disagree about a specific characteristic or how several characteristics can be reduced into a single measure. How do researchers address these challenges and measure how much people’s human rights are protected? ## **What are the Varieties of Democracy project?** In some of our work on democracy, we rely on data published by the [Varieties of Democracy (V-Dem) project](https://www.v-dem.net/vdemds.html).{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Nazifa Alizada, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Sandra Grahn, Allen Hicken, Garry Hindle, Nina Ilchenko, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Ryden, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2022.[ VDem [Country–Year/Country–Date] Dataset v12](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Carl+Henrik+Knutsen%2C+Staffan+I.+Lindberg%2C+Jan+Teorell%2C+Nazifa+Alizada%2C+David+Altman%2C+Michael+Bernhard%2C+Agnes+Cornell%2C+M.+Steven+Fish%2C+Lisa+Gastaldi%2C+Haakon+Gjerl%C3%B8w%2C+Adam+Glynn%2C+Sandra+Grahn%2C+Allen+Hicken%2C+Garry+Hindle%2C+Nina+Ilchenko%2C+Katrin+Kinzelbach%2C+Joshua+Krusell%2C+Kyle+L.+Marquardt%2C+Kelly+McMann%2C+Valeriya+Mechkova%2C+Juraj+Medzihorsky%2C+Pamela+Paxton%2C+Daniel+Pemstein%2C+Josefine+Pernes%2C+Oskar+Ryd%E2%80%80en%2C+Johannes+von+R%C3%B6mer%2C+Brigitte+Seim%2C+Rachel+Sigman%2C+Svend-Erik+Skaaning%2C+Jeffrey+Staton%2C+Aksel+Sundstr%C3%B6m%2C+Eitan+Tzelgov%2C+Yi-ting+Wang%2C+Tore+Wig%2C+Steven+Wilson+and+Daniel+Ziblatt.+2022.+VDem+%5BCountry%E2%80%93Year%2FCountry%E2%80%93Date%5D+Dataset+v12.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG=). Varieties of Democracy (V-Dem) Project.{/ref} The project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is [run](https://www.v-dem.net/about/v-dem-project/) by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators. V-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation. ## **How does V-Dem characterize human rights?** Our team at Our World in Data uses V-Dem’s Civil Liberties Index to measure human rights. V-Dem characterizes civil liberties as three types of freedoms: physical integrity rights, private civil liberties, and political civil liberties. More specifically, this means: * **Physical integrity rights**: people are free and protected from government torture and political killings * **Private civil liberties**: people are free from forced labor, have property rights, and enjoy freedoms of movement (move unrestricted within, to, and from the country) and religion (choose and practice their faith) * **Political civil liberties**: people enjoy freedoms of association (parties and civil society organizations can form and operate freely) and expression (they can voice their views, and the media can present different political perspectives) ## **How are human rights scored?** V-Dem’s Civil Liberties Index scores each country on a spectrum, with some countries protecting human rights more than others. The spectrum ranges from 0 (‘fewest rights’) to 1 (‘most rights’).Measuring the state of human rights across the world helps us understand the extent to which people have fundamental personal and civil rights and freedoms.
\n\n\n\nBut it can be challenging to measure how well-protected these human rights are.. People do not always agree on what rights are fundamental. These rights — such as whether people are free to voice their opinions — are difficult to define and assess. The judgement of experts is to some degree subjective. They may disagree about a specific characteristic or how several characteristics can be reduced into a single measure.
\n\n\n\nHow do researchers address these challenges and measure how much people’s human rights are protected?
\n\n\n\nIn some of our work on democracy, we rely on data published by the Varieties of Democracy (V-Dem) project.{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Nazifa Alizada, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Sandra Grahn, Allen Hicken, Garry Hindle, Nina Ilchenko, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Ryden, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2022. VDem [Country–Year/Country–Date] Dataset v12. Varieties of Democracy (V-Dem) Project.{/ref}
\n\n\n\nThe project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is run by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators.
\n\n\n\nV-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation.
\n\n\n\nOur team at Our World in Data uses V-Dem’s Civil Liberties Index to measure human rights.
\n\n\n\nV-Dem characterizes civil liberties as three types of freedoms: physical integrity rights, private civil liberties, and political civil liberties. More specifically, this means:
\n\n\n\nV-Dem’s Civil Liberties Index scores each country on a spectrum, with some countries protecting human rights more than others.
\n\n\n\nThe spectrum ranges from 0 (‘fewest rights’) to 1 (‘most rights’).
\n\n\n\n\n\n\n\n\n\n\n\nAs of version 12 of the dataset, V-Dem covers 202 countries, going back in time as far as 1789. Many countries have been covered since 1900, including before they became independent from their colonial powers.
\n\n\n\nTo actually measure what it wants to capture, V-Dem assesses the characteristics of human rights mostly through evaluations by experts.{ref} For more details, see: Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
\n\n\n\nThese anonymous experts are primarily academics and members of the media and civil society. They are also often nationals or residents of the country they assess, and therefore know its political system well and can evaluate aspects that are difficult to observe.
\n\n\n\nV-Dem’s own team of researchers supplements the expert evaluations. They code some easier-to-observe rules and laws of the political system, such as whether the legislature has a lower and upper house.
\n\n\n\nV-Dem uses several experts per country, year, and topic, to make its assessments less subjective. In total, around 3,500 country-experts fill surveys for V-Dem every year.
\n\n\n\nWhile there are fewer experts for small countries and for the time before 1900, they rely typically on 25 experts per country and 5 experts per topic.
\n\n\n\nV-Dem also works to make their coders’ assessments comparable across countries and time.
\n\n\n\nThe surveys ask the experts to answer very specific questions on completely explained scales about sub-characteristics of human rights — such as whether women can freely move in their own country — instead of making them rely on their broad impressions.
\n\n\n\nThe surveys are available in English, Arabic, French, Portuguese, Russian, and Spanish to reduce misunderstandings.
\n\n\n\nExperts further evaluate hypothetical countries, many coded several countries, and they denote their own uncertainty and personal demographic information.
\n\n\n\nV-Dem then uses this information to investigate expert biases, which they have found to be limited: they only find that experts from a country tend to be stricter in their assessments. {ref}“We have run extensive tests on how well such individual-level factors predict country-ratings but have found that the only factor consistently associated with country-ratings is country of origin (with “domestic” experts being harsher in their judgments).”
\n\n\n\nCoppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
\n\n\n\nV-Dem uses a statistical model to address any remaining differences between coders.{ref}Specifically, it uses a Bayesian Item-Response Theory estimation strategy.
\n\n\n\nMarquardt, Kyle, and Daniel Pemstein. 2018. IRT Models for Expert-Coded Panel Data. Political Analysis 26(4): 431-456.{/ref}
\n\n\n\nThe model combines the experts’ ratings of actual countries and hypothetical countries, as well as the experts’ stated uncertainties and personal demographics to produce best, upper-, and lower-bound estimates of many characteristics.{ref}Expressed precisely, V-Dem’s measurement model produces a probability distribution over the country-year scores. The best estimate is the distribution’s median, while the upper and lower bound estimates demarcate the interval in which the model places 68 percent of the probability mass.{/ref}
\n\n\n\nV-Dem provides these different estimates for all of its main and supplementary indices, including the Civil Liberties Index and the subindices for physical integrity rights, private civil liberties, and political civil liberties.
\n\n\n\nWith the different estimates, V-Dem explicitly acknowledges that its coders can be uncertain or make errors in their measurement.
\n\n\n\nThe overall Civil Liberties Index score is the result of averaging the three subindices.
\n\n\n\nV-Dem releases its data publicly, and makes it straightforward to download and use.
\n\n\n\nIt publishes the overall scores, the underlying subindices, and several hundred specific questions by country-year, country-date, and coder.
\n\n\n\nV-Dem also releases detailed descriptions of the questions and coding procedures that guide the experts and researchers.
\n\n\n\nIn our work, we expand the years covered by V-Dem further.
\n\n\n\nTo expand the time coverage of today’s countries and include more of the period when they were still non-sovereign territories, we identified the historical entity they were a part of and used that regime’s data whenever available.{ref}For example, V-Dem only provides data since Bangladesh’s independence in 1971. There is, however, data for Pakistan and the colony of India, both of which the current territory of Bangladesh was a part. We, therefore, use the data of Pakistan for Bangladesh from 1947 to 1970, and the data of India from 1789 to 1946. We did so for all countries with a past or current population of more than one million.{/ref}
\n\n\n\nWe also calculated regional and global averages of the Civil Liberties Index and its sub-indices, weighted and unweighted by population.
\n\n\n\nOur code and data are available on GitHub and record our revisions in detail.
\n\n\n\nV-Dem releases a new version of the data each year in March.
\n\n\n\nWe at Our World in Data aim to update our own data within a few weeks of the release.
\n\n\n\nThere are shortcomings in the way that V-Dem’s Civil Liberties Index characterizes and measures human rights.{ref}This and the following section draw on an article summarizing and reviewing some of the leading human rights datasets:
\n\n\n\nCope, Kevin, Charles Crabtree, and Christopher Fariss. 2020. Patterns of disagreement in indicators of state repression. Political Science Research and Methods 8(1): 178-187.{/ref}
\n\n\n\nThe index focuses on human rights as civil liberties and does not account for other characterizations, such as rights to food, health, or education. This means that countries with good health and education outcomes but restricted civil liberties, such as Iran and Singapore in recent years, still receive relatively low scores.
\n\n\n\nThe index also does not tell us anything about how human rights differ across parts of the population, such as between men and women, or between different ethnic groups.{ref}Though some of the index’s specific indicators distinguish between rights for men and women.{/ref}
\n\n\n\nV-Dem also does not cover some countries with very small populations.
\n\n\n\nFurthermore, the index is more difficult to interpret than other measures. The Civil Liberties Index does not identify whether a country grants or protects human rights or not, but only allows us to say whether a country is protecting human rights by comparing it to the range of the index, to other countries, or to the same country at another point in time. And when doing so, it is still difficult to say how large these differences are.{ref}This can be made easier by comparing how a score relates to the index’s overall distribution or its distribution for a specific year.{/ref}
\n\n\n\nThe assessment of the Civil Liberties Index remains to some extent subjective. Its index is built on difficult evaluations by experts and relies less on easier-to-observe characteristics, such as whether forming a civil society organization independent of the state is legal, or the number of allegations made in human rights reports against the government.
\n\n\n\nFinally, the index’s aggregation remains to some extent arbitrary. V-Dem does not say why these specific subindices were chosen, and why the subindices are given the same weight.
\n\n\n\nDespite these shortcomings, the index tells us a lot about how protected human rights are around the world, in the past and today.
\n\n\n\nIts characterization of human rights as people enjoying physical integrity, as well as private and political civil liberties, is commonly recognized to be at the core of human rights.
\n\n\n\nBecause it treats human rights as a spectrum, the index is able to capture both big and small differences in their protection across countries, and to record small changes within countries over time. This allows us to observe whether one country protects human rights more than another, or whether a country has protected human rights more or less over time.
\n\n\n\nThe index also covers many countries and years. With the exception of microstates, it covers all countries in the world. Many countries are covered since 1900 — even while they were colonized by another country — and some of them as far back as 1789.
\n\n\n\nFinally, V-Dem takes many steps to make its assessments valid, precise, comparable across countries and time, and transparent. It relies on many country and subject experts answering detailed surveys to measure aspects of political systems that are often difficult to observe and acknowledges the remaining uncertainty in their assessments.
\n\n\n\nWhether V-Dem’s Civil Liberties Index is a useful measure of human rights will depend on the questions we want to answer.
\n\n\n\nThe index will not give us a satisfying answer if we are interested in an understanding of human rights as also including rights to health or education; in differences in the protection of human rights by gender and ethnicity; and if we are also interested in the political systems of microstates.
\n\n\n\nIn these cases, we may have to rely on other measures.
\n\n\n\nBut if we value a sophisticated measure based on the knowledge of many country experts and are interested in big and small differences in civil liberties, within and across countries, and far into the past, we can learn a lot from this data.
\n\n\n\nIt is for these latter purposes we use the measure in some of our reporting on human rights.
\n\n\n\nI thank Edouard Mathieu for his very helpful comments and ideas about how to improve this article.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""There are several ways to measure human rights. Here is how the Varieties of Democracy project does it, one of the leading sources of global human rights data."", ""protected"": false}, ""date_gmt"": ""2022-12-16T17:33:07"", ""modified"": ""2023-06-08T10:57:14"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre""], ""modified_gmt"": ""2023-06-08T09:57:14"", ""comment_status"": ""closed"", ""featured_media"": 55154, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/12/distribution-human-rights-vdem-150x106.png"", ""medium_large"": ""/app/uploads/2022/12/distribution-human-rights-vdem-768x542.png""}}" 55049,Wild mammals make up only a few percent of the world’s mammals,wild-mammals-birds-biomass,post,publish,"Humans have transformed the mammal kingdom.
A diverse range of mammals once roamed the planet. This changed quickly and dramatically with the arrival of humans. Since then, wild land mammal biomass has declined by an estimated 85%.
Humans are now the dominant species.
We see this when we look at the distribution of mammals across the world today.
There are various ways that we could look at mammal species: we could compare them based on the number of individuals: their abundance. This tends to favor very small animals with large populations and doesn’t necessarily give us an idea of how dominant different species are.
Instead, ecologists often look at a different metric: biomass. This not only takes into account the number of animals but also factors in their size. Each animal is measured in tonnes of carbon, the fundamental building block of life.{ref}To calculate the biomass of a taxonomic group, the researchers multiplied the carbon stock for a single organism by the number of individuals in that group. In humans, for example, they calculate the average carbon quantity of a person and multiply by the human population. If you want to quickly estimate your carbon biomass: calculate 15% of your weight.{/ref} Biomass gives us a measure of the total biological productivity of an ecosystem. It also gives more weight to larger animals at higher levels of the ecological ‘pyramid’: these rely on well-functioning bases below them.
Let’s then look at the breakdown of the global mammal kingdom in 2015. This data is sourced from the study by Yinon Bar-On, Rob Phillips, and Ron Milo.{ref}Bar-On, Y. M., Phillips, R., & Milo, R. (2018). The biomass distribution on Earth. Proceedings of the National Academy of Sciences, 115(25), 6506-6511.{/ref}
Each icon is equivalent to around one million tonnes of carbon. This includes both land and marine wild mammals.
Wild mammals make up just 4% of the mammal kingdom.
The dominance of humans is clear. Alone, we account for around one-third of mammal biomass. Almost ten times greater than wild mammals.
Our livestock then accounts for almost two-thirds. Cattle weigh almost ten times as much as all wild mammals combined. The biomass of all of the world’s wild mammals is about a third of our pigs alone.
When I show people the chart above, one question always comes up: what about chickens? Of course, chickens are not mammals. But we can do a similar comparison between poultry and wild birds.
For birds the distribution is similar: poultry biomass is more than twice that of wild birds. We see this in the chart.
The charts above give us a snapshot of how the mammal kingdom looks in the modern day. But both the distribution and amount of mammal biomass have changed dramatically over time.
In the visualization we can see the total biomass of mammals at four points in time: 100,000 years ago; 10,000 years ago, in the year 1900, and the 2015 snapshot we looked at previously.{ref}These estimates were constructed from three key sources: long historical figures come from the work of Anthony Barnosky (2008); figures for the year 1900 figures from Vaclav Smil (2011); and 2015 figures from Yinon Bar-On, Rob Phillips and Ron Milo (2018).{ref} Barnosky, A. D. (2008). Megafauna biomass tradeoff as a driver of Quaternary and future extinctions. Proceedings of the National Academy of Sciences, 105(Supplement 1), 11543-11548.
Smil, V. (2011). Harvesting the biosphere: What we have taken from nature. MIT Press.
Bar-On, Y. M., Phillips, R., & Milo, R. (2018). The biomass distribution on Earth. Proceedings of the National Academy of Sciences, 115(25), 6506-6511.{/ref}
In the last 100,000 years, as the human population increased, wild mammal biomass has declined by 85%.{ref}This was first driven by hunting: a global population of less than 5 million early humans hunted more than 100 of the largest mammals to extinction. Since the agricultural revolution, the decline in wild mammals has been driven by a mix of hunting but also habitat loss from the expansion of agricultural land.{/ref} I looked at this history in a related article.
The decline of wild mammals is not the only change. At the same time, humans and our livestock have grown significantly – from millions to billions.
What’s interesting is that, while the diversity of the mammal kingdom has decreased, its total size has expanded a lot. Terrestrial mammals weighed in at an estimated 20 million tonnes of carbon 10,000 years ago. This is now around nine times larger.{ref}With the rise and spread of farming, humans started to change the balance of carbon across ecosystems. We cut down forests, releasing carbon, and replaced them with farmlands. We also became much more skilled at balancing other nutrients that are essential for agriculture: until the early 1900s, we slowly improved our ability to balance nitrogen within our soils. This reached a whole new level in the 1920s with the invention of synthetic fertilizers from the Haber-Bosch process.
Around half of the humans alive today owe their existence to synthetic fertilizers.
We were no longer trying to move nitrogen around the biosphere. We were taking nitrogen from the atmosphere and pulling it into the soils and crops where we could use it for food production, and the raising of livestock.
The Industrial Revolution lay at the heart of this change. To grow, we needed energy. Again, we were not just moving carbon around the biosphere. We were adding to it.
Erisman, J. W., Sutton, M. A., Galloway, J., Klimont, Z., & Winiwarter, W. (2008). How a century of ammonia synthesis changed the world. Nature Geoscience, 1(10), 636-639.
Smil, V. (2004). Enriching the Earth: Fritz Haber, Carl Bosch, and the Transformation of World Food Production. MIT Press. ISBN: 9780262194495.
Stewart, W. M., Dibb, D. W., Johnston, A. E., & Smyth, T. J. (2005). The contribution of commercial fertilizer nutrients to food production. Agronomy Journal, 97(1), 1-6.{/ref}
Within centuries, humans have increased the size of the mammal kingdom almost ten-fold.
Many thanks to Max Roser for providing feedback and suggestions on this article and its visualizations.
","{""id"": ""wp-55049"", ""slug"": ""wild-mammals-birds-biomass"", ""content"": {""toc"": [], ""body"": [{""left"": [{""type"": ""text"", ""value"": [{""text"": ""Humans have transformed the mammal kingdom."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A diverse range of mammals once roamed the planet. This changed quickly and dramatically with the arrival of humans. Since then, wild land mammal biomass "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/wild-mammal-decline"", ""children"": [{""text"": ""has declined"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" by an estimated 85%."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Humans are now the dominant species."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We see this when we look at the distribution of mammals across the world today."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are various ways that we could look at mammal species: we could compare them based on the number of individuals: their "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""abundance"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". This tends to favor very small animals with large populations and doesn’t necessarily give us an idea of how dominant different species are."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Instead, ecologists often look at a different metric: biomass. This not only takes into account the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""number"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of animals but also factors in their size. Each animal is measured in tonnes of carbon, the fundamental building block of life.{ref}To calculate the biomass of a taxonomic group, the researchers multiplied the carbon stock for a single organism by the number of individuals in that group. In humans, for example, they calculate the average carbon quantity of a person and multiply by the human population. If you want to quickly estimate your carbon biomass: calculate 15% of your weight.{/ref} Biomass gives us a measure of the total biological productivity of an ecosystem. It also gives more weight to larger animals at higher levels of the ecological ‘pyramid’: these rely on well-functioning bases below them."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Let’s then look at the breakdown of the global mammal kingdom in 2015. This data is sourced from the study by Yinon Bar-On, Rob Phillips, and Ron Milo.{ref}Bar-On, Y. M., Phillips, R., & Milo, R. (2018). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.pnas.org/content/115/25/6506"", ""children"": [{""text"": ""The biomass distribution on Earth"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Proceedings of the National Academy of Sciences"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 115(25), 6506-6511.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Each icon is equivalent to around one million tonnes of carbon. This includes both land and marine wild mammals."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Wild mammals make up just 4% of the mammal kingdom."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The dominance of humans is clear. Alone, we account for around one-third of mammal biomass. Almost ten times greater than wild mammals. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Our livestock then accounts for almost two-thirds. Cattle weigh almost ten times as much as all wild mammals combined. The biomass of all of the world’s wild mammals is about a third of our pigs alone."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Distribution-of-earths-mammals.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Global poultry weighs more than twice that of wild birds"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""When I show people the chart above, one question always comes up: what about chickens? Of course, chickens are not mammals. But we can do a similar comparison between poultry and wild birds."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""For birds the distribution is similar: poultry biomass is more than twice that of wild birds. We see this in the chart."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Distribution-of-mammals-and-birds.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Wild mammals have declined, but the total amount of mammal biomass has increased a lot"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The charts above give us a snapshot of how the mammal kingdom looks in the modern day. But both the distribution and amount of mammal biomass have changed dramatically over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the visualization we can see the total biomass of mammals at four points in time: 100,000 years ago; 10,000 years ago, in the year 1900, and the 2015 snapshot we looked at previously.{ref}These estimates were constructed from three key sources: long historical figures come from the work of Anthony Barnosky (2008); figures for the year 1900 figures from Vaclav Smil (2011); and 2015 figures from Yinon Bar-On, Rob Phillips and Ron Milo (2018).{ref} Barnosky, A. D. (2008). Megafauna biomass tradeoff as a driver of Quaternary and future extinctions. Proceedings of the National Academy of Sciences, 105(Supplement 1), 11543-11548."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Smil, V. (2011). Harvesting the biosphere: What we have taken from nature. MIT Press."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Bar-On, Y. M., Phillips, R., & Milo, R. (2018). The biomass distribution on Earth. Proceedings of the National Academy of Sciences, 115(25), 6506-6511.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the last 100,000 years, as the human population increased, wild mammal biomass has declined by 85%.{ref}This was first driven by hunting: a global population of less than 5 million early humans hunted more than 100 of the largest mammals to extinction. Since the agricultural revolution, the decline in wild mammals has been driven by a mix of hunting but also habitat loss from the expansion of agricultural land.{/ref} I looked at this history in a "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/wild-mammal-decline"", ""children"": [{""children"": [{""text"": ""related article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The decline of wild mammals is not the only change. At the same time, humans and our livestock have grown significantly – from millions to billions. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""What’s interesting is that, while the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""diversity"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of the mammal kingdom has decreased, its total size has expanded a lot. Terrestrial mammals weighed in at an estimated 20 million tonnes of carbon 10,000 years ago. This is now around nine times larger.{ref}With the rise and spread of farming, humans started to change the balance of carbon across ecosystems. We cut down forests, releasing carbon, and replaced them with farmlands. We also became much more skilled at balancing other nutrients that are essential for agriculture: until the early 1900s, we slowly improved our ability to balance nitrogen within our soils. This reached a whole new level in the 1920s with the invention of synthetic fertilizers from the Haber-Bosch process."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Around half of the humans alive today "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/how-many-people-does-synthetic-fertilizer-feed"", ""children"": [{""text"": ""owe their existence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to synthetic fertilizers. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We were no longer trying to move nitrogen around the biosphere. We were taking nitrogen from the atmosphere and pulling it into the soils and crops where we could use it for food production, and the raising of livestock."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Industrial Revolution lay at the heart of this change. To grow, we needed energy. Again, we were not just moving carbon around the biosphere. We were adding to it."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Erisman, J. W., Sutton, M. A., Galloway, J., Klimont, Z., & Winiwarter, W. (2008). How a century of ammonia synthesis changed the world. Nature Geoscience, 1(10), 636-639."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Smil, V. (2004). Enriching the Earth: Fritz Haber, Carl Bosch, and the Transformation of World Food Production. MIT Press. ISBN: 9780262194495."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Stewart, W. M., Dibb, D. W., Johnston, A. E., & Smyth, T. J. (2005). The contribution of commercial fertilizer nutrients to food production. Agronomy Journal, 97(1), 1-6.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Within centuries, humans have increased the size of the mammal kingdom almost ten-fold."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Decline-of-the-worlds-wild-mammals.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Keep reading at Our World in Data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 4, ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/large-mammals-extinction"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/wild-mammal-decline"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/biodiversity"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""text"": [{""text"": ""Acknowledgments"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 4, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many thanks to Max Roser for providing feedback and suggestions on this article and its visualizations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Wild mammals make up only a few percent of the world’s mammals"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""Livestock make up 62% of the world’s mammal biomass; humans account for 34%; and wild mammals are just 4%."", ""dateline"": ""December 15, 2022"", ""subtitle"": ""Livestock make up 62% of the world’s mammal biomass; humans account for 34%; and wild mammals are just 4%."", ""sidebar-toc"": false, ""featured-image"": ""Mammal-distribution-thumbnail.png""}, ""createdAt"": ""2022-12-09T12:12:11.000Z"", ""published"": false, ""updatedAt"": ""2023-08-03T15:11:26.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-12-15T10:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}], ""numBlocks"": 11, ""numErrors"": 3, ""wpTagCounts"": {""image"": 3, ""column"": 6, ""columns"": 3, ""heading"": 4, ""paragraph"": 28, ""owid/prominent-link"": 3}, ""htmlTagCounts"": {""p"": 28, ""h3"": 2, ""h5"": 2, ""div"": 9, ""figure"": 3}}",2022-12-15 10:00:00,2024-03-10 13:04:18,1RJ4x30eYNY1FMG6br7jI8D02z8kb4cQTN-YX-IOm9xs,"[""Hannah Ritchie""]",Livestock make up 62% of the world’s mammal biomass; humans account for 34%; and wild mammals are just 4%.,2022-12-09 12:12:11,2023-08-03 15:11:26,https://ourworldindata.org/wp-content/uploads/2022/11/Mammal-distribution-thumbnail.png,{},"Humans have transformed the mammal kingdom. A diverse range of mammals once roamed the planet. This changed quickly and dramatically with the arrival of humans. Since then, wild land mammal biomass [has declined](https://ourworldindata.org/wild-mammal-decline) by an estimated 85%. Humans are now the dominant species. We see this when we look at the distribution of mammals across the world today. There are various ways that we could look at mammal species: we could compare them based on the number of individuals: their _abundance_. This tends to favor very small animals with large populations and doesn’t necessarily give us an idea of how dominant different species are. Instead, ecologists often look at a different metric: biomass. This not only takes into account the _number_ of animals but also factors in their size. Each animal is measured in tonnes of carbon, the fundamental building block of life.{ref}To calculate the biomass of a taxonomic group, the researchers multiplied the carbon stock for a single organism by the number of individuals in that group. In humans, for example, they calculate the average carbon quantity of a person and multiply by the human population. If you want to quickly estimate your carbon biomass: calculate 15% of your weight.{/ref} Biomass gives us a measure of the total biological productivity of an ecosystem. It also gives more weight to larger animals at higher levels of the ecological ‘pyramid’: these rely on well-functioning bases below them. Let’s then look at the breakdown of the global mammal kingdom in 2015. This data is sourced from the study by Yinon Bar-On, Rob Phillips, and Ron Milo.{ref}Bar-On, Y. M., Phillips, R., & Milo, R. (2018). [The biomass distribution on Earth](https://www.pnas.org/content/115/25/6506). _Proceedings of the National Academy of Sciences_, 115(25), 6506-6511.{/ref} Each icon is equivalent to around one million tonnes of carbon. This includes both land and marine wild mammals. Wild mammals make up just 4% of the mammal kingdom. The dominance of humans is clear. Alone, we account for around one-third of mammal biomass. Almost ten times greater than wild mammals. Our livestock then accounts for almost two-thirds. Cattle weigh almost ten times as much as all wild mammals combined. The biomass of all of the world’s wild mammals is about a third of our pigs alone.Humans have transformed the mammal kingdom.
\n\n\n\nA diverse range of mammals once roamed the planet. This changed quickly and dramatically with the arrival of humans. Since then, wild land mammal biomass has declined by an estimated 85%.
\n\n\n\nHumans are now the dominant species.
\n\n\n\nWe see this when we look at the distribution of mammals across the world today.
\n\n\n\nThere are various ways that we could look at mammal species: we could compare them based on the number of individuals: their abundance. This tends to favor very small animals with large populations and doesn’t necessarily give us an idea of how dominant different species are.
\n\n\n\nInstead, ecologists often look at a different metric: biomass. This not only takes into account the number of animals but also factors in their size. Each animal is measured in tonnes of carbon, the fundamental building block of life.{ref}To calculate the biomass of a taxonomic group, the researchers multiplied the carbon stock for a single organism by the number of individuals in that group. In humans, for example, they calculate the average carbon quantity of a person and multiply by the human population. If you want to quickly estimate your carbon biomass: calculate 15% of your weight.{/ref} Biomass gives us a measure of the total biological productivity of an ecosystem. It also gives more weight to larger animals at higher levels of the ecological ‘pyramid’: these rely on well-functioning bases below them.
\n\n\n\nLet’s then look at the breakdown of the global mammal kingdom in 2015. This data is sourced from the study by Yinon Bar-On, Rob Phillips, and Ron Milo.{ref}Bar-On, Y. M., Phillips, R., & Milo, R. (2018). The biomass distribution on Earth. Proceedings of the National Academy of Sciences, 115(25), 6506-6511.{/ref}
\n\n\n\nEach icon is equivalent to around one million tonnes of carbon. This includes both land and marine wild mammals.
\n\n\n\nWild mammals make up just 4% of the mammal kingdom.
\n\n\n\nThe dominance of humans is clear. Alone, we account for around one-third of mammal biomass. Almost ten times greater than wild mammals.
\n\n\n\nOur livestock then accounts for almost two-thirds. Cattle weigh almost ten times as much as all wild mammals combined. The biomass of all of the world’s wild mammals is about a third of our pigs alone.
\n![\""\""](\""https://owid.cloud/app/uploads/2021/03/Distribution-of-earths-mammals.png\"")
When I show people the chart above, one question always comes up: what about chickens? Of course, chickens are not mammals. But we can do a similar comparison between poultry and wild birds.
\n\n\n\nFor birds the distribution is similar: poultry biomass is more than twice that of wild birds. We see this in the chart.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/Distribution-of-mammals-and-birds.png\"")
The charts above give us a snapshot of how the mammal kingdom looks in the modern day. But both the distribution and amount of mammal biomass have changed dramatically over time.
\n\n\n\nIn the visualization we can see the total biomass of mammals at four points in time: 100,000 years ago; 10,000 years ago, in the year 1900, and the 2015 snapshot we looked at previously.{ref}These estimates were constructed from three key sources: long historical figures come from the work of Anthony Barnosky (2008); figures for the year 1900 figures from Vaclav Smil (2011); and 2015 figures from Yinon Bar-On, Rob Phillips and Ron Milo (2018).{ref} Barnosky, A. D. (2008). Megafauna biomass tradeoff as a driver of Quaternary and future extinctions. Proceedings of the National Academy of Sciences, 105(Supplement 1), 11543-11548.
\n\n\n\nSmil, V. (2011). Harvesting the biosphere: What we have taken from nature. MIT Press.
\n\n\n\nBar-On, Y. M., Phillips, R., & Milo, R. (2018). The biomass distribution on Earth. Proceedings of the National Academy of Sciences, 115(25), 6506-6511.{/ref}
\n\n\n\nIn the last 100,000 years, as the human population increased, wild mammal biomass has declined by 85%.{ref}This was first driven by hunting: a global population of less than 5 million early humans hunted more than 100 of the largest mammals to extinction. Since the agricultural revolution, the decline in wild mammals has been driven by a mix of hunting but also habitat loss from the expansion of agricultural land.{/ref} I looked at this history in a related article.
\n\n\n\nThe decline of wild mammals is not the only change. At the same time, humans and our livestock have grown significantly – from millions to billions.
\n\n\n\nWhat’s interesting is that, while the diversity of the mammal kingdom has decreased, its total size has expanded a lot. Terrestrial mammals weighed in at an estimated 20 million tonnes of carbon 10,000 years ago. This is now around nine times larger.{ref}With the rise and spread of farming, humans started to change the balance of carbon across ecosystems. We cut down forests, releasing carbon, and replaced them with farmlands. We also became much more skilled at balancing other nutrients that are essential for agriculture: until the early 1900s, we slowly improved our ability to balance nitrogen within our soils. This reached a whole new level in the 1920s with the invention of synthetic fertilizers from the Haber-Bosch process.
\n\n\n\nAround half of the humans alive today owe their existence to synthetic fertilizers.
\n\n\n\nWe were no longer trying to move nitrogen around the biosphere. We were taking nitrogen from the atmosphere and pulling it into the soils and crops where we could use it for food production, and the raising of livestock.
\n\n\n\nThe Industrial Revolution lay at the heart of this change. To grow, we needed energy. Again, we were not just moving carbon around the biosphere. We were adding to it.
\n\n\n\nErisman, J. W., Sutton, M. A., Galloway, J., Klimont, Z., & Winiwarter, W. (2008). How a century of ammonia synthesis changed the world. Nature Geoscience, 1(10), 636-639.
\n\n\n\nSmil, V. (2004). Enriching the Earth: Fritz Haber, Carl Bosch, and the Transformation of World Food Production. MIT Press. ISBN: 9780262194495.
\n\n\n\nStewart, W. M., Dibb, D. W., Johnston, A. E., & Smyth, T. J. (2005). The contribution of commercial fertilizer nutrients to food production. Agronomy Journal, 97(1), 1-6.{/ref}
\n\n\n\nWithin centuries, humans have increased the size of the mammal kingdom almost ten-fold.
\n![\""\""](\""https://owid.cloud/app/uploads/2021/03/Decline-of-the-worlds-wild-mammals.png\"")
Many thanks to Max Roser for providing feedback and suggestions on this article and its visualizations.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Livestock make up 62% of the world’s mammal biomass; humans account for 34%; and wild mammals are just 4%."", ""protected"": false}, ""date_gmt"": ""2022-12-15T10:00:00"", ""modified"": ""2023-08-03T16:11:26"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2023-08-03T15:11:26"", ""comment_status"": ""closed"", ""featured_media"": 54667, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/11/Mammal-distribution-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2022/11/Mammal-distribution-thumbnail-768x402.png""}}" 54861,Technology over the long run: zoom out to see how dramatically the world can change within a lifetime,technology-long-run,post,publish,"Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Technology can change the world in ways that are unimaginable, until they happen. Switching on an electric light would have been unimaginable for our medieval ancestors. In their childhood, our grandparents would have struggled to imagine a world connected by smartphones and the Internet.
Similarly, it is hard for us to imagine the arrival of all those technologies that will fundamentally change the world we are used to.
We can remind ourselves that our own future might look very different from the world today by looking back at how rapidly technology has changed our world in the past. That’s what this article is about.
One insight I take away from this long-term perspective is how unusual our time is. Technological change was extremely slow in the past – the technologies that our ancestors got used to in their childhood were still central to their lives in their old age. In stark contrast to those days, we live in a time of extraordinarily fast technological change. For recent generations, it was common for technologies that were unimaginable in their youth to become common later in life.
The big visualization offers a long-term perspective on the history of technology.{ref}The recent speed of technological change makes it difficult to picture the history of technology in one visualization. When you visualize this development on a linear timeline then most of the timeline is almost empty, while all the action is crammed into the right-corner:
In my large visualization here I tried to avoid this problem and instead show the long history of technology in a way that lets you see when each technological breakthrough happened, and how within the last millennia there was a continuous acceleration of technological change.{/ref}
The timeline begins at the center of the spiral. The first use of stone tools, 3.4 million years ago, marks the beginning of this history of technology.{ref}It is of course difficult to assess when exactly the first stone tools were used.
The research by McPherron et al (2010) suggested that it was at least 3.39 million years ago. This is based on two fossilized bones, found in Dikika in Ethiopia, which showed “stone-tool cut marks for flesh removal and percussion marks for marrow access”. These marks were interpreted as being caused by the consumption of meat and provide the first evidence that one of our ancestors, Australopithecus afarensis, used stone tools.
The research by Harmand et al (2015) provided evidence for stone tool use in today’s Kenya 3.3 million years ago.
References:
McPherron et al (2010) – Evidence for stone-tool-assisted consumption of animal tissues before 3.39 million years ago at Dikika, Ethiopia. Published in Nature.
Harmand et al (2015) – 3.3-million-year-old stone tools from Lomekwi 3, West Turkana, Kenya. Published in Nature.{/ref} Each turn of the spiral then represents 200,000 years of history. It took 2.4 million years – 12 turns of the spiral – for our ancestors to control fire and use it for cooking.{ref}Evidence for controlled fire use approximately 1 million years ago is provided by Berna et al (2012) Microstratigraphic evidence of in situ fire in the Acheulean strata of Wonderwerk Cave, Northern Cape province, South Africa, published in PNAS.
The authors write: “The ability to control fire was a crucial turning point in human evolution, but the question of when hominins first developed this ability still remains. Here we show that micromorphological and Fourier transform infrared microspectroscopy (mFTIR) analyses of intact sediments at the site of Wonderwerk Cave, Northern Cape province, South Africa, provide unambiguous evidence—in the form of burned bone and ashed plant remains—that burning took place in the cave during the early Acheulean occupation, approximately 1.0 Ma. To the best of our knowledge, this is the earliest secure evidence for burning in an archaeological context.”{/ref}
To be able to visualize the inventions in the more recent past – the last 12,000 years – I had to unroll the spiral. I needed more space to be able to show when agriculture, writing, and the wheel were invented. During this period, technological change was faster, but it was still relatively slow: several thousand years passed between each of these three inventions.
From 1800 onwards, I stretched out the timeline even further to show the many major inventions that rapidly followed one after the other.
The long-term perspective that this chart provides makes it clear just how unusually fast technological change is in our time.
You can use this visualization to see how technology developed in particular domains. Follow, for example, the history of communication: from writing, to paper, to the printing press, to the telegraph, the telephone, the radio, all the way to the Internet and smartphones.
Or follow the rapid development of human flight. In 1903, the Wright brothers took the first flight in human history (they were in the air for less than a minute), and just 66 years later, we landed on the moon. Many people saw both within their lifetimes: the first plane and the moon landing.
This large visualization also highlights the wide range of technology’s impact on our lives. It includes extraordinarily beneficial innovations, such as the vaccine that allowed humanity to eradicate smallpox, and it includes terrible innovations, like the nuclear bombs that endanger the lives of all of us.
What will the next decades bring?
The red timeline reaches up to the present and then continues in green into the future. Many children born today, even without any further increases in life expectancy, will live well into the 22nd century.
New vaccines, progress in clean, low-carbon energy, better cancer treatments – a range of future innovations could very much improve our living conditions and the environment around us. But, as I argue in a series of articles, there is one technology that could even more profoundly change our world: artificial intelligence (AI).
One reason why artificial intelligence is such an important innovation is that intelligence is the main driver of innovation itself. This fast-paced technological change could speed up even more if it’s not only driven by humanity’s intelligence, but artificial intelligence too. If this happens, the change that is currently stretched out over the course of decades might happen within very brief time spans of just a year. Possibly even faster.{ref}This is what authors like Holden Karnofsky called ‘Process for Automating Scientific and Technological Advancement’ or PASTA. There are some recent developments that go in this direction: DeepMind’s AlphaFold helped to make progress on one of the large problems in biology, and they have also developed an AI system that finds new algorithms that are relevant to building a more powerful AI.{/ref}
I think AI technology could have a fundamentally transformative impact on our world. In many ways, it is already changing our world, as I documented in this companion article. As this technology is becoming more capable in the years and decades to come, it can give immense power to those who control it (and it poses the risk that it could escape our control entirely).
Such systems might seem hard to imagine today, but AI technology is advancing very fast. Many AI experts believe there is a real chance that human-level artificial intelligence will be developed within the next decades, as I documented in this article.
What is familiar to us today – photography, the radio, antibiotics, the Internet, or the International Space Station circling our planet – was unimaginable to our ancestors just a few generations ago. If your great-great-great grandparents could spend a week with you they would be blown away by your everyday life.
What I take away from this history is that I will likely see technologies in my lifetime that appear unimaginable to me today.
In addition to this trend towards increasingly rapid innovation, there is a second long-run trend. Technology has become increasingly powerful. While our ancestors wielded stone tools, we are building globe-spanning AI systems and technologies that can edit our genes.
Because of the immense power that technology gives those who control it, there is little that is as important as the question of which technologies get developed during our lifetimes. Therefore I think it is a mistake to leave the question about the future of technology to the technologists. Which technologies are controlled by whom is one of the most important political questions of our time, because of the enormous power that these technologies convey to those who control them.
We all should strive to gain the knowledge we need to contribute to an intelligent debate about the world we want to live in. To a large part this means gaining the knowledge, and wisdom, on the question of which technologies we want.
Our new topic page on AI.
Acknowledgements: I would like to thank my colleagues Hannah Ritchie, Bastian Herre, Natasha Ahuja, Edouard Mathieu, Daniel Bachler, Charlie Giattino, and Pablo Rosado for their helpful comments to drafts of this essay and the visualization. Thanks also to Lizka Vaintrob and Ben Clifford for a conversation that initiated this visualization.
","{""id"": ""wp-54861"", ""slug"": ""technology-long-run"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Our World in Data presents the data and research to make progress against the world’s largest problems."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""This article draws on data and research discussed in our entry on "", ""spanType"": ""span-simple-text""}, {""children"": [{""url"": ""https://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""Artificial Intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""spanType"": ""span-bold""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Technology can change the world in ways that are unimaginable, until they happen. Switching on an electric light would have been unimaginable for our medieval ancestors. In their childhood, our grandparents would have struggled to imagine a world connected by smartphones and the Internet."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Similarly, it is hard for us to imagine the arrival of all those technologies that will fundamentally change the world we are used to."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We can remind ourselves that our own future might look very different from the world today by looking back at how rapidly technology has changed our world in the past. That’s what this article is about. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One insight I take away from this long-term perspective is how unusual our time is. Technological change was "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""extremely"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" slow in the past – the technologies that our ancestors got used to in their childhood were still central to their lives in their old age. In stark contrast to those days, we live in a time of extraordinarily fast technological change. For recent generations, it was common for technologies that were unimaginable in their youth to become common later in life. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""The long-run perspective on technological change"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The big visualization offers a long-term perspective on the history of technology.{ref}The recent speed of technological change makes it difficult to picture the history of technology in one visualization. When you visualize this development on a linear timeline then most of the timeline is almost empty, while all the action is crammed into the right-corner:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""cH9kcyrR2oopnW-KiDAAnAGVsMA-h0D9L-rPaG7lUr74vHI6xguNWEKZ6Gy047SOBN8lC5sXEbYLvaf_mWLDZ1K_LFoYij_K-e6gOwGjH3p5YGhRQPyWFMPnJAEDM6hhNIfaS2S94pYeaXdW4-risESc3T83GfHXLqf4mv4EHaHdBHw1kuXXTRW8Z-FPrQ"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In my large visualization here I tried to avoid this problem and instead show the long history of technology in a way that lets you see when each technological breakthrough happened, and how within the last millennia there was a continuous acceleration of technological change.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The timeline begins at the center of the spiral. The first use of stone tools, 3.4 million years ago, marks the beginning of this history of technology.{ref}It is of course difficult to assess when exactly the first stone tools were used."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The research by McPherron et al (2010) suggested that it was at least 3.39 million years ago. This is based on two fossilized bones, found in Dikika in Ethiopia, which showed “stone-tool cut marks for flesh removal and percussion marks for marrow access”. These marks were interpreted as being caused by the consumption of meat and provide the first evidence that one of our ancestors, Australopithecus afarensis, used stone tools."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The research by Harmand et al (2015) provided evidence for stone tool use in today’s Kenya 3.3 million years ago."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""References: "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""McPherron et al (2010)"", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/nature09248"", ""children"": [{""text"": "" – Evidence for stone-tool-assisted consumption of animal tissues before 3.39 million years ago at Dikika, Ethiopia"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Published in Nature."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Harmand et al (2015) –"", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/nature14464"", ""children"": [{""text"": "" 3.3-million-year-old stone tools from Lomekwi 3, West Turkana, Kenya"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Published in Nature.{/ref} Each turn of the spiral then represents 200,000 years of history. It took 2.4 million years – 12 turns of the spiral – for our ancestors to control fire and use it for cooking.{ref}Evidence for controlled fire use approximately 1 million years ago is provided by Berna et al (2012)"", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.pnas.org/doi/full/10.1073/pnas.1117620109"", ""children"": [{""text"": "" Microstratigraphic evidence of in situ fire in the Acheulean strata of Wonderwerk Cave, Northern Cape province, South Africa"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", published in PNAS."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The authors write: "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""“The ability to control fire was a crucial turning point in human evolution, but the question of when hominins first developed this ability still remains. Here we show that micromorphological and Fourier transform infrared microspectroscopy (mFTIR) analyses of intact sediments at the site of Wonderwerk Cave, Northern Cape province, South Africa, provide unambiguous evidence—in the form of burned bone and ashed plant remains—that burning took place in the cave during the early Acheulean occupation, approximately 1.0 Ma. To the best of our knowledge, this is the earliest secure evidence for burning in an archaeological context.”"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To be able to visualize the inventions in the more recent past – the last 12,000 years – I had to unroll the spiral. I needed more space to be able to show when agriculture, writing, and the wheel were invented. During this period, technological change was faster, but it was still relatively slow: several thousand years passed between each of these three inventions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""From 1800 onwards, I stretched out the timeline even further to show the many major inventions that rapidly followed one after the other. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The long-term perspective that this chart provides makes it clear just how unusually fast technological change is in our time. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""You can use this visualization to see how technology developed in particular domains. Follow, for example, the history of communication: from writing, to paper, to the printing press, to the telegraph, the telephone, the radio, all the way to the Internet and smartphones."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Or follow the rapid development of human flight. In 1903, the Wright brothers took the first flight in human history (they were in the air for less than a minute), and just 66 years later, we landed on the moon. Many people saw both within their lifetimes: the first plane and the moon landing."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This large visualization also highlights the wide range of technology’s impact on our lives. It includes extraordinarily beneficial innovations, such as the vaccine that allowed humanity to "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/smallpox"", ""children"": [{""text"": ""eradicate smallpox"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", and it includes terrible innovations, like the nuclear bombs that endanger the lives "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/nuclear-weapons-risk"", ""children"": [{""text"": ""of all of us"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""What will the next decades bring? "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The red timeline reaches up to the present and then continues in green into the future. Many children born today, even without any further increases in life expectancy, will live well into the 22nd century. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""New vaccines, progress in clean, low-carbon energy, better cancer treatments – a range of future innovations could very much improve our living conditions and the environment around us. But, as I argue in a "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/artificial-intelligence#research-writing"", ""children"": [{""text"": ""series of articles"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", there is one technology that could even more profoundly change our world: artificial intelligence (AI)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One reason why artificial intelligence is such an important innovation is that intelligence is the main driver of innovation itself. This fast-paced technological change could speed up even more if it’s not only driven by humanity’s intelligence, but artificial intelligence too. If this happens, the change that is currently stretched out over the course of decades might happen within very brief time spans of just a year. Possibly even faster.{ref}This is what authors like Holden Karnofsky "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.cold-takes.com/transformative-ai-timelines-part-1-of-4-what-kind-of-ai/"", ""children"": [{""text"": ""called"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" ‘Process for Automating Scientific and Technological Advancement’ or PASTA. There are some recent developments that go in this direction: DeepMind’s "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/AlphaFold"", ""children"": [{""text"": ""AlphaFold"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" helped to make progress on one of the large problems in biology, and they "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.deepmind.com/blog/discovering-novel-algorithms-with-alphatensor"", ""children"": [{""text"": ""have also"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" developed an AI system that finds new algorithms that are relevant to building a more powerful AI.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I think AI technology could have a fundamentally transformative impact on our world. In many ways, it is already changing our world, as I documented "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/brief-history-of-ai"", ""children"": [{""text"": ""in this companion article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". As this technology is becoming more capable in the years and decades to come, it can give immense power to those who control it (and it poses "", ""spanType"": ""span-simple-text""}, {""url"": ""https://80000hours.org/problem-profiles/artificial-intelligence/"", ""children"": [{""text"": ""the risk"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" that it could escape our control entirely)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Such systems might seem hard to imagine today, but AI technology is advancing very fast. Many AI experts believe there is a real chance that human-level artificial intelligence will be developed within the next decades, as I documented in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""this article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Longterm-timeline-of-technology.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Technology will continue to change the world – we should all make sure that it changes it for the better"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""What is familiar to us today – photography, the radio, antibiotics, the Internet, or the International Space Station circling our planet – was unimaginable to our ancestors just a few generations ago. If your great-great-great grandparents could spend a week with you they would be blown away by your everyday life."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""What I take away from this history is that I will likely see technologies in my lifetime that appear unimaginable to me today. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In addition to this trend towards increasingly rapid innovation, there is a second long-run trend. Technology has become increasingly powerful. While our ancestors wielded stone tools, we are building globe-spanning AI systems and technologies that can edit our genes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Because of the immense power that technology gives those who control it, there is little that is as important as the question of which technologies get developed during our lifetimes. Therefore I think "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-impact"", ""children"": [{""text"": ""it is a mistake"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to leave the question about the future of technology to the technologists. Which technologies are controlled by whom is one of the most important political questions of our time, because of the enormous power that these technologies convey to those who control them."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We all should strive to gain the knowledge we need to contribute to an intelligent debate about the world we want to live in. To a large part this means gaining the knowledge, and wisdom, on the question of which technologies we want."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Our new topic page on AI"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Acknowledgements:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" I would like to thank my colleagues Hannah Ritchie, Bastian Herre, Natasha Ahuja, Edouard Mathieu, Daniel Bachler, Charlie Giattino, and Pablo Rosado for their helpful comments to drafts of this essay and the visualization. Thanks also to Lizka Vaintrob and Ben Clifford for a conversation that initiated this visualization."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Technology over the long run: zoom out to see how dramatically the world can change within a lifetime"", ""authors"": [""Max Roser""], ""excerpt"": ""It is easy to underestimate how much the world can change within a lifetime. Bringing to mind how dramatically the world has changed can help us see how different the world could be in a few years or decades."", ""dateline"": ""February 22, 2023"", ""subtitle"": ""It is easy to underestimate how much the world can change within a lifetime. Bringing to mind how dramatically the world has changed can help us see how different the world could be in a few years or decades."", ""sidebar-toc"": false, ""featured-image"": ""featured-image-Longterm-timeline-of-technology.png""}, ""createdAt"": ""2022-12-02T20:24:58.000Z"", ""published"": false, ""updatedAt"": ""2023-10-11T08:43:29.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-22T11:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""too many figcaption elements after archieml transform"", ""details"": ""Found 10 elements after transforming to archieml""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag spacer""}], ""numBlocks"": 15, ""numErrors"": 7, ""wpTagCounts"": {""html"": 1, ""image"": 2, ""column"": 2, ""spacer"": 1, ""columns"": 1, ""heading"": 2, ""paragraph"": 32, ""separator"": 1, ""owid/prominent-link"": 1}, ""htmlTagCounts"": {""p"": 33, ""h4"": 2, ""hr"": 1, ""div"": 5, ""figure"": 2, ""figcaption"": 1}}",2023-02-22 11:00:00,2024-02-16 14:22:54,1Wln_wppvv0wJNp0yqlKyToN2kokcacMK-nvYC0JSc0E,"[""Max Roser""]",It is easy to underestimate how much the world can change within a lifetime. Bringing to mind how dramatically the world has changed can help us see how different the world could be in a few years or decades.,2022-12-02 20:24:58,2023-10-11 08:43:29,https://ourworldindata.org/wp-content/uploads/2022/12/featured-image-Longterm-timeline-of-technology.png,{},"Our World in Data presents the data and research to make progress against the world’s largest problems. This article draws on data and research discussed in our entry on **[Artificial Intelligence](https://ourworldindata.org/artificial-intelligence)**. Technology can change the world in ways that are unimaginable, until they happen. Switching on an electric light would have been unimaginable for our medieval ancestors. In their childhood, our grandparents would have struggled to imagine a world connected by smartphones and the Internet. Similarly, it is hard for us to imagine the arrival of all those technologies that will fundamentally change the world we are used to. We can remind ourselves that our own future might look very different from the world today by looking back at how rapidly technology has changed our world in the past. That’s what this article is about. One insight I take away from this long-term perspective is how unusual our time is. Technological change was _extremely_ slow in the past – the technologies that our ancestors got used to in their childhood were still central to their lives in their old age. In stark contrast to those days, we live in a time of extraordinarily fast technological change. For recent generations, it was common for technologies that were unimaginable in their youth to become common later in life. ## The long-run perspective on technological change The big visualization offers a long-term perspective on the history of technology.{ref}The recent speed of technological change makes it difficult to picture the history of technology in one visualization. When you visualize this development on a linear timeline then most of the timeline is almost empty, while all the action is crammed into the right-corner:Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Technology can change the world in ways that are unimaginable, until they happen. Switching on an electric light would have been unimaginable for our medieval ancestors. In their childhood, our grandparents would have struggled to imagine a world connected by smartphones and the Internet.
\n\n\n\nSimilarly, it is hard for us to imagine the arrival of all those technologies that will fundamentally change the world we are used to.
\n\n\n\nWe can remind ourselves that our own future might look very different from the world today by looking back at how rapidly technology has changed our world in the past. That’s what this article is about.
\n\n\n\nOne insight I take away from this long-term perspective is how unusual our time is. Technological change was extremely slow in the past – the technologies that our ancestors got used to in their childhood were still central to their lives in their old age. In stark contrast to those days, we live in a time of extraordinarily fast technological change. For recent generations, it was common for technologies that were unimaginable in their youth to become common later in life.
\n\n\n\nThe big visualization offers a long-term perspective on the history of technology.{ref}The recent speed of technological change makes it difficult to picture the history of technology in one visualization. When you visualize this development on a linear timeline then most of the timeline is almost empty, while all the action is crammed into the right-corner:
\n\n\n\n![\""\""/](\""https://lh3.googleusercontent.com/cH9kcyrR2oopnW-KiDAAnAGVsMA-h0D9L-rPaG7lUr74vHI6xguNWEKZ6Gy047SOBN8lC5sXEbYLvaf_mWLDZ1K_LFoYij_K-e6gOwGjH3p5YGhRQPyWFMPnJAEDM6hhNIfaS2S94pYeaXdW4-risESc3T83GfHXLqf4mv4EHaHdBHw1kuXXTRW8Z-FPrQ\"")
In my large visualization here I tried to avoid this problem and instead show the long history of technology in a way that lets you see when each technological breakthrough happened, and how within the last millennia there was a continuous acceleration of technological change.{/ref}
\n\n\n\nThe timeline begins at the center of the spiral. The first use of stone tools, 3.4 million years ago, marks the beginning of this history of technology.{ref}It is of course difficult to assess when exactly the first stone tools were used.
\n\n\n\nThe research by McPherron et al (2010) suggested that it was at least 3.39 million years ago. This is based on two fossilized bones, found in Dikika in Ethiopia, which showed “stone-tool cut marks for flesh removal and percussion marks for marrow access”. These marks were interpreted as being caused by the consumption of meat and provide the first evidence that one of our ancestors, Australopithecus afarensis, used stone tools.
\n\n\n\nThe research by Harmand et al (2015) provided evidence for stone tool use in today’s Kenya 3.3 million years ago.
\n\n\n\nReferences:
\n\n\n\nMcPherron et al (2010) – Evidence for stone-tool-assisted consumption of animal tissues before 3.39 million years ago at Dikika, Ethiopia. Published in Nature.
\n\n\n\nHarmand et al (2015) – 3.3-million-year-old stone tools from Lomekwi 3, West Turkana, Kenya. Published in Nature.{/ref} Each turn of the spiral then represents 200,000 years of history. It took 2.4 million years – 12 turns of the spiral – for our ancestors to control fire and use it for cooking.{ref}Evidence for controlled fire use approximately 1 million years ago is provided by Berna et al (2012) Microstratigraphic evidence of in situ fire in the Acheulean strata of Wonderwerk Cave, Northern Cape province, South Africa, published in PNAS.
\n\n\n\nThe authors write: “The ability to control fire was a crucial turning point in human evolution, but the question of when hominins first developed this ability still remains. Here we show that micromorphological and Fourier transform infrared microspectroscopy (mFTIR) analyses of intact sediments at the site of Wonderwerk Cave, Northern Cape province, South Africa, provide unambiguous evidence—in the form of burned bone and ashed plant remains—that burning took place in the cave during the early Acheulean occupation, approximately 1.0 Ma. To the best of our knowledge, this is the earliest secure evidence for burning in an archaeological context.”{/ref}
\n\n\n\nTo be able to visualize the inventions in the more recent past – the last 12,000 years – I had to unroll the spiral. I needed more space to be able to show when agriculture, writing, and the wheel were invented. During this period, technological change was faster, but it was still relatively slow: several thousand years passed between each of these three inventions.
\n\n\n\nFrom 1800 onwards, I stretched out the timeline even further to show the many major inventions that rapidly followed one after the other.
\n\n\n\nThe long-term perspective that this chart provides makes it clear just how unusually fast technological change is in our time.
\n\n\n\nYou can use this visualization to see how technology developed in particular domains. Follow, for example, the history of communication: from writing, to paper, to the printing press, to the telegraph, the telephone, the radio, all the way to the Internet and smartphones.
\n\n\n\nOr follow the rapid development of human flight. In 1903, the Wright brothers took the first flight in human history (they were in the air for less than a minute), and just 66 years later, we landed on the moon. Many people saw both within their lifetimes: the first plane and the moon landing.
\n\n\n\nThis large visualization also highlights the wide range of technology’s impact on our lives. It includes extraordinarily beneficial innovations, such as the vaccine that allowed humanity to eradicate smallpox, and it includes terrible innovations, like the nuclear bombs that endanger the lives of all of us.
\n\n\n\nWhat will the next decades bring?
\n\n\n\nThe red timeline reaches up to the present and then continues in green into the future. Many children born today, even without any further increases in life expectancy, will live well into the 22nd century.
\n\n\n\nNew vaccines, progress in clean, low-carbon energy, better cancer treatments – a range of future innovations could very much improve our living conditions and the environment around us. But, as I argue in a series of articles, there is one technology that could even more profoundly change our world: artificial intelligence (AI).
\n\n\n\nOne reason why artificial intelligence is such an important innovation is that intelligence is the main driver of innovation itself. This fast-paced technological change could speed up even more if it’s not only driven by humanity’s intelligence, but artificial intelligence too. If this happens, the change that is currently stretched out over the course of decades might happen within very brief time spans of just a year. Possibly even faster.{ref}This is what authors like Holden Karnofsky called ‘Process for Automating Scientific and Technological Advancement’ or PASTA. There are some recent developments that go in this direction: DeepMind’s AlphaFold helped to make progress on one of the large problems in biology, and they have also developed an AI system that finds new algorithms that are relevant to building a more powerful AI.{/ref}
\n\n\n\nI think AI technology could have a fundamentally transformative impact on our world. In many ways, it is already changing our world, as I documented in this companion article. As this technology is becoming more capable in the years and decades to come, it can give immense power to those who control it (and it poses the risk that it could escape our control entirely).
\n\n\n\nSuch systems might seem hard to imagine today, but AI technology is advancing very fast. Many AI experts believe there is a real chance that human-level artificial intelligence will be developed within the next decades, as I documented in this article.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/Longterm-timeline-of-technology.png\"")
What is familiar to us today – photography, the radio, antibiotics, the Internet, or the International Space Station circling our planet – was unimaginable to our ancestors just a few generations ago. If your great-great-great grandparents could spend a week with you they would be blown away by your everyday life.
\n\n\n\nWhat I take away from this history is that I will likely see technologies in my lifetime that appear unimaginable to me today.
\n\n\n\nIn addition to this trend towards increasingly rapid innovation, there is a second long-run trend. Technology has become increasingly powerful. While our ancestors wielded stone tools, we are building globe-spanning AI systems and technologies that can edit our genes.
\n\n\n\nBecause of the immense power that technology gives those who control it, there is little that is as important as the question of which technologies get developed during our lifetimes. Therefore I think it is a mistake to leave the question about the future of technology to the technologists. Which technologies are controlled by whom is one of the most important political questions of our time, because of the enormous power that these technologies convey to those who control them.
\n\n\n\nWe all should strive to gain the knowledge we need to contribute to an intelligent debate about the world we want to live in. To a large part this means gaining the knowledge, and wisdom, on the question of which technologies we want.
\n\n\n\nOur new topic page on AI.
\n\nAcknowledgements: I would like to thank my colleagues Hannah Ritchie, Bastian Herre, Natasha Ahuja, Edouard Mathieu, Daniel Bachler, Charlie Giattino, and Pablo Rosado for their helpful comments to drafts of this essay and the visualization. Thanks also to Lizka Vaintrob and Ben Clifford for a conversation that initiated this visualization.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""It is easy to underestimate how much the world can change within a lifetime. Bringing to mind how dramatically the world has changed can help us see how different the world could be in a few years or decades."", ""protected"": false}, ""date_gmt"": ""2023-02-22T11:00:00"", ""modified"": ""2023-10-11T09:43:29"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Max Roser""], ""modified_gmt"": ""2023-10-11T08:43:29"", ""comment_status"": ""closed"", ""featured_media"": 54867, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/12/featured-image-Longterm-timeline-of-technology-150x86.png"", ""medium_large"": ""/app/uploads/2022/12/featured-image-Longterm-timeline-of-technology-768x440.png""}}" 54836,AI timelines: What do experts in artificial intelligence expect for the future?,ai-timelines,post,publish,"Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Artificial intelligence (AI) that surpasses our own intelligence sounds like the stuff from science-fiction books or films. What do experts in the field of AI research think about such scenarios? Do they dismiss these ideas as fantasy, or are they taking such prospects seriously?
A human-level AI would be a machine, or a network of machines, capable of carrying out the same range of tasks that we humans are capable of. It would be a machine that is “able to learn to do anything that a human can do”, as Norvig and Russell put it in their textbook on AI.{ref}Peter Norvig and Stuart Russell (2021) – Artificial Intelligence: A Modern Approach. Fourth edition. Published by Pearson.{/ref}
It would be able to choose actions that allow the machine to achieve its goals and then carry out those actions. It would be able to do the work of a translator, a doctor, an illustrator, a teacher, a therapist, a driver, or the work of an investor.
In recent years, several research teams contacted AI experts and asked them about their expectations for the future of machine intelligence. Such expert surveys are one of the pieces of information that we can rely on to form an idea of what the future of AI might look like.
The chart shows the answers of 352 experts. This is from the most recent study by Katja Grace and her colleagues, conducted in the summer of 2022.{ref}A total of 4,271 AI experts were contacted; 738 responded (a 17% rate), of which 352 provided complete answers to the human-level AI question.
It’s possible that the respondents were not representative of all the AI experts contacted – that is, that there was “sample bias.” There is not enough data to rule out all potential sources of sample bias. After all, we don’t know what the people who didn’t respond to the survey, or others who weren’t even contacted, believe about AI. However, there is evidence from similar surveys to suggest that at least some potential sources of bias are minimal.
In similar surveys (e.g., Zhang et al. 2022; Grace et al. 2018), the researchers compared the group of respondents with a randomly selected, similarly sized group of non-respondents to see if they differed on measurable demographic characteristics, such as where they were educated, their gender, how many citations they had, years in the field, etc.
In these similar surveys, the researchers found some differences between the respondents and non-respondents, but they were small. So while other, unmeasured sources of sample bias couldn’t be ruled out, large bias due to the demographic characteristics that were measured could be ruled out.{/ref}
Experts were asked when they believe there is a 50% chance that human-level AI exists.{ref}Much of the literature on AI timelines focuses on the 50% probability threshold. I think it would be valuable if this literature would additionally also focus on higher thresholds, say a probability of 80% for the development of a particular technology. In future updates of this article we will aim to broaden the focus and include such higher thresholds.{/ref} Human-level AI was defined as unaided machines being able to accomplish every task better and more cheaply than human workers. More information about the study can be found in the fold-out box at the end of this text.{ref}A discussion of the two most widely used concepts for thinking about the future of powerful AI systems – human-level AI and transformative AI – can be found in this companion article.{/ref}
Each vertical line in this chart represents the answer of one expert. The fact that there are such large differences in answers makes it clear that experts do not agree on how long it will take until such a system might be developed. A few believe that this level of technology will never be developed. Some think that it’s possible, but it will take a long time. And many believe that it will be developed within the next few decades.
As highlighted in the annotations, half of the experts gave a date before 2061, and 90% gave a date within the next 100 years.
Other surveys of AI experts come to similar conclusions. In the following visualization, I have added the timelines from two earlier surveys conducted in 2018 and 2019. It is helpful to look at different surveys, as they differ in how they asked the question and how they defined human-level AI. You can find more details about these studies at the end of this text.
In all three surveys, we see a large disagreement between experts and they also express large uncertainties about their own individual forecasts.{ref}The visualization shows when individual experts gave a 50% chance of human-level machine intelligence. The surveys also include data on when these experts gave much lower chances (e.g., ~10%) as well as much higher ones (~90%), and the spread between the respective dates is often considerable, expressing the AI experts range of their individual uncertainty. For example, the average across individual experts in the Zhang et al study gave a 10% chance of human-level machine intelligence by 2035, a 50% chance by 2060, and a 90% chance by 2105.{/ref}
Expert surveys are one piece of information to consider when we think about the future of AI, but we should not overstate the results of these surveys. Experts in a particular technology are not necessarily experts in making predictions about the future of that technology.
Experts in many fields do not have a good track record in making forecasts about their own field, as researchers including Barbara Mellers, Phil Tetlock, and others have shown.{ref}Mellers, B., Tetlock, P., & Arkes, H. R. (2019). Forecasting tournaments, epistemic humility and attitude depolarization. Cognition, 188, 19-26.
Tetlock, P. (2005) – Expert political judgment: How good is it? How can we know? Princeton, NJ: Princeton University Press
Philip E. Tetlock and Dan Gardner (2015) – Superforecasting: The Art and Science of Prediction.{/ref} The history of flight includes a striking example of such failure. Wilbur Wright is quoted as saying, ""I confess that in 1901, I said to my brother Orville that man would not fly for 50 years."" Two years later, ‘man’ was not only flying, but it was these very men who achieved the feat.{ref}Another example is Ernest Rutherford, father of nuclear physics, calling the possibility of harnessing nuclear energy ""moonshine."" The research paper by John Jenkin discusses why. John G. Jenkin (2011) – Atomic Energy is ‘‘Moonshine’’: What did Rutherford Really Mean?. Published in Physics in Perspective. DOI 10.1007/s00016-010-0038-1{/ref}
Additionally these studies often find large ‘framing effects’, two logically identical questions get answered in very different ways depending on how exactly the questions are worded.{ref}This is discussed in some more detail for the study by Grace et al. in the Appendix.{/ref}
What I do take away from these surveys however, is that the majority of AI experts take the prospect of very powerful AI technology seriously. It is not the case that AI researchers dismiss extremely powerful AI as mere fantasy.
The huge majority thinks that in the coming decades there is an even chance that we will see AI technology which will have a transformative impact on our world. While some have long timelines, many think it is possible that we have very little time before these technologies arrive. Across the three surveys more than half think that there is a 50% chance that a human-level AI would be developed before some point in the 2060s, a time well within the lifetime of today’s young people.
In the big visualization on AI timelines below, I have included the forecast by the Metaculus forecaster community.
The forecasters on the online platform Metaculus.com are not experts in AI but people who dedicate their energy to making good forecasts. Research on forecasting has documented that groups of people can assign surprisingly accurate probabilities to future events when given the right incentives and good feedback.{ref}See the previously cited literature on forecasting by Barbara Mellers, Phil Tetlock, and others.{/ref} To receive this feedback, the online community at Metaculus tracks how well they perform in their forecasts.
What does this group of forecasters expect for the future of AI?
At the time of writing, in November 2022, the forecasters believe that there is a 50/50-chance for an ‘Artificial General Intelligence’ to be ‘devised, tested, and publicly announced’ by the year 2040, less than 20 years from now.
On their page about this specific question, you can find the precise definition of the AI system in question, how the timeline of their forecasts has changed, and the arguments of individual forecasters for how they arrived at their predictions.{ref}There are two other relevant questions on Metaculus. The first one asks for the date when weakly General AI will be publicly known. And the second one is asking for the probability of ‘human/machine intelligence parity’ by 2040.{/ref}
The timelines of the Metaculus community have become much shorter recently. The expected timelines have shortened by about a decade in the spring of 2022, when several impressive AI breakthroughs happened faster than many had anticipated.{ref}Metaculus’s community prediction fell from the year 2058 in March 2022 to the year 2040 in July 2022.{/ref}
The last shown forecast stems from the research by Ajeya Cotra, who works for the nonprofit Open Philanthropy.{ref}Her research was announced in various places, including the AI Alignment Forum: Ajeya Cotra (2020) – Draft report on AI timelines. As far as I know the report itself always remained a ‘draft report’ and was published here on Google Docs.
In 2022 Ajeya Cotra published a Two-year update on my personal AI timelines.{/ref} In 2020 she published a detailed and influential study asking when the world will see transformative AI. Her timeline is not based on surveys, but on the study of long-term trends in the computation used to train AI systems. I present and discuss the long-run trends in training computation in this companion article.
Cotra estimated that there is a 50% chance that a transformative AI system will become possible and affordable by the year 2050. This is her central estimate in her “median scenario.” Cotra emphasizes that there are substantial uncertainties around this median scenario, and also explored two other, more extreme, scenarios. The timelines for these two scenarios – her “most aggressive plausible” scenario and her “most conservative plausible” scenario – are also shown in the visualization. The span from 2040 to 2090 in Cotra’s “plausible” forecasts highlights that she believes that the uncertainty is large.
The visualization also shows that Cotra updated her forecast two years after its initial publication. In 2022 Cotra published an update in which she shortened her median timeline by a full ten years.{ref}Ajeya Cotra’s Two-year update on my personal AI timelines.{/ref}
It is important to note that the definitions of the AI systems in question differ very much across these various studies. For example, the system that Cotra speaks about would have a much more transformative impact on the world than the system that the Metaculus forecasters focus on. More details can be found in the appendix and within the respective studies.
The visualization shows the forecasts of 1128 people – 812 individual AI experts, the aggregated estimates of 315 forecasters from the Metaculus platform, and the findings of the detailed study by Ajeya Cotra.
There are two big takeaways from these forecasts on AI timelines:
The public discourse and the decision-making at major institutions have not caught up with these prospects. In discussions on the future of our world – from the future of our climate, to the future of our economies, to the future of our political institutions – the prospect of transformative AI is rarely central to the conversation. Often it is not mentioned at all, not even in a footnote.
We seem to be in a situation where most people hardly think about the future of artificial intelligence, while the few who dedicate their attention to it find it plausible that one of the biggest transformations in humanity’s history is likely to happen within our lifetimes.
Acknowledgements: I would like to thank my colleagues Natasha Ahuja, Daniel Bachler, Bastian Herre, Edouard Mathieu, Esteban Ortiz-Ospina and Hannah Ritchie for their helpful comments to drafts of this essay.
And I would like to thank my colleague Charlie Giattino who calculated the timelines for individual experts based on the data from the three survey studies and supported the work on this essay. Charlie is also one of the authors of the cited study by Zhang et al. on timelines of AI experts.
The three cited AI experts surveys are:
The surveys were conducted during the following times:
The surveys differ in how the question was asked and how the AI system in question was defined. In the following sections we discuss this in detail for all cited studies.
Survey respondents were given the following text regarding the definition of high-level machine intelligence:
“The following questions ask about ‘high-level machine intelligence’ (HLMI). Say we have ‘high-level machine intelligence’ when unaided machines can accomplish every task better and more cheaply than human workers. Ignore aspects of tasks for which being a human is intrinsically advantageous, e.g., being accepted as a jury member. Think feasibility, not adoption. For the purposes of this question, assume that human scientific activity continues without major negative disruption.”
Each respondent was randomly assigned to give their forecasts under one of two different framings: “fixed-probability” and “fixed-years.”
Those in the fixed-probability framing were asked, “How many years until you expect: A 10% probability of HLMI existing? A 50% probability of HLMI existing? A 90% probability of HLMI existing?” They responded by giving a number of years from the day they took the survey.
Those in the fixed-years framing were asked, “How likely is it that HLMI exists: In 10 years? In 20 years? In 40 years?” They responded by giving a probability of that happening.
Several studies have shown that the framing affects respondents’ timelines, with the fixed-years framing leading to longer timelines (i.e., that HLMI is further in the future). For example, in the previous edition of this survey (which asked identical questions), respondents who got the fixed-years framing gave a 50% chance of HLMI by 2068; those who got fixed-probability gave the year 2054.{ref}Grace et al (2018) Viewpoint: When Will AI Exceed Human Performance? Evidence from AI Experts. Journal of Artificial Intelligence Research. We read both of these numbers of the chart in this publication, these years are not directly reported.{/ref} The framing results from the 2022 edition of the survey have not yet been published.
In addition to this framing effect, there is a larger effect driven by how the concept of HLMI is defined. We can see this in the results from the previous edition of this survey (the result from the 2022 survey hasn’t yet been published). For respondents who were given the HLMI definition above, the average forecast for a 50% chance of HLMI was 2061. A small subset of respondents was instead given another, logically similar question that asked about the full automation of labor; their average forecast for a 50% probability was 2138, a full 77 years later than the first group.
The full automation of labor group was asked: “Say an occupation becomes fully automatable when unaided machines can accomplish it better and more cheaply than human workers. Ignore aspects of occupations for which being a human is intrinsically advantageous, e.g., being accepted as a jury member. Think feasibility, not adoption. Say we have reached ‘full automation of labor’ when all occupations are fully automatable. That is, when for any occupation, machines could be built to carry out the task better and more cheaply than human workers.” This question was asked under both the fixed-probability and fixed-years framings.
Survey respondents were given the following definition of human-level machine intelligence: “Human-level machine intelligence (HLMI) is reached when machines are collectively able to perform almost all tasks (>90% of all tasks) that are economically relevant better than the median human paid to do that task in 2019. You should ignore tasks that are legally or culturally restricted to humans, such as serving on a jury.”
“Economically relevant” tasks were defined as those included in the Occupational Information Network (O*NET) database. O*NET is a widely used dataset of tasks carried out across a wide range of occupations.
As in Grace et al 2022, each survey respondent was randomly assigned to give their forecasts under one of two different framings: “fixed-probability” and “fixed-years.” As was found before, the fixed-years framing resulted in longer timelines on average: the year 2070 for a 50% chance of HLMI, compared to 2050 under the fixed-probability framing.
Survey respondents were asked the following: “These questions will ask your opinion of future AI progress with regard to human tasks. We define human tasks as all unique tasks that humans are currently paid to do. We consider human tasks as different from jobs in that an algorithm may be able to replace humans at some portion of tasks a job requires while not being able to replace humans for all of the job requirements. For example, an AI system(s) may not replace a lawyer entirely but may be able to accomplish 50% of the tasks a lawyer typically performs. In how many years do you expect AI systems to collectively be able to accomplish 99% of human tasks at or above the level of a typical human? Think feasibility.”
We show the results using this definition of AI in the chart, as we judged this definition to be most comparable to the other studies included in the chart.
In addition to this definition, respondents were asked about AI systems that are able to collectively accomplish 50% and 90% of human tasks, as well as “broadly capable AI systems” that are able to accomplish 90% and 99% of human tasks.
All respondents in this survey received a fixed-probability framing.
Cotra’s overall aim was to estimate when we might expect “transformative artificial intelligence” (TAI), defined as “ ‘software’... that has at least as profound an impact on the world’s trajectory as the Industrial Revolution did.”
Cotra focused on “a relatively concrete and easy-to-picture way that TAI could manifest: as a single computer program which performs a large enough diversity of intellectual labor at a high enough level of performance that it alone can drive a transition similar to the Industrial Revolution.”
One intuitive example of such a program is the ‘virtual professional’, “a model that can do roughly everything economically productive that an intelligent and educated human could do remotely from a computer connected to the internet at a hundred-fold speedup, for costs similar to or lower than the costs of employing such a human.”
When might we expect something like a virtual professional to exist?
To answer this, Cotra first estimated the amount of computation that would be required to train such a system using the machine learning architectures and algorithms available to researchers in 2020. She then estimated when that amount of computation would be available at a low enough cost based on extrapolating past trends.
The estimate of training computation relies on an estimate of the amount of computation performed by the human brain each second, combined with different hypotheses for how much training would be required to reach a high enough level of capability.
For example, the “lifetime anchor” hypothesis estimates the total computation performed by the human brain up to age ~32.
Each aspect of these estimates comes with a very high degree of uncertainty. Cotra writes: “The question of whether there is a sensible notion of ‘brain computation’ that can be measured in FLOP/s—and if so, what range of numerical estimates for brain FLOP/s would be reasonable—is conceptually fraught and empirically murky.”
For anyone who is interested in the question of future AI, the study of Cotra is very much worth reading in detail. She lays out good and transparent reasons for her estimates and communicates her reasoning in great detail.
Her research was announced in various places, including the AI Alignment Forum: Ajeya Cotra (2020) – Draft report on AI timelines. As far as I know the report itself always remained a ‘draft report’ and was published here on Google Docs (it is not uncommon in the field of AI research that articles get published in non-standard ways). In 2022 Ajeya Cotra published a Two-year update on my personal AI timelines.
A very different kind of forecast that is also relevant here is the work of David Roodman. In his article Modeling the Human Trajectory he studies the history of global economic output to think about the future. He asks whether it is plausible to see economic growth that could be considered ‘transformative’ – an annual growth rate of the world economy higher than 30% – within this century. One of his conclusions is that ""if the patterns of long-term history continue, some sort of economic explosion will take place again, the most plausible channel being AI.”
And another very different kind of forecast is Tom Davidson’s Report on Semi-informative Priors published in 2021.
Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Artificial intelligence (AI) that surpasses our own intelligence sounds like the stuff from science-fiction books or films. What do experts in the field of AI research think about such scenarios? Do they dismiss these ideas as fantasy, or are they taking such prospects seriously?
\n\n\n\nA human-level AI would be a machine, or a network of machines, capable of carrying out the same range of tasks that we humans are capable of. It would be a machine that is “able to learn to do anything that a human can do”, as Norvig and Russell put it in their textbook on AI.{ref}Peter Norvig and Stuart Russell (2021) – Artificial Intelligence: A Modern Approach. Fourth edition. Published by Pearson.{/ref}
\n\n\n\nIt would be able to choose actions that allow the machine to achieve its goals and then carry out those actions. It would be able to do the work of a translator, a doctor, an illustrator, a teacher, a therapist, a driver, or the work of an investor.
\n\n\n\nIn recent years, several research teams contacted AI experts and asked them about their expectations for the future of machine intelligence. Such expert surveys are one of the pieces of information that we can rely on to form an idea of what the future of AI might look like.
\n\n\n\nThe chart shows the answers of 352 experts. This is from the most recent study by Katja Grace and her colleagues, conducted in the summer of 2022.{ref}A total of 4,271 AI experts were contacted; 738 responded (a 17% rate), of which 352 provided complete answers to the human-level AI question.
It’s possible that the respondents were not representative of all the AI experts contacted – that is, that there was “sample bias.” There is not enough data to rule out all potential sources of sample bias. After all, we don’t know what the people who didn’t respond to the survey, or others who weren’t even contacted, believe about AI. However, there is evidence from similar surveys to suggest that at least some potential sources of bias are minimal.
In similar surveys (e.g., Zhang et al. 2022; Grace et al. 2018), the researchers compared the group of respondents with a randomly selected, similarly sized group of non-respondents to see if they differed on measurable demographic characteristics, such as where they were educated, their gender, how many citations they had, years in the field, etc.
\n\n\n\nIn these similar surveys, the researchers found some differences between the respondents and non-respondents, but they were small. So while other, unmeasured sources of sample bias couldn’t be ruled out, large bias due to the demographic characteristics that were measured could be ruled out.{/ref}
\n\n\n\nExperts were asked when they believe there is a 50% chance that human-level AI exists.{ref}Much of the literature on AI timelines focuses on the 50% probability threshold. I think it would be valuable if this literature would additionally also focus on higher thresholds, say a probability of 80% for the development of a particular technology. In future updates of this article we will aim to broaden the focus and include such higher thresholds.{/ref} Human-level AI was defined as unaided machines being able to accomplish every task better and more cheaply than human workers. More information about the study can be found in the fold-out box at the end of this text.{ref}A discussion of the two most widely used concepts for thinking about the future of powerful AI systems – human-level AI and transformative AI – can be found in this companion article.{/ref}
\n\n\n\nEach vertical line in this chart represents the answer of one expert. The fact that there are such large differences in answers makes it clear that experts do not agree on how long it will take until such a system might be developed. A few believe that this level of technology will never be developed. Some think that it’s possible, but it will take a long time. And many believe that it will be developed within the next few decades.
\n\n\n\nAs highlighted in the annotations, half of the experts gave a date before 2061, and 90% gave a date within the next 100 years.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/When-do-experts-expect-Artificial-General-Intelligence-grace.png\"")
Other surveys of AI experts come to similar conclusions. In the following visualization, I have added the timelines from two earlier surveys conducted in 2018 and 2019. It is helpful to look at different surveys, as they differ in how they asked the question and how they defined human-level AI. You can find more details about these studies at the end of this text.
\n\n\n\nIn all three surveys, we see a large disagreement between experts and they also express large uncertainties about their own individual forecasts.{ref}The visualization shows when individual experts gave a 50% chance of human-level machine intelligence. The surveys also include data on when these experts gave much lower chances (e.g., ~10%) as well as much higher ones (~90%), and the spread between the respective dates is often considerable, expressing the AI experts range of their individual uncertainty. For example, the average across individual experts in the Zhang et al study gave a 10% chance of human-level machine intelligence by 2035, a 50% chance by 2060, and a 90% chance by 2105.{/ref}
\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/When-do-experts-expect-Artificial-General-Intelligence-surveys.png\"")
Expert surveys are one piece of information to consider when we think about the future of AI, but we should not overstate the results of these surveys. Experts in a particular technology are not necessarily experts in making predictions about the future of that technology.
\n\n\n\nExperts in many fields do not have a good track record in making forecasts about their own field, as researchers including Barbara Mellers, Phil Tetlock, and others have shown.{ref}Mellers, B., Tetlock, P., & Arkes, H. R. (2019). Forecasting tournaments, epistemic humility and attitude depolarization. Cognition, 188, 19-26.
\n\n\n\nTetlock, P. (2005) – Expert political judgment: How good is it? How can we know? Princeton, NJ: Princeton University Press
\n\n\n\nPhilip E. Tetlock and Dan Gardner (2015) – Superforecasting: The Art and Science of Prediction.{/ref} The history of flight includes a striking example of such failure. Wilbur Wright is quoted as saying, “I confess that in 1901, I said to my brother Orville that man would not fly for 50 years.” Two years later, ‘man’ was not only flying, but it was these very men who achieved the feat.{ref}Another example is Ernest Rutherford, father of nuclear physics, calling the possibility of harnessing nuclear energy “moonshine.” The research paper by John Jenkin discusses why. John G. Jenkin (2011) – Atomic Energy is ‘‘Moonshine’’: What did Rutherford Really Mean?. Published in Physics in Perspective. DOI 10.1007/s00016-010-0038-1{/ref}
\n\n\n\nAdditionally these studies often find large ‘framing effects’, two logically identical questions get answered in very different ways depending on how exactly the questions are worded.{ref}This is discussed in some more detail for the study by Grace et al. in the Appendix.{/ref}
\n\n\n\nWhat I do take away from these surveys however, is that the majority of AI experts take the prospect of very powerful AI technology seriously. It is not the case that AI researchers dismiss extremely powerful AI as mere fantasy.
\n\n\n\nThe huge majority thinks that in the coming decades there is an even chance that we will see AI technology which will have a transformative impact on our world. While some have long timelines, many think it is possible that we have very little time before these technologies arrive. Across the three surveys more than half think that there is a 50% chance that a human-level AI would be developed before some point in the 2060s, a time well within the lifetime of today’s young people.
\n\n\n\nIn the big visualization on AI timelines below, I have included the forecast by the Metaculus forecaster community.
\n\n\n\nThe forecasters on the online platform Metaculus.com are not experts in AI but people who dedicate their energy to making good forecasts. Research on forecasting has documented that groups of people can assign surprisingly accurate probabilities to future events when given the right incentives and good feedback.{ref}See the previously cited literature on forecasting by Barbara Mellers, Phil Tetlock, and others.{/ref} To receive this feedback, the online community at Metaculus tracks how well they perform in their forecasts.
\n\n\n\nWhat does this group of forecasters expect for the future of AI?
\n\n\n\nAt the time of writing, in November 2022, the forecasters believe that there is a 50/50-chance for an ‘Artificial General Intelligence’ to be ‘devised, tested, and publicly announced’ by the year 2040, less than 20 years from now.
\n\n\n\nOn their page about this specific question, you can find the precise definition of the AI system in question, how the timeline of their forecasts has changed, and the arguments of individual forecasters for how they arrived at their predictions.{ref}There are two other relevant questions on Metaculus. The first one asks for the date when weakly General AI will be publicly known. And the second one is asking for the probability of ‘human/machine intelligence parity’ by 2040.{/ref}
\n\n\n\nThe timelines of the Metaculus community have become much shorter recently. The expected timelines have shortened by about a decade in the spring of 2022, when several impressive AI breakthroughs happened faster than many had anticipated.{ref}Metaculus’s community prediction fell from the year 2058 in March 2022 to the year 2040 in July 2022.{/ref}
\n\n\n\nThe last shown forecast stems from the research by Ajeya Cotra, who works for the nonprofit Open Philanthropy.{ref}Her research was announced in various places, including the AI Alignment Forum: Ajeya Cotra (2020) – Draft report on AI timelines. As far as I know the report itself always remained a ‘draft report’ and was published here on Google Docs.
\n\n\n\nIn 2022 Ajeya Cotra published a Two-year update on my personal AI timelines.{/ref} In 2020 she published a detailed and influential study asking when the world will see transformative AI. Her timeline is not based on surveys, but on the study of long-term trends in the computation used to train AI systems. I present and discuss the long-run trends in training computation in this companion article.
\n\n\n\nCotra estimated that there is a 50% chance that a transformative AI system will become possible and affordable by the year 2050. This is her central estimate in her “median scenario.” Cotra emphasizes that there are substantial uncertainties around this median scenario, and also explored two other, more extreme, scenarios. The timelines for these two scenarios – her “most aggressive plausible” scenario and her “most conservative plausible” scenario – are also shown in the visualization. The span from 2040 to 2090 in Cotra’s “plausible” forecasts highlights that she believes that the uncertainty is large.
\n\n\n\nThe visualization also shows that Cotra updated her forecast two years after its initial publication. In 2022 Cotra published an update in which she shortened her median timeline by a full ten years.{ref}Ajeya Cotra’s Two-year update on my personal AI timelines.{/ref}
\n\n\n\nIt is important to note that the definitions of the AI systems in question differ very much across these various studies. For example, the system that Cotra speaks about would have a much more transformative impact on the world than the system that the Metaculus forecasters focus on. More details can be found in the appendix and within the respective studies.
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/When-do-experts-expect-Artificial-General-Intelligence-big.png\"")
The visualization shows the forecasts of 1128 people – 812 individual AI experts, the aggregated estimates of 315 forecasters from the Metaculus platform, and the findings of the detailed study by Ajeya Cotra.
\n\n\n\nThere are two big takeaways from these forecasts on AI timelines:
\n\n\n\nThe public discourse and the decision-making at major institutions have not caught up with these prospects. In discussions on the future of our world – from the future of our climate, to the future of our economies, to the future of our political institutions – the prospect of transformative AI is rarely central to the conversation. Often it is not mentioned at all, not even in a footnote.
\n\n\n\nWe seem to be in a situation where most people hardly think about the future of artificial intelligence, while the few who dedicate their attention to it find it plausible that one of the biggest transformations in humanity’s history is likely to happen within our lifetimes.
\n\n\n\n\n\n\n\n\n\n\n\nAcknowledgements: I would like to thank my colleagues Natasha Ahuja, Daniel Bachler, Bastian Herre, Edouard Mathieu, Esteban Ortiz-Ospina and Hannah Ritchie for their helpful comments to drafts of this essay.
\n\n\n\nAnd I would like to thank my colleague Charlie Giattino who calculated the timelines for individual experts based on the data from the three survey studies and supported the work on this essay. Charlie is also one of the authors of the cited study by Zhang et al. on timelines of AI experts.
\n\n\n\nThe three cited AI experts surveys are:
\n\n\n\nThe surveys were conducted during the following times:
\n\n\n\nThe surveys differ in how the question was asked and how the AI system in question was defined. In the following sections we discuss this in detail for all cited studies.
\n\n\n\nSurvey respondents were given the following text regarding the definition of high-level machine intelligence:
\n\n\n\n“The following questions ask about ‘high-level machine intelligence’ (HLMI). Say we have ‘high-level machine intelligence’ when unaided machines can accomplish every task better and more cheaply than human workers. Ignore aspects of tasks for which being a human is intrinsically advantageous, e.g., being accepted as a jury member. Think feasibility, not adoption. For the purposes of this question, assume that human scientific activity continues without major negative disruption.”
\n\n\n\nEach respondent was randomly assigned to give their forecasts under one of two different framings: “fixed-probability” and “fixed-years.”
\n\n\n\nThose in the fixed-probability framing were asked, “How many years until you expect: A 10% probability of HLMI existing? A 50% probability of HLMI existing? A 90% probability of HLMI existing?” They responded by giving a number of years from the day they took the survey.
\n\n\n\nThose in the fixed-years framing were asked, “How likely is it that HLMI exists: In 10 years? In 20 years? In 40 years?” They responded by giving a probability of that happening.
\n\n\n\nSeveral studies have shown that the framing affects respondents’ timelines, with the fixed-years framing leading to longer timelines (i.e., that HLMI is further in the future). For example, in the previous edition of this survey (which asked identical questions), respondents who got the fixed-years framing gave a 50% chance of HLMI by 2068; those who got fixed-probability gave the year 2054.{ref}Grace et al (2018) Viewpoint: When Will AI Exceed Human Performance? Evidence from AI Experts. Journal of Artificial Intelligence Research. We read both of these numbers of the chart in this publication, these years are not directly reported.{/ref} The framing results from the 2022 edition of the survey have not yet been published.
\n\n\n\nIn addition to this framing effect, there is a larger effect driven by how the concept of HLMI is defined. We can see this in the results from the previous edition of this survey (the result from the 2022 survey hasn’t yet been published). For respondents who were given the HLMI definition above, the average forecast for a 50% chance of HLMI was 2061. A small subset of respondents was instead given another, logically similar question that asked about the full automation of labor; their average forecast for a 50% probability was 2138, a full 77 years later than the first group.
\n\n\n\nThe full automation of labor group was asked: “Say an occupation becomes fully automatable when unaided machines can accomplish it better and more cheaply than human workers. Ignore aspects of occupations for which being a human is intrinsically advantageous, e.g., being accepted as a jury member. Think feasibility, not adoption. Say we have reached ‘full automation of labor’ when all occupations are fully automatable. That is, when for any occupation, machines could be built to carry out the task better and more cheaply than human workers.” This question was asked under both the fixed-probability and fixed-years framings.
\n\n\n\nSurvey respondents were given the following definition of human-level machine intelligence: “Human-level machine intelligence (HLMI) is reached when machines are collectively able to perform almost all tasks (>90% of all tasks) that are economically relevant better than the median human paid to do that task in 2019. You should ignore tasks that are legally or culturally restricted to humans, such as serving on a jury.”
\n\n\n\n“Economically relevant” tasks were defined as those included in the Occupational Information Network (O*NET) database. O*NET is a widely used dataset of tasks carried out across a wide range of occupations.
\n\n\n\nAs in Grace et al 2022, each survey respondent was randomly assigned to give their forecasts under one of two different framings: “fixed-probability” and “fixed-years.” As was found before, the fixed-years framing resulted in longer timelines on average: the year 2070 for a 50% chance of HLMI, compared to 2050 under the fixed-probability framing.
\n\n\n\nSurvey respondents were asked the following: “These questions will ask your opinion of future AI progress with regard to human tasks. We define human tasks as all unique tasks that humans are currently paid to do. We consider human tasks as different from jobs in that an algorithm may be able to replace humans at some portion of tasks a job requires while not being able to replace humans for all of the job requirements. For example, an AI system(s) may not replace a lawyer entirely but may be able to accomplish 50% of the tasks a lawyer typically performs. In how many years do you expect AI systems to collectively be able to accomplish 99% of human tasks at or above the level of a typical human? Think feasibility.”
\n\n\n\nWe show the results using this definition of AI in the chart, as we judged this definition to be most comparable to the other studies included in the chart.
\n\n\n\nIn addition to this definition, respondents were asked about AI systems that are able to collectively accomplish 50% and 90% of human tasks, as well as “broadly capable AI systems” that are able to accomplish 90% and 99% of human tasks.
\n\n\n\nAll respondents in this survey received a fixed-probability framing.
\n\n\n\nCotra’s overall aim was to estimate when we might expect “transformative artificial intelligence” (TAI), defined as “ ‘software’… that has at least as profound an impact on the world’s trajectory as the Industrial Revolution did.”
\n\n\n\nCotra focused on “a relatively concrete and easy-to-picture way that TAI could manifest: as a single computer program which performs a large enough diversity of intellectual labor at a high enough level of performance that it alone can drive a transition similar to the Industrial Revolution.”
\n\n\n\nOne intuitive example of such a program is the ‘virtual professional’, “a model that can do roughly everything economically productive that an intelligent and educated human could do remotely from a computer connected to the internet at a hundred-fold speedup, for costs similar to or lower than the costs of employing such a human.”
\n\n\n\nWhen might we expect something like a virtual professional to exist?
\n\n\n\nTo answer this, Cotra first estimated the amount of computation that would be required to train such a system using the machine learning architectures and algorithms available to researchers in 2020. She then estimated when that amount of computation would be available at a low enough cost based on extrapolating past trends.
\n\n\n\nThe estimate of training computation relies on an estimate of the amount of computation performed by the human brain each second, combined with different hypotheses for how much training would be required to reach a high enough level of capability.
\n\n\n\nFor example, the “lifetime anchor” hypothesis estimates the total computation performed by the human brain up to age ~32.
\n\n\n\nEach aspect of these estimates comes with a very high degree of uncertainty. Cotra writes: “The question of whether there is a sensible notion of ‘brain computation’ that can be measured in FLOP/s—and if so, what range of numerical estimates for brain FLOP/s would be reasonable—is conceptually fraught and empirically murky.”
\n\n\n\nFor anyone who is interested in the question of future AI, the study of Cotra is very much worth reading in detail. She lays out good and transparent reasons for her estimates and communicates her reasoning in great detail.
\n\n\n\nHer research was announced in various places, including the AI Alignment Forum: Ajeya Cotra (2020) – Draft report on AI timelines. As far as I know the report itself always remained a ‘draft report’ and was published here on Google Docs (it is not uncommon in the field of AI research that articles get published in non-standard ways). In 2022 Ajeya Cotra published a Two-year update on my personal AI timelines.
\n\n\n\nA very different kind of forecast that is also relevant here is the work of David Roodman. In his article Modeling the Human Trajectory he studies the history of global economic output to think about the future. He asks whether it is plausible to see economic growth that could be considered ‘transformative’ – an annual growth rate of the world economy higher than 30% – within this century. One of his conclusions is that “if the patterns of long-term history continue, some sort of economic explosion will take place again, the most plausible channel being AI.”
\n\n\n\nAnd another very different kind of forecast is Tom Davidson’s Report on Semi-informative Priors published in 2021.
\nOur World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Artificial intelligence (AI) technology has steadily become more powerful over the course of the last decades and in recent years it has entered our world in many different domains. In a companion article – the brief history of artificial intelligence – I document this development.
This was achieved despite having relatively few resources. Until recently, investments in terms of capital and scientific efforts were small. In this article I highlight that this has very much changed in recent years. Corporate investment has increased and the scientific field has grown in size.
Given how rapidly AI developed in the past, despite the limited resources, this should make us expect AI technology to continue to become more powerful in the coming decades.
The first chart looks at corporate investment over time.{ref}It is not obvious how to adjust a time-series of AI investments for inflation, and we debated it at some length within the team.
Reporting it in nominal prices (as some do) means it makes little sense to compare observations across time and is therefore not very helpful. To make comparisons over time possible one has to take into account that prices change.
It would be straightforward to adjust the time series for price changes if we knew the prices of the specific goods and services that these investments purchased. This would make it possible to calculate a volume measure of AI investments, and it would tell us how much these investments bought. But such a metric is not available.
(While a comprehensive price index is not available, we know that the cost for some crucial AI technology has fallen rapidly in price. For example, it has become much cheaper to train an AI system.)
In the absence of a comprehensive price index that captures the price of AI-specific goods and services, one has to rely on one of the available metrics for the price of a bundle of goods and services. In the end we decided to use the US Consumer Price Index (CPI).
The US CPI does not provide us with a volume measure of AI goods and services, but it does capture the opportunity costs of these investments. The inflation adjustment of this time series of AI investments therefore lets us understand the size of these investments relative to whatever else these sums of money could have purchased.{/ref}
Until recently, private sector investment was relatively low. But, especially from 2018 onwards, it has increased rapidly.
Investments in 2021 were about 30-times larger than just eight years earlier.
The four following visualizations show that the trends for AI research efforts and the labor market are similar.
Just a decade ago the field was much smaller than it is today. Since then the number of research publications on artificial intelligence has doubled, and AI conferences have become much larger events. This is what the first two charts show.
The third chart documents the increasing importance of AI skills in the labor market. The last chart shows the specialty of computer science PhD students in the US. With an increasing number of students in the field, the number of researchers and research output is set to increase further in years to come.
Additionally, AI itself is contributing to the development of AI: Researchers are beginning to find ways for AI itself to contribute to the development of AI.{ref}One early example for this is Neural architecture search.{/ref}
The developments in the past happened despite the fact that funding and brainpower dedicated to AI were quite limited. As these charts have shown, this has changed. Across a range of metrics, the resources dedicated to AI development have increased substantially.
The fact that the field advanced with relatively small resources, and now has much larger resources at its disposal – leading to rapid advances in the last few years – is one reason why I expect AI technology to continue to develop rapidly and to exert a larger and larger influence on our world.
Continue reading on Our World in Data:
AI timelines: What do experts in artificial intelligence expect for the future?
Acknowledgements: I would like to thank Julia Broden, Charlie Giattino, Joe Hasell, and Edouard Mathieu for their helpful comments to drafts of this essay and the visualizations.
Appendix
Additional charts on the rise of investments into AI technology
","{""id"": ""wp-54814"", ""slug"": ""ai-investments"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Our World in Data presents the data and research to make progress against the world’s largest problems."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""This article draws on data and research discussed in our entry on "", ""spanType"": ""span-simple-text""}, {""children"": [{""url"": ""https://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""Artificial Intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""spanType"": ""span-bold""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Artificial intelligence (AI) technology has steadily become more powerful over the course of the last decades and in recent years it has entered our world in many different domains. In a companion article – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/brief-history-of-ai"", ""children"": [{""text"": ""the brief history of artificial intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" – I document this development."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This was achieved despite having relatively few resources. Until recently, investments in terms of capital and scientific efforts were small. In this article I highlight that this has very much changed in recent years. Corporate investment has increased and the scientific field has grown in size."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Given how rapidly AI developed in the past, despite the limited resources, this should make us expect AI technology to continue to become more powerful in the coming decades. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Investments into AI"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The first chart looks at corporate investment over time.{ref}It is not obvious how to adjust a time-series of AI investments for inflation, and we debated it at some length within the team. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Reporting it in nominal prices (as some do) means it makes little sense to compare observations across time and is therefore not very helpful. To make comparisons over time possible one has to take into account that prices change."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It would be straightforward to adjust the time series for price changes if we knew the prices of the specific goods and services that these investments purchased. This would make it possible to calculate a volume measure of AI investments, and it would tell us how much these investments bought. But such a metric is not available."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""(While a comprehensive price index is not available, we know that the cost for some crucial AI technology has fallen rapidly in price. For example, it has become much cheaper to train an AI system.)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the absence of a comprehensive price index that captures the price of AI-specific goods and services, one has to rely on one of the available metrics for the price of a bundle of goods and services. In the end we decided to use the US Consumer Price Index (CPI)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The US CPI does not provide us with a volume measure of AI goods and services, but it does capture the opportunity costs of these investments. The inflation adjustment of this time series of AI investments therefore lets us understand the size of these investments relative to whatever else these sums of money could have purchased.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Until recently, private sector investment was relatively low. But, especially from 2018 onwards, it has increased rapidly. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Investments in 2021 were about 30-times larger than just eight years earlier."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/corporate-investment-in-artificial-intelligence-by-type"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""Research and the labor market"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The four following visualizations show that the trends for AI research efforts and the labor market are similar. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Just a decade ago the field was much smaller than it is today. Since then the number of research publications on artificial intelligence has doubled, and AI conferences have become much larger events. This is what the first two charts show."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The third chart documents the increasing importance of AI skills in the labor market. The last chart shows the specialty of computer science PhD students in the US. With an increasing number of students in the field, the number of researchers and research output is set to increase further in years to come. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Additionally, AI itself is contributing to the development of AI: Researchers are beginning to find ways for AI itself to contribute to the development of AI.{ref}One early example for this is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Neural_architecture_search"", ""children"": [{""text"": ""Neural architecture search"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/annual-scholarly-publications-on-artificial-intelligence"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/attendance-major-artificial-intelligence-conferences"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/share-artificial-intelligence-job-postings"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/number-new-cs-phds-us-by-specialty"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Conclusion"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The developments in the past happened despite the fact that funding and brainpower dedicated to AI were quite limited. As these charts have shown, this has changed. Across a range of metrics, the resources dedicated to AI development have increased substantially."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The fact that the field advanced with relatively small resources, and now has much larger resources at its disposal – leading to "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/brief-history-of-ai"", ""children"": [{""text"": ""rapid advances"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in the last few years – is one reason why I expect AI technology to continue to develop rapidly and to exert a larger and larger influence on our world. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Continue reading on Our World in Data:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""spanType"": ""span-newline""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""AI timelines: What do experts in artificial intelligence expect for the future?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Acknowledgements:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" I would like to thank Julia Broden, Charlie Giattino, Joe Hasell, and Edouard Mathieu for their helpful comments to drafts of this essay and the visualizations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Appendix"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Additional charts on the rise of investments into AI technology"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Artificial intelligence has advanced despite having few resources dedicated to its development – now investments have increased substantially"", ""authors"": [""Max Roser""], ""excerpt"": ""In the past, relatively few researchers were working on AI technology and there was little commercial interest and funding. Now, the available resources have increased substantially. We should expect that the field continues to advance rapidly."", ""dateline"": ""March 29, 2023"", ""subtitle"": ""In the past, relatively few researchers were working on AI technology and there was little commercial interest and funding. Now, the available resources have increased substantially. We should expect that the field continues to advance rapidly."", ""sidebar-toc"": false, ""featured-image"": ""featured-image-on-ai-investments.png""}, ""createdAt"": ""2022-12-02T17:21:04.000Z"", ""published"": false, ""updatedAt"": ""2023-10-11T08:43:56.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-03-29T09:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag spacer""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag spacer""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}], ""numBlocks"": 28, ""numErrors"": 9, ""wpTagCounts"": {""html"": 6, ""column"": 4, ""spacer"": 2, ""columns"": 2, ""heading"": 3, ""paragraph"": 22, ""separator"": 2, ""owid/prominent-link"": 3}, ""htmlTagCounts"": {""p"": 23, ""h4"": 3, ""hr"": 2, ""div"": 9, ""iframe"": 5}}",2023-03-29 09:00:00,2024-02-16 14:22:55,1IRkR6ijZ2nvYqhqe8BXBtAnZAbo5NGDcLX4CW3oEXnY,"[""Max Roser""]","In the past, relatively few researchers were working on AI technology and there was little commercial interest and funding. Now, the available resources have increased substantially. We should expect that the field continues to advance rapidly.",2022-12-02 17:21:04,2023-10-11 08:43:56,https://ourworldindata.org/wp-content/uploads/2022/12/featured-image-on-ai-investments.png,{},"Our World in Data presents the data and research to make progress against the world’s largest problems. This article draws on data and research discussed in our entry on **[Artificial Intelligence](https://ourworldindata.org/artificial-intelligence)**. Artificial intelligence (AI) technology has steadily become more powerful over the course of the last decades and in recent years it has entered our world in many different domains. In a companion article – [the brief history of artificial intelligence](https://ourworldindata.org/brief-history-of-ai) – I document this development. This was achieved despite having relatively few resources. Until recently, investments in terms of capital and scientific efforts were small. In this article I highlight that this has very much changed in recent years. Corporate investment has increased and the scientific field has grown in size. Given how rapidly AI developed in the past, despite the limited resources, this should make us expect AI technology to continue to become more powerful in the coming decades. ## Investments into AI The first chart looks at corporate investment over time.{ref}It is not obvious how to adjust a time-series of AI investments for inflation, and we debated it at some length within the team. Reporting it in nominal prices (as some do) means it makes little sense to compare observations across time and is therefore not very helpful. To make comparisons over time possible one has to take into account that prices change. It would be straightforward to adjust the time series for price changes if we knew the prices of the specific goods and services that these investments purchased. This would make it possible to calculate a volume measure of AI investments, and it would tell us how much these investments bought. But such a metric is not available. (While a comprehensive price index is not available, we know that the cost for some crucial AI technology has fallen rapidly in price. For example, it has become much cheaper to train an AI system.) In the absence of a comprehensive price index that captures the price of AI-specific goods and services, one has to rely on one of the available metrics for the price of a bundle of goods and services. In the end we decided to use the US Consumer Price Index (CPI). The US CPI does not provide us with a volume measure of AI goods and services, but it does capture the opportunity costs of these investments. The inflation adjustment of this time series of AI investments therefore lets us understand the size of these investments relative to whatever else these sums of money could have purchased.{/ref} Until recently, private sector investment was relatively low. But, especially from 2018 onwards, it has increased rapidly. Investments in 2021 were about 30-times larger than just eight years earlier.Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Artificial intelligence (AI) technology has steadily become more powerful over the course of the last decades and in recent years it has entered our world in many different domains. In a companion article – the brief history of artificial intelligence – I document this development.
\n\n\n\nThis was achieved despite having relatively few resources. Until recently, investments in terms of capital and scientific efforts were small. In this article I highlight that this has very much changed in recent years. Corporate investment has increased and the scientific field has grown in size.
\n\n\n\nGiven how rapidly AI developed in the past, despite the limited resources, this should make us expect AI technology to continue to become more powerful in the coming decades.
\n\n\n\nThe first chart looks at corporate investment over time.{ref}It is not obvious how to adjust a time-series of AI investments for inflation, and we debated it at some length within the team.
\n\n\n\nReporting it in nominal prices (as some do) means it makes little sense to compare observations across time and is therefore not very helpful. To make comparisons over time possible one has to take into account that prices change.
\n\n\n\nIt would be straightforward to adjust the time series for price changes if we knew the prices of the specific goods and services that these investments purchased. This would make it possible to calculate a volume measure of AI investments, and it would tell us how much these investments bought. But such a metric is not available.
\n\n\n\n(While a comprehensive price index is not available, we know that the cost for some crucial AI technology has fallen rapidly in price. For example, it has become much cheaper to train an AI system.)
\n\n\n\nIn the absence of a comprehensive price index that captures the price of AI-specific goods and services, one has to rely on one of the available metrics for the price of a bundle of goods and services. In the end we decided to use the US Consumer Price Index (CPI).
\n\n\n\nThe US CPI does not provide us with a volume measure of AI goods and services, but it does capture the opportunity costs of these investments. The inflation adjustment of this time series of AI investments therefore lets us understand the size of these investments relative to whatever else these sums of money could have purchased.{/ref}
\n\n\n\nUntil recently, private sector investment was relatively low. But, especially from 2018 onwards, it has increased rapidly.
\n\n\n\nInvestments in 2021 were about 30-times larger than just eight years earlier.
\n\n\n\n\n\n\n\nThe four following visualizations show that the trends for AI research efforts and the labor market are similar.
\n\n\n\nJust a decade ago the field was much smaller than it is today. Since then the number of research publications on artificial intelligence has doubled, and AI conferences have become much larger events. This is what the first two charts show.
\n\n\n\nThe third chart documents the increasing importance of AI skills in the labor market. The last chart shows the specialty of computer science PhD students in the US. With an increasing number of students in the field, the number of researchers and research output is set to increase further in years to come.
\n\n\n\nAdditionally, AI itself is contributing to the development of AI: Researchers are beginning to find ways for AI itself to contribute to the development of AI.{ref}One early example for this is Neural architecture search.{/ref}
\n\n\n\nThe developments in the past happened despite the fact that funding and brainpower dedicated to AI were quite limited. As these charts have shown, this has changed. Across a range of metrics, the resources dedicated to AI development have increased substantially.
\n\n\n\nThe fact that the field advanced with relatively small resources, and now has much larger resources at its disposal – leading to rapid advances in the last few years – is one reason why I expect AI technology to continue to develop rapidly and to exert a larger and larger influence on our world.
\n\n\n\nContinue reading on Our World in Data:
AI timelines: What do experts in artificial intelligence expect for the future?
Acknowledgements: I would like to thank Julia Broden, Charlie Giattino, Joe Hasell, and Edouard Mathieu for their helpful comments to drafts of this essay and the visualizations.
\n\n\n\nAppendix
\n\n\n\nAdditional charts on the rise of investments into AI technology
\n\n\nOur World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
To see what the future might look like it is often helpful to study our history. This is what I will do in this article. I retrace the brief history of computers and artificial intelligence to see what we can expect for the future.
How rapidly the world has changed becomes clear by how even quite recent computer technology feels ancient to us today. Mobile phones in the ‘90s were big bricks with tiny green displays. Two decades before that the main storage for computers was punch cards.
In a short period computers evolved so quickly and became such an integral part of our daily lives that it is easy to forget how recent this technology is. The first digital computers were only invented about eight decades ago, as the timeline shows.
Since the early days of this history, some computer scientists have strived to make machines as intelligent as humans. The next timeline shows some of the notable artificial intelligence (AI) systems and describes what they were capable of.
The first system I mention is the Theseus. It was built by Claude Shannon in 1950 and was a remote-controlled mouse that was able to find its way out of a labyrinth and could remember its course.{ref}On the Theseus see Daniel Klein (2019) – Mighty mouse, Published in MIT Technology Review. And this video on YouTube of a presentation by its inventor Claude Shannon.{/ref} In seven decades the abilities of artificial intelligence have come a long way.
The chart shows how we got here by zooming into the last two decades of AI development. The plotted data stems from a number of tests in which human and AI performance were evaluated in five different domains, from handwriting recognition to language understanding.
Within each of the five domains the initial performance of the AI system is set to -100, and human performance in these tests is used as a baseline that is set to zero. This means that when the model’s performance crosses the zero line is when the AI system scored more points in the relevant test than the humans who did the same test.{ref}The chart shows that the speed with which these AI technologies developed increased over time. Systems for which development was started early – handwriting and speech recognition – took more than a decade to approach human-level performance, while more recent AI developments led to systems that overtook humans in the span of only a few years. However one should not overstate this point. To some extent this is dependent on when the researchers started to compare machine and human performance. One could have started evaluating the system for language understanding much earlier and its development would appear much slower in this presentation of the data.{/ref}
Just 10 years ago, no machine could reliably provide language or image recognition at a human level. But, as the chart shows, AI systems have become steadily more capable and are now beating humans in tests in all these domains.{ref}It is important to remember that while these are remarkable achievements — and show very rapid gains — these are the results from specific benchmarking tests. Outside of tests, AI models can fail in surprising ways and do not reliably achieve performance that is comparable with human capabilities.{/ref}
Outside of these standardized tests the performance of these AIs is mixed. In some real-world cases these systems are still performing much worse than humans. On the other hand, some implementations of such AI systems are already so cheap that they are available on the phone in your pocket: image recognition categorizes your photos and speech recognition transcribes what you dictate.
Language and image recognition capabilities of AI systems have improved rapidly{ref}Data from Kiela et al. (2021) – Dynabench: Rethinking Benchmarking in NLP. arXiv:2104.14337v1; https://doi.org/10.48550/arXiv.2104.14337 {/ref}
The previous chart showed the rapid advances in the perceptive abilities of artificial intelligence. AI systems have also become much more capable of generating images.
This series of nine images shows the development over the last nine years. None of the people in these images exist; all of them were generated by an AI system.
The series begins with an image from 2014 in the top left, a primitive image of a pixelated face in black and white. As the first image in the second row shows, just three years later AI systems were already able to generate images that were hard to differentiate from a photograph.
In recent years, the capability of AI systems has become much more impressive still. While the early systems focused on generating images of faces, these newer models broadened their capabilities to text-to-image generation based on almost any prompt. The image in the bottom right shows that even the most challenging prompts – such as “A Pomeranian is sitting on the King’s throne wearing a crown. Two tiger soldiers are standing next to the throne” – are turned into photorealistic images within seconds.{ref}Because these systems have become so powerful, the latest AI systems often don’t allow the user to generate images of human faces to prevent abuse.{/ref}
Timeline of images generated by artificial intelligence{ref}The relevant publications are the following:
2014: Goodfellow et al: Generative Adversarial Networks
2015: Radford, Metz, and Chintala: Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
2016: Liu and Tuzel: Coupled Generative Adversarial Networks
2017: Karras et al: Progressive Growing of GANs for Improved Quality, Stability, and Variation
2018: Karras, Laine, and Aila: A Style-Based Generator Architecture for Generative Adversarial Networks (StyleGAN from NVIDIA)
2019: Karras et al: Analyzing and Improving the Image Quality of StyleGAN
AI-generated faces generated by this technology can be found on thispersondoesnotexist.com.
2020: Ho, Jain, and Abbeel: Denoising Diffusion Probabilistic Models
2021: Ramesh et al: Zero-Shot Text-to-Image Generation (first DALL-E from OpenAI; blog post). See also Ramesh et al (2022) – Hierarchical Text-Conditional Image Generation with CLIP Latents (DALL-E 2 from OpenAI; blog post).
2022: Saharia et al: Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding (Google’s Imagen; blog post){/ref}
Just as striking as the advances of image-generating AIs is the rapid development of systems that parse and respond to human language.
Shown in the image are examples from an AI system developed by Google called PaLM. In these six examples, the system was asked to explain six different jokes. I find the explanation in the bottom right particularly remarkable: the AI explains an anti-joke that is specifically meant to confuse the listener.
AIs that produce language have entered our world in many ways over the last few years. Emails get auto-completed, massive amounts of online texts get translated, videos get automatically transcribed, school children use language models to do their homework, reports get auto-generated, and media outlets publish AI-generated journalism.
AI systems are not yet able to produce long, coherent texts. In the future, we will see whether the recent developments will slow down – or even end – or whether we will one day read a bestselling novel written by an AI.
Output of the AI system PaLM after being asked to interpret six different jokes{ref}From Chowdhery et al (2022) – PaLM: Scaling Language Modeling with Pathways. Published on arXiv on 7 Apr 2022.{/ref}
These rapid advances in AI capabilities have made it possible to use machines in a wide range of new domains:
When you book a flight, it is often an artificial intelligence, and no longer a human, that decides what you pay. When you get to the airport, it is an AI system that monitors what you do at the airport. And once you are on the plane, an AI system assists the pilot in flying you to your destination.
AI systems also increasingly determine whether you get a loan, are eligible for welfare, or get hired for a particular job. Increasingly they help determine who gets released from jail.
Several governments are purchasing autonomous weapons systems for warfare, and some are using AI systems for surveillance and oppression.
AI systems help to program the software you use and translate the texts you read. Virtual assistants, operated by speech recognition, have entered many households over the last decade. Now self-driving cars are becoming a reality.
In the last few years, AI systems helped to make progress on some of the hardest problems in science.
Large AIs called recommender systems determine what you see on social media, which products are shown to you in online shops, and what gets recommended to you on YouTube. Increasingly they are not just recommending the media we consume, but based on their capacity to generate images and texts, they are also creating the media we consume.
Artificial intelligence is no longer a technology of the future; AI is here, and much of what is reality now would have looked like sci-fi just recently. It is a technology that already impacts all of us, and the list above includes just a few of its many applications.
The wide range of listed applications makes clear that this is a very general technology that can be used by people for some extremely good goals – and some extraordinarily bad ones, too. For such ‘dual use technologies’, it is important that all of us develop an understanding of what is happening and how we want the technology to be used.
Just two decades ago the world was very different. What might AI technology be capable of in the future?
The AI systems that we just considered are the result of decades of steady advances in AI technology.
The big chart below brings this history over the last eight decades into perspective. It is based on the dataset produced by Jaime Sevilla and colleagues.{ref}See the footnote on the title of the chart for the references and additional information.{/ref}
Each small circle in this chart represents one AI system. The circle’s position on the horizontal axis indicates when the AI system was built, and its position on the vertical axis shows the amount of computation that was used to train the particular AI system.
Training computation is measured in floating point operations, or FLOP for short. One FLOP is equivalent to one addition, subtraction, multiplication, or division of two decimal numbers.
All AI systems that rely on machine learning need to be trained, and in these systems training computation is one of the three fundamental factors that are driving the capabilities of the system. The other two factors are the algorithms and the input data used for the training. The visualization shows that as training computation has increased, AI systems have become more and more powerful.
The timeline goes back to the 1940s, the very beginning of electronic computers. The first shown AI system is ‘Theseus’, Claude Shannon’s robotic mouse from 1950 that I mentioned at the beginning. Towards the other end of the timeline you find AI systems like DALL-E and PaLM, whose abilities to produce photorealistic images and interpret and generate language we have just seen. They are among the AI systems that used the largest amount of training computation to date.
The training computation is plotted on a logarithmic scale, so that from each grid-line to the next it shows a 100-fold increase. This long-run perspective shows a continuous increase. For the first six decades, training computation increased in line with Moore’s Law, doubling roughly every 20 months. Since about 2010 this exponential growth has sped up further, to a doubling time of just about 6 months. That is an astonishingly fast rate of growth.{ref}At some point in the future, training computation is expected to slow down to the exponential growth rate of Moore's Law. Tamay Besiroglu, Lennart Heim and Jaime Sevilla of the Epoch team estimate in their report that the highest probability for this reversion occuring is in the early 2030s.{/ref}
The fast doubling times have accrued to large increases. PaLM’s training computation was 2.5 billion petaFLOP, more than 5 million times larger than that of AlexNet, the AI with the largest training computation just 10 years earlier.{ref}The training computation of PaLM, developed in 2022, was 2,700,000,000 petaFLOP. The training computation of AlexNet, the AI with the largest training computation up to 2012, was 470 petaFLOP. 2,500,000,000 petaFLOP / 470 petaFLOP = 5,319,148.9. At the same time the amount of training computation required to achieve a given performance has been falling exponentially.
The costs have also increased quickly. The cost to train PaLM is estimated to be in the range of $9–$23 million according to Lennart Heim, a researcher in the Epoch team. See: Lennart Heim (2022) – Estimating PaLM's training cost.{/ref}
Scale-up was already exponential and has sped up substantially over the past decade. What can we learn from this historical development for the future of AI?
The rise of artificial intelligence over the last 8 decades: As training computation has increased, AI systems have become more powerful{ref}The data is taken from Jaime Sevilla, Lennart Heim, Anson Ho, Tamay Besiroglu, Marius Hobbhahn, Pablo Villalobos (2022) – Compute Trends Across Three eras of Machine Learning. Published in arXiv on March 9, 2022. See also their post on the Alignment Forum.
The authors regularly update and extend their dataset, a very helpful service to the AI research community. At Our World in Data my colleague Charlie Giattino regularly updates the interactive version of this chart with the latest data made available by Sevilla and coauthors.
See also these two related charts:
Number of parameters in notable artificial intelligence systems
Number of datapoints used to train notable artificial intelligence systems{/ref}
AI researchers study these long-term trends to see what is possible in the future.{ref}Scaling up the size of neural networks – in terms of the number of parameters and the amount of training data and computation – has led to surprising increases in the capabilities of AI systems. This realization motivated the “scaling hypothesis.” See Gwern Branwen (2020) – The Scaling Hypothesis.{/ref}
Perhaps the most widely discussed study of this kind was published by AI researcher Ajeya Cotra. She studied the increase in training computation to ask at what point in time the computation to train an AI system could match that of the human brain. The idea is that at this point the AI system would match the capabilities of a human brain. In her latest update, Cotra estimated a 50% probability that such “transformative AI” will be developed by the year 2040, less than two decades from now.{ref}Her research was announced in various places, including in the AI Alignment Forum here: Ajeya Cotra (2020) – Draft report on AI timelines. As far as I know the report itself always remained a ‘draft report’ and was published here on Google Docs.
The cited estimate stems from Cotra’s Two-year update on my personal AI timelines, in which she shortened her median timeline by 10 years.
Cotra emphasizes that there are substantial uncertainties around her estimates and therefore communicates her findings in a range of scenarios. She published her big study in 2020 and her median estimate at the time was that around the year 2050 there will be a 50%-probability that the computation required to train such a model may become affordable. In her “most conservative plausible”-scenario this point in time is pushed back to around the year 2090 and in her “most aggressive plausible”-scenario this point in time is reached in 2040.
The same is true for most other forecasters: all emphasize the large uncertainty associated with any of their forecasts.
It is worth emphasizing that the computation of the human brain is highly uncertain. See Joseph Carlsmith's New Report on How Much Computational Power It Takes to Match the Human Brain from 2020.{/ref}
In a related article, I discuss what transformative AI would mean for the world. In short, the idea is that such an AI system would be powerful enough to bring the world into a ‘qualitatively different future’. It could lead to a change at the scale of the two earlier major transformations in human history, the agricultural and industrial revolutions. It would certainly represent the most important global change in our lifetimes.
Cotra’s work is particularly relevant in this context as she based her forecast on the kind of historical long-run trend of training computation that we just studied. But it is worth noting that other forecasters who rely on different considerations arrive at broadly similar conclusions. As I show in my article on AI timelines, many AI experts believe that there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner.
Computers and artificial intelligence have changed our world immensely, but we are still at the early stages of this history. Because this technology feels so familiar, it is easy to forget that all of these technologies that we interact with are very recent innovations, and that most profound changes are yet to come.
Artificial intelligence has already changed what we see, what we know, and what we do. And this is despite the fact that this technology has had only a brief history.
There are no signs that these trends are hitting any limits anytime soon. To the contrary, particularly over the course of the last decade, the fundamental trends have accelerated: investments in AI technology have rapidly increased, and the doubling time of training computation has shortened to just six months.
All major technological innovations lead to a range of positive and negative consequences. This is already true of artificial intelligence. As this technology becomes more and more powerful, we should expect its impact to become greater still.
Because of the importance of AI, we should all be able to form an opinion on where this technology is heading and to understand how this development is changing our world. For this purpose, we are building a repository of AI-related metrics, which you can find on OurWorldinData.org/artificial-intelligence.
We are still in the early stages of this history and much of what will become possible is yet to come. A technological development as powerful as this should be at the center of our attention. Little might be as important for how the future of our world – and the future of our lives – will play out.
Acknowledgements: I would like to thank my colleagues Natasha Ahuja, Daniel Bachler, Julia Broden, Charlie Giattino, Bastian Herre, Edouard Mathieu, and Ike Saunders for their helpful comments to drafts of this essay and their contributions in preparing the visualizations.
","{""id"": ""wp-54765"", ""slug"": ""brief-history-of-ai"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Our World in Data presents the data and research to make progress against the world’s largest problems."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""This article draws on data and research discussed in our entry on "", ""spanType"": ""span-simple-text""}, {""children"": [{""url"": ""https://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""Artificial Intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""spanType"": ""span-bold""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To see what the future might look like it is often helpful to study our history. This is what I will do in this article. I retrace the brief history of computers and artificial intelligence to see what we can expect for the future."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How did we get here?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How rapidly the world has changed becomes clear by how even quite recent computer technology feels ancient to us today. Mobile phones in the ‘90s were big bricks with tiny green displays. Two decades before that the main storage for computers was punch cards. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In a short period computers evolved so quickly and became such an integral part of our daily lives that it is easy to forget how recent this technology is. The first digital computers were only invented about eight decades ago, as the timeline shows."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Computer-timeline.png"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Since the early days of this history, some computer scientists have strived to make machines as intelligent as humans. The next timeline shows some of the notable artificial intelligence (AI) systems and describes what they were capable of. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The first system I mention is the Theseus. It was built by Claude Shannon in 1950 and was a remote-controlled mouse that was able to find its way out of a labyrinth and could remember its course.{ref}On the Theseus see Daniel Klein (2019) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20220125004420/https://www.technologyreview.com/2018/12/19/138508/mighty-mouse/"", ""children"": [{""text"": ""Mighty mouse"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", Published in MIT Technology Review. And "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.youtube.com/watch?v=_9_AEVQ_p74"", ""children"": [{""text"": ""this video on YouTube"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of a presentation by its inventor Claude Shannon.{/ref} In seven decades the abilities of artificial intelligence have come a long way."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""ai-timeline.png"", ""parseErrors"": []}, {""text"": [{""text"": ""The language and image recognition capabilities of AI systems have developed very rapidly"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The chart shows how we got here by zooming into the last two decades of AI development. The plotted data stems from a number of tests in which human and AI performance were evaluated in five different domains, from handwriting recognition to language understanding. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Within each of the five domains the initial performance of the AI system is set to -100, and human performance in these tests is used as a baseline that is set to zero. This means that when the model’s performance crosses the zero line is when the AI system scored more points in the relevant test than the humans who did the same test.{ref}The chart shows that the speed with which these AI technologies developed increased over time. Systems for which development was started early – handwriting and speech recognition – took more than a decade to approach human-level performance, while more recent AI developments led to systems that overtook humans in the span of only a few years. However one should not overstate this point. To some extent this is dependent on when the researchers started to compare machine and human performance. One could have started evaluating the system for language understanding much earlier and its development would appear much slower in this presentation of the data.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Just 10 years ago, no machine could reliably provide language or image recognition at a human level. But, as the chart shows, AI systems have become steadily more capable and are now beating humans in "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""tests"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" in all these domains.{ref}It is important to remember that while these are remarkable achievements — and show very rapid gains — these are the results from specific benchmarking tests. Outside of tests, AI models can fail in surprising ways and do not reliably achieve performance that is comparable with human capabilities.{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Outside of these standardized tests the performance of these AIs is mixed. In some real-world cases these systems are still performing much worse than humans. On the other hand, some implementations of such AI systems are already so cheap that they are available on the phone in your pocket: image recognition categorizes your photos and speech recognition transcribes what you dictate."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Language and image recognition capabilities of AI systems have improved rapidly"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""{ref}Data from Kiela et al. (2021) – Dynabench: Rethinking Benchmarking in NLP. arXiv:2104.14337v1; "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.48550/arXiv.2104.14337"", ""children"": [{""text"": ""https://doi.org/10.48550/arXiv.2104.14337"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""AI-performance_Dynabench-paper.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""From image recognition to image generation"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The previous chart showed the rapid advances in the perceptive abilities of artificial intelligence. AI systems have also become much more capable of generating images. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This series of nine images shows the development over the last nine years. None of the people in these images exist; all of them were generated by an AI system."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The series begins with an image from 2014 in the top left, a primitive image of a pixelated face in black and white. As the first image in the second row shows, just three years later AI systems were already able to generate images that were hard to differentiate from a photograph."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In recent years, the capability of AI systems has become much more impressive still. While the early systems focused on generating images of faces, these newer models broadened their capabilities to text-to-image generation based on almost any prompt. The image in the bottom right shows that even the most challenging prompts – such as "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""“A Pomeranian is sitting on the King’s throne wearing a crown. Two tiger soldiers are standing next to the throne”"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" – are turned into photorealistic images within seconds.{ref}Because these systems have become so powerful, the latest AI systems often don’t allow the user to generate images of human faces to prevent abuse.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Timeline of images generated by artificial intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""{ref}The relevant publications are the following:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2014: Goodfellow et al:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/1406.2661"", ""children"": [{""text"": "" Generative Adversarial Networks"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2015: Radford, Metz, and Chintala:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/1511.06434"", ""children"": [{""text"": "" Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2016: Liu and Tuzel:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/1606.07536"", ""children"": [{""text"": "" Coupled Generative Adversarial Networks"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2017: Karras et al:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/1710.10196"", ""children"": [{""text"": "" Progressive Growing of GANs for Improved Quality, Stability, and Variation"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2018: Karras, Laine, and Aila:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/1812.04948"", ""children"": [{""text"": "" A Style-Based Generator Architecture for Generative Adversarial Networks"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (StyleGAN from NVIDIA)"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2019: Karras et al:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/1912.04958"", ""children"": [{""text"": "" Analyzing and Improving the Image Quality of StyleGAN"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""AI-generated faces generated by this technology can be found on "", ""spanType"": ""span-simple-text""}, {""url"": ""https://thispersondoesnotexist.com/"", ""children"": [{""text"": ""thispersondoesnotexist.com"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2020: Ho, Jain, and Abbeel:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/2006.11239"", ""children"": [{""text"": "" Denoising Diffusion Probabilistic Models"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2021: Ramesh et al:"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/2102.12092"", ""children"": [{""text"": "" Zero-Shot Text-to-Image Generation"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (first DALL-E from OpenAI;"", ""spanType"": ""span-simple-text""}, {""url"": ""https://openai.com/blog/dall-e/"", ""children"": [{""text"": "" blog post"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""). See also Ramesh et al (2022) –"", ""spanType"": ""span-simple-text""}, {""url"": ""https://cdn.openai.com/papers/dall-e-2.pdf"", ""children"": [{""text"": "" Hierarchical Text-Conditional Image Generation with CLIP Latents"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (DALL-E 2 from OpenAI;"", ""spanType"": ""span-simple-text""}, {""url"": ""https://openai.com/dall-e-2/"", ""children"": [{""text"": "" blog post"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "")."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""2022: Saharia et al: "", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/2205.11487"", ""children"": [{""text"": ""Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (Google’s Imagen;"", ""spanType"": ""span-simple-text""}, {""url"": ""https://imagen.research.google/"", ""children"": [{""text"": "" blog post"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""){/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Timeline-of-AI-generated-faces.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Language recognition and production is developing fast"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Just as striking as the advances of image-generating AIs is the rapid development of systems that parse and respond to human language. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Shown in the image are examples from an AI system developed by Google called PaLM. In these six examples, the system was asked to explain six different jokes. I find the explanation in the bottom right particularly remarkable: the AI explains an anti-joke that is specifically meant to confuse the listener."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""AIs that produce language have entered our world in many ways over the last few years. Emails get auto-completed, massive amounts of online texts get translated, videos get automatically transcribed, school children use language models to do their homework, reports get auto-generated, and media outlets "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Automated_journalism"", ""children"": [{""text"": ""publish"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" AI-generated journalism."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""AI systems are not yet able to produce long, coherent texts. In the future, we will see whether the recent developments will slow down – or even end – or whether we will one day read a bestselling novel written by an AI."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Output of the AI system PaLM after being asked to interpret six different jokes"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""{ref}From Chowdhery et al (2022) –"", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/pdf/2204.02311v2.pdf"", ""children"": [{""text"": "" PaLM: Scaling Language Modeling with Pathways"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Published on arXiv on 7 Apr 2022.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""jJ58V3vrTBw9cg2lzM-w2xiU7ExufRO7WNJUCp7a3ZGf6c79LjqrusGmFaF8nMUEtn-gH3K7J1CHHG3jMW4WotJnNjGYhCVwt3Ou5g66geZDw81yiwez1OxPB80E6TmoRNHa9dNcch9TLDj5ruQBHDvkPpl5Hpl6TtnPVXvdh0mp4jFmmiBu0Wmxsp1mKA"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Where we are now: AI is here"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These rapid advances in AI capabilities have made it possible to use machines in a wide range of new domains:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""When you book a flight, it is often an artificial intelligence, and no longer a human, that "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.bloomberg.com/news/articles/2022-10-20/artificial-intelligence-helps-airlines-find-the-right-prices-for-flight-tickets"", ""children"": [{""text"": ""decides"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" what you pay. When you get to the airport, it is an AI system that "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.sourcesecurity.com/news/co-2166-ga.132.html"", ""children"": [{""text"": ""monitors"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" what you do at the airport. And once you are on the plane, an AI system assists the pilot in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.airbus.com/en/innovation/industry-4-0/artificial-intelligence"", ""children"": [{""text"": ""flying"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" you to your destination. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""AI systems also increasingly determine whether you "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.brookings.edu/research/reducing-bias-in-ai-based-financial-services/"", ""children"": [{""text"": ""get a loan"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", are "", ""spanType"": ""span-simple-text""}, {""url"": ""https://theconversation.com/ai-algorithms-intended-to-root-out-welfare-fraud-often-end-up-punishing-the-poor-instead-131625"", ""children"": [{""text"": ""eligible"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" for welfare, or get "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G"", ""children"": [{""text"": ""hired"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" for a particular job. Increasingly they help determine who gets "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.technologyreview.com/2019/01/21/137783/algorithms-criminal-justice-ai/"", ""children"": [{""text"": ""released from jail"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Several governments are purchasing "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Lethal_autonomous_weapon"", ""children"": [{""text"": ""autonomous weapons systems"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" for warfare, and some are using AI systems for "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nytimes.com/2019/04/14/technology/china-surveillance-artificial-intelligence-racial-profiling.html"", ""children"": [{""text"": ""surveillance and oppression"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""AI systems "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/GitHub_Copilot"", ""children"": [{""text"": ""help"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to program the software you use and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Machine_translation"", ""children"": [{""text"": ""translate"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" the texts you read. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Virtual_assistant"", ""children"": [{""text"": ""Virtual assistants"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", operated by speech recognition, have entered many households over the last decade. Now "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Self-driving_car"", ""children"": [{""text"": ""self-driving cars"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" are becoming a reality. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the last few years, AI systems "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/s42254-022-00518-3"", ""children"": [{""text"": ""helped"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://www.quantamagazine.org/how-artificial-intelligence-is-changing-science-20190311/"", ""children"": [{""text"": ""to"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://www.nature.com/articles/d41586-022-00997-5"", ""children"": [{""text"": ""make"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://www.nature.com/articles/d41586-022-03209-2"", ""children"": [{""text"": ""progress"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on some of the hardest problems in science."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Large AIs called "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Recommender_system"", ""children"": [{""text"": ""recommender systems"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" determine what you see on social media, which products are shown to you in online shops, and what gets recommended to you on YouTube. Increasingly they are not just recommending the media we consume, but based on their capacity to generate images and texts, they are also "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/d41586-021-00530-0"", ""children"": [{""text"": ""creating"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" the media we consume. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Artificial intelligence is no longer a technology of the future; AI is here, and much of what is reality now would have looked like sci-fi just recently. It is a technology that already impacts all of us, and the list above includes just a few of its "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Applications_of_artificial_intelligence"", ""children"": [{""text"": ""many applications"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The wide range of listed applications makes clear that this is a very general technology that can be used by people for some extremely good goals – and some extraordinarily bad ones, too. For such ‘dual use technologies’, it is important that all of us develop an understanding of what is happening and how we want the technology to be used."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Just two decades ago the world was very different. What might AI technology be capable of in the future?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What is next? "", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The AI systems that we just considered are the result of decades of steady advances in AI technology. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The big chart below brings this history over the last eight decades into perspective. It is based on the dataset produced by Jaime Sevilla and colleagues.{ref}See the footnote on the title of the chart for the references and additional information.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Each small circle in this chart represents one AI system. The circle’s position on the horizontal axis indicates when the AI system was built, and its position on the vertical axis shows the amount of computation that was used to train the particular AI system."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Training computation is measured in "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""floating point operations"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", or FLOP for short. One FLOP is equivalent to one addition, subtraction, multiplication, or division of two decimal numbers. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""All AI systems that rely on machine learning need to be trained, and in these systems training computation is one of the three fundamental factors that are driving the capabilities of the system. The other two factors are the algorithms and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/artificial-intelligence-number-training-datapoints"", ""children"": [{""text"": ""the input data used"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" for the training. The visualization shows that as training computation has increased, AI systems have become more and more powerful."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The timeline goes back to the 1940s, the very beginning of electronic computers. The first shown AI system is ‘Theseus’, Claude Shannon’s robotic mouse from 1950 that I mentioned at the beginning. Towards the other end of the timeline you find AI systems like DALL-E and PaLM, whose abilities to produce photorealistic images and interpret and generate language we have just seen. They are among the AI systems that used the largest amount of training computation to date."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The training computation is plotted on a logarithmic scale, so that from each grid-line to the next it shows a 100-fold increase. This long-run perspective shows a continuous increase. For the first six decades, training computation increased in line with "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/technological-change#moore-s-law-the-exponential-increase-of-the-number-of-transistors-on-integrated-circuits"", ""children"": [{""text"": ""Moore’s Law"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", doubling roughly every 20 months. Since about 2010 this exponential growth has sped up further, to a doubling time of just about 6 months. That is an astonishingly fast rate of growth.{ref}At some point in the future, training computation is expected to slow down to the exponential growth rate of Moore's Law. Tamay Besiroglu, Lennart Heim and Jaime Sevilla of the Epoch team estimate in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://epochai.org/blog/projecting-compute-trends"", ""children"": [{""text"": ""their report"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" that the highest probability for this reversion occuring is in the early 2030s.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The fast doubling times have accrued to large increases. PaLM’s training computation was 2.5 billion petaFLOP, more than 5 million times larger than that of AlexNet, the AI with the largest training computation just 10 years earlier.{ref}The training computation of PaLM, developed in 2022, was 2,700,000,000 petaFLOP. The training computation of AlexNet, the AI with the largest training computation up to 2012, was 470 petaFLOP. 2,500,000,000 petaFLOP / 470 petaFLOP = 5,319,148.9. At the same time the amount of training computation required to achieve a given performance has been falling exponentially."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The costs have also increased quickly. The cost to train PaLM is estimated to be in the range of $9–$23 million according to Lennart Heim, a researcher in the Epoch team. See: Lennart Heim (2022) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://blog.heim.xyz/palm-training-cost/"", ""children"": [{""text"": ""Estimating PaLM's training cost"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Scale-up was already exponential and has sped up substantially over the past decade. What can we learn from this historical development for the future of AI?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""The rise of artificial intelligence over the last 8 decades: As training computation has increased, AI systems have become more powerful"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""{ref}The data is taken from Jaime Sevilla, Lennart Heim, Anson Ho, Tamay Besiroglu, Marius Hobbhahn, Pablo Villalobos (2022) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/pdf/2202.05924.pdf"", ""children"": [{""text"": ""Compute Trends Across Three eras of Machine Learning"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Published in arXiv on March 9, 2022. See also "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.alignmentforum.org/posts/XKtybmbjhC6mXDm5z/compute-trends-across-three-eras-of-machine-learning"", ""children"": [{""text"": ""their post on the Alignment Forum"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The authors regularly update and extend their dataset, a very helpful service to the AI research community. At Our World in Data my colleague Charlie Giattino regularly updates "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/ai-training-computation"", ""children"": [{""text"": ""the interactive version of this chart"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" with the latest data made available by Sevilla and coauthors."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""See also these two related charts:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/grapher/artificial-intelligence-parameter-count"", ""children"": [{""text"": ""Number of parameters in notable artificial intelligence systems"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/grapher/artificial-intelligence-number-training-datapoints"", ""children"": [{""text"": ""Number of datapoints used to train notable artificial intelligence systems"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""ai-training-computation-3.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Studying the long-run trends to predict the future of AI"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""AI researchers study these long-term trends to see what is possible in the future.{ref}Scaling up the size of neural networks – in terms of the number of parameters and the amount of training data and computation – has led to surprising increases in the capabilities of AI systems. This realization motivated the “scaling hypothesis.” See Gwern Branwen (2020) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.gwern.net/Scaling-hypothesis"", ""children"": [{""text"": ""The Scaling Hypothesis"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Perhaps the most widely discussed study of this kind was published by AI researcher Ajeya Cotra. She studied the increase in training computation to ask at what point in time the computation to train an AI system could match that of the human brain. The idea is that at this point the AI system would match the capabilities of a human brain. In her latest update, Cotra estimated a 50% probability that such “transformative AI” will be developed by the year 2040, less than two decades from now.{ref}Her research was announced in various places, including in the AI Alignment Forum here: Ajeya Cotra (2020) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.alignmentforum.org/posts/KrJfoZzpSDpnrv9va/draft-report-on-ai-timelines"", ""children"": [{""text"": ""Draft report on AI timelines"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". As far as I know the report itself always remained a ‘draft report’ and was published "", ""spanType"": ""span-simple-text""}, {""url"": ""https://drive.google.com/drive/u/1/folders/15ArhEPZSTYU8f012bs6ehPS6-xmhtBPP"", ""children"": [{""text"": ""here on Google Docs"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The cited estimate stems from Cotra’s "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.alignmentforum.org/posts/AfH2oPHCApdKicM4m/two-year-update-on-my-personal-ai-timelines"", ""children"": [{""text"": ""Two-year update on my personal AI timelines"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", in which she shortened her median timeline by 10 years."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Cotra emphasizes that there are substantial uncertainties around her estimates and therefore communicates her findings in a range of scenarios. She published her big study in 2020 and her median estimate at the time was that around the year 2050 there will be a 50%-probability that the computation required to train such a model may become affordable. In her “most conservative plausible”-scenario this point in time is pushed back to around the year 2090 and in her “most aggressive plausible”-scenario this point in time is reached in 2040."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The same is true for most other forecasters: all emphasize the large uncertainty associated with any of "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""their forecasts"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It is worth emphasizing that the computation of the human brain is highly uncertain. See Joseph Carlsmith's "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.openphilanthropy.org/research/new-report-on-how-much-computational-power-it-takes-to-match-the-human-brain/"", ""children"": [{""text"": ""New Report on How Much Computational Power It Takes to Match the Human Brain"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" from 2020.{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-impact"", ""children"": [{""text"": ""a related article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", I discuss what transformative AI would mean for the world. In short, the idea is that such an AI system would be powerful enough to bring the world into a ‘qualitatively different future’. It could lead to a change at the scale of the two earlier major transformations in human history, the agricultural and industrial revolutions. It would certainly represent the most important global change in our lifetimes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Cotra’s work is particularly relevant in this context as she based her forecast on the kind of historical long-run trend of training computation that we just studied. But it is worth noting that other forecasters who rely on different considerations arrive at broadly similar conclusions. As I show in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""my article on AI timelines"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", many AI experts believe that there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Building a public resource to enable the necessary public conversation"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Computers and artificial intelligence have changed our world immensely, but we are still at the early stages of this history. Because this technology feels so familiar, it is easy to forget that all of these technologies that we interact with are very recent innovations, and that most profound changes are yet to come."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Artificial intelligence has already changed what we see, what we know, and what we do. And this is despite the fact that this technology has had only a brief history. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are no signs that these trends are hitting any limits anytime soon. To the contrary, particularly over the course of the last decade, the fundamental trends have accelerated: investments in AI technology have "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-investments"", ""children"": [{""text"": ""rapidly increased"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", and the doubling time of training computation has shortened to just six months."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""All major technological innovations lead to a range of positive and negative consequences. This is already true of artificial intelligence. As this technology becomes more and more powerful, we should expect its impact to become greater still. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Because of the importance of AI, we should all be able to form an opinion on where this technology is heading and to understand how this development is changing our world. For this purpose, we are building a repository of AI-related metrics, which you can find on "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""OurWorldinData.org/artificial-intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We are still in the early stages of this history and much of what will become possible is yet to come. A technological development as powerful as this should be at the center of our attention. Little might be as important for how the future of our world – and the future of our lives – will play out."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Acknowledgements:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" I would like to thank my colleagues Natasha Ahuja, Daniel Bachler, Julia Broden, Charlie Giattino, Bastian Herre, Edouard Mathieu, and Ike Saunders for their helpful comments to drafts of this essay and their contributions in preparing the visualizations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""The brief history of artificial intelligence: The world has changed fast – what might be next?"", ""authors"": [""Max Roser""], ""excerpt"": ""Despite their brief history, computers and AI have fundamentally changed what we see, what we know, and what we do. Little is as important for the future of the world, and our own lives, as how this history continues."", ""dateline"": ""December 6, 2022"", ""subtitle"": ""Despite their brief history, computers and AI have fundamentally changed what we see, what we know, and what we do. Little is as important for the future of the world, and our own lives, as how this history continues."", ""sidebar-toc"": false, ""featured-image"": ""featured-image-ai-training-computation.png""}, ""createdAt"": ""2022-12-02T12:33:16.000Z"", ""published"": false, ""updatedAt"": ""2023-07-31T13:05:20.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-12-06T01:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag spacer""}], ""numBlocks"": 45, ""numErrors"": 9, ""wpTagCounts"": {""html"": 1, ""image"": 6, ""column"": 12, ""spacer"": 1, ""columns"": 6, ""heading"": 8, ""paragraph"": 71, ""separator"": 1}, ""htmlTagCounts"": {""p"": 72, ""h3"": 2, ""h4"": 6, ""hr"": 1, ""div"": 20, ""figure"": 6}}",2022-12-06 01:00:00,2024-02-16 14:22:54,1WGJaY95A4hVjybBPzHOA80iGygKUKcgx0kSs-G_b7aA,"[""Max Roser""]","Despite their brief history, computers and AI have fundamentally changed what we see, what we know, and what we do. Little is as important for the future of the world, and our own lives, as how this history continues.",2022-12-02 12:33:16,2023-07-31 13:05:20,https://ourworldindata.org/wp-content/uploads/2022/12/featured-image-ai-training-computation.png,{},"Our World in Data presents the data and research to make progress against the world’s largest problems. This article draws on data and research discussed in our entry on **[Artificial Intelligence](https://ourworldindata.org/artificial-intelligence)**. To see what the future might look like it is often helpful to study our history. This is what I will do in this article. I retrace the brief history of computers and artificial intelligence to see what we can expect for the future. ### How did we get here? How rapidly the world has changed becomes clear by how even quite recent computer technology feels ancient to us today. Mobile phones in the ‘90s were big bricks with tiny green displays. Two decades before that the main storage for computers was punch cards. In a short period computers evolved so quickly and became such an integral part of our daily lives that it is easy to forget how recent this technology is. The first digital computers were only invented about eight decades ago, as the timeline shows.Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
To see what the future might look like it is often helpful to study our history. This is what I will do in this article. I retrace the brief history of computers and artificial intelligence to see what we can expect for the future.
\n\n\n\nHow rapidly the world has changed becomes clear by how even quite recent computer technology feels ancient to us today. Mobile phones in the ‘90s were big bricks with tiny green displays. Two decades before that the main storage for computers was punch cards.
\n\n\n\nIn a short period computers evolved so quickly and became such an integral part of our daily lives that it is easy to forget how recent this technology is. The first digital computers were only invented about eight decades ago, as the timeline shows.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/Computer-timeline.png\"")
Since the early days of this history, some computer scientists have strived to make machines as intelligent as humans. The next timeline shows some of the notable artificial intelligence (AI) systems and describes what they were capable of.
\n\n\n\nThe first system I mention is the Theseus. It was built by Claude Shannon in 1950 and was a remote-controlled mouse that was able to find its way out of a labyrinth and could remember its course.{ref}On the Theseus see Daniel Klein (2019) – Mighty mouse, Published in MIT Technology Review. And this video on YouTube of a presentation by its inventor Claude Shannon.{/ref} In seven decades the abilities of artificial intelligence have come a long way.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/ai-timeline.png\"")
The chart shows how we got here by zooming into the last two decades of AI development. The plotted data stems from a number of tests in which human and AI performance were evaluated in five different domains, from handwriting recognition to language understanding.
\n\n\n\nWithin each of the five domains the initial performance of the AI system is set to -100, and human performance in these tests is used as a baseline that is set to zero. This means that when the model’s performance crosses the zero line is when the AI system scored more points in the relevant test than the humans who did the same test.{ref}The chart shows that the speed with which these AI technologies developed increased over time. Systems for which development was started early – handwriting and speech recognition – took more than a decade to approach human-level performance, while more recent AI developments led to systems that overtook humans in the span of only a few years. However one should not overstate this point. To some extent this is dependent on when the researchers started to compare machine and human performance. One could have started evaluating the system for language understanding much earlier and its development would appear much slower in this presentation of the data.{/ref}
\n\n\n\nJust 10 years ago, no machine could reliably provide language or image recognition at a human level. But, as the chart shows, AI systems have become steadily more capable and are now beating humans in tests in all these domains.{ref}It is important to remember that while these are remarkable achievements — and show very rapid gains — these are the results from specific benchmarking tests. Outside of tests, AI models can fail in surprising ways and do not reliably achieve performance that is comparable with human capabilities.{/ref}
\n\n\n\nOutside of these standardized tests the performance of these AIs is mixed. In some real-world cases these systems are still performing much worse than humans. On the other hand, some implementations of such AI systems are already so cheap that they are available on the phone in your pocket: image recognition categorizes your photos and speech recognition transcribes what you dictate.
\nLanguage and image recognition capabilities of AI systems have improved rapidly{ref}Data from Kiela et al. (2021) – Dynabench: Rethinking Benchmarking in NLP. arXiv:2104.14337v1; https://doi.org/10.48550/arXiv.2104.14337 {/ref}
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/AI-performance_Dynabench-paper.png\"")
The previous chart showed the rapid advances in the perceptive abilities of artificial intelligence. AI systems have also become much more capable of generating images.
\n\n\n\nThis series of nine images shows the development over the last nine years. None of the people in these images exist; all of them were generated by an AI system.
\n\n\n\nThe series begins with an image from 2014 in the top left, a primitive image of a pixelated face in black and white. As the first image in the second row shows, just three years later AI systems were already able to generate images that were hard to differentiate from a photograph.
\n\n\n\nIn recent years, the capability of AI systems has become much more impressive still. While the early systems focused on generating images of faces, these newer models broadened their capabilities to text-to-image generation based on almost any prompt. The image in the bottom right shows that even the most challenging prompts – such as “A Pomeranian is sitting on the King’s throne wearing a crown. Two tiger soldiers are standing next to the throne” – are turned into photorealistic images within seconds.{ref}Because these systems have become so powerful, the latest AI systems often don’t allow the user to generate images of human faces to prevent abuse.{/ref}
\nTimeline of images generated by artificial intelligence{ref}The relevant publications are the following:
\n\n\n\n2014: Goodfellow et al: Generative Adversarial Networks
\n\n\n\n2015: Radford, Metz, and Chintala: Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
\n\n\n\n2016: Liu and Tuzel: Coupled Generative Adversarial Networks
\n\n\n\n2017: Karras et al: Progressive Growing of GANs for Improved Quality, Stability, and Variation
\n\n\n\n2018: Karras, Laine, and Aila: A Style-Based Generator Architecture for Generative Adversarial Networks (StyleGAN from NVIDIA)
\n\n\n\n2019: Karras et al: Analyzing and Improving the Image Quality of StyleGAN
\n\n\n\nAI-generated faces generated by this technology can be found on thispersondoesnotexist.com.
\n\n\n\n2020: Ho, Jain, and Abbeel: Denoising Diffusion Probabilistic Models
\n\n\n\n2021: Ramesh et al: Zero-Shot Text-to-Image Generation (first DALL-E from OpenAI; blog post). See also Ramesh et al (2022) – Hierarchical Text-Conditional Image Generation with CLIP Latents (DALL-E 2 from OpenAI; blog post).
\n\n\n\n2022: Saharia et al: Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding (Google’s Imagen; blog post){/ref}
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/Timeline-of-AI-generated-faces.png\"")
Just as striking as the advances of image-generating AIs is the rapid development of systems that parse and respond to human language.
\n\n\n\nShown in the image are examples from an AI system developed by Google called PaLM. In these six examples, the system was asked to explain six different jokes. I find the explanation in the bottom right particularly remarkable: the AI explains an anti-joke that is specifically meant to confuse the listener.
\n\n\n\nAIs that produce language have entered our world in many ways over the last few years. Emails get auto-completed, massive amounts of online texts get translated, videos get automatically transcribed, school children use language models to do their homework, reports get auto-generated, and media outlets publish AI-generated journalism.
\n\n\n\nAI systems are not yet able to produce long, coherent texts. In the future, we will see whether the recent developments will slow down – or even end – or whether we will one day read a bestselling novel written by an AI.
\nOutput of the AI system PaLM after being asked to interpret six different jokes{ref}From Chowdhery et al (2022) – PaLM: Scaling Language Modeling with Pathways. Published on arXiv on 7 Apr 2022.{/ref}
\n\n\n\n![\""\""/](\""https://lh3.googleusercontent.com/jJ58V3vrTBw9cg2lzM-w2xiU7ExufRO7WNJUCp7a3ZGf6c79LjqrusGmFaF8nMUEtn-gH3K7J1CHHG3jMW4WotJnNjGYhCVwt3Ou5g66geZDw81yiwez1OxPB80E6TmoRNHa9dNcch9TLDj5ruQBHDvkPpl5Hpl6TtnPVXvdh0mp4jFmmiBu0Wmxsp1mKA\"")
These rapid advances in AI capabilities have made it possible to use machines in a wide range of new domains:
\n\n\n\nWhen you book a flight, it is often an artificial intelligence, and no longer a human, that decides what you pay. When you get to the airport, it is an AI system that monitors what you do at the airport. And once you are on the plane, an AI system assists the pilot in flying you to your destination.
\n\n\n\nAI systems also increasingly determine whether you get a loan, are eligible for welfare, or get hired for a particular job. Increasingly they help determine who gets released from jail.
\n\n\n\nSeveral governments are purchasing autonomous weapons systems for warfare, and some are using AI systems for surveillance and oppression.
\n\n\n\nAI systems help to program the software you use and translate the texts you read. Virtual assistants, operated by speech recognition, have entered many households over the last decade. Now self-driving cars are becoming a reality.
\n\n\n\nIn the last few years, AI systems helped to make progress on some of the hardest problems in science.
\n\n\n\nLarge AIs called recommender systems determine what you see on social media, which products are shown to you in online shops, and what gets recommended to you on YouTube. Increasingly they are not just recommending the media we consume, but based on their capacity to generate images and texts, they are also creating the media we consume.
\n\n\n\nArtificial intelligence is no longer a technology of the future; AI is here, and much of what is reality now would have looked like sci-fi just recently. It is a technology that already impacts all of us, and the list above includes just a few of its many applications.
\n\n\n\nThe wide range of listed applications makes clear that this is a very general technology that can be used by people for some extremely good goals – and some extraordinarily bad ones, too. For such ‘dual use technologies’, it is important that all of us develop an understanding of what is happening and how we want the technology to be used.
\n\n\n\nJust two decades ago the world was very different. What might AI technology be capable of in the future?
\n\n\n\nThe AI systems that we just considered are the result of decades of steady advances in AI technology.
\n\n\n\nThe big chart below brings this history over the last eight decades into perspective. It is based on the dataset produced by Jaime Sevilla and colleagues.{ref}See the footnote on the title of the chart for the references and additional information.{/ref}
\n\n\n\nEach small circle in this chart represents one AI system. The circle’s position on the horizontal axis indicates when the AI system was built, and its position on the vertical axis shows the amount of computation that was used to train the particular AI system.
\n\n\n\nTraining computation is measured in floating point operations, or FLOP for short. One FLOP is equivalent to one addition, subtraction, multiplication, or division of two decimal numbers.
\n\n\n\nAll AI systems that rely on machine learning need to be trained, and in these systems training computation is one of the three fundamental factors that are driving the capabilities of the system. The other two factors are the algorithms and the input data used for the training. The visualization shows that as training computation has increased, AI systems have become more and more powerful.
\n\n\n\nThe timeline goes back to the 1940s, the very beginning of electronic computers. The first shown AI system is ‘Theseus’, Claude Shannon’s robotic mouse from 1950 that I mentioned at the beginning. Towards the other end of the timeline you find AI systems like DALL-E and PaLM, whose abilities to produce photorealistic images and interpret and generate language we have just seen. They are among the AI systems that used the largest amount of training computation to date.
\n\n\n\nThe training computation is plotted on a logarithmic scale, so that from each grid-line to the next it shows a 100-fold increase. This long-run perspective shows a continuous increase. For the first six decades, training computation increased in line with Moore’s Law, doubling roughly every 20 months. Since about 2010 this exponential growth has sped up further, to a doubling time of just about 6 months. That is an astonishingly fast rate of growth.{ref}At some point in the future, training computation is expected to slow down to the exponential growth rate of Moore’s Law. Tamay Besiroglu, Lennart Heim and Jaime Sevilla of the Epoch team estimate in their report that the highest probability for this reversion occuring is in the early 2030s.{/ref}
\n\n\n\nThe fast doubling times have accrued to large increases. PaLM’s training computation was 2.5 billion petaFLOP, more than 5 million times larger than that of AlexNet, the AI with the largest training computation just 10 years earlier.{ref}The training computation of PaLM, developed in 2022, was 2,700,000,000 petaFLOP. The training computation of AlexNet, the AI with the largest training computation up to 2012, was 470 petaFLOP. 2,500,000,000 petaFLOP / 470 petaFLOP = 5,319,148.9. At the same time the amount of training computation required to achieve a given performance has been falling exponentially.
\n\n\n\nThe costs have also increased quickly. The cost to train PaLM is estimated to be in the range of $9–$23 million according to Lennart Heim, a researcher in the Epoch team. See: Lennart Heim (2022) – Estimating PaLM’s training cost.{/ref}
\n\n\n\nScale-up was already exponential and has sped up substantially over the past decade. What can we learn from this historical development for the future of AI?
\nThe rise of artificial intelligence over the last 8 decades: As training computation has increased, AI systems have become more powerful{ref}The data is taken from Jaime Sevilla, Lennart Heim, Anson Ho, Tamay Besiroglu, Marius Hobbhahn, Pablo Villalobos (2022) – Compute Trends Across Three eras of Machine Learning. Published in arXiv on March 9, 2022. See also their post on the Alignment Forum.
\n\n\n\nThe authors regularly update and extend their dataset, a very helpful service to the AI research community. At Our World in Data my colleague Charlie Giattino regularly updates the interactive version of this chart with the latest data made available by Sevilla and coauthors.
\n\n\n\nSee also these two related charts:
\n\n\n\nNumber of parameters in notable artificial intelligence systems
\n\n\n\nNumber of datapoints used to train notable artificial intelligence systems{/ref}
\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/ai-training-computation-3.png\"")
AI researchers study these long-term trends to see what is possible in the future.{ref}Scaling up the size of neural networks – in terms of the number of parameters and the amount of training data and computation – has led to surprising increases in the capabilities of AI systems. This realization motivated the “scaling hypothesis.” See Gwern Branwen (2020) – The Scaling Hypothesis.{/ref}
\n\n\n\nPerhaps the most widely discussed study of this kind was published by AI researcher Ajeya Cotra. She studied the increase in training computation to ask at what point in time the computation to train an AI system could match that of the human brain. The idea is that at this point the AI system would match the capabilities of a human brain. In her latest update, Cotra estimated a 50% probability that such “transformative AI” will be developed by the year 2040, less than two decades from now.{ref}Her research was announced in various places, including in the AI Alignment Forum here: Ajeya Cotra (2020) – Draft report on AI timelines. As far as I know the report itself always remained a ‘draft report’ and was published here on Google Docs.
\n\n\n\nThe cited estimate stems from Cotra’s Two-year update on my personal AI timelines, in which she shortened her median timeline by 10 years.
\n\n\n\nCotra emphasizes that there are substantial uncertainties around her estimates and therefore communicates her findings in a range of scenarios. She published her big study in 2020 and her median estimate at the time was that around the year 2050 there will be a 50%-probability that the computation required to train such a model may become affordable. In her “most conservative plausible”-scenario this point in time is pushed back to around the year 2090 and in her “most aggressive plausible”-scenario this point in time is reached in 2040.
\n\n\n\nThe same is true for most other forecasters: all emphasize the large uncertainty associated with any of their forecasts.
\n\n\n\nIt is worth emphasizing that the computation of the human brain is highly uncertain. See Joseph Carlsmith’s New Report on How Much Computational Power It Takes to Match the Human Brain from 2020.{/ref}
\n\n\n\nIn a related article, I discuss what transformative AI would mean for the world. In short, the idea is that such an AI system would be powerful enough to bring the world into a ‘qualitatively different future’. It could lead to a change at the scale of the two earlier major transformations in human history, the agricultural and industrial revolutions. It would certainly represent the most important global change in our lifetimes.
\n\n\n\nCotra’s work is particularly relevant in this context as she based her forecast on the kind of historical long-run trend of training computation that we just studied. But it is worth noting that other forecasters who rely on different considerations arrive at broadly similar conclusions. As I show in my article on AI timelines, many AI experts believe that there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner.
\n\n\n\nComputers and artificial intelligence have changed our world immensely, but we are still at the early stages of this history. Because this technology feels so familiar, it is easy to forget that all of these technologies that we interact with are very recent innovations, and that most profound changes are yet to come.
\n\n\n\nArtificial intelligence has already changed what we see, what we know, and what we do. And this is despite the fact that this technology has had only a brief history.
\n\n\n\nThere are no signs that these trends are hitting any limits anytime soon. To the contrary, particularly over the course of the last decade, the fundamental trends have accelerated: investments in AI technology have rapidly increased, and the doubling time of training computation has shortened to just six months.
\n\n\n\nAll major technological innovations lead to a range of positive and negative consequences. This is already true of artificial intelligence. As this technology becomes more and more powerful, we should expect its impact to become greater still.
\n\n\n\nBecause of the importance of AI, we should all be able to form an opinion on where this technology is heading and to understand how this development is changing our world. For this purpose, we are building a repository of AI-related metrics, which you can find on OurWorldinData.org/artificial-intelligence.
\n\n\n\nWe are still in the early stages of this history and much of what will become possible is yet to come. A technological development as powerful as this should be at the center of our attention. Little might be as important for how the future of our world – and the future of our lives – will play out.
\n\n\n\nAcknowledgements: I would like to thank my colleagues Natasha Ahuja, Daniel Bachler, Julia Broden, Charlie Giattino, Bastian Herre, Edouard Mathieu, and Ike Saunders for their helpful comments to drafts of this essay and their contributions in preparing the visualizations.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Despite their brief history, computers and AI have fundamentally changed what we see, what we know, and what we do. Little is as important for the future of the world, and our own lives, as how this history continues."", ""protected"": false}, ""date_gmt"": ""2022-12-06T01:00:00"", ""modified"": ""2023-07-31T14:05:20"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Max Roser""], ""modified_gmt"": ""2023-07-31T13:05:20"", ""comment_status"": ""closed"", ""featured_media"": 54806, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/12/featured-image-ai-training-computation-150x86.png"", ""medium_large"": ""/app/uploads/2022/12/featured-image-ai-training-computation-768x440.png""}}" 54759,Artificial intelligence is transforming our world — it is on all of us to make sure that it goes well,ai-impact,post,publish,"Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Why should you care about the development of artificial intelligence?
Think about what the alternative would look like. If you and the wider public do not get informed and engaged, then we leave it to a few entrepreneurs and engineers to decide how this technology will transform our world.
That is the status quo. This small number of people at a few tech firms directly working on artificial intelligence (AI) do understand how extraordinarily powerful this technology is becoming. If the rest of society does not become engaged, then it will be this small elite who decides how this technology will change our lives.
To change this status quo, I want to answer three questions in this article: Why is it hard to take the prospect of a world transformed by AI seriously? How can we imagine such a world? And what is at stake as this technology becomes more powerful?
In some way, it should be obvious how technology can fundamentally transform the world. We just have to look at how much the world has already changed. If you could invite a family of hunter-gatherers from 20,000 years ago on your next flight, they would be pretty surprised. Technology has changed our world already, so we should expect that it can happen again.
But while we have seen the world transform before, we have seen these transformations play out over the course of generations. What is different now is how very rapid these technological changes have become. In the past, the technologies that our ancestors used in their childhood were still central to their lives in their old age. This has not been the case anymore for recent generations. Instead, it has become common that technologies unimaginable in one's youth become ordinary in later life.
This is the first reason we might not take the prospect seriously: it is easy to underestimate the speed at which technology can change the world.
The second reason why it is difficult to take the possibility of transformative AI – potentially even AI as intelligent as humans – seriously is that it is an idea that we first heard in the cinema. It is not surprising that for many of us, the first reaction to a scenario in which machines have human-like capabilities is the same as if you had asked us to take seriously a future in which vampires, werewolves, or zombies roam the planet.{ref}This problem becomes even larger when we try to imagine how a future with a human-level AI might play out. Any particular scenario will not only involve the idea that this powerful AI exists, but a whole range of additional assumptions about the future context in which this happens. It is therefore hard to communicate a scenario of a world with human-level AI that does not sound contrived, bizarre or even silly.{/ref}
But, it is plausible that it is both the stuff of sci-fi fantasy and the central invention that could arrive in our, or our children’s, lifetimes.
The third reason why it is difficult to take this prospect seriously is by failing to see that powerful AI could lead to very large changes. This is also understandable. It is difficult to form an idea of a future that is very different from our own time. There are two concepts that I find helpful in imagining a very different future with artificial intelligence. Let’s look at both of them.
When thinking about the future of artificial intelligence, I find it helpful to consider two different concepts in particular: human-level AI, and transformative AI.{ref}Both of these concepts are widely used in the scientific literature on artificial intelligence. For example, questions about the timelines for the development of future AI are often framed using these terms. See my article on this topic.{/ref} The first concept highlights the AI’s capabilities and anchors them to a familiar benchmark, while transformative AI emphasizes the impact that this technology would have on the world.
From where we are today, much of this may sound like science fiction. It is therefore worth keeping in mind that the majority of surveyed AI experts believe there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner.
One way to think about human-level artificial intelligence is to contrast it with the current state of AI technology. While today’s AI systems often have capabilities similar to a particular, limited part of the human mind, a human-level AI would be a machine that is capable of carrying out the same range of intellectual tasks that we humans are capable of.{ref}The fact that humans are capable of a range of intellectual tasks means that you arrive at different definitions of intelligence depending on which aspect within that range you focus on (the Wikipedia entry on intelligence, for example, lists a number of definitions from various researchers and different disciplines). As a consequence there are also various definitions of ‘human-level AI’.
There are also several closely related terms: Artificial General Intelligence, High-Level Machine Intelligence, Strong AI, or Full AI are sometimes synonymously used, and sometimes defined in similar, yet different ways. In specific discussions, it is necessary to define this concept more narrowly; for example, in studies on AI timelines researchers offer more precise definitions of what human-level AI refers to in their particular study.{/ref} It is a machine that would be “able to learn to do anything that a human can do,” as Norvig and Russell put it in their textbook on AI.{ref}Peter Norvig and Stuart Russell (2021) — Artificial Intelligence: A Modern Approach. Fourth edition. Published by Pearson.{/ref}
Taken together, the range of abilities that characterize intelligence gives humans the ability to solve problems and achieve a wide variety of goals. A human-level AI would therefore be a system that could solve all those problems that we humans can solve, and do the tasks that humans do today. Such a machine, or collective of machines, would be able to do the work of a translator, an accountant, an illustrator, a teacher, a therapist, a truck driver, or the work of a trader on the world’s financial markets. Like us, it would also be able to do research and science, and to develop new technologies based on that.
The concept of human-level AI has some clear advantages. Using the familiarity of our own intelligence as a reference provides us with some clear guidance on how to imagine the capabilities of this technology.
However, it also has clear disadvantages. Anchoring the imagination of future AI systems to the familiar reality of human intelligence carries the risk that it obscures the very real differences between them.
Some of these differences are obvious. For example, AI systems will have the immense memory of computer systems, against which our own capacity to store information pales. Another obvious difference is the speed at which a machine can absorb and process information. But information storage and processing speed are not the only differences. The domains in which machines already outperform humans is steadily increasing: in chess, after matching the level of the best human players in the late 90s, AI systems reached superhuman levels more than a decade ago. In other games like Go or complex strategy games, this has happened more recently.{ref}The AI system AlphaGo, and its various successors, won against Go masters. The AI system Pluribus beat humans at no-limit Texas hold 'em poker. The AI system Cicero can strategize and use human language to win the strategy game Diplomacy. See: Meta Fundamental AI Research Diplomacy Team (FAIR), Anton Bakhtin, Noam Brown, Emily Dinan, Gabriele Farina, Colin Flaherty, Daniel Fried, et al. (2022) – ‘Human-Level Play in the Game of Diplomacy by Combining Language Models with Strategic Reasoning’. In Science 0, no. 0 (22 November 2022): eade9097. https://doi.org/10.1126/science.ade9097.{/ref}
These differences mean that an AI that is at least as good as humans in every domain would overall be much more powerful than the human mind. Even the first “human-level AI” would therefore be quite superhuman in many ways.{ref}This also poses a problem when we evaluate how the intelligence of a machine compares with the intelligence of humans. If intelligence was a general ability, a single capacity, then we could easily compare and evaluate it, but the fact that it is a range of skills makes it much more difficult to compare across machine and human intelligence. Tests for AI systems are therefore comprising a wide range of tasks. See for example Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, Jacob Steinhardt (2020) – Measuring Massive Multitask Language Understanding or the definition of what would qualify as artificial general intelligence in this Metaculus prediction.{/ref}
Human intelligence is also a bad metaphor for machine intelligence in other ways. The way we think is often very different from machines, and as a consequence the output of thinking machines can be very alien to us.
Most perplexing and most concerning are the strange and unexpected ways in which machine intelligence can fail. The AI-generated image of the horse below provides an example: on the one hand, AIs can do what no human can do – produce an image of anything, in any style (here photorealistic), in mere seconds – but on the other hand it can fail in ways that no human would fail.{ref}An overview of how AI systems can fail can be found in Charles Choi – 7 Revealing Ways AIs Fail. It is also worth reading through the AIAAIC Repository which “details recent incidents and controversies driven by or relating to AI, algorithms, and automation.""{/ref} No human would make the mistake of drawing a horse with five legs.{ref}I have taken this example from AI researcher François Chollet, who published it here.{/ref}
Imagining a powerful future AI as just another human would therefore likely be a mistake. The differences might be so large that it will be a misnomer to call such systems “human-level.”
AI-generated image of a horse{ref}Via François Chollet, who published it here. Based on Chollet’s comments it seems that this image was created by the AI system ‘Stable Diffusion’.{/ref}
In contrast, the concept of transformative AI is not based on a comparison with human intelligence. This has the advantage of sidestepping the problems that the comparisons with our own mind bring. But it has the disadvantage that it is harder to imagine what such a system would look like and be capable of. It requires more from us. It requires us to imagine a world with intelligent actors that are potentially very different from ourselves.
Transformative AI is not defined by any specific capabilities, but by the real-world impact that the AI would have. To qualify as transformative, researchers think of it as AI that is “powerful enough to bring us into a new, qualitatively different future.”{ref}This quote is from Holden Karnofsky (2021) – AI Timelines: Where the Arguments, and the ""Experts,"" Stand. For Holden Karnofsky’s earlier thinking on this conceptualization of AI see his 2016 article ‘Some Background on Our Views Regarding Advanced Artificial Intelligence’.
Ajeya Cotra, whose research on AI timelines I discuss in other articles of this series, attempts to give a quantitative definition of what would qualify as transformative AI. in her widely cited report on AI timelines she defines it as a change in software technology that brings the growth rate of gross world product ""to 20%-30% per year"". Several other researchers define TAI in similar terms.{/ref}
In humanity’s history, there have been two cases of such major transformations, the agricultural and the industrial revolutions.
Transformative AI becoming a reality would be an event on that scale. Like the arrival of agriculture 10,000 years ago, or the transition from hand- to machine-manufacturing, it would be an event that would change the world for billions of people around the globe and for the entire trajectory of humanity’s future.
Technologies that fundamentally change how a wide range of goods or services are produced are called ‘general-purpose technologies’. The two previous transformative events were caused by the discovery of two particularly significant general-purpose technologies: the change in food production as humanity transitioned from hunting and gathering to farming, and the rise of machine manufacturing in the industrial revolution. Based on the evidence and arguments presented in this series on AI development, I believe it is plausible that powerful AI could represent the introduction of a similarly significant general-purpose technology.
Timeline of the three transformative events in world history
The two concepts are closely related, but they are not the same. The creation of a human-level AI would certainly have a transformative impact on our world. If the work of most humans could be carried out by an AI, the lives of millions of people would change.{ref}Human-level AI is typically defined as a software system that can carry out at least 90% or 99% of all economically relevant tasks that humans carry out. A lower-bar definition would be an AI system that can carry out all those tasks that can currently be done by another human who is working remotely on a computer.{/ref}
The opposite, however, is not true: we might see transformative AI without developing human-level AI. Since the human mind is in many ways a poor metaphor for the intelligence of machines, we might plausibly develop transformative AI before we develop human-level AI. Depending on how this goes, this might mean that we will never see any machine intelligence for which human intelligence is a helpful comparison.
When and if AI systems might reach either of these levels is of course difficult to predict. In my companion article on this question, I give an overview of what researchers in this field currently believe. Many AI experts believe there is a real chance that such systems will be developed within the next decades, and some believe that they will exist much sooner.
All major technological innovations lead to a range of positive and negative consequences. For AI, the spectrum of possible outcomes – from the most negative to the most positive – is extraordinarily wide.
That the use of AI technology can cause harm is clear, because it is already happening.
AI systems can cause harm when people use them maliciously. For example, when they are used in politically-motivated disinformation campaigns or to enable mass surveillance.{ref}On the use of AI in politically-motivated disinformation campaigns see for example John Villasenor (November 2020) – How to deal with AI-enabled disinformation. More generally on this topic see Brundage and Avin et al. (2018) – The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation, published at maliciousaireport.com. A starting point for literature and reporting on mass surveillance by governments is the relevant Wikipedia entry.{/ref}
But AI systems can also cause unintended harm, when they act differently than intended or fail. For example, in the Netherlands the authorities used an AI system which falsely claimed that an estimated 26,000 parents made fraudulent claims for child care benefits. The false allegations led to hardship for many poor families, and also resulted in the resignation of the Dutch government in 2021.{ref}See for example the Wikipedia entry on the ‘Dutch childcare benefits scandal’ and Melissa Heikkilä (2022) – ‘Dutch scandal serves as a warning for Europe over risks of using algorithms’, in Politico. The technology can also reinforce discrimination in terms of race and gender. See Brian Christian’s book The Alignment Problem and the reports of the AI Now Institute.{/ref}
As AI becomes more powerful, the possible negative impacts could become much larger. Many of these risks have rightfully received public attention: more powerful AI could lead to mass labor displacement, or extreme concentrations of power and wealth. In the hands of autocrats, it could empower totalitarianism through its suitability for mass surveillance and control.
The so-called alignment problem of AI is another extreme risk. This is the concern that nobody would be able to control a powerful AI system, even if the AI takes actions that harm us humans, or humanity as a whole. This risk is unfortunately receiving little attention from the wider public, but it is seen as an extremely large risk by many leading AI researchers.{ref}Overviews are provided in Stuart Russell (2019) – Human Compatible (especially chapter 5) and Brian Christian’s 2020 book The Alignment Problem. Christian presents the thinking of many leading AI researchers from the earliest days up to now and presents an excellent overview of this problem. It is also seen as a large risk by some of the leading private firms who work towards powerful AI – see OpenAI's article ""Our approach to alignment research"" from August 2022.{/ref}
How could an AI possibly escape human control and end up harming humans?
The risk is not that an AI becomes self-aware, develops bad intentions, and “chooses” to do this. The risk is that we try to instruct the AI to pursue some specific goal – even a very worthwhile one – and in the pursuit of that goal it ends up harming humans. It is about unintended consequences. The AI does what we told it to do, but not what we wanted it to do.
Can’t we just tell the AI to not do those things? It is definitely possible to build an AI that avoids any particular problem we foresee, but it is hard to foresee all the possible harmful unintended consequences. The alignment problem arises because of “the impossibility of defining true human purposes correctly and completely,” as AI researcher Stuart Russell puts it.{ref}Stuart Russell (2019) – Human Compatible{/ref}
Can’t we then just switch off the AI? This might also not be possible. That is because a powerful AI would know two things: it faces a risk that humans could turn it off, and it can’t achieve its goals once it has been turned off. As a consequence, the AI will pursue a very fundamental goal of ensuring that it won’t be switched off. This is why, once we realize that an extremely intelligent AI is causing unintended harm in the pursuit of some specific goal, it might not be possible to turn it off or change what the system does.{ref}A question that follows from this is, why build such a powerful AI in the first place?
The incentives are very high. As I emphasize below, this innovation has the potential to lead to very positive developments. In addition to the large social benefits there are also large incentives for those who develop it – the governments that can use it for their goals, the individuals who can use it to become more powerful and wealthy. Additionally, it is of scientific interest and might help us to understand our own mind and intelligence better. And lastly, even if we wanted to stop building powerful AIs, it is likely very hard to actually achieve it. It is very hard to coordinate across the whole world and agree to stop building more advanced AI – countries around the world would have to agree and then find ways to actually implement it.{/ref}
This risk – that humanity might not be able to stay in control once AI becomes very powerful, and that this might lead to an extreme catastrophe – has been recognized right from the early days of AI research more than 70 years ago.{ref}In 1950 the computer science pioneer Alan Turing put it like this: “If a machine can think, it might think more intelligently than we do, and then where should we be? … [T]his new danger is much closer. If it comes at all it will almost certainly be within the next millennium. It is remote but not astronomically remote, and is certainly something which can give us anxiety. It is customary, in a talk or article on this subject, to offer a grain of comfort, in the form of a statement that some particularly human characteristic could never be imitated by a machine. … I cannot offer any such comfort, for I believe that no such bounds can be set.” Alan. M. Turing (1950) – Computing Machinery and Intelligence, In Mind, Volume LIX, Issue 236, October 1950, Pages 433–460.
Norbert Wiener is another pioneer who saw the alignment problem very early. One way he put it was “If we use, to achieve our purposes, a mechanical agency with whose operation we cannot interfere effectively … we had better be quite sure that the purpose put into the machine is the purpose which we really desire.” quoted from Norbert Wiener (1960) – Some Moral and Technical Consequences of Automation: As machines learn they may develop unforeseen strategies at rates that baffle their programmers. In Science.
In 1950 – the same year in which Turing published the cited article – Wiener published his book The Human Use of Human Beings, whose front-cover blurb reads: “The ‘mechanical brain’ and similar machines can destroy human values or enable us to realize them as never before.”{/ref} The very rapid development of AI in recent years has made a solution to this problem much more urgent.
I have tried to summarize some of the risks of AI, but a short article is not enough space to address all possible questions. Especially on the very worst risks of AI systems, and what we can do now to reduce them, I recommend reading the book The Alignment Problem by Brian Christian and Benjamin Hilton’s article ‘Preventing an AI-related catastrophe’.
If we manage to avoid these risks, transformative AI could also lead to very positive consequences. Advances in science and technology were crucial to the many positive developments in humanity’s history. If artificial ingenuity can augment our own, it could help us make progress on the many large problems we face: from cleaner energy, to the replacement of unpleasant work, to much better healthcare.
This extremely large contrast between the possible positives and negatives makes clear that the stakes are unusually high with this technology. Reducing the negative risks and solving the alignment problem could mean the difference between a healthy, flourishing, and wealthy future for humanity – and the destruction of the same.
Making sure that the development of artificial intelligence goes well is not just one of the most crucial questions of our time, but likely one of the most crucial questions in human history. This needs public resources – public funding, public attention, and public engagement.
Currently, almost all resources that are dedicated to AI aim to speed up the development of this technology. Efforts that aim to increase the safety of AI systems, on the other hand, do not receive the resources they need. Researcher Toby Ord estimated that in 2020 between $10 to $50 million was spent on work to address the alignment problem.{ref}Toby Ord – The Precipice. He makes this projection in footnote 55 of chapter 2. It is based on the 2017 estimate by Farquhar.{/ref} Corporate AI investment in the same year was more than 2000-times larger, it summed up to $153 billion.
This is not only the case for the AI alignment problem. The work on the entire range of negative social consequences from AI is under-resourced compared to the large investments to increase the power and use of AI systems.
It is frustrating and concerning for society as a whole that AI safety work is extremely neglected and that little public funding is dedicated to this crucial field of research. On the other hand, for each individual person this neglect means that they have a good chance to actually make a positive difference, if they dedicate themselves to this problem now. And while the field of AI safety is small, it does provide good resources on what you can do concretely if you want to work on this problem.
I hope that more people dedicate their individual careers to this cause, but it needs more than individual efforts. A technology that is transforming our society needs to be a central interest of all of us. As a society we have to think more about the societal impact of AI, become knowledgeable about the technology, and understand what is at stake.
When our children look back at today, I imagine that they will find it difficult to understand how little attention and resources we dedicated to the development of safe AI. I hope that this changes in the coming years, and that we begin to dedicate more resources to making sure that powerful AI gets developed in a way that benefits us and the next generations.
If we fail to develop this broad-based understanding, then it will remain the small elite that finances and builds this technology that will determine how one of the – or plausibly the – most powerful technology in human history will transform our world.
With our work at Our World in Data we want to do our small part to enable a better informed public conversation on AI and the future we want to live in. You can find these resources on OurWorldinData.org/artificial-intelligence
Acknowledgements: I would like to thank my colleagues Daniel Bachler, Charlie Giattino, and Edouard Mathieu for their helpful comments to drafts of this essay.
","{""id"": ""wp-54759"", ""slug"": ""ai-impact"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Our World in Data presents the data and research to make progress against the world’s largest problems."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""This article draws on data and research discussed in our entry on "", ""spanType"": ""span-simple-text""}, {""children"": [{""url"": ""https://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""Artificial Intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""spanType"": ""span-bold""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Why should you care about the development of artificial intelligence?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Think about what the alternative would look like. If you and the wider public do not get informed and engaged, then we leave it to a few entrepreneurs and engineers to decide how this technology will transform our world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That is the status quo. This small number of people at a few tech firms directly working on artificial intelligence (AI) do understand how extraordinarily powerful this technology is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/brief-history-of-ai"", ""children"": [{""text"": ""becoming"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". If the rest of society does not become engaged, then it will be this small elite who decides how this technology will change our lives."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To change this status quo, I want to answer three questions in this article: Why is it hard to take the prospect of a world transformed by AI seriously? How can we imagine such a world? And what is at stake as this technology becomes more powerful?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Why is it hard to take the prospect of a world transformed by artificial intelligence seriously?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In some way, it should be obvious how technology can fundamentally transform the world. We just have to look at how much the world has already changed. If you could invite a family of hunter-gatherers from 20,000 years ago on your next flight, they would be pretty surprised. Technology has changed our world already, so we should expect that it can happen again."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But while we have seen the world transform before, we have seen these transformations play out over the course of generations. What is different now is how very rapid these technological changes have become. In the past, the technologies that our ancestors used in their childhood were still central to their lives in their old age. This has not been the case anymore for recent generations. Instead, it has "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/technology-long-run"", ""children"": [{""text"": ""become common"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" that technologies unimaginable in one's youth become ordinary in later life."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is the first reason we might not take the prospect seriously: it is easy to underestimate the speed at which technology can change the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The second reason why it is difficult to take the possibility of transformative AI – potentially even AI as intelligent as humans – seriously is that it is an idea that we first heard in the cinema. It is not surprising that for many of us, the first reaction to a scenario in which machines have human-like capabilities is the same as if you had asked us to take seriously a future in which vampires, werewolves, or zombies roam the planet.{ref}This problem becomes even larger when we try to imagine how a future with a human-level AI might play out. Any "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""particular"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" scenario will not only involve the idea that this powerful AI exists, but a whole range of additional assumptions about the future context in which this happens. It is therefore hard to communicate a scenario of a world with human-level AI that does not sound contrived, bizarre or even silly.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But, it is plausible that it is both the stuff of sci-fi fantasy "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""and"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" the central invention that could arrive in our, or our children’s, lifetimes. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The third reason why it is difficult to take this prospect seriously is by failing to see that powerful AI could lead to very large changes. This is also understandable. It is difficult to form an idea of a future that is very different from our own time. There are two concepts that I find helpful in imagining a very different future with artificial intelligence. Let’s look at both of them."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How to develop an idea of what the future of artificial intelligence might look like?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""When thinking about the future of artificial intelligence, I find it helpful to consider two different concepts in particular: human-level AI, and transformative AI.{ref}Both of these concepts are widely used in the scientific literature on artificial intelligence. For example, questions about the timelines for the development of future AI are often framed using these terms. See "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""my article on this topic"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref} The first concept highlights the AI’s capabilities and anchors them to a familiar benchmark, while transformative AI emphasizes the impact that this technology would have on the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""From where we are today, much of this may sound like science fiction. It is therefore worth keeping in mind that the majority of surveyed AI experts "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""believe"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""The advantages and disadvantages of comparing machine and human intelligence"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One way to think about human-level artificial intelligence is to contrast it with the current state of AI technology. While today’s AI systems often have capabilities similar to a particular, limited part of the human mind, a human-level AI would be a machine that is capable of carrying out the same "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""range"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of intellectual tasks that we humans are capable of.{ref}The fact that humans are capable of a "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""range"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of intellectual tasks means that you arrive at different definitions of intelligence depending on which aspect within that range you focus on (the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Intelligence"", ""children"": [{""text"": ""Wikipedia entry on intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", for example, lists a number of definitions from various researchers and different disciplines). As a consequence there are also various definitions of ‘human-level AI’. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are also several closely related terms: Artificial General Intelligence, High-Level Machine Intelligence, Strong AI, or Full AI are sometimes synonymously used, and sometimes defined in similar, yet different ways. In specific discussions, it is necessary to define this concept more narrowly; for example, in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""studies on AI timelines"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" researchers offer more precise definitions of what human-level AI refers to in their particular study.{/ref} It is a machine that would be “able to learn to do anything that a human can do,” as Norvig and Russell put it in their textbook on AI.{ref}Peter Norvig and Stuart Russell (2021) — Artificial Intelligence: A Modern Approach. Fourth edition. Published by Pearson.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Taken together, the range of abilities that characterize intelligence gives humans the ability to solve problems and achieve a wide variety of goals. A human-level AI would therefore be a system that could solve all those problems that we humans can solve, and do the tasks that humans do today. Such a machine, or collective of machines, would be able to do the work of a translator, an accountant, an illustrator, a teacher, a therapist, a truck driver, or the work of a trader on the world’s financial markets. Like us, it would also be able to do research and science, and to develop new technologies based on that."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The concept of human-level AI has some clear advantages. Using the familiarity of our own intelligence as a reference provides us with some clear guidance on how to imagine the capabilities of this technology. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""However, it also has clear disadvantages. Anchoring the imagination of future AI systems to the familiar reality of human intelligence carries the risk that it obscures the very real differences between them. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Some of these differences are obvious. For example, AI systems will have the immense memory of computer systems, against which our own capacity to store information pales. Another obvious difference is the speed at which a machine can absorb and process information. But information storage and processing speed are not the only differences. The domains in which machines already outperform humans is steadily increasing: in chess, after matching the level of the best human players in the late 90s, AI systems "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/computer-chess-ability"", ""children"": [{""text"": ""reached"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" superhuman levels more than a decade ago. In other games like Go or complex strategy games, this has happened more recently.{ref}The AI system "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/AlphaGo"", ""children"": [{""text"": ""AlphaGo"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", and its various successors, won against Go masters. The AI system "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Pluribus_(poker_bot)"", ""children"": [{""text"": ""Pluribus"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" beat humans at no-limit Texas hold 'em poker. The AI system Cicero can strategize and use human language to win the strategy game Diplomacy. See: Meta Fundamental AI Research Diplomacy Team (FAIR), Anton Bakhtin, Noam Brown, Emily Dinan, Gabriele Farina, Colin Flaherty, Daniel Fried, et al. (2022) – ‘Human-Level Play in the Game of Diplomacy by Combining Language Models with Strategic Reasoning’. In "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" 0, no. 0 (22 November 2022): eade9097."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1126/science.ade9097"", ""children"": [{""text"": "" https://doi.org/10.1126/science.ade9097"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These differences mean that an AI that is at least as good as humans in every domain would overall be much more powerful than the human mind. Even the first “human-level AI” would therefore be quite superhuman in many ways.{ref}This also poses a problem when we evaluate how the intelligence of a machine compares with the intelligence of humans. If intelligence was a general ability, a single capacity, then we could easily compare and evaluate it, but the fact that it is a range of skills makes it much more difficult to compare across machine and human intelligence. Tests for AI systems are therefore comprising a wide range of tasks. See for example Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, Jacob Steinhardt (2020) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://arxiv.org/abs/2009.03300"", ""children"": [{""text"": ""Measuring Massive Multitask Language Understanding"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" or the definition of what would qualify as artificial general intelligence in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/"", ""children"": [{""text"": ""this Metaculus prediction"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Human intelligence is also a bad metaphor for machine intelligence in other ways. The way we think is often very different from machines, and as a consequence the output of thinking machines can be very alien to us."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Most perplexing and most concerning are the strange and unexpected ways in which machine intelligence can fail. The AI-generated image of the horse below provides an example: on the one hand, AIs can do what no human can do – produce an image of anything, in any style (here photorealistic), in mere seconds – but on the other hand it can fail in ways that no human would fail.{ref}An overview of how AI systems can fail can be found in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://spectrum.ieee.org/ai-failures"", ""children"": [{""text"": ""Charles Choi – 7 Revealing Ways AIs Fail"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". It is also worth reading through the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.aiaaic.org/aiaaic-repository/ai-and-algorithmic-incidents-and-controversies"", ""children"": [{""text"": ""AIAAIC Repository"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" which “details recent incidents and controversies driven by or relating to AI, algorithms, and automation.\""{/ref} No human would make the mistake of drawing a horse with five legs.{ref}I have taken this example from "", ""spanType"": ""span-simple-text""}, {""url"": ""https://fchollet.com/"", ""children"": [{""text"": ""AI researcher François Chollet"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", who published it "", ""spanType"": ""span-simple-text""}, {""url"": ""https://twitter.com/fchollet/status/1573752180720312320?s=46&t=qPwLwDgLdJrLlXxa878BDQ"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Imagining a powerful future AI as just another human would therefore likely be a mistake. The differences might be so large that it will be a misnomer to call such systems “human-level.”"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""AI-generated image of a horse"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""{ref}Via "", ""spanType"": ""span-simple-text""}, {""url"": ""https://fchollet.com/"", ""children"": [{""text"": ""François Chollet"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", who published it "", ""spanType"": ""span-simple-text""}, {""url"": ""https://twitter.com/fchollet/status/1573752180720312320?s=46&t=qPwLwDgLdJrLlXxa878BDQ"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Based on Chollet’s comments it seems that this image was created by the AI system ‘Stable Diffusion’.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""alt"": ""A brown horse running in a grassy field. The horse appears to have five legs."", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""ai-generated-image-of-a-horse.png"", ""parseErrors"": []}, {""text"": [{""text"": ""Transformative artificial intelligence is defined by the impact this technology would have on the world"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In contrast, the concept of transformative AI is not based on a comparison with human intelligence. This has the advantage of sidestepping the problems that the comparisons with our own mind bring. But it has the disadvantage that it is harder to imagine what such a system would look like and be capable of. It requires more from us. It requires us to imagine a world with intelligent actors that are potentially very different from ourselves."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Transformative AI is not defined by any specific capabilities, but by the real-world impact that the AI would have. To qualify as transformative, researchers think of it as AI that is “powerful enough to bring us into a new, qualitatively different future.”{ref}This quote is from Holden Karnofsky (2021) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.cold-takes.com/where-ai-forecasting-stands-today/"", ""children"": [{""text"": ""AI Timelines: Where the Arguments, and the \""Experts,\"" Stand"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". For Holden Karnofsky’s earlier thinking on this conceptualization of AI see his 2016 article "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.openphilanthropy.org/research/some-background-on-our-views-regarding-advanced-artificial-intelligence/#Sec1"", ""children"": [{""text"": ""‘Some Background on Our Views Regarding Advanced Artificial Intelligence’"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Ajeya Cotra, whose research on AI timelines I discuss in other articles of this series, attempts to give a quantitative definition of what would qualify as transformative AI. in her widely cited "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.alignmentforum.org/posts/KrJfoZzpSDpnrv9va/draft-report-on-ai-timelines"", ""children"": [{""text"": ""report on AI timelines"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" she defines it as a change in software technology that brings the growth rate of gross world product \""to 20%-30% per year\"". Several other researchers define TAI in similar terms.{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In humanity’s history, there have been two cases of such major transformations, the agricultural and the industrial revolutions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Transformative AI becoming a reality would be an event on that scale. Like the arrival of agriculture 10,000 years ago, or the transition from hand- to machine-manufacturing, it would be an event that would change the world for billions of people around the globe and for the entire trajectory of "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/longtermism"", ""children"": [{""text"": ""humanity’s future"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Technologies that fundamentally change how a wide range of goods or services are produced are called ‘general-purpose technologies’. The two previous transformative events were caused by the discovery of two particularly significant general-purpose technologies: the change in food production as humanity transitioned from hunting and gathering to farming, and the rise of machine manufacturing in the industrial revolution. Based on the evidence and arguments presented in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/artificial-intelligence#research-writing"", ""children"": [{""text"": ""this series"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on AI development, I believe it is plausible that powerful AI could represent the introduction of a similarly significant general-purpose technology."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Timeline of the three transformative events in world history"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Timeline-of-Transformative-Events.png"", ""parseErrors"": []}, {""text"": [{""text"": ""A future of human-level or transformative AI?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The two concepts are closely related, but they are not the same. The creation of a human-level AI would certainly have a transformative impact on our world. If the work of most humans could be carried out by an AI, the lives of millions of people would change.{ref}Human-level AI is typically defined as a software system that can carry out at least 90% or 99% of all economically relevant tasks that humans carry out. A lower-bar definition would be an AI system that can carry out all those tasks that can currently be done by another human who is working remotely on a computer.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The opposite, however, is not true: we might see transformative AI without developing human-level AI. Since the human mind is in many ways a poor metaphor for the intelligence of machines, we might plausibly develop transformative AI before we develop human-level AI. Depending on how this goes, this might mean that we will never see any machine intelligence for which human intelligence is a helpful comparison."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""When and if AI systems might reach either of these levels is of course difficult to predict. In my "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ai-timelines"", ""children"": [{""text"": ""companion article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on this question, I give an overview of what researchers in this field currently believe. Many AI experts believe there is a real chance that such systems will be developed within the next decades, and some believe that they will exist much sooner."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What is at stake as artificial intelligence becomes more powerful?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""All major technological innovations lead to a range of positive and negative consequences. For AI, the spectrum of possible outcomes – from the most negative to the most positive – is extraordinarily wide. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That the use of AI technology can cause harm is clear, because it is already happening. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""AI systems can cause harm when people use them maliciously. For example, when they are used in politically-motivated disinformation campaigns or to enable mass surveillance.{ref}On the use of AI in politically-motivated disinformation campaigns see for example John Villasenor (November 2020) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20220907044354/https://www.brookings.edu/research/how-to-deal-with-ai-enabled-disinformation/"", ""children"": [{""text"": ""How to deal with AI-enabled disinformation"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". More generally on this topic see Brundage and Avin et al. (2018) – The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation, published at "", ""spanType"": ""span-simple-text""}, {""url"": ""https://maliciousaireport.com/"", ""children"": [{""text"": ""maliciousaireport.com"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". A starting point for literature and reporting on mass surveillance by governments is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/List_of_government_mass_surveillance_projects"", ""children"": [{""text"": ""the relevant Wikipedia entry"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But AI systems can also cause unintended harm, when they act differently than intended or fail. For example, in the Netherlands the authorities used an AI system which falsely claimed that an estimated 26,000 parents made fraudulent claims for child care benefits. The false allegations led to hardship for many poor families, and also resulted in the resignation of the Dutch government in 2021.{ref}See for example the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/Dutch_childcare_benefits_scandal"", ""children"": [{""text"": ""Wikipedia entry"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on the ‘Dutch childcare benefits scandal’ and Melissa Heikkilä (2022) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20221117053636/https://www.politico.eu/article/dutch-scandal-serves-as-a-warning-for-europe-over-risks-of-using-algorithms/"", ""children"": [{""text"": ""‘Dutch scandal serves as a warning for Europe over risks of using algorithms’"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", in Politico. The technology can also reinforce discrimination in terms of race and gender. See Brian Christian’s book The Alignment Problem and the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ainowinstitute.org/reports.html"", ""children"": [{""text"": ""reports of the AI Now Institute"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As AI becomes more powerful, the possible negative impacts could become much larger. Many of these risks have rightfully received public attention: more powerful AI could lead to mass labor displacement, or extreme concentrations of power and wealth. In the hands of autocrats, it could empower totalitarianism through its suitability for mass surveillance and control. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The so-called "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""alignment problem"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of AI is another extreme risk. This is the concern that "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""nobody"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" would be able to control a powerful AI system, even if the AI takes actions that harm us humans, or humanity as a whole. This risk is unfortunately receiving little attention from the wider public, but it is seen as an extremely large risk by many leading AI researchers.{ref}Overviews are provided in Stuart Russell (2019) – Human Compatible (especially chapter 5) and Brian Christian’s 2020 book "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/The_Alignment_Problem"", ""children"": [{""text"": ""The Alignment Problem"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Christian presents the thinking of many leading AI researchers from the earliest days up to now and presents an excellent overview of this problem. It is also seen as a large risk by some of the leading private firms who work towards powerful AI – see OpenAI's article \"""", ""spanType"": ""span-simple-text""}, {""url"": ""https://openai.com/blog/our-approach-to-alignment-research/"", ""children"": [{""text"": ""Our approach to alignment research"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""\"" from August 2022.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How could an AI possibly escape human control and end up harming humans?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The risk is not that an AI becomes self-aware, develops bad intentions, and “chooses” to do this. The risk is that we try to instruct the AI to pursue some specific goal – even a very worthwhile one – and in the pursuit of that goal it ends up harming humans. It is about unintended consequences. The AI does what we told it to do, but not what we wanted it to do."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Can’t we just tell the AI to not do those things? It is definitely possible to build an AI that avoids any particular problem we foresee, but it is hard to foresee "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""all"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" the possible harmful unintended consequences. The alignment problem arises because of “the impossibility of defining true human purposes correctly and completely,” as AI researcher Stuart Russell puts it.{ref}Stuart Russell (2019) – Human Compatible{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Can’t we then just switch off the AI? This might also not be possible. That is because a powerful AI would know two things: it faces a risk that humans could turn it off, and it can’t achieve its goals once it has been turned off. As a consequence, the AI will pursue a very fundamental goal of ensuring that it won’t be switched off. This is why, once we realize that an extremely intelligent AI is causing unintended harm in the pursuit of some specific goal, it might not be possible to turn it off or change what the system does.{ref}A question that follows from this is, why build such a powerful AI in the first place? "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The incentives are very high. As I emphasize below, this innovation has the potential to lead to very positive developments. In addition to the large social benefits there are also large incentives for those who develop it – the governments that can use it for their goals, the individuals who can use it to become more powerful and wealthy. Additionally, it is of scientific interest and might help us to understand our own mind and intelligence better. And lastly, even if we wanted to stop building powerful AIs, it is likely very hard to actually achieve it. It is very hard to coordinate across the whole world and agree to stop building more advanced AI – countries around the world would have to agree and then find ways to actually implement it.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This risk – that humanity might not be able to stay in control once AI becomes very powerful, and that this might lead to an extreme catastrophe – has been recognized right from the early days of AI research more than 70 years ago.{ref}In 1950 the computer science pioneer Alan Turing put it like this: "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""“If a machine can think, it might think more intelligently than we do, and then where should we be? … [T]his new danger is much closer. If it comes at all it will almost certainly be within the next millennium. It is remote but not astronomically remote, and is certainly something which can give us anxiety. It is customary, in a talk or article on this subject, to offer a grain of comfort, in the form of a statement that some particularly human characteristic could never be imitated by a machine. … I cannot offer any such comfort, for I believe that no such bounds can be set.”"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" Alan. M. Turing (1950) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1093/mind/LIX.236.433"", ""children"": [{""text"": ""Computing Machinery and Intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", In Mind, Volume LIX, Issue 236, October 1950, Pages 433–460."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Norbert Wiener is another pioneer who saw the alignment problem very early. One way he put it was “If we use, to achieve our purposes, a mechanical agency with whose operation we cannot interfere effectively … we had better be quite sure that the purpose put into the machine is the purpose which we really desire.” quoted from Norbert Wiener (1960) – Some Moral and Technical Consequences of Automation: As machines learn they may develop unforeseen strategies at rates that baffle their programmers. In Science."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In 1950 – the same year in which Turing published the cited article – Wiener published his book The Human Use of Human Beings, whose front-cover blurb reads: “The ‘mechanical brain’ and similar machines can destroy human values or enable us to realize them as never before.”{/ref} The very rapid development of AI in recent years has made a solution to this problem much more urgent."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I have tried to summarize some of the risks of AI, but a short article is not enough space to address all possible questions. Especially on the very worst risks of AI systems, and what we can do now to reduce them, I recommend reading the book "", ""spanType"": ""span-simple-text""}, {""url"": ""https://brianchristian.org/the-alignment-problem/"", ""children"": [{""text"": ""The Alignment Problem"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" by Brian Christian and Benjamin Hilton’s article "", ""spanType"": ""span-simple-text""}, {""url"": ""https://80000hours.org/problem-profiles/artificial-intelligence"", ""children"": [{""text"": ""‘Preventing an AI-related catastrophe’"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If we manage to avoid these risks, transformative AI could also lead to very positive consequences. Advances in science and technology were crucial to "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/a-history-of-global-living-conditions-in-5-charts"", ""children"": [{""text"": ""the many positive developments"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in humanity’s history. If artificial ingenuity can augment our own, it could help us make progress on the many large problems we face: from cleaner energy, to the replacement of unpleasant work, to much better healthcare."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This extremely large contrast between the possible positives and negatives makes clear that the stakes are unusually high with this technology. Reducing the negative risks and solving the alignment problem could mean the difference between a healthy, flourishing, and wealthy future for humanity – and the destruction of the same. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How can we make sure that the development of AI goes well?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Making sure that the development of artificial intelligence goes well is not just one of the most crucial questions of our time, but likely one of the most crucial questions in human history. This needs public resources – public funding, public attention, and public engagement."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Currently, almost all resources that are dedicated to AI aim to speed up the development of this technology. Efforts that aim to increase the safety of AI systems, on the other hand, do not receive the resources they need. Researcher Toby Ord estimated that in 2020 between $10 to $50 million was spent on work to address the alignment problem.{ref}Toby Ord – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://theprecipice.com/"", ""children"": [{""text"": ""The Precipice"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". He makes this projection in footnote 55 of chapter 2. It is based on the 2017 estimate by Farquhar.{/ref} Corporate AI investment in the same year was more than 2000-times larger, it "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/corporate-investment-in-artificial-intelligence-by-type"", ""children"": [{""text"": ""summed up"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to $153 billion. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is not only the case for the AI alignment problem. The work on the entire range of negative social consequences from AI is under-resourced compared to the large investments to increase the power and use of AI systems."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It is frustrating and concerning for society as a whole that AI safety work is extremely neglected and that little public funding is dedicated to this crucial field of research. On the other hand, for each "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""individual"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" person this neglect means that they have a good chance to actually make a positive difference, if they dedicate themselves to this problem now. And while the field of AI safety is small, it does provide "", ""spanType"": ""span-simple-text""}, {""url"": ""https://80000hours.org/problem-profiles/artificial-intelligence/#what-can-you-do-concretely-to-help"", ""children"": [{""text"": ""good resources"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on what you can do concretely if you want to work on this problem."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I hope that more people dedicate their individual careers to this cause, but it needs more than individual efforts. A technology that is transforming our society needs to be a central interest of all of us. As a society we have to think more about the societal impact of AI, become knowledgeable about the technology, and understand what is at stake. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""When our children look back at today, I imagine that they will find it difficult to understand how little attention and resources we dedicated to the development of safe AI. I hope that this changes in the coming years, and that we begin to dedicate more resources to making sure that powerful AI gets developed in a way that benefits us and the next generations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If we fail to develop this broad-based understanding, then it will remain the small elite that finances and builds this technology that will determine how one of the – or plausibly "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""the"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" – most powerful technology in human history will transform our world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""With our work at Our World in Data we want to do our small part to enable a better informed public conversation on AI and the future we want to live in. You can find these resources on "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/artificial-intelligence"", ""children"": [{""text"": ""OurWorldinData.org/artificial-intelligence"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Acknowledgements:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" I would like to thank my colleagues Daniel Bachler, Charlie Giattino, and Edouard Mathieu for their helpful comments to drafts of this essay."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Artificial intelligence is transforming our world — it is on all of us to make sure that it goes well"", ""authors"": [""Max Roser""], ""excerpt"": ""How AI gets built is currently decided by a small group of technologists. As this technology is transforming our lives, it should be in all of our interest to become informed and engaged."", ""dateline"": ""December 15, 2022"", ""subtitle"": ""How AI gets built is currently decided by a small group of technologists. As this technology is transforming our lives, it should be in all of our interest to become informed and engaged."", ""sidebar-toc"": false, ""featured-image"": ""featured-image-Timeline-of-Transformative-Events-1.png""}, ""createdAt"": ""2022-12-02T12:25:34.000Z"", ""published"": false, ""updatedAt"": ""2023-10-11T08:43:04.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-12-15T05:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag spacer""}], ""numBlocks"": 69, ""numErrors"": 5, ""wpTagCounts"": {""html"": 1, ""image"": 2, ""column"": 4, ""spacer"": 1, ""columns"": 2, ""heading"": 7, ""paragraph"": 59, ""separator"": 1}, ""htmlTagCounts"": {""p"": 60, ""h4"": 4, ""h5"": 3, ""hr"": 1, ""div"": 8, ""figure"": 2}}",2022-12-15 05:00:00,2024-02-16 14:22:54,1rzaBWNe1jarUZIhWvqJOB3Om4RzgJ1rlYu20Ic7O5Og,"[""Max Roser""]","How AI gets built is currently decided by a small group of technologists. As this technology is transforming our lives, it should be in all of our interest to become informed and engaged.",2022-12-02 12:25:34,2023-10-11 08:43:04,https://ourworldindata.org/wp-content/uploads/2022/12/featured-image-Timeline-of-Transformative-Events-1.png,{},"Our World in Data presents the data and research to make progress against the world’s largest problems. This article draws on data and research discussed in our entry on **[Artificial Intelligence](https://ourworldindata.org/artificial-intelligence)**. Why should you care about the development of artificial intelligence? Think about what the alternative would look like. If you and the wider public do not get informed and engaged, then we leave it to a few entrepreneurs and engineers to decide how this technology will transform our world. That is the status quo. This small number of people at a few tech firms directly working on artificial intelligence (AI) do understand how extraordinarily powerful this technology is [becoming](https://ourworldindata.org/brief-history-of-ai). If the rest of society does not become engaged, then it will be this small elite who decides how this technology will change our lives. To change this status quo, I want to answer three questions in this article: Why is it hard to take the prospect of a world transformed by AI seriously? How can we imagine such a world? And what is at stake as this technology becomes more powerful? ## Why is it hard to take the prospect of a world transformed by artificial intelligence seriously? In some way, it should be obvious how technology can fundamentally transform the world. We just have to look at how much the world has already changed. If you could invite a family of hunter-gatherers from 20,000 years ago on your next flight, they would be pretty surprised. Technology has changed our world already, so we should expect that it can happen again. But while we have seen the world transform before, we have seen these transformations play out over the course of generations. What is different now is how very rapid these technological changes have become. In the past, the technologies that our ancestors used in their childhood were still central to their lives in their old age. This has not been the case anymore for recent generations. Instead, it has [become common](https://ourworldindata.org/technology-long-run) that technologies unimaginable in one's youth become ordinary in later life. This is the first reason we might not take the prospect seriously: it is easy to underestimate the speed at which technology can change the world. The second reason why it is difficult to take the possibility of transformative AI – potentially even AI as intelligent as humans – seriously is that it is an idea that we first heard in the cinema. It is not surprising that for many of us, the first reaction to a scenario in which machines have human-like capabilities is the same as if you had asked us to take seriously a future in which vampires, werewolves, or zombies roam the planet.{ref}This problem becomes even larger when we try to imagine how a future with a human-level AI might play out. Any _particular_ scenario will not only involve the idea that this powerful AI exists, but a whole range of additional assumptions about the future context in which this happens. It is therefore hard to communicate a scenario of a world with human-level AI that does not sound contrived, bizarre or even silly.{/ref} But, it is plausible that it is both the stuff of sci-fi fantasy _and_ the central invention that could arrive in our, or our children’s, lifetimes. The third reason why it is difficult to take this prospect seriously is by failing to see that powerful AI could lead to very large changes. This is also understandable. It is difficult to form an idea of a future that is very different from our own time. There are two concepts that I find helpful in imagining a very different future with artificial intelligence. Let’s look at both of them. ## How to develop an idea of what the future of artificial intelligence might look like? When thinking about the future of artificial intelligence, I find it helpful to consider two different concepts in particular: human-level AI, and transformative AI.{ref}Both of these concepts are widely used in the scientific literature on artificial intelligence. For example, questions about the timelines for the development of future AI are often framed using these terms. See [my article on this topic](https://ourworldindata.org/ai-timelines).{/ref} The first concept highlights the AI’s capabilities and anchors them to a familiar benchmark, while transformative AI emphasizes the impact that this technology would have on the world. From where we are today, much of this may sound like science fiction. It is therefore worth keeping in mind that the majority of surveyed AI experts [believe](https://ourworldindata.org/ai-timelines) there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner. ### The advantages and disadvantages of comparing machine and human intelligence One way to think about human-level artificial intelligence is to contrast it with the current state of AI technology. While today’s AI systems often have capabilities similar to a particular, limited part of the human mind, a human-level AI would be a machine that is capable of carrying out the same _range_ of intellectual tasks that we humans are capable of.{ref}The fact that humans are capable of a _range_ of intellectual tasks means that you arrive at different definitions of intelligence depending on which aspect within that range you focus on (the [Wikipedia entry on intelligence](https://en.wikipedia.org/wiki/Intelligence), for example, lists a number of definitions from various researchers and different disciplines). As a consequence there are also various definitions of ‘human-level AI’. There are also several closely related terms: Artificial General Intelligence, High-Level Machine Intelligence, Strong AI, or Full AI are sometimes synonymously used, and sometimes defined in similar, yet different ways. In specific discussions, it is necessary to define this concept more narrowly; for example, in [studies on AI timelines](https://ourworldindata.org/ai-timelines) researchers offer more precise definitions of what human-level AI refers to in their particular study.{/ref} It is a machine that would be “able to learn to do anything that a human can do,” as Norvig and Russell put it in their textbook on AI.{ref}Peter Norvig and Stuart Russell (2021) — Artificial Intelligence: A Modern Approach. Fourth edition. Published by Pearson.{/ref} Taken together, the range of abilities that characterize intelligence gives humans the ability to solve problems and achieve a wide variety of goals. A human-level AI would therefore be a system that could solve all those problems that we humans can solve, and do the tasks that humans do today. Such a machine, or collective of machines, would be able to do the work of a translator, an accountant, an illustrator, a teacher, a therapist, a truck driver, or the work of a trader on the world’s financial markets. Like us, it would also be able to do research and science, and to develop new technologies based on that. The concept of human-level AI has some clear advantages. Using the familiarity of our own intelligence as a reference provides us with some clear guidance on how to imagine the capabilities of this technology. However, it also has clear disadvantages. Anchoring the imagination of future AI systems to the familiar reality of human intelligence carries the risk that it obscures the very real differences between them. Some of these differences are obvious. For example, AI systems will have the immense memory of computer systems, against which our own capacity to store information pales. Another obvious difference is the speed at which a machine can absorb and process information. But information storage and processing speed are not the only differences. The domains in which machines already outperform humans is steadily increasing: in chess, after matching the level of the best human players in the late 90s, AI systems [reached](https://ourworldindata.org/grapher/computer-chess-ability) superhuman levels more than a decade ago. In other games like Go or complex strategy games, this has happened more recently.{ref}The AI system [AlphaGo](https://en.wikipedia.org/wiki/AlphaGo), and its various successors, won against Go masters. The AI system [Pluribus](https://en.wikipedia.org/wiki/Pluribus_(poker_bot)) beat humans at no-limit Texas hold 'em poker. The AI system Cicero can strategize and use human language to win the strategy game Diplomacy. See: Meta Fundamental AI Research Diplomacy Team (FAIR), Anton Bakhtin, Noam Brown, Emily Dinan, Gabriele Farina, Colin Flaherty, Daniel Fried, et al. (2022) – ‘Human-Level Play in the Game of Diplomacy by Combining Language Models with Strategic Reasoning’. In _Science_ 0, no. 0 (22 November 2022): eade9097.[ https://doi.org/10.1126/science.ade9097](https://doi.org/10.1126/science.ade9097).{/ref} These differences mean that an AI that is at least as good as humans in every domain would overall be much more powerful than the human mind. Even the first “human-level AI” would therefore be quite superhuman in many ways.{ref}This also poses a problem when we evaluate how the intelligence of a machine compares with the intelligence of humans. If intelligence was a general ability, a single capacity, then we could easily compare and evaluate it, but the fact that it is a range of skills makes it much more difficult to compare across machine and human intelligence. Tests for AI systems are therefore comprising a wide range of tasks. See for example Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, Jacob Steinhardt (2020) – [Measuring Massive Multitask Language Understanding](https://arxiv.org/abs/2009.03300) or the definition of what would qualify as artificial general intelligence in [this Metaculus prediction](https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/).{/ref} Human intelligence is also a bad metaphor for machine intelligence in other ways. The way we think is often very different from machines, and as a consequence the output of thinking machines can be very alien to us. Most perplexing and most concerning are the strange and unexpected ways in which machine intelligence can fail. The AI-generated image of the horse below provides an example: on the one hand, AIs can do what no human can do – produce an image of anything, in any style (here photorealistic), in mere seconds – but on the other hand it can fail in ways that no human would fail.{ref}An overview of how AI systems can fail can be found in [Charles Choi – 7 Revealing Ways AIs Fail](https://spectrum.ieee.org/ai-failures). It is also worth reading through the [AIAAIC Repository](https://www.aiaaic.org/aiaaic-repository/ai-and-algorithmic-incidents-and-controversies) which “details recent incidents and controversies driven by or relating to AI, algorithms, and automation.""{/ref} No human would make the mistake of drawing a horse with five legs.{ref}I have taken this example from [AI researcher François Chollet](https://fchollet.com/), who published it [here](https://twitter.com/fchollet/status/1573752180720312320?s=46&t=qPwLwDgLdJrLlXxa878BDQ).{/ref} Imagining a powerful future AI as just another human would therefore likely be a mistake. The differences might be so large that it will be a misnomer to call such systems “human-level.” **AI-generated image of a horse**{ref}Via [François Chollet](https://fchollet.com/), who published it [here](https://twitter.com/fchollet/status/1573752180720312320?s=46&t=qPwLwDgLdJrLlXxa878BDQ). Based on Chollet’s comments it seems that this image was created by the AI system ‘Stable Diffusion’.{/ref}Our World in Data presents the data and research to make progress against the world’s largest problems.
This article draws on data and research discussed in our entry on Artificial Intelligence.
Why should you care about the development of artificial intelligence?
\n\n\n\nThink about what the alternative would look like. If you and the wider public do not get informed and engaged, then we leave it to a few entrepreneurs and engineers to decide how this technology will transform our world.
\n\n\n\nThat is the status quo. This small number of people at a few tech firms directly working on artificial intelligence (AI) do understand how extraordinarily powerful this technology is becoming. If the rest of society does not become engaged, then it will be this small elite who decides how this technology will change our lives.
\n\n\n\nTo change this status quo, I want to answer three questions in this article: Why is it hard to take the prospect of a world transformed by AI seriously? How can we imagine such a world? And what is at stake as this technology becomes more powerful?
\n\n\n\nIn some way, it should be obvious how technology can fundamentally transform the world. We just have to look at how much the world has already changed. If you could invite a family of hunter-gatherers from 20,000 years ago on your next flight, they would be pretty surprised. Technology has changed our world already, so we should expect that it can happen again.
\n\n\n\nBut while we have seen the world transform before, we have seen these transformations play out over the course of generations. What is different now is how very rapid these technological changes have become. In the past, the technologies that our ancestors used in their childhood were still central to their lives in their old age. This has not been the case anymore for recent generations. Instead, it has become common that technologies unimaginable in one’s youth become ordinary in later life.
\n\n\n\nThis is the first reason we might not take the prospect seriously: it is easy to underestimate the speed at which technology can change the world.
\n\n\n\nThe second reason why it is difficult to take the possibility of transformative AI – potentially even AI as intelligent as humans – seriously is that it is an idea that we first heard in the cinema. It is not surprising that for many of us, the first reaction to a scenario in which machines have human-like capabilities is the same as if you had asked us to take seriously a future in which vampires, werewolves, or zombies roam the planet.{ref}This problem becomes even larger when we try to imagine how a future with a human-level AI might play out. Any particular scenario will not only involve the idea that this powerful AI exists, but a whole range of additional assumptions about the future context in which this happens. It is therefore hard to communicate a scenario of a world with human-level AI that does not sound contrived, bizarre or even silly.{/ref}
\n\n\n\nBut, it is plausible that it is both the stuff of sci-fi fantasy and the central invention that could arrive in our, or our children’s, lifetimes.
\n\n\n\nThe third reason why it is difficult to take this prospect seriously is by failing to see that powerful AI could lead to very large changes. This is also understandable. It is difficult to form an idea of a future that is very different from our own time. There are two concepts that I find helpful in imagining a very different future with artificial intelligence. Let’s look at both of them.
\n\n\n\nWhen thinking about the future of artificial intelligence, I find it helpful to consider two different concepts in particular: human-level AI, and transformative AI.{ref}Both of these concepts are widely used in the scientific literature on artificial intelligence. For example, questions about the timelines for the development of future AI are often framed using these terms. See my article on this topic.{/ref} The first concept highlights the AI’s capabilities and anchors them to a familiar benchmark, while transformative AI emphasizes the impact that this technology would have on the world.
\n\n\n\nFrom where we are today, much of this may sound like science fiction. It is therefore worth keeping in mind that the majority of surveyed AI experts believe there is a real chance that human-level artificial intelligence will be developed within the next decades, and some believe that it will exist much sooner.
\n\n\n\nOne way to think about human-level artificial intelligence is to contrast it with the current state of AI technology. While today’s AI systems often have capabilities similar to a particular, limited part of the human mind, a human-level AI would be a machine that is capable of carrying out the same range of intellectual tasks that we humans are capable of.{ref}The fact that humans are capable of a range of intellectual tasks means that you arrive at different definitions of intelligence depending on which aspect within that range you focus on (the Wikipedia entry on intelligence, for example, lists a number of definitions from various researchers and different disciplines). As a consequence there are also various definitions of ‘human-level AI’.
\n\n\n\nThere are also several closely related terms: Artificial General Intelligence, High-Level Machine Intelligence, Strong AI, or Full AI are sometimes synonymously used, and sometimes defined in similar, yet different ways. In specific discussions, it is necessary to define this concept more narrowly; for example, in studies on AI timelines researchers offer more precise definitions of what human-level AI refers to in their particular study.{/ref} It is a machine that would be “able to learn to do anything that a human can do,” as Norvig and Russell put it in their textbook on AI.{ref}Peter Norvig and Stuart Russell (2021) — Artificial Intelligence: A Modern Approach. Fourth edition. Published by Pearson.{/ref}
\n\n\n\nTaken together, the range of abilities that characterize intelligence gives humans the ability to solve problems and achieve a wide variety of goals. A human-level AI would therefore be a system that could solve all those problems that we humans can solve, and do the tasks that humans do today. Such a machine, or collective of machines, would be able to do the work of a translator, an accountant, an illustrator, a teacher, a therapist, a truck driver, or the work of a trader on the world’s financial markets. Like us, it would also be able to do research and science, and to develop new technologies based on that.
\n\n\n\nThe concept of human-level AI has some clear advantages. Using the familiarity of our own intelligence as a reference provides us with some clear guidance on how to imagine the capabilities of this technology.
\n\n\n\nHowever, it also has clear disadvantages. Anchoring the imagination of future AI systems to the familiar reality of human intelligence carries the risk that it obscures the very real differences between them.
\n\n\n\nSome of these differences are obvious. For example, AI systems will have the immense memory of computer systems, against which our own capacity to store information pales. Another obvious difference is the speed at which a machine can absorb and process information. But information storage and processing speed are not the only differences. The domains in which machines already outperform humans is steadily increasing: in chess, after matching the level of the best human players in the late 90s, AI systems reached superhuman levels more than a decade ago. In other games like Go or complex strategy games, this has happened more recently.{ref}The AI system AlphaGo, and its various successors, won against Go masters. The AI system Pluribus beat humans at no-limit Texas hold ’em poker. The AI system Cicero can strategize and use human language to win the strategy game Diplomacy. See: Meta Fundamental AI Research Diplomacy Team (FAIR), Anton Bakhtin, Noam Brown, Emily Dinan, Gabriele Farina, Colin Flaherty, Daniel Fried, et al. (2022) – ‘Human-Level Play in the Game of Diplomacy by Combining Language Models with Strategic Reasoning’. In Science 0, no. 0 (22 November 2022): eade9097. https://doi.org/10.1126/science.ade9097.{/ref}
\n\n\n\nThese differences mean that an AI that is at least as good as humans in every domain would overall be much more powerful than the human mind. Even the first “human-level AI” would therefore be quite superhuman in many ways.{ref}This also poses a problem when we evaluate how the intelligence of a machine compares with the intelligence of humans. If intelligence was a general ability, a single capacity, then we could easily compare and evaluate it, but the fact that it is a range of skills makes it much more difficult to compare across machine and human intelligence. Tests for AI systems are therefore comprising a wide range of tasks. See for example Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, Jacob Steinhardt (2020) – Measuring Massive Multitask Language Understanding or the definition of what would qualify as artificial general intelligence in this Metaculus prediction.{/ref}
\n\n\n\nHuman intelligence is also a bad metaphor for machine intelligence in other ways. The way we think is often very different from machines, and as a consequence the output of thinking machines can be very alien to us.
\n\n\n\nMost perplexing and most concerning are the strange and unexpected ways in which machine intelligence can fail. The AI-generated image of the horse below provides an example: on the one hand, AIs can do what no human can do – produce an image of anything, in any style (here photorealistic), in mere seconds – but on the other hand it can fail in ways that no human would fail.{ref}An overview of how AI systems can fail can be found in Charles Choi – 7 Revealing Ways AIs Fail. It is also worth reading through the AIAAIC Repository which “details recent incidents and controversies driven by or relating to AI, algorithms, and automation.”{/ref} No human would make the mistake of drawing a horse with five legs.{ref}I have taken this example from AI researcher François Chollet, who published it here.{/ref}
\n\n\n\nImagining a powerful future AI as just another human would therefore likely be a mistake. The differences might be so large that it will be a misnomer to call such systems “human-level.”
\n\n\n\nAI-generated image of a horse{ref}Via François Chollet, who published it here. Based on Chollet’s comments it seems that this image was created by the AI system ‘Stable Diffusion’.{/ref}
\n\n\n\n![\""A](\""https://owid.cloud/app/uploads/2022/12/ai-generated-image-of-a-horse.png\"")
In contrast, the concept of transformative AI is not based on a comparison with human intelligence. This has the advantage of sidestepping the problems that the comparisons with our own mind bring. But it has the disadvantage that it is harder to imagine what such a system would look like and be capable of. It requires more from us. It requires us to imagine a world with intelligent actors that are potentially very different from ourselves.
\n\n\n\nTransformative AI is not defined by any specific capabilities, but by the real-world impact that the AI would have. To qualify as transformative, researchers think of it as AI that is “powerful enough to bring us into a new, qualitatively different future.”{ref}This quote is from Holden Karnofsky (2021) – AI Timelines: Where the Arguments, and the “Experts,” Stand. For Holden Karnofsky’s earlier thinking on this conceptualization of AI see his 2016 article ‘Some Background on Our Views Regarding Advanced Artificial Intelligence’.
\n\n\n\nAjeya Cotra, whose research on AI timelines I discuss in other articles of this series, attempts to give a quantitative definition of what would qualify as transformative AI. in her widely cited report on AI timelines she defines it as a change in software technology that brings the growth rate of gross world product “to 20%-30% per year”. Several other researchers define TAI in similar terms.{/ref}
\n\n\n\nIn humanity’s history, there have been two cases of such major transformations, the agricultural and the industrial revolutions.
\n\n\n\nTransformative AI becoming a reality would be an event on that scale. Like the arrival of agriculture 10,000 years ago, or the transition from hand- to machine-manufacturing, it would be an event that would change the world for billions of people around the globe and for the entire trajectory of humanity’s future.
\n\n\n\nTechnologies that fundamentally change how a wide range of goods or services are produced are called ‘general-purpose technologies’. The two previous transformative events were caused by the discovery of two particularly significant general-purpose technologies: the change in food production as humanity transitioned from hunting and gathering to farming, and the rise of machine manufacturing in the industrial revolution. Based on the evidence and arguments presented in this series on AI development, I believe it is plausible that powerful AI could represent the introduction of a similarly significant general-purpose technology.
\n\n\n\nTimeline of the three transformative events in world history
\n![\""\""](\""https://owid.cloud/app/uploads/2022/12/Timeline-of-Transformative-Events.png\"")
The two concepts are closely related, but they are not the same. The creation of a human-level AI would certainly have a transformative impact on our world. If the work of most humans could be carried out by an AI, the lives of millions of people would change.{ref}Human-level AI is typically defined as a software system that can carry out at least 90% or 99% of all economically relevant tasks that humans carry out. A lower-bar definition would be an AI system that can carry out all those tasks that can currently be done by another human who is working remotely on a computer.{/ref}
\n\n\n\nThe opposite, however, is not true: we might see transformative AI without developing human-level AI. Since the human mind is in many ways a poor metaphor for the intelligence of machines, we might plausibly develop transformative AI before we develop human-level AI. Depending on how this goes, this might mean that we will never see any machine intelligence for which human intelligence is a helpful comparison.
\n\n\n\nWhen and if AI systems might reach either of these levels is of course difficult to predict. In my companion article on this question, I give an overview of what researchers in this field currently believe. Many AI experts believe there is a real chance that such systems will be developed within the next decades, and some believe that they will exist much sooner.
\n\n\n\nAll major technological innovations lead to a range of positive and negative consequences. For AI, the spectrum of possible outcomes – from the most negative to the most positive – is extraordinarily wide.
\n\n\n\nThat the use of AI technology can cause harm is clear, because it is already happening.
\n\n\n\nAI systems can cause harm when people use them maliciously. For example, when they are used in politically-motivated disinformation campaigns or to enable mass surveillance.{ref}On the use of AI in politically-motivated disinformation campaigns see for example John Villasenor (November 2020) – How to deal with AI-enabled disinformation. More generally on this topic see Brundage and Avin et al. (2018) – The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation, published at maliciousaireport.com. A starting point for literature and reporting on mass surveillance by governments is the relevant Wikipedia entry.{/ref}
\n\n\n\nBut AI systems can also cause unintended harm, when they act differently than intended or fail. For example, in the Netherlands the authorities used an AI system which falsely claimed that an estimated 26,000 parents made fraudulent claims for child care benefits. The false allegations led to hardship for many poor families, and also resulted in the resignation of the Dutch government in 2021.{ref}See for example the Wikipedia entry on the ‘Dutch childcare benefits scandal’ and Melissa Heikkilä (2022) – ‘Dutch scandal serves as a warning for Europe over risks of using algorithms’, in Politico. The technology can also reinforce discrimination in terms of race and gender. See Brian Christian’s book The Alignment Problem and the reports of the AI Now Institute.{/ref}
\n\n\n\nAs AI becomes more powerful, the possible negative impacts could become much larger. Many of these risks have rightfully received public attention: more powerful AI could lead to mass labor displacement, or extreme concentrations of power and wealth. In the hands of autocrats, it could empower totalitarianism through its suitability for mass surveillance and control.
\n\n\n\nThe so-called alignment problem of AI is another extreme risk. This is the concern that nobody would be able to control a powerful AI system, even if the AI takes actions that harm us humans, or humanity as a whole. This risk is unfortunately receiving little attention from the wider public, but it is seen as an extremely large risk by many leading AI researchers.{ref}Overviews are provided in Stuart Russell (2019) – Human Compatible (especially chapter 5) and Brian Christian’s 2020 book The Alignment Problem. Christian presents the thinking of many leading AI researchers from the earliest days up to now and presents an excellent overview of this problem. It is also seen as a large risk by some of the leading private firms who work towards powerful AI – see OpenAI’s article “Our approach to alignment research” from August 2022.{/ref}
\n\n\n\nHow could an AI possibly escape human control and end up harming humans?
\n\n\n\nThe risk is not that an AI becomes self-aware, develops bad intentions, and “chooses” to do this. The risk is that we try to instruct the AI to pursue some specific goal – even a very worthwhile one – and in the pursuit of that goal it ends up harming humans. It is about unintended consequences. The AI does what we told it to do, but not what we wanted it to do.
\n\n\n\nCan’t we just tell the AI to not do those things? It is definitely possible to build an AI that avoids any particular problem we foresee, but it is hard to foresee all the possible harmful unintended consequences. The alignment problem arises because of “the impossibility of defining true human purposes correctly and completely,” as AI researcher Stuart Russell puts it.{ref}Stuart Russell (2019) – Human Compatible{/ref}
\n\n\n\nCan’t we then just switch off the AI? This might also not be possible. That is because a powerful AI would know two things: it faces a risk that humans could turn it off, and it can’t achieve its goals once it has been turned off. As a consequence, the AI will pursue a very fundamental goal of ensuring that it won’t be switched off. This is why, once we realize that an extremely intelligent AI is causing unintended harm in the pursuit of some specific goal, it might not be possible to turn it off or change what the system does.{ref}A question that follows from this is, why build such a powerful AI in the first place?
\n\n\n\nThe incentives are very high. As I emphasize below, this innovation has the potential to lead to very positive developments. In addition to the large social benefits there are also large incentives for those who develop it – the governments that can use it for their goals, the individuals who can use it to become more powerful and wealthy. Additionally, it is of scientific interest and might help us to understand our own mind and intelligence better. And lastly, even if we wanted to stop building powerful AIs, it is likely very hard to actually achieve it. It is very hard to coordinate across the whole world and agree to stop building more advanced AI – countries around the world would have to agree and then find ways to actually implement it.{/ref}
\n\n\n\nThis risk – that humanity might not be able to stay in control once AI becomes very powerful, and that this might lead to an extreme catastrophe – has been recognized right from the early days of AI research more than 70 years ago.{ref}In 1950 the computer science pioneer Alan Turing put it like this: “If a machine can think, it might think more intelligently than we do, and then where should we be? … [T]his new danger is much closer. If it comes at all it will almost certainly be within the next millennium. It is remote but not astronomically remote, and is certainly something which can give us anxiety. It is customary, in a talk or article on this subject, to offer a grain of comfort, in the form of a statement that some particularly human characteristic could never be imitated by a machine. … I cannot offer any such comfort, for I believe that no such bounds can be set.” Alan. M. Turing (1950) – Computing Machinery and Intelligence, In Mind, Volume LIX, Issue 236, October 1950, Pages 433–460.
\n\n\n\nNorbert Wiener is another pioneer who saw the alignment problem very early. One way he put it was “If we use, to achieve our purposes, a mechanical agency with whose operation we cannot interfere effectively … we had better be quite sure that the purpose put into the machine is the purpose which we really desire.” quoted from Norbert Wiener (1960) – Some Moral and Technical Consequences of Automation: As machines learn they may develop unforeseen strategies at rates that baffle their programmers. In Science.
\n\n\n\nIn 1950 – the same year in which Turing published the cited article – Wiener published his book The Human Use of Human Beings, whose front-cover blurb reads: “The ‘mechanical brain’ and similar machines can destroy human values or enable us to realize them as never before.”{/ref} The very rapid development of AI in recent years has made a solution to this problem much more urgent.
\n\n\n\nI have tried to summarize some of the risks of AI, but a short article is not enough space to address all possible questions. Especially on the very worst risks of AI systems, and what we can do now to reduce them, I recommend reading the book The Alignment Problem by Brian Christian and Benjamin Hilton’s article ‘Preventing an AI-related catastrophe’.
\n\n\n\nIf we manage to avoid these risks, transformative AI could also lead to very positive consequences. Advances in science and technology were crucial to the many positive developments in humanity’s history. If artificial ingenuity can augment our own, it could help us make progress on the many large problems we face: from cleaner energy, to the replacement of unpleasant work, to much better healthcare.
\n\n\n\nThis extremely large contrast between the possible positives and negatives makes clear that the stakes are unusually high with this technology. Reducing the negative risks and solving the alignment problem could mean the difference between a healthy, flourishing, and wealthy future for humanity – and the destruction of the same.
\n\n\n\nMaking sure that the development of artificial intelligence goes well is not just one of the most crucial questions of our time, but likely one of the most crucial questions in human history. This needs public resources – public funding, public attention, and public engagement.
\n\n\n\nCurrently, almost all resources that are dedicated to AI aim to speed up the development of this technology. Efforts that aim to increase the safety of AI systems, on the other hand, do not receive the resources they need. Researcher Toby Ord estimated that in 2020 between $10 to $50 million was spent on work to address the alignment problem.{ref}Toby Ord – The Precipice. He makes this projection in footnote 55 of chapter 2. It is based on the 2017 estimate by Farquhar.{/ref} Corporate AI investment in the same year was more than 2000-times larger, it summed up to $153 billion.
\n\n\n\nThis is not only the case for the AI alignment problem. The work on the entire range of negative social consequences from AI is under-resourced compared to the large investments to increase the power and use of AI systems.
\n\n\n\nIt is frustrating and concerning for society as a whole that AI safety work is extremely neglected and that little public funding is dedicated to this crucial field of research. On the other hand, for each individual person this neglect means that they have a good chance to actually make a positive difference, if they dedicate themselves to this problem now. And while the field of AI safety is small, it does provide good resources on what you can do concretely if you want to work on this problem.
\n\n\n\nI hope that more people dedicate their individual careers to this cause, but it needs more than individual efforts. A technology that is transforming our society needs to be a central interest of all of us. As a society we have to think more about the societal impact of AI, become knowledgeable about the technology, and understand what is at stake.
\n\n\n\nWhen our children look back at today, I imagine that they will find it difficult to understand how little attention and resources we dedicated to the development of safe AI. I hope that this changes in the coming years, and that we begin to dedicate more resources to making sure that powerful AI gets developed in a way that benefits us and the next generations.
\n\n\n\nIf we fail to develop this broad-based understanding, then it will remain the small elite that finances and builds this technology that will determine how one of the – or plausibly the – most powerful technology in human history will transform our world.
\n\n\n\nWith our work at Our World in Data we want to do our small part to enable a better informed public conversation on AI and the future we want to live in. You can find these resources on OurWorldinData.org/artificial-intelligence
\n\n\n\nAcknowledgements: I would like to thank my colleagues Daniel Bachler, Charlie Giattino, and Edouard Mathieu for their helpful comments to drafts of this essay.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""How AI gets built is currently decided by a small group of technologists. As this technology is transforming our lives, it should be in all of our interest to become informed and engaged."", ""protected"": false}, ""date_gmt"": ""2022-12-15T05:00:00"", ""modified"": ""2023-10-11T09:43:04"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Max Roser""], ""modified_gmt"": ""2023-10-11T08:43:04"", ""comment_status"": ""closed"", ""featured_media"": 54894, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/12/featured-image-Timeline-of-Transformative-Events-1-150x86.png"", ""medium_large"": ""/app/uploads/2022/12/featured-image-Timeline-of-Transformative-Events-1-768x440.png""}}" 54733,We published a redesign of our work on the Environmental Impacts of Food,redesign-environmental-impacts-food,post,publish,,"{""id"": ""wp-54733"", ""slug"": ""redesign-environmental-impacts-food"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We published a redesign of our work on the Environmental Impacts of Food"", ""authors"": [""Hannah Ritchie"", ""Pablo Rosado""], ""excerpt"": ""We published a major redesign of our work on the Environmental Impacts of Food. Explore all of our data and research in one place."", ""dateline"": ""February 23, 2023"", ""subtitle"": ""We published a major redesign of our work on the Environmental Impacts of Food. Explore all of our data and research in one place."", ""sidebar-toc"": false, ""featured-image"": ""Environmental-Impacts-of-Food-Production-Thumbnail.png""}, ""createdAt"": ""2022-12-02T05:31:12.000Z"", ""published"": false, ""updatedAt"": ""2023-07-10T16:19:02.000Z"", ""revisionId"": null, ""publishedAt"": ""2023-02-23T12:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2023-02-23 12:00:00,2024-02-16 14:22:54,,"[""Hannah Ritchie"", ""Pablo Rosado""]",We published a major redesign of our work on the Environmental Impacts of Food. Explore all of our data and research in one place.,2022-12-02 05:31:12,2023-07-10 16:19:02,https://ourworldindata.org/wp-content/uploads/2022/11/Environmental-Impacts-of-Food-Production-Thumbnail.png,{},,"{""id"": 54733, ""date"": ""2023-02-23T12:00:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=54733""}, ""link"": ""https://owid.cloud/redesign-environmental-impacts-food"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""redesign-environmental-impacts-food"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We published a redesign of our work on the Environmental Impacts of Food""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54733""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=54733"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=54733"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=54733"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=54733""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54733/revisions"", ""count"": 4}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/54675"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 56277, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54733/revisions/56277""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""We published a major redesign of our work on the Environmental Impacts of Food. Explore all of our data and research in one place."", ""protected"": false}, ""date_gmt"": ""2023-02-23T12:00:00"", ""modified"": ""2023-07-10T17:19:02"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Pablo Rosado""], ""modified_gmt"": ""2023-07-10T16:19:02"", ""comment_status"": ""closed"", ""featured_media"": 54675, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/11/Environmental-Impacts-of-Food-Production-Thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2022/11/Environmental-Impacts-of-Food-Production-Thumbnail-768x403.png""}}" 54697,The state of the world's elephant populations,elephant-populations,post,publish,"Elephants are the world’s largest living land animals, weighing in at up to 7.5 tonnes.{ref}African elephants are typically bigger than the Asian species. African elephants can weigh up to 7.5 tonnes; Asian elephants up to 5 tonnes.{/ref}
Their size has made them a prime target for poaching. History has shown us that it is usually the largest mammals that are most at risk from human hunting. Elephants are no different. We hunt them for their meat, their trunks, and their lucrative tusks.
There are around 450,000 elephants in the world. But this is a tiny fraction of how many there used to be.
In this article we look at the state of elephant populations, and how these populations have changed over time.
There are two species of elephant: the African (with the official name Loxodonta africana) and the Asian (Elephas maximus) elephant.
If you want to tell the difference between them, look at their ears: the African elephant has much bigger ears, very similar in shape to the African continent; the Asian elephant has much smaller, rounded ears.{ref}Elephants dissipate heat via their ears, and so use them for temperature regulation. The reason that African elephants have larger ears is that they live in warmer climates and therefore need to dissipate more heat.{/ref} Their tusks are also a useful indicator: both male and female African elephants can grow tusks, but only male Asian ones can.
In the table, I have summarized the status of their populations.
There are around ten times as many African than Asian elephants in the world. In 2015, there were around 415,000 African elephants left. For the Asian species, this is in the range of 40,000 to 50,000.
The Asian elephant is classified as ‘endangered’, one level down from ‘critically endangered’ before extinction, on the IUCN Red List.{ref}Choudhury, A., Lahiri Choudhury, D.K., Desai, A., Duckworth, J.W., Easa, P.S., Johnsingh, A.J.T., Fernando, P., Hedges, S., Gunawardena, M., Kurt, F., Karanth, U., Lister, A., Menon, V., Riddle, H., Rübel, A. & Wikramanayake, E. (IUCN SSC Asian Elephant Specialist Group) 2008. Elephas maximus. The IUCN Red List of Threatened Species 2008.{/ref} The African elephant was previously treated as a single species, but has recently been separated into the African forest elephant (Loxodonta cyclotis) and African savanna elephant (Loxodonta africana) for evaluation. The forest elephant is listed as ‘critically endangered’ and the savanna elephant as ‘endangered’. The populations of both species are declining.
| Elephant species | Population (latest estimate) | Extinction risk | Population trend |
| African elephant (Loxodonta africana) | 415,000 | By subspecies below | By subspecies below |
| African forest elephant (Loxodonta cyclotis) | Critically endangered | Decreasing | |
| African savanna elephant (Loxodonta africana) | Endangered | Decreasing | |
| Asian elephant (Elephas maximus) | 40,000 - 50,000 | Endangered | Decreasing |
To understand the vulnerability of elephant populations, knowing the the number of animals alive today is not enough. We also need to know the direction and rate of change. If population numbers are falling quickly, we should be concerned even if there are hundreds of thousands left.
Let’s take a look at the African and Asian elephant species one by one.
There are ten times as many African elephants as Asian elephants in the world. That makes them seem abundant. But their numbers are a tiny fraction of what they were in the past.
African elephant populations have shrunk by 98% since 1500. We see this in the chart.{ref}The main sources of data for African elephant populations are the Great Elephant Census and the IUCN SSC African Elephant Specialist Group (AfESG). The AfESG gathers population estimates by country every few years, publishing it in an African Elephant Database, and African Elephant Status Report. You can find the latest African Elephant Status Report (2016) here.
It should be noted that long historical estimates in particular are crude and come with significant uncertainty. Nonetheless they are accurate enough to provide a sense of magnitude for the relative change over time.{/ref}
In 1500, there were over 25 million elephants in Africa. By 1900 this had fallen to around 10 million, and by 1979 down to 1.3 million.
There was a rapid decline in population size over the 1970s and 1980s such that by the mid-1990s numbers had fallen below 300,000. Over the following decades, conservation efforts across some countries managed to restore populations to over 470,000 in 2008. But increased poaching rates over the past decade have sent numbers back into decline.
Another piece of evidence we have that populations have been in decline comes from a metric called the ‘carcass ratio’.
During population surveys, researchers don’t only count the number of alive elephants, they also count the number of dead elephants (carcasses). The carcass ratio is the number of dead elephants observed during surveys, given as a percentage of the total population.
The carcass ratio across Africa as a whole was 11.9%.{ref}Chase, M. J., Schlossberg, S., Griffin, C. R., Bouché, P. J., Djene, S. W., Elkan, P. W., ... & Omondi, P. (2016). Continent-wide survey reveals massive decline in African savannah elephants. PeerJ, 4, e2354.{/ref} This means that for every 100 live elephants, there were around 12 dead elephants. A carcass ratio greater than 8% usually means the population is shrinking, because this will be greater than the replacement rate.{ref}Douglas-Hamilton, I., & Burrill, A. (1991). Using elephant carcass ratios to determine population trends. African wildlife: research and management, 98-105.{/ref}The overall population of African elephants has been falling in recent years. But this varies significantly across countries. In some, the carcass ratio was very high: in Cameroon, it was 83%, it was 32% in Mozambique, and 30% in Angola.{ref}Data is only available for the years 2007 and 2015. So even countries which show an increase in over this decade – Cameroon, for example – might have seen a decline in very recent years, which is reflected in carcass ratio data.{/ref}
There are fewer estimates of Asian Elephant populations. This is more worrying because the Asian elephant is at a higher risk of extinction. We should be tracking these numbers more, not less, closely.
We do have some data for select countries, and some longer-term estimates.
The IUCN estimates that the total population of Asian elephants has more than halved over the past century. It estimates that there were 100,000 animals in the early 1900s; today that figure is in the range of 40,000 to 50,000.
Population data over time is available for some countries in Asia: the Indian government, for example, has published estimates periodically since 1970.
In the map, you can explore the latest population estimates from the Asian IUCN SSC Asian Elephant Specialist Group (AsESG) for each country.
In India, populations have been steadily increasing since 1980, rising from around 16,000 to over 27,000 in 2017. This shows that it’s possible to protect these species and help their populations rebuild.
However, the lack of data over time for many countries makes it difficult to properly assess the health of Asian elephant populations.
By far the biggest threat to both African and Asian elephants is poaching. Elephants are killed for their trunks and their tusks. Ivory is a lucrative business.
It’s not just elephants that are under pressure.
Poaching is the leading threat to all large mammals. But as we’ve seen from some country-level examples: protecting these species is possible: India has managed to protect and restore elephant populations. Namibia, Zimbabwe, and Angola have also managed to turn the trend.
","{""id"": ""wp-54697"", ""slug"": ""elephant-populations"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Elephants are the world’s largest living land animals, weighing in at"", ""spanType"": ""span-simple-text""}, {""url"": ""https://d2ouvy59p0dg6k.cloudfront.net/downloads/african_elephant_factsheet2007w.pdf"", ""children"": [{""text"": "" up to 7.5 tonnes"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}African elephants are typically bigger than the Asian species. African elephants can weigh up to 7.5 tonnes; Asian elephants"", ""spanType"": ""span-simple-text""}, {""url"": ""http://wwf.panda.org/knowledge_hub/endangered_species/elephants/asian_elephants/"", ""children"": [{""text"": "" up to 5 tonnes"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Their size has made them a prime target for poaching. History has shown us that it is usually the largest mammals that are most at risk from human hunting. Elephants are no different. We hunt them for their meat, their trunks, and their lucrative tusks."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are around 450,000 elephants in the world. But this is a tiny fraction of how many there used to be."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this article we look at the state of elephant populations, and how these populations have changed over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""State of elephant populations today"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""There are two species of elephant: the African (with the official name "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Loxodonta africana"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "") and the Asian ("", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Elephas maximus"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "") elephant. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If you want to tell the difference between them, look at their ears: the African elephant has much bigger ears, very similar in shape to the African continent; the Asian elephant has much smaller, rounded ears.{ref}Elephants dissipate heat via their ears, and so use them for temperature regulation. The reason that African elephants have larger ears is that they live in warmer climates and therefore need to dissipate more heat.{/ref} Their tusks are also a useful indicator: both male and female African elephants can grow tusks, but only male Asian ones can."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the table, I have summarized the status of their populations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are around ten times as many African than Asian elephants in the world. In 2015, there were"", ""spanType"": ""span-simple-text""}, {""url"": ""http://africanelephantdatabase.org/report/2016/Africa"", ""children"": [{""text"": "" around 415,000"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" African elephants left. For the Asian species, this is in the range of"", ""spanType"": ""span-simple-text""}, {""url"": ""http://wwf.panda.org/knowledge_hub/endangered_species/elephants/asian_elephants/"", ""children"": [{""text"": "" 40,000 to 50,000"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Asian elephant is"", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.iucnredlist.org/species/7140/12828813"", ""children"": [{""text"": "" classified as"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" ‘endangered’, one level down from ‘critically endangered’ before extinction, on the IUCN Red List.{ref}Choudhury, A., Lahiri Choudhury, D.K., Desai, A., Duckworth, J.W., Easa, P.S., Johnsingh, A.J.T., Fernando, P., Hedges, S., Gunawardena, M., Kurt, F., Karanth, U., Lister, A., Menon, V., Riddle, H., Rübel, A. & Wikramanayake, E. (IUCN SSC Asian Elephant Specialist Group) 2008."", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.iucnredlist.org/species/7140/12828813"", ""children"": [{""text"": "" "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Elephas maximus"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""The IUCN Red List of Threatened Species"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" 2008.{/ref} The African elephant was previously treated as a single species, but has"", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.iucn.org/news/species/202103/african-elephant-species-now-endangered-and-critically-endangered-iucn-red-list"", ""children"": [{""text"": "" recently been separated"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" into the African forest elephant (Loxodonta cyclotis) and African savanna elephant (Loxodonta africana) for evaluation. The forest elephant is listed as ‘critically endangered’ and the savanna elephant as ‘endangered’. The populations of both species are declining."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""type"": ""html"", ""value"": ""| Elephant species | Population (latest estimate) | Extinction risk | Population trend |
| African elephant (Loxodonta africana) | 415,000 | By subspecies below | By subspecies below |
| African forest elephant (Loxodonta cyclotis) | Critically endangered | Decreasing | |
| African savanna elephant (Loxodonta africana) | Endangered | Decreasing | |
| Asian elephant (Elephas maximus) | 40,000 - 50,000 | Endangered | Decreasing |
| Elephant species | Population (latest estimate) | Extinction risk | Population trend |
| African elephant (Loxodonta africana) | 415,000 | By subspecies below | By subspecies below |
| African forest elephant (Loxodonta cyclotis) | Critically endangered | Decreasing | |
| African savanna elephant (Loxodonta africana) | Endangered | Decreasing | |
| Asian elephant (Elephas maximus) | 40,000 - 50,000 | Endangered | Decreasing |
Elephants are the world’s largest living land animals, weighing in at up to 7.5 tonnes.{ref}African elephants are typically bigger than the Asian species. African elephants can weigh up to 7.5 tonnes; Asian elephants up to 5 tonnes.{/ref}
\n\n\n\nTheir size has made them a prime target for poaching. History has shown us that it is usually the largest mammals that are most at risk from human hunting. Elephants are no different. We hunt them for their meat, their trunks, and their lucrative tusks.
\n\n\n\nThere are around 450,000 elephants in the world. But this is a tiny fraction of how many there used to be.
\n\n\n\nIn this article we look at the state of elephant populations, and how these populations have changed over time.
\n\n\n\nThere are two species of elephant: the African (with the official name Loxodonta africana) and the Asian (Elephas maximus) elephant.
\n\n\n\nIf you want to tell the difference between them, look at their ears: the African elephant has much bigger ears, very similar in shape to the African continent; the Asian elephant has much smaller, rounded ears.{ref}Elephants dissipate heat via their ears, and so use them for temperature regulation. The reason that African elephants have larger ears is that they live in warmer climates and therefore need to dissipate more heat.{/ref} Their tusks are also a useful indicator: both male and female African elephants can grow tusks, but only male Asian ones can.
\n\n\n\nIn the table, I have summarized the status of their populations.
\n\n\n\nThere are around ten times as many African than Asian elephants in the world. In 2015, there were around 415,000 African elephants left. For the Asian species, this is in the range of 40,000 to 50,000.
\n\n\n\nThe Asian elephant is classified as ‘endangered’, one level down from ‘critically endangered’ before extinction, on the IUCN Red List.{ref}Choudhury, A., Lahiri Choudhury, D.K., Desai, A., Duckworth, J.W., Easa, P.S., Johnsingh, A.J.T., Fernando, P., Hedges, S., Gunawardena, M., Kurt, F., Karanth, U., Lister, A., Menon, V., Riddle, H., Rübel, A. & Wikramanayake, E. (IUCN SSC Asian Elephant Specialist Group) 2008. Elephas maximus. The IUCN Red List of Threatened Species 2008.{/ref} The African elephant was previously treated as a single species, but has recently been separated into the African forest elephant (Loxodonta cyclotis) and African savanna elephant (Loxodonta africana) for evaluation. The forest elephant is listed as ‘critically endangered’ and the savanna elephant as ‘endangered’. The populations of both species are declining.
\n| Elephant species | Population (latest estimate) | Extinction risk | Population trend |
| African elephant (Loxodonta africana) | 415,000 | By subspecies below | By subspecies below |
| African forest elephant (Loxodonta cyclotis) | Critically endangered | Decreasing | |
| African savanna elephant (Loxodonta africana) | Endangered | Decreasing | |
| Asian elephant (Elephas maximus) | 40,000 – 50,000 | Endangered | Decreasing |
To understand the vulnerability of elephant populations, knowing the the number of animals alive today is not enough. We also need to know the direction and rate of change. If population numbers are falling quickly, we should be concerned even if there are hundreds of thousands left.
\n\n\n\nLet’s take a look at the African and Asian elephant species one by one.
\n\n\n\nThere are ten times as many African elephants as Asian elephants in the world. That makes them seem abundant. But their numbers are a tiny fraction of what they were in the past.
\n\n\n\nAfrican elephant populations have shrunk by 98% since 1500. We see this in the chart.{ref}The main sources of data for African elephant populations are the Great Elephant Census and the IUCN SSC African Elephant Specialist Group (AfESG). The AfESG gathers population estimates by country every few years, publishing it in an African Elephant Database, and African Elephant Status Report. You can find the latest African Elephant Status Report (2016) here.
\n\n\n\nIt should be noted that long historical estimates in particular are crude and come with significant uncertainty. Nonetheless they are accurate enough to provide a sense of magnitude for the relative change over time.{/ref}
\n\n\n\nIn 1500, there were over 25 million elephants in Africa. By 1900 this had fallen to around 10 million, and by 1979 down to 1.3 million.
\n\n\n\nThere was a rapid decline in population size over the 1970s and 1980s such that by the mid-1990s numbers had fallen below 300,000. Over the following decades, conservation efforts across some countries managed to restore populations to over 470,000 in 2008. But increased poaching rates over the past decade have sent numbers back into decline.
\n\n\n\nAnother piece of evidence we have that populations have been in decline comes from a metric called the ‘carcass ratio’.
\n\n\n\nDuring population surveys, researchers don’t only count the number of alive elephants, they also count the number of dead elephants (carcasses). The carcass ratio is the number of dead elephants observed during surveys, given as a percentage of the total population.
\n\n\n\nThe carcass ratio across Africa as a whole was 11.9%.{ref}Chase, M. J., Schlossberg, S., Griffin, C. R., Bouché, P. J., Djene, S. W., Elkan, P. W., … & Omondi, P. (2016). Continent-wide survey reveals massive decline in African savannah elephants. PeerJ, 4, e2354.{/ref} This means that for every 100 live elephants, there were around 12 dead elephants. A carcass ratio greater than 8% usually means the population is shrinking, because this will be greater than the replacement rate.{ref}Douglas-Hamilton, I., & Burrill, A. (1991). Using elephant carcass ratios to determine population trends. African wildlife: research and management, 98-105.{/ref}The overall population of African elephants has been falling in recent years. But this varies significantly across countries. In some, the carcass ratio was very high: in Cameroon, it was 83%, it was 32% in Mozambique, and 30% in Angola.{ref}Data is only available for the years 2007 and 2015. So even countries which show an increase in over this decade – Cameroon, for example – might have seen a decline in very recent years, which is reflected in carcass ratio data.{/ref}
\n\n\n\nThere are fewer estimates of Asian Elephant populations. This is more worrying because the Asian elephant is at a higher risk of extinction. We should be tracking these numbers more, not less, closely.
\n\n\n\nWe do have some data for select countries, and some longer-term estimates.
\n\n\n\nThe IUCN estimates that the total population of Asian elephants has more than halved over the past century. It estimates that there were 100,000 animals in the early 1900s; today that figure is in the range of 40,000 to 50,000.
\n\n\n\nPopulation data over time is available for some countries in Asia: the Indian government, for example, has published estimates periodically since 1970.
\n\n\n\nIn the map, you can explore the latest population estimates from the Asian IUCN SSC Asian Elephant Specialist Group (AsESG) for each country.
\n\n\n\nIn India, populations have been steadily increasing since 1980, rising from around 16,000 to over 27,000 in 2017. This shows that it’s possible to protect these species and help their populations rebuild.
\n\n\n\nHowever, the lack of data over time for many countries makes it difficult to properly assess the health of Asian elephant populations.
\nBy far the biggest threat to both African and Asian elephants is poaching. Elephants are killed for their trunks and their tusks. Ivory is a lucrative business.
\n\n\n\nIt’s not just elephants that are under pressure.
\n\n\n\nPoaching is the leading threat to all large mammals. But as we’ve seen from some country-level examples: protecting these species is possible: India has managed to protect and restore elephant populations. Namibia, Zimbabwe, and Angola have also managed to turn the trend.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""How have elephant populations changed over time? What species are at risk of extinction today?"", ""protected"": false}, ""date_gmt"": ""2022-12-01T11:55:45"", ""modified"": ""2022-12-16T13:28:06"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2022-12-16T13:28:06"", ""comment_status"": ""closed"", ""featured_media"": 54668, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/11/Elephants-150x79.png"", ""medium_large"": ""/app/uploads/2022/11/Elephants-768x404.png""}}" 54652,There have been five mass extinctions in Earth's history,mass-extinctions,post,publish,"There have been five big mass extinctions in Earth's history – these are called the ""Big Five"". Understanding the reasons and timelines of these events is important to understand the speed and scale of species extinctions today.
When and why did these mass extinction events happen?
First, we must be clear on what we mean by ""mass extinction"". Extinctions are a normal part of evolution: they occur naturally and periodically over time.{ref}Jablonski D (1986) Mass and background extinctions: the alternation of macroevolutionary regimes. Science 231:129–133{/ref}
There’s a natural background rate to the timing and frequency of extinctions: 10% of species are lost every million years, 30% every 10 million years, and 65% every 100 million years.{ref}Raup DM (1991) A kill curve for Phanerozoic marine species. Paleobiology. 17:37–48.{/ref} It would be wrong to assume that species going extinct is out of line with what we would expect. Evolution occurs through the balance of extinction – the end of species – and speciation – the creation of new ones.
Extinctions occur periodically at what we would call the ""background rate"". We can therefore identify periods of history when extinctions were happening much faster than this background rate – this would tell us that there was an additional environmental or ecological pressure creating more extinctions than we would expect.
However, mass extinctions are periods with much higher extinction rates than normal. They are defined by both magnitude and rate. Magnitude is the percentage of species that are lost. Rate is how quickly this happens. These metrics are inevitably linked, but we need both to qualify as a mass extinction.
In a mass extinction, at least 75% of species go extinct within a relatively (by geological standard) short period of time.{ref}We can see a 75% reduction in species in two ways: high extinction or very low speciation rates. If speciation – the creation of new species – slows down a lot, the extinction rate does not need to be as high as we would expect in order to deplete species numbers by 75%. These events are sometimes called ""mass depletions"" but are treated the same way as mass extinctions.{/ref} Typically less than two million years.
There have been five mass extinction events in Earth’s history, at least since 500 million years ago. We know very little about extinction events in the Precambrian and early Cambrian earlier, which predate this.{ref}Jenkins RJF (1989) The supposed terminal Precambrian extinction event in relation to the Cnidaria. Memoirs of the Association of Australasian Paleontologists 8:307–317.{/ref} These are called the ""Big Five"" for obvious reasons.
In the chart, we see the timing of events in Earth’s history.{ref}This data and detail comes from multiple sources:
Barnosky, A. D., Matzke, N., Tomiya, S., Wogan, G. O., Swartz, B., Quental, T. B., ... & Ferrer, E. A. (2011). Has the Earth’s sixth mass extinction already arrived? Nature, 471(7336), 51-57.
McCallum, M. L. (2015). Vertebrate biodiversity losses point to a sixth mass extinction. Biodiversity and Conservation, 24(10), 2497-2519.
Howard Hughes Medical Institute.{/ref} It shows the changing extinction rate (measured as the number of families that went extinct per million years). Again, note that this number was never zero: background extinction rates were low – typically less than 5 families per million years – but ever-present.
We see the spikes in extinction rates marked as the five events:
Finally, at the end of the timeline, we have the question of what will come. Perhaps we are headed for a sixth mass extinction. But we are currently far from that point.
There are a range of trajectories that the extinction rate could take in the decades and centuries to follow; which one we follow is determined by us.
All of the ""Big Five"" were caused by some combination of rapid and dramatic changes in climate, combined with significant changes in the composition of environments on land or the ocean (such as ocean acidification or acid rain from intense volcanic activity).
In the table here, I detail the proposed causes for each of the five extinction events.{ref}Barnosky, A. D., Matzke, N., Tomiya, S., Wogan, G. O., Swartz, B., Quental, T. B., ... & Ferrer, E. A. (2011). Has the Earth’s sixth mass extinction already arrived?. Nature, 471(7336), 51-57.{/ref}
| Extinction Event | Age(mya) | Percentage of species lost | Cause of extinction |
| End Ordovician | 444 | 86% | Intense glacial and interglacial periods created large sea-level swings and moved shorelines dramatically. The tectonic uplift of the Appalachian mountains created lots of weathering, sequestration of CO2, and with it, changes in climate and ocean chemistry. |
| Late Devonian | 360 | 75% | Rapid growth and diversification of land plants generated rapid and severe global cooling. |
| End Permian | 250 | 96% | Intense volcanic activity in Siberia. This caused global warming. Elevated CO2 and sulfur (H2S) levels from volcanoes caused ocean acidification, acid rain, and other changes in ocean and land chemistry. |
| End Triassic | 200 | 80% | Underwater volcanic activity in the Central Atlantic Magmatic Province (CAMP) caused global warming and a dramatic change in the chemical composition of the oceans. |
| End Cretaceous | 65 | 76% | Asteroid impact in Yucatán, Mexico. This caused a global cataclysm and rapid cooling. Some changes may have already pre-dated this asteroid, with intense volcanic activity and tectonic uplift. |
| Extinction Event | Age(mya) | Percentage of species lost | Cause of extinction |
| End Ordovician | 444 | 86% | Intense glacial and interglacial periods created large sea-level swings and moved shorelines dramatically. The tectonic uplift of the Appalachian mountains created lots of weathering, sequestration of CO2, and with it, changes in climate and ocean chemistry. |
| Late Devonian | 360 | 75% | Rapid growth and diversification of land plants generated rapid and severe global cooling. |
| End Permian | 250 | 96% | Intense volcanic activity in Siberia. This caused global warming. Elevated CO2 and sulfur (H2S) levels from volcanoes caused ocean acidification, acid rain, and other changes in ocean and land chemistry. |
| End Triassic | 200 | 80% | Underwater volcanic activity in the Central Atlantic Magmatic Province (CAMP) caused global warming and a dramatic change in the chemical composition of the oceans. |
| End Cretaceous | 65 | 76% | Asteroid impact in Yucatán, Mexico. This caused a global cataclysm and rapid cooling. Some changes may have already pre-dated this asteroid, with intense volcanic activity and tectonic uplift. |
| Extinction Event | Age(mya) | Percentage of species lost | Cause of extinction |
| End Ordovician | 444 | 86% | Intense glacial and interglacial periods created large sea-level swings and moved shorelines dramatically. The tectonic uplift of the Appalachian mountains created lots of weathering, sequestration of CO2, and with it, changes in climate and ocean chemistry. |
| Late Devonian | 360 | 75% | Rapid growth and diversification of land plants generated rapid and severe global cooling. |
| End Permian | 250 | 96% | Intense volcanic activity in Siberia. This caused global warming. Elevated CO2 and sulfur (H2S) levels from volcanoes caused ocean acidification, acid rain, and other changes in ocean and land chemistry. |
| End Triassic | 200 | 80% | Underwater volcanic activity in the Central Atlantic Magmatic Province (CAMP) caused global warming and a dramatic change in the chemical composition of the oceans. |
| End Cretaceous | 65 | 76% | Asteroid impact in Yucatán, Mexico. This caused a global cataclysm and rapid cooling. Some changes may have already pre-dated this asteroid, with intense volcanic activity and tectonic uplift. |
There have been five big mass extinctions in Earth’s history – these are called the “Big Five”. Understanding the reasons and timelines of these events is important to understand the speed and scale of species extinctions today.
\n\n\n\nWhen and why did these mass extinction events happen?
\n\n\n\nFirst, we must be clear on what we mean by “mass extinction”. Extinctions are a normal part of evolution: they occur naturally and periodically over time.{ref}Jablonski D (1986) Mass and background extinctions: the alternation of macroevolutionary regimes. Science 231:129–133{/ref}
\n\n\n\nThere’s a natural background rate to the timing and frequency of extinctions: 10% of species are lost every million years, 30% every 10 million years, and 65% every 100 million years.{ref}Raup DM (1991) A kill curve for Phanerozoic marine species. Paleobiology. 17:37–48.{/ref} It would be wrong to assume that species going extinct is out of line with what we would expect. Evolution occurs through the balance of extinction – the end of species – and speciation – the creation of new ones.
\n\n\n\nExtinctions occur periodically at what we would call the “background rate”. We can therefore identify periods of history when extinctions were happening much faster than this background rate – this would tell us that there was an additional environmental or ecological pressure creating more extinctions than we would expect.
\n\n\n\nHowever, mass extinctions are periods with much higher extinction rates than normal. They are defined by both magnitude and rate. Magnitude is the percentage of species that are lost. Rate is how quickly this happens. These metrics are inevitably linked, but we need both to qualify as a mass extinction.
\n\n\n\nIn a mass extinction, at least 75% of species go extinct within a relatively (by geological standard) short period of time.{ref}We can see a 75% reduction in species in two ways: high extinction or very low speciation rates. If speciation – the creation of new species – slows down a lot, the extinction rate does not need to be as high as we would expect in order to deplete species numbers by 75%. These events are sometimes called “mass depletions” but are treated the same way as mass extinctions.{/ref} Typically less than two million years.
\n\n\n\nThere have been five mass extinction events in Earth’s history, at least since 500 million years ago. We know very little about extinction events in the Precambrian and early Cambrian earlier, which predate this.{ref}Jenkins RJF (1989) The supposed terminal Precambrian extinction event in relation to the Cnidaria. Memoirs of the Association of Australasian Paleontologists 8:307–317.{/ref} These are called the “Big Five” for obvious reasons.
\n\n\n\nIn the chart, we see the timing of events in Earth’s history.{ref}This data and detail comes from multiple sources:
Barnosky, A. D., Matzke, N., Tomiya, S., Wogan, G. O., Swartz, B., Quental, T. B., … & Ferrer, E. A. (2011). Has the Earth’s sixth mass extinction already arrived? Nature, 471(7336), 51-57.
McCallum, M. L. (2015). Vertebrate biodiversity losses point to a sixth mass extinction. Biodiversity and Conservation, 24(10), 2497-2519.
Howard Hughes Medical Institute.{/ref} It shows the changing extinction rate (measured as the number of families that went extinct per million years). Again, note that this number was never zero: background extinction rates were low – typically less than 5 families per million years – but ever-present.
We see the spikes in extinction rates marked as the five events:
\n\n\n\nFinally, at the end of the timeline, we have the question of what will come. Perhaps we are headed for a sixth mass extinction. But we are currently far from that point.
\n\n\n\nThere are a range of trajectories that the extinction rate could take in the decades and centuries to follow; which one we follow is determined by us.
\n![\""\""](\""https://owid.cloud/app/uploads/2021/03/Big-Five-Mass-Extinctions-800x511.png\"")
All of the “Big Five” were caused by some combination of rapid and dramatic changes in climate, combined with significant changes in the composition of environments on land or the ocean (such as ocean acidification or acid rain from intense volcanic activity).
\n\n\n\nIn the table here, I detail the proposed causes for each of the five extinction events.{ref}Barnosky, A. D., Matzke, N., Tomiya, S., Wogan, G. O., Swartz, B., Quental, T. B., … & Ferrer, E. A. (2011). Has the Earth’s sixth mass extinction already arrived?. Nature, 471(7336), 51-57.{/ref}
\n| Extinction Event | Age(mya) | Percentage of species lost | Cause of extinction |
| End Ordovician | 444 | 86% | Intense glacial and interglacial periods created large sea-level swings and moved shorelines dramatically. The tectonic uplift of the Appalachian mountains created lots of weathering, sequestration of CO2, and with it, changes in climate and ocean chemistry. |
| Late Devonian | 360 | 75% | Rapid growth and diversification of land plants generated rapid and severe global cooling. |
| End Permian | 250 | 96% | Intense volcanic activity in Siberia. This caused global warming. Elevated CO2 and sulfur (H2S) levels from volcanoes caused ocean acidification, acid rain, and other changes in ocean and land chemistry. |
| End Triassic | 200 | 80% | Underwater volcanic activity in the Central Atlantic Magmatic Province (CAMP) caused global warming and a dramatic change in the chemical composition of the oceans. |
| End Cretaceous | 65 | 76% | Asteroid impact in Yucatán, Mexico. This caused a global cataclysm and rapid cooling. Some changes may have already pre-dated this asteroid, with intense volcanic activity and tectonic uplift. |
How many species do we share our planet with? It's such a basic and fundamental question to understanding the world around us.
It's almost unthinkable that we would not know this number or at least have a good estimate. But the truth is that it's a question that continues to escape the world's taxonomists.
An important distinction is how many species we have identified and described and how many species there actually are. We've only identified a small fraction of the world's species, so these numbers are very different.
Before we look at estimates of how many species there are in total, we should first ask the question of how many species we know that we know. Species that we have identified and named.
The IUCN Red List tracks the number of described species and updates this figure annually based on the latest work of taxonomists. In 2021, it listed 2.13 million species on the planet. In the chart, we see the breakdown across a range of taxonomic groups – 1.05 million insects, over 11,000 birds, over 11,000 reptiles, and over 6,000 mammals.
These figures – particularly for lesser-known groups such as plants or fungi – might be a bit too high. This is because some described species are synonyms – the description of already-known species, simply given a separate name.{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth's species before they go extinct?. Science, 339(6118), 413-416.{/ref} There is a continual evaluation process to remove synonyms (and most are removed eventually), but often species are added at a faster rate than synonyms can be found and removed.{ref}Solow, A. R., Mound, L. A., & Gaston, K. J. (1995). Estimating the rate of synonymy. Systematic Biology, 44(1), 93-96.{/ref} To give a sense of how large this effect might be, in a study published in Science, Costello et al. (2013) estimated that around 20% of the described species were undiscovered synonyms (in other words, duplicates).{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth's species before they go extinct?. Science, 339(6118), 413-416.{/ref} They estimated that the 1.9 million described species at the time were actually closer to 1.5 million unique species.
If we were to assume this ""20% synonym"" figure held true, our 2.12 million described species might actually be closer to 1.7 million.
Regardless, we know that all these figures underestimate the actual number of species. The fact that there are so many species that we've yet to discover has real consequences for our ability to understand changes in global biodiversity and the rate of species extinctions.
If we don't know that certain species exist, we also don't know that they might have, or will soon, go extinct. Some species will inevitably go extinct before we realize that they used to exist.
As Robert May summarised in a paper published in Science{ref}May, R. M. (2010). Tropical arthropod species, more or less?. Science, 329(5987), 41-42.{/ref}:
If some alien version of the Starship Enterprise visited Earth, what might be the visitors' first question? I think it would be: “How many distinct life forms—species—does your planet have?” Embarrassingly, our best-guess answer would be in the range of 5 to 10 million eukaryotes (never mind the viruses and bacteria), but we could defend numbers exceeding 100 million, or as low as 3 million.
Researchers have come up with wide-ranging estimates for how many species there are. As May points out, this ranges anywhere from 3 to well over 100 million – many orders of magnitude of difference. Some more recent studies estimate that this figure is as much as one trillion.
One of the most widely cited figures comes from Camilo Mora and colleagues; they estimated that there are around 8.7 million species on Earth today.{ref}Mora, C., Tittensor, D. P., Adl, S., Simpson, A. G., & Worm, B. (2011). How many species are there on Earth and in the ocean?. PLoS Biol, 9(8), e1001127.{/ref} Costello et al. (2013) estimate 5 ± 3 million species; Chapman (2009) estimate 11 million; and after reviewing the range in the literature, Scheffers et al. (2012) choose not to give a concrete figure at all.{ref}
Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth’s species before they go extinct?. Science, 339(6118), 413-416.
A. D. Chapman, Numbers of Living Species in Australia and the World (Biodiversity Information Services, Toowoomba, Australia, 2009).
Scheffers, B. R., Joppa, L. N., Pimm, S. L., & Laurance, W. F. (2012). What we know and don’t know about Earth’s missing biodiversity. Trends in ecology & evolution, 27(9), 501-510.{/ref} More recent studies suggest that the true number is in the billions.
Why is there such large disagreement on the number of species?
The first challenge is even defining what a ‘species’ is. Even in well-known taxonomic groups – such as birds or reptiles – the delineation of species can change over time.{ref}Tobias, J. A., Seddon, N., Spottiswoode, C. N., Pilgrim, J. D., Fishpool, L. D., & Collar, N. J. (2010). Quantitative criteria for species delimitation. Ibis, 152(4), 724-746.{/ref} Our scientific understanding of organisms, and their relationship to others is still improving. That can mean ‘splitting’ a species into multiple, or combining ‘separate species’ into a single one. A clear example of this was when a BirdLife International Review split the Red-bellied Pitta – which was formerly a single bird species – into twelve separate species.
The second challenge is coming up with estimates for groups that are less well-studied than mammals, birds, and reptiles. Most of the disagreement lies is in insects, fungi, and other smaller microbial species. Reaching a consensus on such small and inaccessible lifeforms is undoubtedly hard. There are between 6000 to 7000 known mammal species, but 350,000 - 400,000 of described species of beetles.{ref}Stork, N. E., McBroom, J., Gely, C., & Hamilton, A. J. (2015). New approaches narrow global species estimates for beetles, insects, and terrestrial arthropods. Proceedings of the National Academy of Sciences, 112(24), 7519-7523.{/ref}
The biggest area of uncertainty in species estimates is for bacteria and archaea. This can range from mere thousands to billions.{ref}Dykhuizen, D. (2005). Species numbers in bacteria. Proceedings. California Academy of Sciences, 56(6 Suppl 1), 62.{/ref} A 2017 paper by Larsen et al. estimates that there are 1 to 6 billion species on Earth, and bacteria make up 70% to 90% of them.{ref}Larsen, B. B., Miller, E. C., Rhodes, M. K., & Wiens, J. J. (2017). Inordinate fondness multiplied and redistributed: the number of species on earth and the new pie of life. The Quarterly Review of Biology, 92(3), 229-265.{/ref}
The honest answer to the question, “How many species are there?” is that we don’t really know. Estimates span several orders of magnitude, from a few million to billions. Most recent estimates lean towards the higher range. The biggest uncertainty is in the small lifeforms – bacteria and archaea – where we’ve only described a small percentage of the total.
Update: This article was updated in February 2024 with more discussion on the uncertainty of estimates on the number of species globally.
","{""id"": ""wp-54648"", ""slug"": ""how-many-species-are-there"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""How many species do we share our planet with? It's such a basic and fundamental question to understanding the world around us. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It's almost unthinkable that we would not know this number or at least have a good estimate. But the truth is that it's a question that continues to escape the world's taxonomists."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""An important distinction is "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""how many species we have identified and described"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" and how many species there actually are. We've only identified a small fraction of the world's species, so these numbers are very different."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How many species have we "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""described"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Before we look at estimates of how many species there are in total, we should first ask the question of how many species we "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""know that we know. "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""Species that we have identified and named."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.iucnredlist.org/resources/summary-statistics"", ""children"": [{""text"": ""IUCN Red List"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" tracks the number of described species and updates this figure annually based on the latest work of taxonomists. In 2021, it listed 2.13 million species on the planet. In the chart, we see the breakdown across a range of taxonomic groups – 1.05 million insects, over 11,000 birds, over 11,000 reptiles, and over 6,000 mammals."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These figures – particularly for lesser-known groups such as plants or fungi – might be a bit too high. This is because some described species are synonyms – the description of already-known species, simply given a separate name.{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth's species before they go extinct?. Science, 339(6118), 413-416.{/ref} There is a continual evaluation process to remove synonyms (and most are removed eventually), but often species are added at a faster rate than synonyms can be found and removed.{ref}Solow, A. R., Mound, L. A., & Gaston, K. J. (1995). Estimating the rate of synonymy. Systematic Biology, 44(1), 93-96.{/ref} To give a sense of how large this effect might be, in a study published in Science, Costello et al. (2013) estimated that around 20% of the described species were undiscovered synonyms (in other words, duplicates).{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth's species before they go extinct?. Science, 339(6118), 413-416.{/ref} They estimated that the 1.9 million described species at the time were actually closer to 1.5 million unique species."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If we were to assume this \""20% synonym\"" figure held true, our 2.12 million described species might actually be closer to 1.7 million."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Regardless, we know that all these figures underestimate the actual number of species. The fact that there are so many species that we've yet to discover has real consequences for our ability to understand changes in global biodiversity and the rate of species extinctions. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If we don't know that certain species exist, we also don't know that they might have, or will soon, go extinct. Some species will inevitably go extinct before we realize that they used to exist."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/number-of-described-species?country=Fishes~Insects~Reptiles~Mammals~Birds~Amphibians~Corals~Molluscs~Crustaceans~Arachnids~All+groups~Fungi+and+Protists"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""How many species are there really?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As Robert May summarised in a paper published in "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""{ref}May, R. M. (2010). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://science.sciencemag.org/content/329/5987/41.summary"", ""children"": [{""text"": ""Tropical arthropod species, more or less?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""329"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(5987), 41-42.{/ref}:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""type"": ""text"", ""value"": [{""text"": ""If some alien version of the Starship Enterprise visited Earth, what might be the visitors' first question? I think it would be: “How many distinct life forms—species—does your planet have?” Embarrassingly, our best-guess answer would be in the range of 5 to 10 million eukaryotes (never mind the viruses and bacteria), but we could defend numbers exceeding 100 million, or as low as 3 million."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""blockquote"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Researchers have come up with wide-ranging estimates for how many species there are. As May points out, this ranges anywhere from 3 to well over 100 million – many orders of magnitude of difference. Some more recent studies estimate that this figure is as much as one trillion."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One of the most widely cited figures comes from Camilo Mora and colleagues; they estimated that there are around 8.7 million species on Earth today.{ref}Mora, C., Tittensor, D. P., Adl, S., Simpson, A. G., & Worm, B. (2011). How many species are there on Earth and in the ocean?. PLoS Biol, 9(8), e1001127.{/ref} Costello et al. (2013) estimate 5 ± 3 million species; Chapman (2009) estimate 11 million; and after reviewing the range in the literature, Scheffers et al. (2012) choose not to give a concrete figure at all.{ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth’s species before they go extinct?. Science, 339(6118), 413-416."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A. D. Chapman, Numbers of Living Species in Australia and the World (Biodiversity Information Services, Toowoomba, Australia, 2009)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Scheffers, B. R., Joppa, L. N., Pimm, S. L., & Laurance, W. F. (2012). What we know and don’t know about Earth’s missing biodiversity. Trends in ecology & evolution, 27(9), 501-510.{/ref} More recent studies suggest that the true number is in the billions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Why is there such large disagreement on the number of species?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The first challenge is even defining what a ‘species’ is. Even in well-known taxonomic groups – such as birds or reptiles – the delineation of species can change over time.{ref}Tobias, J. A., Seddon, N., Spottiswoode, C. N., Pilgrim, J. D., Fishpool, L. D., & Collar, N. J. (2010). Quantitative criteria for species delimitation. Ibis, 152(4), 724-746.{/ref} Our scientific understanding of organisms, and their relationship to others is still improving. That can mean ‘splitting’ a species into multiple, or combining ‘separate species’ into a single one. A "", ""spanType"": ""span-simple-text""}, {""url"": ""https://britishbirds.co.uk/content/bb%C2%A0eye%C2%A0do-taxonomic-changes-affect-conservation"", ""children"": [{""text"": ""clear example"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of this was when a BirdLife International Review split the Red-bellied Pitta – which was formerly a single bird species – into twelve separate species."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The second challenge is coming up with estimates for groups that are less well-studied than mammals, birds, and reptiles. Most of the disagreement lies is in insects, fungi, and other smaller microbial species. Reaching a consensus on such small and inaccessible lifeforms is undoubtedly hard. There are between 6000 to 7000 known mammal species, but 350,000 - 400,000 of described species of beetles.{ref}Stork, N. E., McBroom, J., Gely, C., & Hamilton, A. J. (2015). New approaches narrow global species estimates for beetles, insects, and terrestrial arthropods. Proceedings of the National Academy of Sciences, 112(24), 7519-7523.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The biggest area of uncertainty in species estimates is for bacteria and archaea. This can range from mere thousands to billions.{ref}Dykhuizen, D. (2005). Species numbers in bacteria. Proceedings. California Academy of Sciences, 56(6 Suppl 1), 62.{/ref} A 2017 paper by Larsen et al. estimates that there are 1 to 6 billion species on Earth, and bacteria make up 70% to 90% of them.{ref}Larsen, B. B., Miller, E. C., Rhodes, M. K., & Wiens, J. J. (2017). Inordinate fondness multiplied and redistributed: the number of species on earth and the new pie of life. The Quarterly Review of Biology, 92(3), 229-265.{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The honest answer to the question, “How many species are there?” is that we don’t really know. Estimates span several orders of magnitude, from a few million to billions. Most recent estimates lean towards the higher range. The biggest uncertainty is in the small lifeforms – bacteria and archaea – where we’ve only described a small percentage of the total."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""children"": [{""text"": ""Update: This article was updated in February 2024 with more discussion on the uncertainty of estimates on the number of species globally."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""How many species are there?"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""How many species do we share our planet with? How many of these species have we found and identified?"", ""dateline"": ""November 30, 2022"", ""subtitle"": ""How many species do we share our planet with? How many of these species have we found and identified?"", ""sidebar-toc"": false, ""featured-image"": ""Number-of-species-thumbnail.png""}, ""createdAt"": ""2022-11-30T11:42:49.000Z"", ""published"": false, ""updatedAt"": ""2024-02-05T17:42:57.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-11-30T11:42:49.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag quote""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}], ""numBlocks"": 19, ""numErrors"": 3, ""wpTagCounts"": {""html"": 1, ""quote"": 1, ""column"": 4, ""columns"": 2, ""heading"": 2, ""paragraph"": 21, ""separator"": 1}, ""htmlTagCounts"": {""p"": 22, ""h3"": 2, ""hr"": 1, ""div"": 6, ""iframe"": 1, ""blockquote"": 1}}",2022-11-30 11:42:49,2024-03-09 18:11:26,1iHlA_srDqieTz8-l5HHfHdLMRju6BlWQ5FAnXBw5ak8,"[""Hannah Ritchie""]",How many species do we share our planet with? How many of these species have we found and identified?,2022-11-30 11:42:49,2024-02-05 17:42:57,https://ourworldindata.org/wp-content/uploads/2022/11/Number-of-species-thumbnail.png,{},"How many species do we share our planet with? It's such a basic and fundamental question to understanding the world around us. It's almost unthinkable that we would not know this number or at least have a good estimate. But the truth is that it's a question that continues to escape the world's taxonomists. An important distinction is **how many species we have identified and described** and how many species there actually are. We've only identified a small fraction of the world's species, so these numbers are very different. ## How many species have we _described_? Before we look at estimates of how many species there are in total, we should first ask the question of how many species we _know that we know. _Species that we have identified and named. The [IUCN Red List](https://www.iucnredlist.org/resources/summary-statistics) tracks the number of described species and updates this figure annually based on the latest work of taxonomists. In 2021, it listed 2.13 million species on the planet. In the chart, we see the breakdown across a range of taxonomic groups – 1.05 million insects, over 11,000 birds, over 11,000 reptiles, and over 6,000 mammals. These figures – particularly for lesser-known groups such as plants or fungi – might be a bit too high. This is because some described species are synonyms – the description of already-known species, simply given a separate name.{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth's species before they go extinct?. Science, 339(6118), 413-416.{/ref} There is a continual evaluation process to remove synonyms (and most are removed eventually), but often species are added at a faster rate than synonyms can be found and removed.{ref}Solow, A. R., Mound, L. A., & Gaston, K. J. (1995). Estimating the rate of synonymy. Systematic Biology, 44(1), 93-96.{/ref} To give a sense of how large this effect might be, in a study published in Science, Costello et al. (2013) estimated that around 20% of the described species were undiscovered synonyms (in other words, duplicates).{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth's species before they go extinct?. Science, 339(6118), 413-416.{/ref} They estimated that the 1.9 million described species at the time were actually closer to 1.5 million unique species. If we were to assume this ""20% synonym"" figure held true, our 2.12 million described species might actually be closer to 1.7 million. Regardless, we know that all these figures underestimate the actual number of species. The fact that there are so many species that we've yet to discover has real consequences for our ability to understand changes in global biodiversity and the rate of species extinctions. If we don't know that certain species exist, we also don't know that they might have, or will soon, go extinct. Some species will inevitably go extinct before we realize that they used to exist.How many species do we share our planet with? It’s such a basic and fundamental question to understanding the world around us.
\n\n\n\nIt’s almost unthinkable that we would not know this number or at least have a good estimate. But the truth is that it’s a question that continues to escape the world’s taxonomists.
\n\n\n\nAn important distinction is how many species we have identified and described and how many species there actually are. We’ve only identified a small fraction of the world’s species, so these numbers are very different.
\n\n\n\nBefore we look at estimates of how many species there are in total, we should first ask the question of how many species we know that we know. Species that we have identified and named.
\n\n\n\nThe IUCN Red List tracks the number of described species and updates this figure annually based on the latest work of taxonomists. In 2021, it listed 2.13 million species on the planet. In the chart, we see the breakdown across a range of taxonomic groups – 1.05 million insects, over 11,000 birds, over 11,000 reptiles, and over 6,000 mammals.
\n\n\n\nThese figures – particularly for lesser-known groups such as plants or fungi – might be a bit too high. This is because some described species are synonyms – the description of already-known species, simply given a separate name.{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth’s species before they go extinct?. Science, 339(6118), 413-416.{/ref} There is a continual evaluation process to remove synonyms (and most are removed eventually), but often species are added at a faster rate than synonyms can be found and removed.{ref}Solow, A. R., Mound, L. A., & Gaston, K. J. (1995). Estimating the rate of synonymy. Systematic Biology, 44(1), 93-96.{/ref} To give a sense of how large this effect might be, in a study published in Science, Costello et al. (2013) estimated that around 20% of the described species were undiscovered synonyms (in other words, duplicates).{ref}Costello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth’s species before they go extinct?. Science, 339(6118), 413-416.{/ref} They estimated that the 1.9 million described species at the time were actually closer to 1.5 million unique species.
\n\n\n\nIf we were to assume this “20% synonym” figure held true, our 2.12 million described species might actually be closer to 1.7 million.
\n\n\n\nRegardless, we know that all these figures underestimate the actual number of species. The fact that there are so many species that we’ve yet to discover has real consequences for our ability to understand changes in global biodiversity and the rate of species extinctions.
\n\n\n\nIf we don’t know that certain species exist, we also don’t know that they might have, or will soon, go extinct. Some species will inevitably go extinct before we realize that they used to exist.
\nAs Robert May summarised in a paper published in Science{ref}May, R. M. (2010). Tropical arthropod species, more or less?. Science, 329(5987), 41-42.{/ref}:
\n\n\n\n\n\n\n\nIf some alien version of the Starship Enterprise visited Earth, what might be the visitors’ first question? I think it would be: “How many distinct life forms—species—does your planet have?” Embarrassingly, our best-guess answer would be in the range of 5 to 10 million eukaryotes (never mind the viruses and bacteria), but we could defend numbers exceeding 100 million, or as low as 3 million.
Researchers have come up with wide-ranging estimates for how many species there are. As May points out, this ranges anywhere from 3 to well over 100 million – many orders of magnitude of difference. Some more recent studies estimate that this figure is as much as one trillion.
\n\n\n\nOne of the most widely cited figures comes from Camilo Mora and colleagues; they estimated that there are around 8.7 million species on Earth today.{ref}Mora, C., Tittensor, D. P., Adl, S., Simpson, A. G., & Worm, B. (2011). How many species are there on Earth and in the ocean?. PLoS Biol, 9(8), e1001127.{/ref} Costello et al. (2013) estimate 5 ± 3 million species; Chapman (2009) estimate 11 million; and after reviewing the range in the literature, Scheffers et al. (2012) choose not to give a concrete figure at all.{ref}
\n\n\n\nCostello, M. J., May, R. M., & Stork, N. E. (2013). Can we name Earth’s species before they go extinct?. Science, 339(6118), 413-416.
\n\n\n\nA. D. Chapman, Numbers of Living Species in Australia and the World (Biodiversity Information Services, Toowoomba, Australia, 2009).
\n\n\n\nScheffers, B. R., Joppa, L. N., Pimm, S. L., & Laurance, W. F. (2012). What we know and don’t know about Earth’s missing biodiversity. Trends in ecology & evolution, 27(9), 501-510.{/ref} More recent studies suggest that the true number is in the billions.
\n\n\n\nWhy is there such large disagreement on the number of species?
\n\n\n\nThe first challenge is even defining what a ‘species’ is. Even in well-known taxonomic groups – such as birds or reptiles – the delineation of species can change over time.{ref}Tobias, J. A., Seddon, N., Spottiswoode, C. N., Pilgrim, J. D., Fishpool, L. D., & Collar, N. J. (2010). Quantitative criteria for species delimitation. Ibis, 152(4), 724-746.{/ref} Our scientific understanding of organisms, and their relationship to others is still improving. That can mean ‘splitting’ a species into multiple, or combining ‘separate species’ into a single one. A clear example of this was when a BirdLife International Review split the Red-bellied Pitta – which was formerly a single bird species – into twelve separate species.
\n\n\n\nThe second challenge is coming up with estimates for groups that are less well-studied than mammals, birds, and reptiles. Most of the disagreement lies is in insects, fungi, and other smaller microbial species. Reaching a consensus on such small and inaccessible lifeforms is undoubtedly hard. There are between 6000 to 7000 known mammal species, but 350,000 – 400,000 of described species of beetles.{ref}Stork, N. E., McBroom, J., Gely, C., & Hamilton, A. J. (2015). New approaches narrow global species estimates for beetles, insects, and terrestrial arthropods. Proceedings of the National Academy of Sciences, 112(24), 7519-7523.{/ref}
\n\n\n\nThe biggest area of uncertainty in species estimates is for bacteria and archaea. This can range from mere thousands to billions.{ref}Dykhuizen, D. (2005). Species numbers in bacteria. Proceedings. California Academy of Sciences, 56(6 Suppl 1), 62.{/ref} A 2017 paper by Larsen et al. estimates that there are 1 to 6 billion species on Earth, and bacteria make up 70% to 90% of them.{ref}Larsen, B. B., Miller, E. C., Rhodes, M. K., & Wiens, J. J. (2017). Inordinate fondness multiplied and redistributed: the number of species on earth and the new pie of life. The Quarterly Review of Biology, 92(3), 229-265.{/ref}
\n\n\n\nThe honest answer to the question, “How many species are there?” is that we don’t really know. Estimates span several orders of magnitude, from a few million to billions. Most recent estimates lean towards the higher range. The biggest uncertainty is in the small lifeforms – bacteria and archaea – where we’ve only described a small percentage of the total.
\nUpdate: This article was updated in February 2024 with more discussion on the uncertainty of estimates on the number of species globally.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""How many species do we share our planet with? How many of these species have we found and identified?"", ""protected"": false}, ""date_gmt"": ""2022-11-30T11:42:49"", ""modified"": ""2024-02-05T17:42:57"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2024-02-05T17:42:57"", ""comment_status"": ""closed"", ""featured_media"": 54664, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/11/Number-of-species-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2022/11/Number-of-species-thumbnail-768x402.png""}}" 54645,Did humans cause the Quaternary megafauna extinction?,quaternary-megafauna-extinction,post,publish,"Humans have had such a profound impact on the planet’s ecosystems and climate that Earth might be defined by a new geological epoch: the Anthropocene (where “anthropo” means “human”).
Some think this new epoch should start during the Industrial Revolution, and some at the advent of agriculture 10,000 to 15,000 years ago. This feeds into the popular notion that environmental destruction is a recent phenomenon.
The lives of our ancestors are often romanticized. Many think they lived in balance with nature, unlike modern society where we fight against it. But when we look at the evidence of human impacts over millennia, it's hard to see how this was true.
Our ancient ancestors contributed to the extinction of many of the world's largest mammals ('megafauna'). This was during an event known as the Quaternary megafauna extinction (QME).
The extent of these extinctions across continents is shown in the chart. Between 52,000 and 9,000 BCE, more than 178 species of the world’s largest mammals were killed off. These were mammals heavier than 44 kilograms, ranging from mammals the size of sheep to mammoths.
There is strong evidence to suggest that these were largely driven by humans – we look at this in more detail later.
Africa was the least hard-hit, losing only 21% of its megafauna. Humans evolved in Africa, and hominins had already interacted with mammals for a long time. The same is also likely to be true across Eurasia, where 35% of megafauna were lost. But Australia, North America, and South America were particularly hard-hit; very soon after humans arrived, most large mammals were gone. Australia lost 88%; North America lost 83%; and South America, 72%.
Far from being in balance with ecosystems, tiny populations of hunter-gatherers changed them forever. By 8,000 BCE – almost at the end of the QME – there were only around 5 million people in the world.
The driver of the QME has been debated for centuries. Climate and human impacts have been proposed as potential drivers.
One possibility is that both played some role in the downfall of the mammals. This could be through overhunting, the reshaping of landscapes through fire, or the introduction of invasive species.
There are several reasons why we think our ancestors were at least partly responsible.
Extinction timings closely match the timing of human arrival. The timing of megafauna extinctions was not consistent across the world; instead, the timing of their demise coincided closely with the arrival of humans on each continent. The timing of human arrivals and extinction events is shown on the map below.
Humans reached Australia somewhere between 65 to 44,000 years ago.{ref}Andermann, T., Faurby, S., Turvey, S. T., Antonelli, A., & Silvestro, D. (2020). The past and future human impact on mammalian diversity. Science Advances, 6(36), eabb2313.
Smith, F. A., Smith, R. E. E., Lyons, S. K., & Payne, J. L. (2018). Body size downgrading of mammals over the late Quaternary. Science, 360(6386), 310-313.{/ref} Between 50 and 40,000 years ago, 82% of megafauna had been wiped out. It was tens of thousands of years before the extinctions in North and South America occurred. And several more before these occurred in Madagascar and the Caribbean islands. Elephant birds in Madagascar were still present eight millennia after the mammoth and mastodon were killed off in America. Extinction events followed in man’s footsteps.
QME selectively impacted large mammals. There have been many extinction events in Earth’s history. There have been five big mass extinction events and several smaller ones. These events don’t usually target specific groups of animals. Large ecological changes tend to impact everything from large to small mammals, reptiles, birds, and fish. During times of high climate variability over the past 66 million years (the ‘Cenozoic period’), neither small nor large mammals were more vulnerable to extinction.{ref}Smith, F. A., Smith, R. E. E., Lyons, S. K., & Payne, J. L. (2018). Body size downgrading of mammals over the late Quaternary. Science, 360(6386), 310-313.{/ref}
The QME was different and unique in the fossil record: it selectively killed off large mammals. Now there are obvious reasons why larger mammals are at a greater risk of extinction from any cause: they are slower to reproduce, so declines or crashes in their populations take longer to recover from.
But there is also a strong bias to human pressures: humans selectively hunt the larger ones.
Islands were more heavily impacted than Africa. As we saw previously, Africa was less heavily impacted than other continents during this period. We might expect this since hominids had been interacting with mammals for a long time before this. These interactions between species would have impacted mammal populations more gradually and to a lesser extent. They may have already reached some form of equilibrium. When humans arrived on other continents – such as Australia or the Americas – these interactions were new and represented a step-change in the dynamics of the ecosystem. Humans were efficient new predators.
There have now been many studies focused on the question of whether humans were a key driver of the QME. Many suggest that the answer is yes. Climatic changes might have driven an initial decline in large mammal populations – small population crashes – but human pressures are likely to have thwarted their recovery. Large mammals survived previous periods of climatic change, but the arrival of humans put pressure on already-depleted populations.
Human impact on ecosystems dates back tens of thousands of years, despite the Anthropocene paradigm implying this is a recent phenomenon. We’ve not only been in direct competition with other mammals, but we’ve also reshaped the landscape beyond recognition.
Humans have had such a profound impact on the planet’s ecosystems and climate that Earth might be defined by a new geological epoch: the Anthropocene (where “anthropo” means “human”).
\n\n\n\nSome think this new epoch should start during the Industrial Revolution, and some at the advent of agriculture 10,000 to 15,000 years ago. This feeds into the popular notion that environmental destruction is a recent phenomenon.
\n\n\n\nThe lives of our ancestors are often romanticized. Many think they lived in balance with nature, unlike modern society where we fight against it. But when we look at the evidence of human impacts over millennia, it’s hard to see how this was true.
\n\n\n\nOur ancient ancestors contributed to the extinction of many of the world’s largest mammals (‘megafauna’). This was during an event known as the Quaternary megafauna extinction (QME).
\n\n\n\nThe extent of these extinctions across continents is shown in the chart. Between 52,000 and 9,000 BCE, more than 178 species of the world’s largest mammals were killed off. These were mammals heavier than 44 kilograms, ranging from mammals the size of sheep to mammoths.
\n\n\n\nThere is strong evidence to suggest that these were largely driven by humans – we look at this in more detail later.
\n\n\n\nAfrica was the least hard-hit, losing only 21% of its megafauna. Humans evolved in Africa, and hominins had already interacted with mammals for a long time. The same is also likely to be true across Eurasia, where 35% of megafauna were lost. But Australia, North America, and South America were particularly hard-hit; very soon after humans arrived, most large mammals were gone. Australia lost 88%; North America lost 83%; and South America, 72%.
\n\n\n\nFar from being in balance with ecosystems, tiny populations of hunter-gatherers changed them forever. By 8,000 BCE – almost at the end of the QME – there were only around 5 million people in the world.
\n![\""\""](\""https://owid.cloud/app/uploads/2023/09/QME-Extinctions-800x524.png\"")
The driver of the QME has been debated for centuries. Climate and human impacts have been proposed as potential drivers.
\n\n\n\nOne possibility is that both played some role in the downfall of the mammals. This could be through overhunting, the reshaping of landscapes through fire, or the introduction of invasive species.
\n\n\n\nThere are several reasons why we think our ancestors were at least partly responsible.
\n\n\n\nExtinction timings closely match the timing of human arrival. The timing of megafauna extinctions was not consistent across the world; instead, the timing of their demise coincided closely with the arrival of humans on each continent. The timing of human arrivals and extinction events is shown on the map below.
\n\n\n\nHumans reached Australia somewhere between 65 to 44,000 years ago.{ref}Andermann, T., Faurby, S., Turvey, S. T., Antonelli, A., & Silvestro, D. (2020). The past and future human impact on mammalian diversity. Science Advances, 6(36), eabb2313.
Smith, F. A., Smith, R. E. E., Lyons, S. K., & Payne, J. L. (2018). Body size downgrading of mammals over the late Quaternary. Science, 360(6386), 310-313.{/ref} Between 50 and 40,000 years ago, 82% of megafauna had been wiped out. It was tens of thousands of years before the extinctions in North and South America occurred. And several more before these occurred in Madagascar and the Caribbean islands. Elephant birds in Madagascar were still present eight millennia after the mammoth and mastodon were killed off in America. Extinction events followed in man’s footsteps.
QME selectively impacted large mammals. There have been many extinction events in Earth’s history. There have been five big mass extinction events and several smaller ones. These events don’t usually target specific groups of animals. Large ecological changes tend to impact everything from large to small mammals, reptiles, birds, and fish. During times of high climate variability over the past 66 million years (the ‘Cenozoic period’), neither small nor large mammals were more vulnerable to extinction.{ref}Smith, F. A., Smith, R. E. E., Lyons, S. K., & Payne, J. L. (2018). Body size downgrading of mammals over the late Quaternary. Science, 360(6386), 310-313.{/ref}
\n\n\n\nThe QME was different and unique in the fossil record: it selectively killed off large mammals. Now there are obvious reasons why larger mammals are at a greater risk of extinction from any cause: they are slower to reproduce, so declines or crashes in their populations take longer to recover from.
\n\n\n\nBut there is also a strong bias to human pressures: humans selectively hunt the larger ones.
\n\n\n\nIslands were more heavily impacted than Africa. As we saw previously, Africa was less heavily impacted than other continents during this period. We might expect this since hominids had been interacting with mammals for a long time before this. These interactions between species would have impacted mammal populations more gradually and to a lesser extent. They may have already reached some form of equilibrium. When humans arrived on other continents – such as Australia or the Americas – these interactions were new and represented a step-change in the dynamics of the ecosystem. Humans were efficient new predators.
\n\n\n\nThere have now been many studies focused on the question of whether humans were a key driver of the QME. Many suggest that the answer is yes. Climatic changes might have driven an initial decline in large mammal populations – small population crashes – but human pressures are likely to have thwarted their recovery. Large mammals survived previous periods of climatic change, but the arrival of humans put pressure on already-depleted populations.
\n\n\n\nHuman impact on ecosystems dates back tens of thousands of years, despite the Anthropocene paradigm implying this is a recent phenomenon. We’ve not only been in direct competition with other mammals, but we’ve also reshaped the landscape beyond recognition.
\n![\""\""](\""https://owid.cloud/app/uploads/2023/09/Human-arrival-map.png\"")
Weighing in at over 150 tonnes, Blue Whales are the largest animal to have lived on Earth. But their size hasn’t protected them from human exploitation. In fact, their size made them an incredibly valuable source of oil, meat and blubber. This fat gave us a rich fuel for energy and many other industrial uses.
Human hunting has resulted in a massive decline in global whale populations. Some have been pushed to the brink of extinction. But, as we will see, a combination of technological change, economic incentives and international policies have brought global whaling to very low levels. While some populations are still very small, we have been successful in preserving these beautiful animals.
Whaling is a practice that dates back to Medieval times; fortunately at low levels, meaning this early practice had little impact on whale populations.{ref}Birnie, P. W. (1989). International legal issues in the management and protection of the whale: a review of four decades of experience. Natural Resources Journal, 29(4), 903-934.{/ref}
While global whaling was a feature of the 1900s, the peak of whaling in the United States was a century earlier. In the 18th and 19th centuries, whaling was a key industry in the US.
Although the uses of whale oil later diversified, Americans mainly used it for lighting. Whale oil was used for the lighting of not only homes but also outdoor street lighting, lighthouses and miner’s headlamps – a range of uses that had a big impact on civilization during that period.{ref}Starbuck, A., 1878, (reprinted 1989), History of the American whale fishery: Secaucus, NJ., Castle.{/ref}
As we see in the chart, US production of whale and sperm oil followed the classic inverted-U curve, peaking in the mid-19th century. At this point, petroleum oil had been discovered, and kerosene – which was cheaper than whale oil – began to replace it in lighting. As whale hunting became increasingly unprofitable, production soon declined.{ref}Coleman, J. L. (1995). The American whale oil industry: A look back to the future of the American petroleum industry?. Nonrenewable Resources, 4(3), 273-288.{/ref}
While the United States dominated whaling during the 19th century, it didn’t really reach the global stage until the 20th century.
Towards the end of the 1800s, new technologies were being developed that could catch whales in much larger numbers. Rather than the classic sail- or oar-powered boats that the Americans had been using, the Norwegians developed mechanized, steam-powered vessels with cannons and harpoons. This made whaling much more efficient. Not only could we catch more whales, it also allowed us to catch species – like Blue and Fin Whales – that were too fast for our old technologies.
This marked the start of the ‘modern’ whaling era, from around 1890 onwards. It had a dramatic impact.
Not all species were impacted equally by whaling. Some of the largest – the Blue and Fin Whale in particular – were prime targets. This meant that the decline of the biomass of whales over this period was even greater.
We see this discrimination towards particular species in the chart. All species declined, but the extent was wildly different. The Minke Whale saw a fall of just 20%.{ref}We can calculate this as: [(637,000 - 507,000) / 637,000 = 20%].{/ref} The Blue Whale was almost plundered into extinction. Its populations fell from 340,000 to just 5,000. A reduction of 98.5%.{ref}We can calculate this as: [(340,000 - 5,000) / 340,000 = 98.5%].{/ref} The Fin Whale lost 85% of its numbers.{ref}We can calculate this as: [(762,000 - 110,000) / 762,000 = 85%].{/ref}
The 20th century didn’t only bring technological innovations on how to track and hunt whales, it also brought advances in how we could use the oils, blubber, and bone that they provided. Both demand and supply rocketed.
Whale oils were initially used for lighting, but their market soon expanded. Sperm oil is special because it maintains its lubricating qualities at very high temperatures – this made it a vital ingredient for machinery, engines, guns and watches during the Industrial Revolution. Advancements in cosmetics and food chemistry meant that its by-products were soon used for soaps, textiles, and even margarine.
Ambergris – a substance found in the intestine of sperm whales – was, and still is, used to make perfume. You will find it in the luxury perfume, Chanel No.5. Whales made it into the fashion industry too. Instead of teeth, baleen whales have long strips of keratin (the substance found in human nails and hair) which hang from their mouths – these plates were used in everything from skirts and women’s corsets, to umbrellas, parasols, fishing poles and crossbows.{ref}Tønnessen, J.N. and Johnsen, A.O. The history of modern whaling. University of California Press, Berkeley.{/ref}
In the chart, we see how this rising demand for whale products affected hunting rates across the world. This shows the number of whales killed each year, from 1900 onwards.
For decades, tens of thousands of whales were killed each year. Only World War II gave these animals some reprieve – notice the drop in kills during the early-1940s.
At its peak in the 1960s, we were hunting 80,000 whales every year.
It didn’t go unnoticed that this rate of whale catch was unsustainable. In fact, a number of countries formed the International Whaling Commission (IWC) in 1946 to work out how to manage international whaling stocks. Quotas were introduced, but countries weren’t adhering to them.{ref}Clapham , P.J. and B aker, C.S. (2002). Modern whaling. In: Perrin , W.F., Würsig, B. & T hewissen, J.G .M. (eds.), Encyclopedia of Marine Mammals, pp. 1328-1332. Academic Press, New York.{/ref} Whaling rates continued to rise, even after the lull of World War II. By the mid-20th century, many species were pushed to the brink of extinction.
By the mid-20th century, the prospects for whales was looking bleak. And yet, the world managed to turn things around.
In the chart we see the number of whales killed per decade.
We see a gradual rise over the first half of the century – with the exception of the war-stricken 1940s – with whale catch reaching its peak in the 1960s. 703,000 whales were killed during the 1960s.
But, we've seen a decline since then. What caused this?
There were a number of factors at play. By the 1960s, whale populations had become increasingly depleted – this scarcity had an impact on the economic incentives to hunt them. Whales were becoming harder to find and catch. Technological advances also meant that substitutes for whale oil and bone in the cosmetic, food, and textile industries were becoming cheaper and more accessible. The whaling industry was losing its profitability.
There was another massive policy change in the 1980s. More and more countries were becoming members of the International Whaling Commission.
In the map, we see the timeline by which countries joined the IWC.
After many decades of failed quota agreements, the IWC agreed to a global moratorium. This made commercial whaling illegal, with only a few exceptions.{ref}The moratorium only applies to commercial whaling, so whaling classified for scientific research purposes and aboriginal-subsistence provisions is still allowed.{/ref} This came into action in 1987, and we see a dramatic drop-off in whale catch in the decades since then.
You will notice that we haven’t eliminated whaling completely. Some countries – such as Japan, Norway and Iceland – have resumed limited whaling outside the guidelines of the IWC.{ref}The moratorium only applies to commercial whaling, so whaling classified for scientific research purposes and aboriginal-subsistence provisions is still allowed.{/ref} In 2019, Japan withdrew from the international agreement. These countries typically hunt species that are not considered to be critically endangered.
Some species will still take many decades to recover. Populations of North Atlantic right whale, the Arctic bowhead, and the Pacific blue whale are still critical.
Yet the story of whaling overall is a conservation success: an animal that was once in high demand across the world has been saved from extinction through substitution and international cooperation. Many species are in a similar position to whales in the 1960s. But the decline of whaling should give us hope that we can turn things around once again.
Weighing in at over 150 tonnes, Blue Whales are the largest animal to have lived on Earth. But their size hasn’t protected them from human exploitation. In fact, their size made them an incredibly valuable source of oil, meat and blubber. This fat gave us a rich fuel for energy and many other industrial uses.
\n\n\n\nHuman hunting has resulted in a massive decline in global whale populations. Some have been pushed to the brink of extinction. But, as we will see, a combination of technological change, economic incentives and international policies have brought global whaling to very low levels. While some populations are still very small, we have been successful in preserving these beautiful animals.
\n\n\n\nWhaling is a practice that dates back to Medieval times; fortunately at low levels, meaning this early practice had little impact on whale populations.{ref}Birnie, P. W. (1989). International legal issues in the management and protection of the whale: a review of four decades of experience. Natural Resources Journal, 29(4), 903-934.{/ref}
\n\n\n\nWhile global whaling was a feature of the 1900s, the peak of whaling in the United States was a century earlier. In the 18th and 19th centuries, whaling was a key industry in the US.
\n\n\n\nAlthough the uses of whale oil later diversified, Americans mainly used it for lighting. Whale oil was used for the lighting of not only homes but also outdoor street lighting, lighthouses and miner’s headlamps – a range of uses that had a big impact on civilization during that period.{ref}Starbuck, A., 1878, (reprinted 1989), History of the American whale fishery: Secaucus, NJ., Castle.{/ref}
\n\n\n\nAs we see in the chart, US production of whale and sperm oil followed the classic inverted-U curve, peaking in the mid-19th century. At this point, petroleum oil had been discovered, and kerosene – which was cheaper than whale oil – began to replace it in lighting. As whale hunting became increasingly unprofitable, production soon declined.{ref}Coleman, J. L. (1995). The American whale oil industry: A look back to the future of the American petroleum industry?. Nonrenewable Resources, 4(3), 273-288.{/ref}
\n![\""\""](\""https://owid.cloud/app/uploads/2021/03/Whale-oil-production-in-the-US-562x550.png\"")
While the United States dominated whaling during the 19th century, it didn’t really reach the global stage until the 20th century.
\n\n\n\nTowards the end of the 1800s, new technologies were being developed that could catch whales in much larger numbers. Rather than the classic sail- or oar-powered boats that the Americans had been using, the Norwegians developed mechanized, steam-powered vessels with cannons and harpoons. This made whaling much more efficient. Not only could we catch more whales, it also allowed us to catch species – like Blue and Fin Whales – that were too fast for our old technologies.
\n\n\n\nThis marked the start of the ‘modern’ whaling era, from around 1890 onwards. It had a dramatic impact.
\n\n\n\nNot all species were impacted equally by whaling. Some of the largest – the Blue and Fin Whale in particular – were prime targets. This meant that the decline of the biomass of whales over this period was even greater.
\n\n\n\nWe see this discrimination towards particular species in the chart. All species declined, but the extent was wildly different. The Minke Whale saw a fall of just 20%.{ref}We can calculate this as: [(637,000 – 507,000) / 637,000 = 20%].{/ref} The Blue Whale was almost plundered into extinction. Its populations fell from 340,000 to just 5,000. A reduction of 98.5%.{ref}We can calculate this as: [(340,000 – 5,000) / 340,000 = 98.5%].{/ref} The Fin Whale lost 85% of its numbers.{ref}We can calculate this as: [(762,000 – 110,000) / 762,000 = 85%].{/ref}
\nThe 20th century didn’t only bring technological innovations on how to track and hunt whales, it also brought advances in how we could use the oils, blubber, and bone that they provided. Both demand and supply rocketed.
\n\n\n\nWhale oils were initially used for lighting, but their market soon expanded. Sperm oil is special because it maintains its lubricating qualities at very high temperatures – this made it a vital ingredient for machinery, engines, guns and watches during the Industrial Revolution. Advancements in cosmetics and food chemistry meant that its by-products were soon used for soaps, textiles, and even margarine.
\n\n\n\nAmbergris – a substance found in the intestine of sperm whales – was, and still is, used to make perfume. You will find it in the luxury perfume, Chanel No.5. Whales made it into the fashion industry too. Instead of teeth, baleen whales have long strips of keratin (the substance found in human nails and hair) which hang from their mouths – these plates were used in everything from skirts and women’s corsets, to umbrellas, parasols, fishing poles and crossbows.{ref}Tønnessen, J.N. and Johnsen, A.O. The history of modern whaling. University of California Press, Berkeley.{/ref}
\n\n\n\nIn the chart, we see how this rising demand for whale products affected hunting rates across the world. This shows the number of whales killed each year, from 1900 onwards.
\n\n\n\nFor decades, tens of thousands of whales were killed each year. Only World War II gave these animals some reprieve – notice the drop in kills during the early-1940s.
\n\n\n\nAt its peak in the 1960s, we were hunting 80,000 whales every year.
\n\n\n\nIt didn’t go unnoticed that this rate of whale catch was unsustainable. In fact, a number of countries formed the International Whaling Commission (IWC) in 1946 to work out how to manage international whaling stocks. Quotas were introduced, but countries weren’t adhering to them.{ref}Clapham , P.J. and B aker, C.S. (2002). Modern whaling. In: Perrin , W.F., Würsig, B. & T hewissen, J.G .M. (eds.), Encyclopedia of Marine Mammals, pp. 1328-1332. Academic Press, New York.{/ref} Whaling rates continued to rise, even after the lull of World War II. By the mid-20th century, many species were pushed to the brink of extinction.
\nBy the mid-20th century, the prospects for whales was looking bleak. And yet, the world managed to turn things around.
\n\n\n\nIn the chart we see the number of whales killed per decade.
\n\n\n\nWe see a gradual rise over the first half of the century – with the exception of the war-stricken 1940s – with whale catch reaching its peak in the 1960s. 703,000 whales were killed during the 1960s.
\n\n\n\nBut, we’ve seen a decline since then. What caused this?
\n\n\n\nThere were a number of factors at play. By the 1960s, whale populations had become increasingly depleted – this scarcity had an impact on the economic incentives to hunt them. Whales were becoming harder to find and catch. Technological advances also meant that substitutes for whale oil and bone in the cosmetic, food, and textile industries were becoming cheaper and more accessible. The whaling industry was losing its profitability.
\n\n\n\nThere was another massive policy change in the 1980s. More and more countries were becoming members of the International Whaling Commission.
\n\n\n\nIn the map, we see the timeline by which countries joined the IWC.
\n\n\n\nAfter many decades of failed quota agreements, the IWC agreed to a global moratorium. This made commercial whaling illegal, with only a few exceptions.{ref}The moratorium only applies to commercial whaling, so whaling classified for scientific research purposes and aboriginal-subsistence provisions is still allowed.{/ref} This came into action in 1987, and we see a dramatic drop-off in whale catch in the decades since then.
\n\n\n\nYou will notice that we haven’t eliminated whaling completely. Some countries – such as Japan, Norway and Iceland – have resumed limited whaling outside the guidelines of the IWC.{ref}The moratorium only applies to commercial whaling, so whaling classified for scientific research purposes and aboriginal-subsistence provisions is still allowed.{/ref} In 2019, Japan withdrew from the international agreement. These countries typically hunt species that are not considered to be critically endangered.
\n\n\n\nSome species will still take many decades to recover. Populations of North Atlantic right whale, the Arctic bowhead, and the Pacific blue whale are still critical.
\n\n\n\nYet the story of whaling overall is a conservation success: an animal that was once in high demand across the world has been saved from extinction through substitution and international cooperation. Many species are in a similar position to whales in the 1960s. But the decline of whaling should give us hope that we can turn things around once again.
\nThe largest mammals are at the greatest risk of extinction. This has been the case since the arrival of humans, and is still true today.
Weighing in at several tonnes, rhinos are some of the world’s largest mammals.{ref}The average mass of rhinos vary by species and sex. Male White and Indian rhinos can weigh up to 2700 kilograms; at the other end, Sumatran rhinos have a lower range of around 600 kilograms.{/ref} With extravagant prices for rhino horns and body parts, they’re also one of the most threatened: three of the five rhino species are ‘critically endangered’.
In this article we take a look at the status of rhino populations today, and whether they are in decline or on the path to recovery. They offer us both stories of concern and reasons for optimism.
There are five species of rhino. Africa is home to the White and Black Rhino. The remaining three – the Sumatran, Javan and Indian – exist in Asia.
Some of these species have dangerously low population levels.
The Sumatran and the Javan Rhino can now only be found in Indonesia. They are Critically Endangered with less than 100 individuals left in the wild. The Black Rhino, despite having a population in the thousands, is also Critically Endangered. It experienced a rapid decline over the 20th century as we will see later.
The Indian (Greater One-Horned) offers us an important success story; it has seen a significant recovery in recent decades. The White Rhino too has shown an impressive recovery. But it has a dark side: the Southern White Rhino might be doing well but the Northern sub-species is on the brink of extinction. There are only two Northern White Rhinos left, and both of them are female.
The health of these rhino populations is not just determined by how many animals are left today. The direction and pace that these populations are changing matters too.
For each species I have built a time-series of populations globally and by country. This data is aggregated from multiple sources: the main one being the African and Asian Rhino Specialist Groups (AfrSG) and TRAFFIC, which collate statistics on all rhino populations, and submit them to the IUCN.{ref}Emslie, R.H. et al., 2019. African and Asian rhinoceroses - status, conservation and trade. A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.
Other sources used to build these time-series include:
Pusparini, W., Sievert, P. R., Fuller, T. K., Randhir, T. O., & Andayani, N. (2015). Rhinos in the parks: an island-wide survey of the last wild population of the Sumatran rhinoceros. PloS one, 10(9), e0136643.
Emslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino
Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.
Thapa, K., Nepal, S., Thapa, G., Bhatta, S. R., & Wikramanayake, E. (2013). Past, present and future conservation of the greater one-horned rhinoceros Rhinoceros unicornis in Nepal. Oryx, 47(3), 345-351.{/ref}
Let’s take a look at each of the five species one-by-one.
Overall, the story of the White rhino is a positive one. But this tale also has a darker side. There are two main subspecies: the Northern White and Southern White rhino. A century ago, the Northern was much more abundant than the Southern. Now the opposite is true.
In the chart we see the population trend of the Southern White Rhino. It’s touted as one of the world’s greatest conservation success stories.{ref}Emslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.{/ref} We don’t have precise estimates but it’s reported that by the mid-19th century it was still abundant. However, intense poaching by the Europeans and killings in the conversion of land for agriculture meant that by the late 19th century it was close to extinction. By 1900 there were only 20 left. All were in the Hluhluwe–iMfolozi Park in South Africa – now a nature reserve.
Over the course of the 20th century, severe protection of these species – particularly in African nature reserves – led to a significant and rapid increase. Populations were restored to more than 21,000. Numbers grew 1000-fold within a century.
Over the last few years, increases in poaching rates have unfortunately led to another decline. South Africa is home to around three-quarters of Southern White Rhinos. You can explore population estimates by country in the chart.
If the story of the Southern White Rhino is one of our greatest successes, the story of its Northern cousin must be one of our biggest failures. The Northern White Rhino is on the brink of extinction. There are only two individuals left. Both are female.
In the chart we see its demise over the second-half of the 20th century. In 1960, it’s estimated there were more than 2,000, predominantly in Sudan and the Democratic Republic of Congo. Intense poaching, and challenges for protection during civil unrest in stronghold countries, has led to a rapid decline.
Sudan, the last remaining male died in Kenya in 2018. There are now only two female rhinos left: Najin and her daughter, Fatu. Both are guarded in a semi-wild enclosure, and have had their horns sawed off to deter poachers. Scientists are investigating ways to continue reproduction from the last females, including stem cell treatment and hybrid embryos from Northern White Rhino eggs and Southern White Rhino sperm.{ref}Saragusty, J., Diecke, S., Drukker, M., Durrant, B., Friedrich Ben‐Nun, I., Galli, C., ... & Johnson, S. (2016). Rewinding the process of mammalian extinction. Zoo Biology, 35(4), 280-292.
Callaway, E. (2016). Stem-cell plan aims to bring rhino back from brink of extinction. Nature News, 533(7601), 20.{/ref}
In 2021, researchers aim to implant embryos in both to try to get them to reproduce. Scientific innovation is now the only way to save these sub-species. Even if reproduction is successful, population numbers will be incredibly low for a long time. They will have to be closely guarded for a long time.
The story of the Black Rhino is similar to that of the Northern White. The once abundant species has seen a dramatic decline over the 20th century.
Even after intense poaching by European settlers over the 19th and early 20th century, in 1960 there were still around 100,000 Black rhinos in Africa.{ref}Emslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.{/ref} The decline which followed was rapid and dramatic: the number fell by more than 80% in only two decades. The population reached its low point in the early 1990s at around 2,500, and have since began to recover. The number of Black Rhinos has more than doubled to around 6,000. Nonetheless, continued poaching has still limited population growth in recent decades. The Black Rhino is still classified as Critically Endangered.{ref}Emslie, R.H. et al., 2019. African and Asian rhinoceroses - status, conservation and trade. A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.{/ref}
The Indian (also called the Greater One-Horned) Rhino provides one success story in the restoration of wild mammals.
By the mid-20th century, there were very few Indian Rhinos left in the world. It’s estimated (with larger uncertainty since it predates regular surveying) that there were only 40 individuals left.
Since the mid-1960s, populations have increased nearly 100-fold. Latest estimates, taken in 2021, suggest there are now over 4,000 rhinos in the wild. This was the result of impressive conservation efforts to reduce poaching in both India and Nepal.
Unsurprising based on the name, India is home to more than 80% of the species. But Nepal also has a steady population. With only a few being spotted in recent decades, they are thought to now be extinct in Pakistan.
The Javan Rhino is one of the world’s most endangered mammals. It is estimated there were only 76 rhinos left in 2021. This makes it Critically Endangered.
In the chart you see how Javan Rhino populations have changed over time.
In recent decades it has existed in two countries: Indonesia and Vietnam. But by 2010 it had gone extinct in Vietnam. Indonesia is now its only remaining home.
Its total population has, however, increased from 50 years ago. In the 1960s there were only 20 to 30 Javan Rhinos left in the world. From then until the 1990s, the population approximately doubled. The latest estimate puts this figure at 76 Javan Rhinos.
Like the Javan species, Sumatran rhinos are also one of the most endangered mammals. They’re Critically Endangered. There were only 41 left in the world in 2021.{ref}Estimates range from 34 to 47.
Emslie, R.H. et al., 2019. African and Asian rhinoceroses - status, conservation and trade. A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.{/ref}
In the chart you can see how its population has changed over time.
Once found in both Malaysia and Indonesia, the Sumatran Rhino is thought to be extinct in Malaysia; none have been recorded in 2016 and 2018 surveys.
But in contrast to the Javan, the number of Sumatran Rhinos in the world has been falling in recent decades, from 600 individuals in the mid-1980s to around one-tenth of that figure today. Both Malaysian and Indonesian rhino populations have contributed to this loss.
The biggest threat to rhinos is poaching. Rhino horns are still seen as luxury goods and can sell for a lot of money in illegal wildlife markets.
But rhinos are not the only species at risk from poaching. It’s the leading threat for most large mammals. But some success stories – such as the restoration of Southern White and Indian Rhino populations – shows us that their demise is not inevitable. With the right approach we have the opportunity to turn things around. In a follow-up article we will look at the scale of global poaching, and what we can learn from the countries that have been successful in bringing it to an end.
","{""id"": ""wp-54629"", ""slug"": ""rhino-populations"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""The largest mammals are at the greatest risk of extinction. This has been the case since the arrival of humans, and is still true today."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Weighing in at several tonnes, rhinos are some of the world’s largest mammals.{ref}The "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20230130123018/https://rhinos.org/wp-content/uploads/2020/10/IRF_factsheet_digital_Mar2020-1.pdf"", ""children"": [{""text"": ""average mass of rhinos"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" vary by species and sex. Male White and Indian rhinos can weigh up to 2700 kilograms; at the other end, Sumatran rhinos have a lower range of around 600 kilograms.{/ref} With extravagant prices for rhino horns and body parts, they’re also one of the most threatened: three of the five rhino species are ‘critically endangered’. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this article we take a look at the status of rhino populations today, and whether they are in decline or on the path to recovery. They offer us both stories of concern and reasons for optimism."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""State of rhino populations today"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are five species of rhino. Africa is home to the White and Black Rhino. The remaining three – the Sumatran, Javan and Indian – exist in Asia. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Some of these species have dangerously low population levels."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Sumatran and the Javan Rhino can now only be found in Indonesia. They are Critically Endangered with less than 100 individuals left in the wild. The Black Rhino, despite having a population in the thousands, is also Critically Endangered. It experienced a rapid decline over the 20th century as we will see later."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Indian (Greater One-Horned) offers us an important success story; it has seen a significant recovery in recent decades. The White Rhino too has shown an impressive recovery. But it has a dark side: the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Southern"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" White Rhino might be doing well but the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Northern"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" sub-species is on the brink of extinction. There are only two Northern White Rhinos left, and both of them are female."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Are rhino populations increasing or decreasing?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The health of these rhino populations is not just determined by how many animals are left today. The direction and pace that these populations are changing matters too. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""For each species I have built a time-series of populations globally and by country. This data is aggregated from multiple sources: the main one being the African and Asian Rhino Specialist Groups (AfrSG) and TRAFFIC, which collate statistics on all rhino populations, and submit them to the IUCN.{ref}Emslie, R.H. et al., 2019. "", ""spanType"": ""span-simple-text""}, {""url"": ""http://www.rhinoresourcecenter.com/pdf_files/156/1560170144.pdf"", ""children"": [{""text"": ""African and Asian rhinoceroses - status, conservation and trade"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Other sources used to build these time-series include:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Pusparini, W., Sievert, P. R., Fuller, T. K., Randhir, T. O., & Andayani, N. (2015). Rhinos in the parks: an island-wide survey of the last wild population of the Sumatran rhinoceros. PloS one, 10(9), e0136643."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Emslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Thapa, K., Nepal, S., Thapa, G., Bhatta, S. R., & Wikramanayake, E. (2013). Past, present and future conservation of the greater one-horned rhinoceros Rhinoceros unicornis in Nepal. Oryx, 47(3), 345-351.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Let’s take a look at each of the five species one-by-one."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""White rhino ("", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Ceratotherium simum"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "")"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Overall, the story of the White rhino is a positive one. But this tale also has a darker side. There are two main subspecies: the Northern White and Southern White rhino. A century ago, the Northern was much more abundant than the Southern. Now the opposite is true."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart we see the population trend of the Southern White Rhino. It’s touted as one of the world’s greatest conservation success stories.{ref}Emslie, R. and Brooks, M. (1999) "", ""spanType"": ""span-simple-text""}, {""url"": ""https://portals.iucn.org/library/sites/library/files/documents/1999-049.pdf"", ""children"": [{""text"": ""African Rhino. Status Survey and Conservation Action Plan"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.{/ref} We don’t have precise estimates but it’s reported that by the mid-19th century it was still abundant. However, intense poaching by the Europeans and killings in the conversion of land for agriculture meant that by the late 19th century it was close to extinction. By 1900 there were only 20 left. All were in the Hluhluwe–iMfolozi Park in South Africa – now a nature reserve. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Over the course of the 20th century, severe protection of these species – particularly in African nature reserves – led to a significant and rapid increase. Populations were restored to more than 21,000. Numbers grew 1000-fold within a century."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Over the last few years, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/number-of-rhinos-poached"", ""children"": [{""text"": ""increases in poaching rates"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" have unfortunately led to another decline. South Africa is home to around three-quarters of Southern White Rhinos. You can explore population estimates by country in the chart."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/southern-white-rhinos?country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/southern-white-rhinos?tab=map"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If the story of the Southern White Rhino is one of our greatest successes, the story of its Northern cousin must be one of our biggest failures. The "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Northern White Rhino"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" is on the brink of extinction. There are only two individuals left. Both are female. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart we see its demise over the second-half of the 20th century. In 1960, it’s estimated there were "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/northern-white-rhinos"", ""children"": [{""text"": ""more than 2,000"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", predominantly in Sudan and the Democratic Republic of Congo. Intense poaching, and challenges for protection during civil unrest in stronghold countries, has led to a rapid decline."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Sudan, the last remaining male died in Kenya in 2018. There are now only two female rhinos left: Najin and her daughter, Fatu. Both are guarded in a semi-wild enclosure, and have had their horns sawed off to deter poachers. Scientists are investigating ways to continue reproduction from the last females, including stem cell treatment and hybrid embryos from Northern White Rhino eggs and Southern White Rhino sperm.{ref}Saragusty, J., Diecke, S., Drukker, M., Durrant, B., Friedrich Ben‐Nun, I., Galli, C., ... & Johnson, S. (2016). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=doi.org%2F10.1002%2Fzoo.21284&btnG="", ""children"": [{""text"": ""Rewinding the process of mammalian extinction"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Zoo Biology"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 35(4), 280-292."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""Callaway, E. (2016). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/news/stem-cell-plan-aims-to-bring-rhino-back-from-brink-of-extinction-1.19849"", ""children"": [{""text"": ""Stem-cell plan aims to bring rhino back from brink of extinction"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Nature News"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 533(7601), 20.{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In 2021, researchers "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.newscientist.com/article/2264951-embryos-set-to-be-implanted-in-the-last-two-northern-white-rhinos/"", ""children"": [{""text"": ""aim to implant"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" embryos in both to try to get them to reproduce. Scientific innovation is now the only way to save these sub-species. Even if reproduction is successful, population numbers will be incredibly low for a long time. They will have to be closely guarded for a long time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/northern-white-rhinos"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/northern-white-rhinos?tab=map"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Black Rhino ("", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Diceros bicornis"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "")"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The story of the Black Rhino is similar to that of the Northern White. The once abundant species has seen a dramatic decline over the 20th century."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Even after intense poaching by European settlers over the 19th and early 20th century, in 1960 there were still around 100,000 Black rhinos in Africa.{ref}Emslie, R. and Brooks, M. (1999) "", ""spanType"": ""span-simple-text""}, {""url"": ""https://portals.iucn.org/library/sites/library/files/documents/1999-049.pdf"", ""children"": [{""text"": ""African Rhino. Status Survey and Conservation Action Plan"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.{/ref} The decline which followed was rapid and dramatic: the number fell by more than 80% in only two decades. The population reached its low point in the early 1990s at around 2,500, and have since began to recover. The number of Black Rhinos has more than doubled to around 6,000. Nonetheless, continued poaching has still limited population growth in recent decades. The Black Rhino is still classified as Critically Endangered.{ref}Emslie, R.H. et al., 2019. "", ""spanType"": ""span-simple-text""}, {""url"": ""http://www.rhinoresourcecenter.com/pdf_files/156/1560170144.pdf"", ""children"": [{""text"": ""African and Asian rhinoceroses - status, conservation and trade"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/black-rhinos?tab=chart"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/black-rhinos?tab=map"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Indian / Greater One-Horned Rhino ("", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Rhinoceros unicornis"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "")"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Indian (also called the Greater One-Horned) Rhino provides one success story in the restoration of wild mammals."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""By the mid-20th century, there were very few Indian Rhinos left in the world. It’s estimated (with larger uncertainty since it predates regular surveying) that there were only 40 individuals left. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Since the mid-1960s, populations have increased nearly 100-fold. Latest estimates, taken in 2021, suggest there are now over 4,000 rhinos in the wild. This was the result of impressive conservation efforts to reduce poaching in both India and Nepal."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Unsurprising based on the name, India is home to more than 80% of the species. But Nepal also has a steady population. With only a few being spotted in recent decades, they are thought to now be extinct in Pakistan."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/indian-rhinos"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/indian-rhinos?tab=map"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Javan Rhino ("", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Rhinoceros sondaicus"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "")"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Javan Rhino is one of the world’s most endangered mammals. It is estimated there were only 76 rhinos left in 2021. This makes it Critically Endangered."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart you see how Javan Rhino populations have changed over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In recent decades it has existed in two countries: Indonesia and Vietnam. But by 2010 it had gone extinct in Vietnam. Indonesia is now its only remaining home."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Its total population has, however, increased from 50 years ago. In the 1960s there were only 20 to 30 Javan Rhinos left in the world. From then until the 1990s, the population approximately doubled. The latest estimate puts this figure at 76 Javan Rhinos."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/javan-rhinos"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/javan-rhinos?tab=map"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Sumatran rhino ("", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Dicerorhinus sumatrensis"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "")"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Like the Javan species, Sumatran rhinos are also one of the most endangered mammals. They’re Critically Endangered. There were only 41 left in the world in 2021.{ref}Estimates range from 34 to 47. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Emslie, R.H. et al., 2019. "", ""spanType"": ""span-simple-text""}, {""url"": ""http://www.rhinoresourcecenter.com/pdf_files/156/1560170144.pdf"", ""children"": [{""text"": ""African and Asian rhinoceroses - status, conservation and trade"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart you can see how its population has changed over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Once found in both Malaysia and Indonesia, the Sumatran Rhino is thought to be extinct in Malaysia; none have been recorded in 2016 and 2018 surveys."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""But in contrast to the Javan, the number of Sumatran Rhinos in the world has been falling in recent decades, from 600 individuals in the mid-1980s to around one-tenth of that figure today. Both Malaysian and Indonesian rhino populations "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/sumatran-rhinos?time=2007..2018&country=IDN+MYS"", ""children"": [{""text"": ""have contributed"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to this loss."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/sumatran-rhinos"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/sumatran-rhinos?tab=map"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""How do we protect rhino populations?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The biggest threat to rhinos is poaching. Rhino horns are still seen as luxury goods and can sell for a lot of money in illegal wildlife markets."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But rhinos are not the only species at risk from poaching. It’s the leading threat for most large mammals. But some success stories – such as the restoration of Southern White and Indian Rhino populations – shows us that their demise is not inevitable. With the right approach we have the opportunity to turn things around. In a follow-up article we will look at the scale of global poaching, and what we can learn from the countries that have been successful in bringing it to an end."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""The state of the world's rhino populations"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""How have rhino populations changed over time? What species are at risk of extinction today?"", ""dateline"": ""November 30, 2022"", ""subtitle"": ""How have rhino populations changed over time? What species are at risk of extinction today?"", ""sidebar-toc"": false, ""featured-image"": ""Rhino-thumbnail.png""}, ""createdAt"": ""2022-11-30T10:37:36.000Z"", ""published"": false, ""updatedAt"": ""2023-01-30T12:36:59.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-11-30T10:38:19.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 53, ""numErrors"": 0, ""wpTagCounts"": {""html"": 12, ""column"": 14, ""columns"": 7, ""heading"": 8, ""paragraph"": 39}, ""htmlTagCounts"": {""p"": 39, ""h3"": 2, ""h4"": 6, ""div"": 21, ""iframe"": 12}}",2022-11-30 10:38:19,2024-03-06 14:59:49,1G89eli-ne-k-GFvfTQ0F5kHWtyx-90oCrTjwqJ8cNXQ,"[""Hannah Ritchie""]",How have rhino populations changed over time? What species are at risk of extinction today?,2022-11-30 10:37:36,2023-01-30 12:36:59,https://ourworldindata.org/wp-content/uploads/2022/11/Rhino-thumbnail.png,{},"The largest mammals are at the greatest risk of extinction. This has been the case since the arrival of humans, and is still true today. Weighing in at several tonnes, rhinos are some of the world’s largest mammals.{ref}The [average mass of rhinos](https://web.archive.org/web/20230130123018/https://rhinos.org/wp-content/uploads/2020/10/IRF_factsheet_digital_Mar2020-1.pdf) vary by species and sex. Male White and Indian rhinos can weigh up to 2700 kilograms; at the other end, Sumatran rhinos have a lower range of around 600 kilograms.{/ref} With extravagant prices for rhino horns and body parts, they’re also one of the most threatened: three of the five rhino species are ‘critically endangered’. In this article we take a look at the status of rhino populations today, and whether they are in decline or on the path to recovery. They offer us both stories of concern and reasons for optimism. ## State of rhino populations today There are five species of rhino. Africa is home to the White and Black Rhino. The remaining three – the Sumatran, Javan and Indian – exist in Asia. Some of these species have dangerously low population levels. The Sumatran and the Javan Rhino can now only be found in Indonesia. They are Critically Endangered with less than 100 individuals left in the wild. The Black Rhino, despite having a population in the thousands, is also Critically Endangered. It experienced a rapid decline over the 20th century as we will see later. The Indian (Greater One-Horned) offers us an important success story; it has seen a significant recovery in recent decades. The White Rhino too has shown an impressive recovery. But it has a dark side: the _Southern_ White Rhino might be doing well but the _Northern_ sub-species is on the brink of extinction. There are only two Northern White Rhinos left, and both of them are female. ## Are rhino populations increasing or decreasing? The health of these rhino populations is not just determined by how many animals are left today. The direction and pace that these populations are changing matters too. For each species I have built a time-series of populations globally and by country. This data is aggregated from multiple sources: the main one being the African and Asian Rhino Specialist Groups (AfrSG) and TRAFFIC, which collate statistics on all rhino populations, and submit them to the IUCN.{ref}Emslie, R.H. et al., 2019. [African and Asian rhinoceroses - status, conservation and trade](http://www.rhinoresourcecenter.com/pdf_files/156/1560170144.pdf). A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38. Other sources used to build these time-series include: Pusparini, W., Sievert, P. R., Fuller, T. K., Randhir, T. O., & Andayani, N. (2015). Rhinos in the parks: an island-wide survey of the last wild population of the Sumatran rhinoceros. PloS one, 10(9), e0136643. Emslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp. Thapa, K., Nepal, S., Thapa, G., Bhatta, S. R., & Wikramanayake, E. (2013). Past, present and future conservation of the greater one-horned rhinoceros Rhinoceros unicornis in Nepal. Oryx, 47(3), 345-351.{/ref} Let’s take a look at each of the five species one-by-one. ### White rhino (_Ceratotherium simum_) Overall, the story of the White rhino is a positive one. But this tale also has a darker side. There are two main subspecies: the Northern White and Southern White rhino. A century ago, the Northern was much more abundant than the Southern. Now the opposite is true. In the chart we see the population trend of the Southern White Rhino. It’s touted as one of the world’s greatest conservation success stories.{ref}Emslie, R. and Brooks, M. (1999) [African Rhino. Status Survey and Conservation Action Plan](https://portals.iucn.org/library/sites/library/files/documents/1999-049.pdf). IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.{/ref} We don’t have precise estimates but it’s reported that by the mid-19th century it was still abundant. However, intense poaching by the Europeans and killings in the conversion of land for agriculture meant that by the late 19th century it was close to extinction. By 1900 there were only 20 left. All were in the Hluhluwe–iMfolozi Park in South Africa – now a nature reserve. Over the course of the 20th century, severe protection of these species – particularly in African nature reserves – led to a significant and rapid increase. Populations were restored to more than 21,000. Numbers grew 1000-fold within a century. Over the last few years, [increases in poaching rates](https://ourworldindata.org/grapher/number-of-rhinos-poached) have unfortunately led to another decline. South Africa is home to around three-quarters of Southern White Rhinos. You can explore population estimates by country in the chart.The largest mammals are at the greatest risk of extinction. This has been the case since the arrival of humans, and is still true today.
\n\n\n\nWeighing in at several tonnes, rhinos are some of the world’s largest mammals.{ref}The average mass of rhinos vary by species and sex. Male White and Indian rhinos can weigh up to 2700 kilograms; at the other end, Sumatran rhinos have a lower range of around 600 kilograms.{/ref} With extravagant prices for rhino horns and body parts, they’re also one of the most threatened: three of the five rhino species are ‘critically endangered’.
\n\n\n\nIn this article we take a look at the status of rhino populations today, and whether they are in decline or on the path to recovery. They offer us both stories of concern and reasons for optimism.
\n\n\n\nThere are five species of rhino. Africa is home to the White and Black Rhino. The remaining three – the Sumatran, Javan and Indian – exist in Asia.
\n\n\n\nSome of these species have dangerously low population levels.
\n\n\n\nThe Sumatran and the Javan Rhino can now only be found in Indonesia. They are Critically Endangered with less than 100 individuals left in the wild. The Black Rhino, despite having a population in the thousands, is also Critically Endangered. It experienced a rapid decline over the 20th century as we will see later.
\n\n\n\nThe Indian (Greater One-Horned) offers us an important success story; it has seen a significant recovery in recent decades. The White Rhino too has shown an impressive recovery. But it has a dark side: the Southern White Rhino might be doing well but the Northern sub-species is on the brink of extinction. There are only two Northern White Rhinos left, and both of them are female.
\n\n\n\nThe health of these rhino populations is not just determined by how many animals are left today. The direction and pace that these populations are changing matters too.
\n\n\n\nFor each species I have built a time-series of populations globally and by country. This data is aggregated from multiple sources: the main one being the African and Asian Rhino Specialist Groups (AfrSG) and TRAFFIC, which collate statistics on all rhino populations, and submit them to the IUCN.{ref}Emslie, R.H. et al., 2019. African and Asian rhinoceroses – status, conservation and trade. A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.
\n\n\n\nOther sources used to build these time-series include:
\n\n\n\nPusparini, W., Sievert, P. R., Fuller, T. K., Randhir, T. O., & Andayani, N. (2015). Rhinos in the parks: an island-wide survey of the last wild population of the Sumatran rhinoceros. PloS one, 10(9), e0136643.
\n\n\n\nEmslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino
\n\n\n\nSpecialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.
\n\n\n\nThapa, K., Nepal, S., Thapa, G., Bhatta, S. R., & Wikramanayake, E. (2013). Past, present and future conservation of the greater one-horned rhinoceros Rhinoceros unicornis in Nepal. Oryx, 47(3), 345-351.{/ref}
\n\n\n\nLet’s take a look at each of the five species one-by-one.
\n\n\n\nOverall, the story of the White rhino is a positive one. But this tale also has a darker side. There are two main subspecies: the Northern White and Southern White rhino. A century ago, the Northern was much more abundant than the Southern. Now the opposite is true.
\n\n\n\nIn the chart we see the population trend of the Southern White Rhino. It’s touted as one of the world’s greatest conservation success stories.{ref}Emslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.{/ref} We don’t have precise estimates but it’s reported that by the mid-19th century it was still abundant. However, intense poaching by the Europeans and killings in the conversion of land for agriculture meant that by the late 19th century it was close to extinction. By 1900 there were only 20 left. All were in the Hluhluwe–iMfolozi Park in South Africa – now a nature reserve.
\n\n\n\nOver the course of the 20th century, severe protection of these species – particularly in African nature reserves – led to a significant and rapid increase. Populations were restored to more than 21,000. Numbers grew 1000-fold within a century.
\n\n\n\nOver the last few years, increases in poaching rates have unfortunately led to another decline. South Africa is home to around three-quarters of Southern White Rhinos. You can explore population estimates by country in the chart.
\nIf the story of the Southern White Rhino is one of our greatest successes, the story of its Northern cousin must be one of our biggest failures. The Northern White Rhino is on the brink of extinction. There are only two individuals left. Both are female.
\n\n\n\nIn the chart we see its demise over the second-half of the 20th century. In 1960, it’s estimated there were more than 2,000, predominantly in Sudan and the Democratic Republic of Congo. Intense poaching, and challenges for protection during civil unrest in stronghold countries, has led to a rapid decline.
\n\n\n\nSudan, the last remaining male died in Kenya in 2018. There are now only two female rhinos left: Najin and her daughter, Fatu. Both are guarded in a semi-wild enclosure, and have had their horns sawed off to deter poachers. Scientists are investigating ways to continue reproduction from the last females, including stem cell treatment and hybrid embryos from Northern White Rhino eggs and Southern White Rhino sperm.{ref}Saragusty, J., Diecke, S., Drukker, M., Durrant, B., Friedrich Ben‐Nun, I., Galli, C., … & Johnson, S. (2016). Rewinding the process of mammalian extinction. Zoo Biology, 35(4), 280-292.
Callaway, E. (2016). Stem-cell plan aims to bring rhino back from brink of extinction. Nature News, 533(7601), 20.{/ref}
In 2021, researchers aim to implant embryos in both to try to get them to reproduce. Scientific innovation is now the only way to save these sub-species. Even if reproduction is successful, population numbers will be incredibly low for a long time. They will have to be closely guarded for a long time.
\n\n\n\nThe story of the Black Rhino is similar to that of the Northern White. The once abundant species has seen a dramatic decline over the 20th century.
\n\n\n\nEven after intense poaching by European settlers over the 19th and early 20th century, in 1960 there were still around 100,000 Black rhinos in Africa.{ref}Emslie, R. and Brooks, M. (1999) African Rhino. Status Survey and Conservation Action Plan. IUCN/SSC African Rhino Specialist Group. IUCN, Gland, Switzerland and Cambridge, UK. ix + 92 pp.{/ref} The decline which followed was rapid and dramatic: the number fell by more than 80% in only two decades. The population reached its low point in the early 1990s at around 2,500, and have since began to recover. The number of Black Rhinos has more than doubled to around 6,000. Nonetheless, continued poaching has still limited population growth in recent decades. The Black Rhino is still classified as Critically Endangered.{ref}Emslie, R.H. et al., 2019. African and Asian rhinoceroses – status, conservation and trade. A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.{/ref}
\n\n\n\nThe Indian (also called the Greater One-Horned) Rhino provides one success story in the restoration of wild mammals.
\n\n\n\nBy the mid-20th century, there were very few Indian Rhinos left in the world. It’s estimated (with larger uncertainty since it predates regular surveying) that there were only 40 individuals left.
\n\n\n\nSince the mid-1960s, populations have increased nearly 100-fold. Latest estimates, taken in 2021, suggest there are now over 4,000 rhinos in the wild. This was the result of impressive conservation efforts to reduce poaching in both India and Nepal.
\n\n\n\nUnsurprising based on the name, India is home to more than 80% of the species. But Nepal also has a steady population. With only a few being spotted in recent decades, they are thought to now be extinct in Pakistan.
\n\n\n\nThe Javan Rhino is one of the world’s most endangered mammals. It is estimated there were only 76 rhinos left in 2021. This makes it Critically Endangered.
\n\n\n\nIn the chart you see how Javan Rhino populations have changed over time.
\n\n\n\nIn recent decades it has existed in two countries: Indonesia and Vietnam. But by 2010 it had gone extinct in Vietnam. Indonesia is now its only remaining home.
\n\n\n\nIts total population has, however, increased from 50 years ago. In the 1960s there were only 20 to 30 Javan Rhinos left in the world. From then until the 1990s, the population approximately doubled. The latest estimate puts this figure at 76 Javan Rhinos.
\n\n\n\nLike the Javan species, Sumatran rhinos are also one of the most endangered mammals. They’re Critically Endangered. There were only 41 left in the world in 2021.{ref}Estimates range from 34 to 47.
\n\n\n\nEmslie, R.H. et al., 2019. African and Asian rhinoceroses – status, conservation and trade. A report from the IUCN Species Survival Commission (IUCN/SSC) African and Asian Rhino Specialist Groups and TRAFFIC to the CITES Secretariat pursuant to Resolution Conf. 9.14 (Rev. CoP17). Report to CITES 17th meeting (Colombo, June 2019), CoP 18 Doc.83.1 annex 3: 1-38.{/ref}
\n\n\n\nIn the chart you can see how its population has changed over time.
\n\n\n\nOnce found in both Malaysia and Indonesia, the Sumatran Rhino is thought to be extinct in Malaysia; none have been recorded in 2016 and 2018 surveys.
But in contrast to the Javan, the number of Sumatran Rhinos in the world has been falling in recent decades, from 600 individuals in the mid-1980s to around one-tenth of that figure today. Both Malaysian and Indonesian rhino populations have contributed to this loss.
The biggest threat to rhinos is poaching. Rhino horns are still seen as luxury goods and can sell for a lot of money in illegal wildlife markets.
\n\n\n\nBut rhinos are not the only species at risk from poaching. It’s the leading threat for most large mammals. But some success stories – such as the restoration of Southern White and Indian Rhino populations – shows us that their demise is not inevitable. With the right approach we have the opportunity to turn things around. In a follow-up article we will look at the scale of global poaching, and what we can learn from the countries that have been successful in bringing it to an end.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""How have rhino populations changed over time? What species are at risk of extinction today?"", ""protected"": false}, ""date_gmt"": ""2022-11-30T10:38:19"", ""modified"": ""2023-01-30T12:36:59"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2023-01-30T12:36:59"", ""comment_status"": ""closed"", ""featured_media"": 54669, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/11/Rhino-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2022/11/Rhino-thumbnail-768x402.png""}}" 54596,We just published a new topic page on Artificial Intelligence,artificial-intelligence-launch,post,publish,,"{""id"": ""wp-54596"", ""slug"": ""artificial-intelligence-launch"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We just published a new topic page on Artificial Intelligence"", ""authors"": [""Charlie Giattino"", ""Edouard Mathieu"", ""Julia Broden"", ""Max Roser""], ""excerpt"": ""AI is already having a large impact on our world. Explore research and data to understand the trajectory of this important technology."", ""dateline"": ""December 6, 2022"", ""subtitle"": ""AI is already having a large impact on our world. Explore research and data to understand the trajectory of this important technology."", ""sidebar-toc"": false, ""featured-image"": ""artificial-intelligence-featured-image.png""}, ""createdAt"": ""2022-11-28T11:25:03.000Z"", ""published"": false, ""updatedAt"": ""2022-12-06T11:34:03.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-12-06T00:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2022-12-06 00:00:00,2024-02-16 14:22:54,,"[""Charlie Giattino"", ""Edouard Mathieu"", ""Julia Broden"", ""Max Roser""]",AI is already having a large impact on our world. Explore research and data to understand the trajectory of this important technology.,2022-11-28 11:25:03,2022-12-06 11:34:03,https://ourworldindata.org/wp-content/uploads/2022/11/artificial-intelligence-featured-image.png,{},,"{""id"": 54596, ""date"": ""2022-12-06T00:00:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=54596""}, ""link"": ""https://owid.cloud/artificial-intelligence-launch"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""artificial-intelligence-launch"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We just published a new topic page on Artificial Intelligence""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54596""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/44"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=54596"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=54596"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=54596"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=54596""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54596/revisions"", ""count"": 4}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/54594"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 54725, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54596/revisions/54725""}]}, ""author"": 44, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""AI is already having a large impact on our world. Explore research and data to understand the trajectory of this important technology."", ""protected"": false}, ""date_gmt"": ""2022-12-06T00:00:00"", ""modified"": ""2022-12-06T11:34:03"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Charlie Giattino"", ""Edouard Mathieu"", ""Julia Broden"", ""Max Roser""], ""modified_gmt"": ""2022-12-06T11:34:03"", ""comment_status"": ""closed"", ""featured_media"": 54594, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/11/artificial-intelligence-featured-image-150x79.png"", ""medium_large"": ""/app/uploads/2022/11/artificial-intelligence-featured-image-768x403.png""}}" 54575,Artificial Intelligence,artificial-intelligence,page,publish,"Artificial intelligence (AI) systems already greatly impact our lives — they increasingly shape what we see, believe, and do. Based on the steady advances in AI technology and the large recent increases in investment, we should expect AI technology to become even more powerful and impactful in the following years and decades.
It is easy to underestimate how much the world can change within a lifetime, so it is worth taking seriously what those who work on AI expect for the future. Many AI experts believe there is a real chance that human-level artificial intelligence will be developed within the following decades, and some think it will exist much sooner.
How such powerful AI systems are built and used will be very important for the future of our world and our own lives. All technologies have positive and negative consequences, but with AI, the range of these consequences is extraordinarily large: the technology has immense potential for good. Still, it comes with significant downsides and high risks.
A technology that has such an enormous impact needs to be of central interest to people across our entire society. But currently, the question of how this technology will get developed and used is left to a small group of entrepreneurs and engineers.
With our publications on artificial intelligence, we want to help change this status quo and support a broader societal engagement.
On this page, you will find articles and an overview of AI-related metrics that let you monitor what is happening and where we might be heading. We hope that this work will be helpful for the growing and necessary public conversation on AI.
Related topics
How AI gets built is currently decided by a small group of technologists. As this technology is transforming our lives, it should be in all of our interest to become informed and engaged.
Max Roser
Despite their brief history, computers and AI have fundamentally changed what we see, what we know, and what we do. Little is as important for the future of the world, and our own lives, as how this history continues.
Max Roser
Artificial intelligence (AI) systems already greatly impact our lives — they increasingly shape what we see, believe, and do. Based on the steady advances in AI technology and the large recent increases in investment, we should expect AI technology to become even more powerful and impactful in the following years and decades.
\n\n\n\nIt is easy to underestimate how much the world can change within a lifetime, so it is worth taking seriously what those who work on AI expect for the future. Many AI experts believe there is a real chance that human-level artificial intelligence will be developed within the following decades, and some think it will exist much sooner.
\n\n\n\nHow such powerful AI systems are built and used will be very important for the future of our world and our own lives. All technologies have positive and negative consequences, but with AI, the range of these consequences is extraordinarily large: the technology has immense potential for good. Still, it comes with significant downsides and high risks.
\n\n\n\nA technology that has such an enormous impact needs to be of central interest to people across our entire society. But currently, the question of how this technology will get developed and used is left to a small group of entrepreneurs and engineers.
\n\n\n\nWith our publications on artificial intelligence, we want to help change this status quo and support a broader societal engagement.
\n\n\n\nOn this page, you will find articles and an overview of AI-related metrics that let you monitor what is happening and where we might be heading. We hope that this work will be helpful for the growing and necessary public conversation on AI.
\nRelated topics
\n\n\n\n\nMax Roser
\nMax Roser
\nThe Living Planet Index (LPI) is a measure of biodiversity that is often misinterpreted. In my accompanying article on the Living Planet Index, I describe these misinterpretations in detail, and what it actually tells us about the state of global biodiversity.
In this short page I answer some commonly-asked questions about this measure of biodiversity.
The Living Planet Index (LPI) provides a measure of wildlife abundance. It measures the average decline in population size since 1970 across a wide range of species.{ref}WWF (2022) Living Planet Report 2022 – Building a nature-positive society. Almond, R.E.A., Grooten, M., Juffe Bignoli, D. & Petersen, T. (Eds). WWF, Gland, Switzerland.{/ref}
The Living Planet Index does not measure:
The LPI only includes data from vertebrate populations. This includes mammals, birds, reptiles, amphibians, and fish. Taxonomic groups including insects, corals, fungi, and plants are not included.
In its latest report, published in 2022, 31,821 populations across 5,230 species were included. It includes species and populations across all continents. This breakdown is given in the table.
| Region | Number of species included |
| Global | 5,230 |
| Africa | 510 |
| North America | 952 |
| Latin America & Caribbean | 1,261 |
| Asia-Pacific | 729 |
| Europe and Central Asia | 627 |
Only vertebrate species are included in the LPI: this includes mammals, birds, reptiles, amphibians and fish. Taxonomic groups including insects, corals, fungi and plants are not included.
Only a small percentage of known species in these groups are included. The number of populations and species from each group is shown in the table. This also details the percentage of known species that are included.
| Taxonomic group | Number of populations included | Number of species included | Percentage of known species that are included |
| Birds | 12,995 | 1,802 | 16% |
| Mammals | 6,171 | 751 | 11% |
| Fish | 11,282 | 2,116 | 6% |
| Amphibians and Reptiles | 1,373 | 561 | 3% |
The underlying data for the LPI comes from a combination of published scientific articles, online databases and government reports. To be included, data points must contain a time series of vertebrate populations spanning any number of years from 1970 onwards.
The latest results from the LPI indicate an average decline in the studied wildlife populations of 69% between 1970 and 2018.
Note that this does not mean that we have lost 69% of wildlife over this period. For a clear example of why this is the wrong conclusion, and how the LPI is calculated, see our example here.
No, the results of the LPI show that around half of the studied populations are increasing in abundance, while half are declining. The fact that there is such a large average decline suggests that the magnitude of the decline across many populations is much larger than magnitude for increasing populations.
","{""id"": ""wp-54367"", ""slug"": ""faq-living-planet-index"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""The Living Planet Index (LPI) is a measure of biodiversity that is often misinterpreted. In my "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/living-planet-index-decline"", ""children"": [{""text"": ""accompanying article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on the Living Planet Index, I describe these misinterpretations in detail, and what it actually tells us about the state of global biodiversity."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this short page I answer some commonly-asked questions about this measure of biodiversity."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What does the Living Planet Index (LPI) measure?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Living Planet Index (LPI) provides a measure of wildlife "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""abundance"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". It measures the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""average"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" decline in population size since 1970 across a wide range of species.{ref}WWF (2022) "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Living Planet Report 2022 – Building a nature-positive society"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". Almond, R.E.A., Grooten, M., Juffe Bignoli, D. & Petersen, T. (Eds). WWF, Gland, Switzerland.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""What does the Living Planet Index (LPI) "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""not"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" measure?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Living Planet Index does not measure:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""text"": ""Number of species lost"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Number of populations or individuals that have been lost"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Number or percentage of species or populations that are declining"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Number of extinctions"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""What types of species are included?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The LPI only includes data from vertebrate populations. This includes mammals, birds, reptiles, amphibians, and fish. Taxonomic groups including insects, corals, fungi, and plants are not included."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How many species does it cover? What is the geographical range of this coverage?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In its latest report, published in 2022, 31,821 populations across 5,230 species were included. It includes species and populations across all continents. This breakdown is given in the table."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""html"", ""value"": ""| Region | Number of species included |
| Global | 5,230 |
| Africa | 510 |
| North America | 952 |
| Latin America & Caribbean | 1,261 |
| Asia-Pacific | 729 |
| Europe and Central Asia | 627 |
| Taxonomic group | Number of populations included | Number of species included | Percentage of known species that are included |
| Birds | 12,995 | 1,802 | 16% |
| Mammals | 6,171 | 751 | 11% |
| Fish | 11,282 | 2,116 | 6% |
| Amphibians and Reptiles | 1,373 | 561 | 3% |
| Region | Number of species included |
| Global | 5,230 |
| Africa | 510 |
| North America | 952 |
| Latin America & Caribbean | 1,261 |
| Asia-Pacific | 729 |
| Europe and Central Asia | 627 |
| Taxonomic group | Number of populations included | Number of species included | Percentage of known species that are included |
| Birds | 12,995 | 1,802 | 16% |
| Mammals | 6,171 | 751 | 11% |
| Fish | 11,282 | 2,116 | 6% |
| Amphibians and Reptiles | 1,373 | 561 | 3% |
The Living Planet Index (LPI) is a measure of biodiversity that is often misinterpreted. In my accompanying article on the Living Planet Index, I describe these misinterpretations in detail, and what it actually tells us about the state of global biodiversity.
\n\n\n\nIn this short page I answer some commonly-asked questions about this measure of biodiversity.
\n\n\n\nThe Living Planet Index (LPI) provides a measure of wildlife abundance. It measures the average decline in population size since 1970 across a wide range of species.{ref}WWF (2022) Living Planet Report 2022 – Building a nature-positive society. Almond, R.E.A., Grooten, M., Juffe Bignoli, D. & Petersen, T. (Eds). WWF, Gland, Switzerland.{/ref}
\n\n\n\nThe Living Planet Index does not measure:
\n\n\n\nThe LPI only includes data from vertebrate populations. This includes mammals, birds, reptiles, amphibians, and fish. Taxonomic groups including insects, corals, fungi, and plants are not included.
\n\n\n\nIn its latest report, published in 2022, 31,821 populations across 5,230 species were included. It includes species and populations across all continents. This breakdown is given in the table.
\n\n\n\n| Region | Number of species included |
| Global | 5,230 |
| Africa | 510 |
| North America | 952 |
| Latin America & Caribbean | 1,261 |
| Asia-Pacific | 729 |
| Europe and Central Asia | 627 |
Only vertebrate species are included in the LPI: this includes mammals, birds, reptiles, amphibians and fish. Taxonomic groups including insects, corals, fungi and plants are not included.
\n\n\n\nOnly a small percentage of known species in these groups are included. The number of populations and species from each group is shown in the table. This also details the percentage of known species that are included.
\n\n\n\n| Taxonomic group | Number of populations included | Number of species included | Percentage of known species that are included |
| Birds | 12,995 | 1,802 | 16% |
| Mammals | 6,171 | 751 | 11% |
| Fish | 11,282 | 2,116 | 6% |
| Amphibians and Reptiles | 1,373 | 561 | 3% |
The underlying data for the LPI comes from a combination of published scientific articles, online databases and government reports. To be included, data points must contain a time series of vertebrate populations spanning any number of years from 1970 onwards.
\n\n\n\nThe latest results from the LPI indicate an average decline in the studied wildlife populations of 69% between 1970 and 2018.
Note that this does not mean that we have lost 69% of wildlife over this period. For a clear example of why this is the wrong conclusion, and how the LPI is calculated, see our example here.
No, the results of the LPI show that around half of the studied populations are increasing in abundance, while half are declining. The fact that there is such a large average decline suggests that the magnitude of the decline across many populations is much larger than magnitude for increasing populations.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""The Living Planet Index is one of the most common measures used in biodiversity monitoring. But what is it, and where does this data come from?"", ""protected"": false}, ""date_gmt"": ""2022-10-13T12:59:00"", ""modified"": ""2023-06-15T10:45:46"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2023-06-15T09:45:46"", ""comment_status"": ""closed"", ""featured_media"": 42160, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2021/03/Living-Planet-Index-thumbnail-150x79.png"", ""medium_large"": ""/app/uploads/2021/03/Living-Planet-Index-thumbnail-768x404.png""}}" 54361,How does the Living Planet Index vary by region?,living-planet-index-region,post,publish,"The Living Planet Index (LPI) is a measure of biodiversity that is often misinterpreted. In my accompanying article on the Living Planet Index, I describe these misinterpretations in detail, and what it actually tells us about the state of global biodiversity.
In brief, the LPI measures the average change in the size of studied animal populations. It does not tell us the total change in animal populations; the share of populations that have been lost; or anything about extinctions.
The main headline result of the 2022 Living Planet Index report was that studied animal populations globally have seen an average decline of 69% between 1970 and 2018.{ref}WWF (2022) Living Planet Report 2022 – Building a nature-positive society. Almond, R.E.A., Grooten, M., Juffe Bignoli, D. & Petersen, T. (Eds). WWF, Gland, Switzerland.{/ref}
But these results vary a lot from region to region.
In the chart we see the LPI given by region. This is measured relative to 1970, which is assigned a value of one.
Latin America and the Caribbean
Latin America and the Caribbean have seen the most severe decline of any region. It experienced an average decline of 94% across its studied populations. When we think about the changes in some of the leading drivers of biodiversity loss, we should not be surprised that Latin America has been badly affected.
In recent decades it has experienced intense deforestation and expansion of agricultural land: the primary driver of habitat loss. It is also a biodiversity hotspot, being home to many endemic (unique) tropical species. These species are often highly specialized, and don’t adapt well to changes in local conditions.
The LPI highlights the conversion of grasslands, forests, wetlands and the harvesting of species for hunting and poaching as the main contributors to this decline. It was most severe for fish, reptiles and amphibians.
North America
The trend for North America is one of two halves. Since 1970, it has experienced an average decline of 20%. But, most of this occurred before the year 2000. After a steady decline through the 1970s, 80s and 90s, the trend seems to have stabilized from the turn of the millennium. In fact, the average trend might be slightly increasing.
Europe and Central Asia
Europe has seen the smallest decline of all of the continents. An average decline of 18% since 1970. This paints a less severe picture of biodiversity in Europe and Central Asia, and is partly attributed to successful conservation efforts. For example, most countries in Europe are afforesting and restoring wild ecosystems.
It’s also true that most of the continent’s transformation – the expansion of agriculture, deforestation, and destruction of habitats – occurred well before 1970. It is a history that goes back centuries. In other words, it’s biodiversity was already in a significantly depleted state. We’d expect that any declines in the region would be less severe.
Africa
Africa is still rich in biodiversity. It’s the only continent with a significant number of large mammals left. Unfortunately, it has dramatically declined in recent decades—an average decline of two-thirds (66%) since 1970.
Like Latin America, the main drivers of this are habitat loss for expanding agriculture and the overexploitation of animals for hunting, poaching, and wildlife trade. Looking at initial trends in the continent’s subregions, the decline has been most severe in West, Central, and East Africa. Populations have been more stable in the North and South.
Asia-Pacific
The Asia-Pacific spans a wide range of countries and habitats. It includes large terrestrial landscapes but also many small island states. This means it’s home to many endemic species and unique ecosystems.
Since 1970, it has experienced an average decline of 55%. But there are initial signs that this is changing. We’ve seen a positive trend since 2010. This has been seen clearly in population trends for several species of reptiles and amphibians.
There is one particular environment that has experienced an incredibly severe decline: the world’s freshwaters.
Almost one-third of freshwater species are threatened with extinction.{ref}Collen, B., Whitton, F., Dyer, E. E., Baillie, J. E., Cumberlidge, N., Darwall, W. R., ... & Böhm, M. (2014). Global patterns of freshwater species diversity, threat and endemism. Global Ecology and Biogeography, 23(1), 40-51.{/ref} Freshwater species are at higher risk of extinction compared to terrestrial species. This bleak reality is also reflected in trends within the LPI.
The LPI covers 3,741 freshwater populations across 944 species. Not a small, limited sample. Since 1970, the average decline has been 83%. That’s a 3.6% decline every year. Most of this decline has come from reptiles, amphibians, and fishes, particularly across Latin America and the Caribbean.
Overfishing and overharvesting of freshwater amphibians and reptiles – which are often traded in markets for their body parts, or for medicinal uses – is one of the biggest threats to these species.
Freshwaters are also polluted by many different types of waste: agricultural waste from manure and fertilizers, industrial discharge, and domestic waste. This can very quickly tip the balance of species within freshwaters. The diversion of rivers and use of dams can also alter the habitats for these species.
Conservation efforts in these environments are also more difficult because it usually requires input from multiple sectors. The allocation of responsibility for water is not as straightforward as for land. Just like terrestrial mammals, it’s the megafauna (the largest species) that are at greatest risk.
Freshwater megafauna are species that grow to more than 30 kilograms and include animals such as Mekong giant catfish, river dolphins, otters, beavers and hippos. All of these are targeted by humans. What makes them especially vulnerable is that they reproduce at much slower rates and have less offspring.{ref}Cardillo, M., Mace, G. M., Jones, K. E., Bielby, J., Bininda-Emonds, O. R., Sechrest, W., ... & Purvis, A. (2005). Multiple causes of high extinction risk in large mammal species. Science, 309(5738), 1239-1241.{/ref} This makes it more difficult for them to restore their populations that have been depleted.
The Living Planet Index (LPI) is a measure of biodiversity that is often misinterpreted. In my accompanying article on the Living Planet Index, I describe these misinterpretations in detail, and what it actually tells us about the state of global biodiversity.
\n\n\n\nIn brief, the LPI measures the average change in the size of studied animal populations. It does not tell us the total change in animal populations; the share of populations that have been lost; or anything about extinctions.
\n\n\n\nThe main headline result of the 2022 Living Planet Index report was that studied animal populations globally have seen an average decline of 69% between 1970 and 2018.{ref}WWF (2022) Living Planet Report 2022 – Building a nature-positive society. Almond, R.E.A., Grooten, M., Juffe Bignoli, D. & Petersen, T. (Eds). WWF, Gland, Switzerland.{/ref}
\n\n\n\nBut these results vary a lot from region to region.
\n\n\n\nIn the chart we see the LPI given by region. This is measured relative to 1970, which is assigned a value of one.
\n\n\n\nLatin America and the Caribbean
\n\n\n\nLatin America and the Caribbean have seen the most severe decline of any region. It experienced an average decline of 94% across its studied populations. When we think about the changes in some of the leading drivers of biodiversity loss, we should not be surprised that Latin America has been badly affected.
\n\n\n\nIn recent decades it has experienced intense deforestation and expansion of agricultural land: the primary driver of habitat loss. It is also a biodiversity hotspot, being home to many endemic (unique) tropical species. These species are often highly specialized, and don’t adapt well to changes in local conditions.
\n\n\n\nThe LPI highlights the conversion of grasslands, forests, wetlands and the harvesting of species for hunting and poaching as the main contributors to this decline. It was most severe for fish, reptiles and amphibians.
\n\n\n\nNorth America
\n\n\n\nThe trend for North America is one of two halves. Since 1970, it has experienced an average decline of 20%. But, most of this occurred before the year 2000. After a steady decline through the 1970s, 80s and 90s, the trend seems to have stabilized from the turn of the millennium. In fact, the average trend might be slightly increasing.
\n\n\n\nEurope and Central Asia
\n\n\n\nEurope has seen the smallest decline of all of the continents. An average decline of 18% since 1970. This paints a less severe picture of biodiversity in Europe and Central Asia, and is partly attributed to successful conservation efforts. For example, most countries in Europe are afforesting and restoring wild ecosystems.
\n\n\n\nIt’s also true that most of the continent’s transformation – the expansion of agriculture, deforestation, and destruction of habitats – occurred well before 1970. It is a history that goes back centuries. In other words, it’s biodiversity was already in a significantly depleted state. We’d expect that any declines in the region would be less severe.
\n\n\n\nAfrica
\n\n\n\nAfrica is still rich in biodiversity. It’s the only continent with a significant number of large mammals left. Unfortunately, it has dramatically declined in recent decades—an average decline of two-thirds (66%) since 1970.
\n\n\n\nLike Latin America, the main drivers of this are habitat loss for expanding agriculture and the overexploitation of animals for hunting, poaching, and wildlife trade. Looking at initial trends in the continent’s subregions, the decline has been most severe in West, Central, and East Africa. Populations have been more stable in the North and South.
\n\n\n\nAsia-Pacific
\n\n\n\nThe Asia-Pacific spans a wide range of countries and habitats. It includes large terrestrial landscapes but also many small island states. This means it’s home to many endemic species and unique ecosystems.
\n\n\n\nSince 1970, it has experienced an average decline of 55%. But there are initial signs that this is changing. We’ve seen a positive trend since 2010. This has been seen clearly in population trends for several species of reptiles and amphibians.
\nThere is one particular environment that has experienced an incredibly severe decline: the world’s freshwaters.
\n\n\n\nAlmost one-third of freshwater species are threatened with extinction.{ref}Collen, B., Whitton, F., Dyer, E. E., Baillie, J. E., Cumberlidge, N., Darwall, W. R., … & Böhm, M. (2014). Global patterns of freshwater species diversity, threat and endemism. Global Ecology and Biogeography, 23(1), 40-51.{/ref} Freshwater species are at higher risk of extinction compared to terrestrial species. This bleak reality is also reflected in trends within the LPI.
\n\n\n\nThe LPI covers 3,741 freshwater populations across 944 species. Not a small, limited sample. Since 1970, the average decline has been 83%. That’s a 3.6% decline every year. Most of this decline has come from reptiles, amphibians, and fishes, particularly across Latin America and the Caribbean.
\n\n\n\nOverfishing and overharvesting of freshwater amphibians and reptiles – which are often traded in markets for their body parts, or for medicinal uses – is one of the biggest threats to these species.
\n\n\n\nFreshwaters are also polluted by many different types of waste: agricultural waste from manure and fertilizers, industrial discharge, and domestic waste. This can very quickly tip the balance of species within freshwaters. The diversion of rivers and use of dams can also alter the habitats for these species.
\n\n\n\nConservation efforts in these environments are also more difficult because it usually requires input from multiple sectors. The allocation of responsibility for water is not as straightforward as for land. Just like terrestrial mammals, it’s the megafauna (the largest species) that are at greatest risk.
\n\n\n\nFreshwater megafauna are species that grow to more than 30 kilograms and include animals such as Mekong giant catfish, river dolphins, otters, beavers and hippos. All of these are targeted by humans. What makes them especially vulnerable is that they reproduce at much slower rates and have less offspring.{ref}Cardillo, M., Mace, G. M., Jones, K. E., Bielby, J., Bininda-Emonds, O. R., Sechrest, W., … & Purvis, A. (2005). Multiple causes of high extinction risk in large mammal species. Science, 309(5738), 1239-1241.{/ref} This makes it more difficult for them to restore their populations that have been depleted.
\nWe source the data for the Global Health Explorer from three main sources:
The Global Health Estimates are primarily calculated using cause-of-death statistics that are reported to the WHO by individual countries.
These vital registration (VR) statistics are submitted to the WHO Mortality Database on an annual basis by country, year, cause, age and sex. This data is included in the Global Health Estimates if it meets criteria assessing completeness and quality. Since many countries don't meet these criteria, the GHE does not incorporate VR statistics for every country.
There are a number of specialist WHO groups and UN agencies that collect topic or disease-specific data. The dataset on HIV and AIDS – collected and published by UN AIDS – is one example of this. The estimates based on the VR data are compared to the data from specialist WHO groups and UN agencies and adjustments are made if necessary.
Where the VR data is not usable and there is no other nationally representative cause of death data, then the World Health Organization adopts the IHME’s Global Burden of Disease data to fill these gaps. Full details on how the WHO’s Global Health Estimates are calculated are available here.
The WHO’s Global Health Observatory brings together a large number of variables produced by the WHO and specialist UN agencies.
The variables found within the Global Health Observatory are limited to the leading causes of death and injury. This means the WHO GHO is less ‘complete’ than the IHME Global Burden of Disease study and WHO GHE in terms of the diseases included.
The method used to produce each of these variables is different and tailored to the specific cause of death or injury. This is a different approach to both the Global Burden of Disease (IHME) and Global Health Estimates (WHO) which attempt to use a consistent modeling approach for all causes of death and injuries. Additionally, variables in the Global Health Observatory are not consistently disaggregated by age or sex. Instead, they show what data is available and is most relevant for the given variable.
The full list of variables available on the Global Health Observatory are available here, each has their own associated metadata and method.
In the Global Burden of Disease study, the IHME uses a wide range of input data. This includes, but is not limited to census data, household surveys, civil registration and vital statistics, disease registries, health service use, air pollution monitors, satellite imaging, disease notifications, and other sources. These data are available through scientific journals, reports, online databases, books, news reports, and other resources. For most diseases and injuries the data are used as input into a series of standardized models which are used to generate estimates of each disease or injury for all age-groups, sexes, locations, and years.
Since the publication of its Global Burden of Disease study in 2017, the IHME has used its own population estimates. These differ from those used by the UN Population Division, which are used by the World Health Organization (WHO). This key difference propagates through all of the resulting health datasets. Even if the IHME and WHO assumed the same rates of health and mortality burden, this difference in population estimates means many of their population-adjusted figures would be different.
Full details on how the IHME’s Global Burden of Disease is calculated are available here and here.
Collecting precise data on global health is difficult: we can never know exactly which diseases or injuries affect people across the world at any given point in time. In the absence of perfect data, health researchers have a number of ways by which they try to estimate the burden of global health outcomes. These estimates are what we present in the Global Health Explorer.
Each of these sources have slightly different methods for estimating the burden of disease and causes of death, and in some cases the definition of the diseases and injuries in question differ slightly as our Explorer shows.
The diseases and injuries presented in the Global Health Explorer are typically aggregates of multiple similar causes of death. For example, ‘Deaths from falls’ includes 20 different types of fall which are combined into this one variable.
Individual causes of death have an associated International Classification of Disease (ICD) code. The tenth version of the ICD codes are used in the most recent Global Health Estimates and Global Burden of Disease and this is referred to as ICD-10. For example ‘Fall from tree’ has the ICD-10 code ‘W14’, and the ICD-10 codes for all ‘Deaths from falls’ are ‘W00-W19’.
Both IHME and WHO follow the same definition for ‘Deaths from falls’ but for some other causes of death, their definitions of the same aggregate cause of death doesn’t include exactly the same ICD-10 codes. For example, ICD-10 code ‘F02.3 - A dementia developing in the course of established Parkinson disease’ is included as ‘Dementia’ in the WHO’s Global Health Estimates but as ‘Parkinson’s disease’ in the IHME’s Global Burden of Disease.
These slight differences in the definition of diseases may also contribute to the different values we see when looking at the same metric from different sources. The ICD-10 codes used by IHME to define causes of death can be found here and the ICD-10 codes used by the WHO can be found here.
","{""id"": ""wp-54262"", ""slug"": ""global-health-explorer-faq"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""Data sources and methodological differences"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We source the data for the Global Health Explorer from three main sources: "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""text"": ""World Health Organization’s Global Health Estimates"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""World Health Organization’s Global Health Observatory"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Institute of Health Metrics and Evaluation’s Global Burden of Disease Study"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Global Health Estimates (WHO GHE)"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Global Health Estimates are primarily calculated using cause-of-death statistics that are reported to the WHO by individual countries. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These vital registration (VR) statistics are submitted to the WHO Mortality Database on an annual basis by country, year, cause, age and sex. This data is included in the Global Health Estimates if it meets criteria assessing completeness and quality. Since many countries don't meet these criteria, the GHE does not incorporate VR statistics for every country."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are a number of specialist WHO groups and UN agencies that collect topic or disease-specific data. The dataset on HIV and AIDS – collected and published by UN AIDS – is one example of this. The estimates based on the VR data are compared to the data from specialist WHO groups and UN agencies and adjustments are made if necessary. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Where the VR data is not usable and there is no other nationally representative cause of death data, then the World Health Organization adopts the IHME’s Global Burden of Disease data to fill these gaps. Full details on how the WHO’s Global Health Estimates are calculated are available "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20221028092355/https://cdn.who.int/media/docs/default-source/gho-documents/global-health-estimates/ghe2019_cod_methods.pdf?sfvrsn=37bcfacc_5"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Global Health Observatory (WHO GHO)"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The WHO’s Global Health Observatory brings together a large number of variables produced by the WHO and specialist UN agencies. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The variables found within the Global Health Observatory are limited to the leading causes of death and injury. This means the WHO GHO is less ‘complete’ than the IHME Global Burden of Disease study and WHO GHE in terms of the diseases included."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The method used to produce each of these variables is different and tailored to the specific cause of death or injury. This is a different approach to both the Global Burden of Disease (IHME) and Global Health Estimates (WHO) which attempt to use a consistent modeling approach for all causes of death and injuries. Additionally, variables in the Global Health Observatory are not consistently disaggregated by age or sex. Instead, they show what data is available and is most relevant for the given variable. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The full list of variables available on the Global Health Observatory are available "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.who.int/data/gho/data/indicators"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", each has their own associated metadata and method."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Global Burden of Disease (IHME)"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the Global Burden of Disease study, the IHME uses a wide range of input data. This includes, but is not limited to census data, household surveys, civil registration and vital statistics, disease registries, health service use, air pollution monitors, satellite imaging, disease notifications, and other sources. These data are available through scientific journals, reports, online databases, books, news reports, and other resources. For most diseases and injuries the data are used as input into a series of standardized models which are used to generate estimates of each disease or injury for all age-groups, sexes, locations, and years. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Since the publication of its Global Burden of Disease study in 2017, the IHME has used its "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ghdx.healthdata.org/record/ihme-data/gbd-2019-population-estimates-1950-2019"", ""children"": [{""text"": ""own population estimates"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". These differ from those used by the UN Population Division, which are used by the World Health Organization (WHO). This key difference propagates through all of the resulting health datasets. Even if the IHME and WHO assumed the same rates of health and mortality burden, this difference in population estimates means many of their population-adjusted figures would be different."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Full details on how the IHME’s Global Burden of Disease is calculated are available "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20221031093116/https://www.healthdata.org/sites/default/files/files/Projects/GBD/March2020_GBD%20Protocol_v4.pdf"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20221031141613/https://openresearch.lsbu.ac.uk/download/5509e9b91b2aa850bee0b22228faf9118f2db71edf109660eff4fd2d75d22f06/1927952/PIIS0140673620309259.pdf"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Why do different sources disagree on figures for the same metric?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Collecting precise data on global health is difficult: we can never know exactly which diseases or injuries affect people across the world at any given point in time. In the absence of perfect data, health researchers have a number of ways by which they try to estimate the burden of global health outcomes. These estimates are what we present in the Global Health Explorer. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Each of these sources have slightly different methods for estimating the burden of disease and causes of death, and in some cases the definition of the diseases and injuries in question differ slightly as our Explorer shows."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Differences in disease definition between sources"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The diseases and injuries presented in the Global Health Explorer are typically aggregates of multiple similar causes of death. For example, ‘Deaths from falls’ includes 20 different types of fall which are combined into this one variable. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Individual causes of death have an associated "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/International_Classification_of_Diseases"", ""children"": [{""text"": ""International Classification of Disease"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (ICD) code. The tenth version of the ICD codes are used in the most recent Global Health Estimates and Global Burden of Disease and this is referred to as ICD-10. For example ‘Fall from tree’ has the ICD-10 code ‘W14’, and the ICD-10 codes for all ‘Deaths from falls’ are ‘W00-W19’."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Both IHME and WHO follow the same definition for ‘Deaths from falls’ but for some other causes of death, their definitions of the same aggregate cause of death doesn’t include exactly the same ICD-10 codes. For example, ICD-10 code ‘F02.3 - A dementia developing in the course of established Parkinson disease’ is included as ‘Dementia’ in the WHO’s Global Health Estimates but as ‘Parkinson’s disease’ in the IHME’s Global Burden of Disease."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These slight differences in the definition of diseases may also contribute to the different values we see when looking at the same metric from different sources. The ICD-10 codes used by IHME to define causes of death can be found "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ghdx.healthdata.org/record/ihme-data/gbd-2019-cause-icd-code-mappings"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and the ICD-10 codes used by the WHO can be found "", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20221031140839/https://cdn.who.int/media/docs/default-source/gho-documents/global-health-estimates/ghe2019_cod_methods.pdf"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Global Health Explorer FAQ"", ""authors"": [null], ""dateline"": ""November 1, 2022"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2022-11-01T11:35:57.000Z"", ""published"": false, ""updatedAt"": ""2022-11-02T16:28:46.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-11-01T11:31:37.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 25, ""numErrors"": 1, ""wpTagCounts"": {""list"": 1, ""heading"": 6, ""paragraph"": 18}, ""htmlTagCounts"": {""p"": 18, ""h3"": 1, ""h4"": 5, ""ul"": 1}}",2022-11-01 11:31:37,2024-02-16 14:23:03,,[null],,2022-11-01 11:35:57,2022-11-02 16:28:46,,{},"## Data sources and methodological differences We source the data for the Global Health Explorer from three main sources: * World Health Organization’s Global Health Estimates * World Health Organization’s Global Health Observatory * Institute of Health Metrics and Evaluation’s Global Burden of Disease Study ### Global Health Estimates (WHO GHE) The Global Health Estimates are primarily calculated using cause-of-death statistics that are reported to the WHO by individual countries. These vital registration (VR) statistics are submitted to the WHO Mortality Database on an annual basis by country, year, cause, age and sex. This data is included in the Global Health Estimates if it meets criteria assessing completeness and quality. Since many countries don't meet these criteria, the GHE does not incorporate VR statistics for every country. There are a number of specialist WHO groups and UN agencies that collect topic or disease-specific data. The dataset on HIV and AIDS – collected and published by UN AIDS – is one example of this. The estimates based on the VR data are compared to the data from specialist WHO groups and UN agencies and adjustments are made if necessary. Where the VR data is not usable and there is no other nationally representative cause of death data, then the World Health Organization adopts the IHME’s Global Burden of Disease data to fill these gaps. Full details on how the WHO’s Global Health Estimates are calculated are available [here](https://web.archive.org/web/20221028092355/https://cdn.who.int/media/docs/default-source/gho-documents/global-health-estimates/ghe2019_cod_methods.pdf?sfvrsn=37bcfacc_5). ### Global Health Observatory (WHO GHO) The WHO’s Global Health Observatory brings together a large number of variables produced by the WHO and specialist UN agencies. The variables found within the Global Health Observatory are limited to the leading causes of death and injury. This means the WHO GHO is less ‘complete’ than the IHME Global Burden of Disease study and WHO GHE in terms of the diseases included. The method used to produce each of these variables is different and tailored to the specific cause of death or injury. This is a different approach to both the Global Burden of Disease (IHME) and Global Health Estimates (WHO) which attempt to use a consistent modeling approach for all causes of death and injuries. Additionally, variables in the Global Health Observatory are not consistently disaggregated by age or sex. Instead, they show what data is available and is most relevant for the given variable. The full list of variables available on the Global Health Observatory are available [here](https://www.who.int/data/gho/data/indicators), each has their own associated metadata and method. ### Global Burden of Disease (IHME) In the Global Burden of Disease study, the IHME uses a wide range of input data. This includes, but is not limited to census data, household surveys, civil registration and vital statistics, disease registries, health service use, air pollution monitors, satellite imaging, disease notifications, and other sources. These data are available through scientific journals, reports, online databases, books, news reports, and other resources. For most diseases and injuries the data are used as input into a series of standardized models which are used to generate estimates of each disease or injury for all age-groups, sexes, locations, and years. Since the publication of its Global Burden of Disease study in 2017, the IHME has used its [own population estimates](https://ghdx.healthdata.org/record/ihme-data/gbd-2019-population-estimates-1950-2019). These differ from those used by the UN Population Division, which are used by the World Health Organization (WHO). This key difference propagates through all of the resulting health datasets. Even if the IHME and WHO assumed the same rates of health and mortality burden, this difference in population estimates means many of their population-adjusted figures would be different. Full details on how the IHME’s Global Burden of Disease is calculated are available [here](https://web.archive.org/web/20221031093116/https://www.healthdata.org/sites/default/files/files/Projects/GBD/March2020_GBD%20Protocol_v4.pdf) and [here](https://web.archive.org/web/20221031141613/https://openresearch.lsbu.ac.uk/download/5509e9b91b2aa850bee0b22228faf9118f2db71edf109660eff4fd2d75d22f06/1927952/PIIS0140673620309259.pdf). ### Why do different sources disagree on figures for the same metric? Collecting precise data on global health is difficult: we can never know exactly which diseases or injuries affect people across the world at any given point in time. In the absence of perfect data, health researchers have a number of ways by which they try to estimate the burden of global health outcomes. These estimates are what we present in the Global Health Explorer. Each of these sources have slightly different methods for estimating the burden of disease and causes of death, and in some cases the definition of the diseases and injuries in question differ slightly as our Explorer shows. ### Differences in disease definition between sources The diseases and injuries presented in the Global Health Explorer are typically aggregates of multiple similar causes of death. For example, ‘Deaths from falls’ includes 20 different types of fall which are combined into this one variable. Individual causes of death have an associated [International Classification of Disease](https://en.wikipedia.org/wiki/International_Classification_of_Diseases) (ICD) code. The tenth version of the ICD codes are used in the most recent Global Health Estimates and Global Burden of Disease and this is referred to as ICD-10. For example ‘Fall from tree’ has the ICD-10 code ‘W14’, and the ICD-10 codes for all ‘Deaths from falls’ are ‘W00-W19’. Both IHME and WHO follow the same definition for ‘Deaths from falls’ but for some other causes of death, their definitions of the same aggregate cause of death doesn’t include exactly the same ICD-10 codes. For example, ICD-10 code ‘F02.3 - A dementia developing in the course of established Parkinson disease’ is included as ‘Dementia’ in the WHO’s Global Health Estimates but as ‘Parkinson’s disease’ in the IHME’s Global Burden of Disease. These slight differences in the definition of diseases may also contribute to the different values we see when looking at the same metric from different sources. The ICD-10 codes used by IHME to define causes of death can be found [here](https://ghdx.healthdata.org/record/ihme-data/gbd-2019-cause-icd-code-mappings) and the ICD-10 codes used by the WHO can be found [here](https://web.archive.org/web/20221031140839/https://cdn.who.int/media/docs/default-source/gho-documents/global-health-estimates/ghe2019_cod_methods.pdf).","{""data"": {""wpBlock"": {""content"": ""\nWe source the data for the Global Health Explorer from three main sources:
\n\n\n\nThe Global Health Estimates are primarily calculated using cause-of-death statistics that are reported to the WHO by individual countries.
\n\n\n\nThese vital registration (VR) statistics are submitted to the WHO Mortality Database on an annual basis by country, year, cause, age and sex. This data is included in the Global Health Estimates if it meets criteria assessing completeness and quality. Since many countries don’t meet these criteria, the GHE does not incorporate VR statistics for every country.
\n\n\n\nThere are a number of specialist WHO groups and UN agencies that collect topic or disease-specific data. The dataset on HIV and AIDS – collected and published by UN AIDS – is one example of this. The estimates based on the VR data are compared to the data from specialist WHO groups and UN agencies and adjustments are made if necessary.
\n\n\n\nWhere the VR data is not usable and there is no other nationally representative cause of death data, then the World Health Organization adopts the IHME’s Global Burden of Disease data to fill these gaps. Full details on how the WHO’s Global Health Estimates are calculated are available here.
\n\n\n\nThe WHO’s Global Health Observatory brings together a large number of variables produced by the WHO and specialist UN agencies.
\n\n\n\nThe variables found within the Global Health Observatory are limited to the leading causes of death and injury. This means the WHO GHO is less ‘complete’ than the IHME Global Burden of Disease study and WHO GHE in terms of the diseases included.
\n\n\n\nThe method used to produce each of these variables is different and tailored to the specific cause of death or injury. This is a different approach to both the Global Burden of Disease (IHME) and Global Health Estimates (WHO) which attempt to use a consistent modeling approach for all causes of death and injuries. Additionally, variables in the Global Health Observatory are not consistently disaggregated by age or sex. Instead, they show what data is available and is most relevant for the given variable.
\n\n\n\nThe full list of variables available on the Global Health Observatory are available here, each has their own associated metadata and method.
\n\n\n\nIn the Global Burden of Disease study, the IHME uses a wide range of input data. This includes, but is not limited to census data, household surveys, civil registration and vital statistics, disease registries, health service use, air pollution monitors, satellite imaging, disease notifications, and other sources. These data are available through scientific journals, reports, online databases, books, news reports, and other resources. For most diseases and injuries the data are used as input into a series of standardized models which are used to generate estimates of each disease or injury for all age-groups, sexes, locations, and years.
\n\n\n\nSince the publication of its Global Burden of Disease study in 2017, the IHME has used its own population estimates. These differ from those used by the UN Population Division, which are used by the World Health Organization (WHO). This key difference propagates through all of the resulting health datasets. Even if the IHME and WHO assumed the same rates of health and mortality burden, this difference in population estimates means many of their population-adjusted figures would be different.
\n\n\n\nFull details on how the IHME’s Global Burden of Disease is calculated are available here and here.
\n\n\n\nCollecting precise data on global health is difficult: we can never know exactly which diseases or injuries affect people across the world at any given point in time. In the absence of perfect data, health researchers have a number of ways by which they try to estimate the burden of global health outcomes. These estimates are what we present in the Global Health Explorer.
\n\n\n\nEach of these sources have slightly different methods for estimating the burden of disease and causes of death, and in some cases the definition of the diseases and injuries in question differ slightly as our Explorer shows.
\n\n\n\nThe diseases and injuries presented in the Global Health Explorer are typically aggregates of multiple similar causes of death. For example, ‘Deaths from falls’ includes 20 different types of fall which are combined into this one variable.
\n\n\n\nIndividual causes of death have an associated International Classification of Disease (ICD) code. The tenth version of the ICD codes are used in the most recent Global Health Estimates and Global Burden of Disease and this is referred to as ICD-10. For example ‘Fall from tree’ has the ICD-10 code ‘W14’, and the ICD-10 codes for all ‘Deaths from falls’ are ‘W00-W19’.
\n\n\n\nBoth IHME and WHO follow the same definition for ‘Deaths from falls’ but for some other causes of death, their definitions of the same aggregate cause of death doesn’t include exactly the same ICD-10 codes. For example, ICD-10 code ‘F02.3 – A dementia developing in the course of established Parkinson disease’ is included as ‘Dementia’ in the WHO’s Global Health Estimates but as ‘Parkinson’s disease’ in the IHME’s Global Burden of Disease.
\n\n\n\nThese slight differences in the definition of diseases may also contribute to the different values we see when looking at the same metric from different sources. The ICD-10 codes used by IHME to define causes of death can be found here and the ICD-10 codes used by the WHO can be found here.
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 54245,The 'Varieties of Democracy' data: how do researchers measure democracy?,vdem-electoral-democracy-data,post,publish,"Measuring the state of democracy across the world helps us understand the extent to which people have political rights and freedoms.
But measuring democracy comes with many challenges. People do not always agree on what characteristics define a democracy. These characteristics — such as whether an election was free and fair — are difficult to define and assess. The judgment of experts is to some degree subjective. They may disagree about a specific characteristic or how something as complex as a political system can be reduced into a single measure.
How do researchers address these challenges and measure democracy?
In some of our work on democracy, we rely on data published by the Varieties of Democracy (V-Dem) project.{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. V-Dem [Country-Year/Country-Date] Dataset v13. Varieties of Democracy (V-Dem) Project.
Pemstein, Daniel, Kyle L. Marquardt, Eitan Tzelgov, Yi-ting Wang, Juraj Medzihorsky, Joshua Krusell, Farhad Miri, and Johannes von Römer. 2023. The V-Dem Measurement Model: Latent Variable Analysis for Cross-National and Cross-Temporal Expert-Coded Data. V-Dem Working Paper No. 21. University of Gothenburg: Varieties of Democracy Institute.{/ref}
The project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is run by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators.
V-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation.
True to its name, the Varieties of Democracy project acknowledges that democracy can be characterized differently, and measures electoral, liberal, participatory, deliberative, and egalitarian characterizations of democracy.
At Our World in Data we primarily use V-Dem’s Electoral Democracy Index to measure democracy.{ref}The index is sometimes also called the Polyarchy Index.{/ref} The index is used in all of V-Dem’s other democracy indices because V-Dem considers there to be no democracy without elections. The other aspects can therefore be thought of as measuring the quality of a democracy.
V-Dem characterizes electoral democracy as a political system in which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed. More specifically, this means:
You can find data on the other democracy indices, electoral democracy’s characteristics, and other derived measures in our Democracy Data Explorer.
The Electoral Democracy Index scores each country on a spectrum, with some countries being more democratic than others.
The spectrum ranges from 0 (‘highly undemocratic’) to 1 (‘highly democratic’).
This scoring thereby differs from other approaches such as ‘Regimes of the World’ and other projects, which classify countries as a binary: either they are a democracy or not.
As of version 13 of the dataset, V-Dem covers 202 countries, going back in time as far as 1789. Many countries have been covered since 1900, including before they became independent from their colonial powers.
To actually measure what it wants to capture, V-Dem assesses the characteristics of democracy mostly through evaluations by experts.{ref} For more details, see: Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
These anonymous experts are primarily academics and members of the media and civil society. They are also often nationals or residents of the country they assess, and therefore know its political system well and can evaluate aspects that are difficult to observe.
V-Dem’s own team of researchers supplements the expert evaluations. They code some easier-to-observe rules and laws of the political system, such as whether the legislature has a lower and upper house.
V-Dem uses several experts per country, year, and topic, to make its assessments less subjective. In total, around 3,500 country-experts fill surveys for V-Dem every year.
While there are fewer experts for small countries and for the time before 1900, they rely typically on 25 experts per country and 5 experts per topic.
V-Dem also works to make their coders’ assessments comparable across countries and time.
The surveys ask the experts to answer very specific questions on completely explained scales about sub-characteristics of political systems — such as the presence or absence of election fraud — instead of making them rely on their broad impressions.
The surveys are available in English, Arabic, French, Portuguese, Russian, and Spanish to reduce misunderstandings.
Experts further evaluate hypothetical countries, many coded several countries, and they denote their own uncertainty and personal demographic information.
V-Dem then uses this information to investigate expert biases, which they have found to be limited: they only find that experts from a country tend to be stricter in their assessments. {ref}“We have run extensive tests on how well such individual-level factors predict country-ratings but have found that the only factor consistently associated with country-ratings is country of origin (with “domestic” experts being harsher in their judgments).”
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
V-Dem uses a statistical model to address any remaining differences between coders.{ref}Specifically, it uses a Bayesian Item-Response Theory estimation strategy.
Marquardt, Kyle, and Daniel Pemstein. 2018. IRT Models for Expert-Coded Panel Data. Political Analysis 26(4): 431-456.{/ref}
The model combines the experts’ ratings of actual countries and hypothetical countries, as well as the experts’ stated uncertainties and personal demographics to produce best, upper-, and lower-bound estimates of many characteristics.{ref}Expressed precisely, V-Dem’s measurement model produces a probability distribution over the country-year scores. The best estimate is the distribution’s median, while the upper and lower bound estimates demarcate the interval in which the model places 68 percent of the probability mass.{/ref}
V-Dem provides these different estimates for all of its main and supplementary indices, including the Electoral Democracy Index and the subindices for free and fair elections, freedom of association, and freedom of expression.
With the different estimates, V-Dem explicitly acknowledges that its coders can be uncertain or make errors in their measurement.
The overall Electoral Democracy Index score is the result of weighing, multiplying, and adding up the subindices.{ref} The precise formula is:
electoral democracy index = 0.5 * multiplicative polyarchy index + 0.5 * additive polyarchy index; with
multiplicative polyarchy index = elected officials * free and fair elections * comprehensive suffrage * freedom of association * freedom of expression; and
additive polyarchy index = 0.125 * elected officials + 0.25 * free and fair elections + 0.125 * comprehensive suffrage + 0.25 * freedom of association + 0.25 * freedom of expression{/ref}
The subindices are weighted because V-Dem considers some of them as more important than others: elected officials and voting rights are weighted less because they capture more formal requirements, as opposed to free and fair elections and the freedoms of association and expression that rely more on expert assessments.
The subindices are partially multiplied and partially added up because V-Dem wants the subindices to partially compensate for one another, and partially for them to reinforce each other. An example of compensation is voting rights partially making up for a lack of rights to assemble and protest, whereas an example of reinforcement is voting rights mattering more if voters can also choose opposition candidates.
V-Dem releases its data publicly, and makes it straightforward to download and use.
It publishes the overall scores, the underlying subindices, and several hundred specific questions by country-year, country-date, and coder.
V-Dem also releases detailed descriptions of how they characterize democracy, the questions and coding procedures that guide the experts and researchers, as well as why it weighs, adds, and multiplies the scores for specific characteristics.
In our work, we expand the years covered by V-Dem further.
To expand the time coverage of today’s countries and include more of the period when they were still non-sovereign territories, we identified the historical entity they were a part of and used that regime’s data whenever available.{ref}For example, V-Dem only provides regime data since Bangladesh’s independence in 1971. There is, however, regime data for Pakistan and the colony of India, both of which the current territory of Bangladesh was a part. We, therefore, use the regime data of Pakistan for Bangladesh from 1947 to 1970, and the regime data of India from 1789 to 1946. We did so for all countries with a past or current population of more than one million.{/ref}
We also calculated regional and global averages of the Electoral Democracy Index and its sub-indices, weighted and unweighted by population.
Our code and data are available on GitHub and record our revisions in detail.
V-Dem releases a new version of the data each year in March.
We at Our World in Data aim to update our own data within a few weeks of the release.
There are shortcomings in the way the Electoral Democracy Index characterizes and measures democracy.{ref}This and the following section draw on several very helpful other articles summarizing and reviewing some of the leading democracy datasets:
Boese, Vanessa. 2019. How (not) to measure democracy. International Area Studies Review 22(2): 95-127.
Coppedge, Michael, John Gerring, Staffan I. Lindberg, Svend-Erik Skaaning, and Jan Teorell. 2017. V-Dem Comparisons and Contrasts with Other Measurement Projects. V-Dem Working Paper 45.
Møller, Jørgen and Svend-Erik Skaaning. 2021. Varieties of Measurement: A Comparative Assessment of Relatively New Democracy Ratings based on Original Data. V-Dem Working Paper 123.Skaaning, Svend-Erik. 2018. Different Types of Data and the Validity of Democracy Measures. Politics and Governance 6(1): 105-116.{/ref}
The index focuses on an electoral understanding of democracy and does not account for other characterizations, such as democracies as egalitarian political systems, in which political power is equally distributed to allow everyone to participate. This means that some of the most economically-unequal countries in the world, such as Brazil and South Africa, are classified as broadly democratic in recent years.{ref}True to its name, however, V-Dem provides several democracy indices in addition to the Electoral Democracy Index, and also measures liberal, participatory, deliberative, and egalitarian characterizations of democracy.{/ref}
V-Dem also does not cover some countries with very small populations.
Furthermore, the index is more difficult to interpret than other measures. Measures that group countries into democracies and autocracies, such as the Regimes of the World classification, make it possible to say which country was a democracy.
The Electoral Democracy Index makes no clear assessment there, and only allows us to say whether a country is relatively democratic by comparing it to the range of the index, to other countries, or to the same country at another point in time. And when doing so, it is still difficult to say how large these differences are.{ref}This can be made easier by comparing how a score relates to the index’s overall distribution or its distribution for a specific year.{/ref}
The assessment of the Electoral Democracy Index remains to some extent subjective. Its index is built on difficult evaluations by experts that rely less on easier-to-observe characteristics, such as whether regular elections are held.
Finally, the index’s aggregation remains to some extent arbitrary. It is unclear why these specific subindices were chosen; and why two subindices, elected officials and voting rights, are weighted less than the others.
Despite these shortcomings, the index tells us a lot about how democratic the world was in the past and today.
Its characterization of democracy as an electoral political system, in which citizens get to participate in free and fair elections, is commonly recognized as the basic principle of democracy and shared by all of the leading approaches of measuring democracy.
Because it treats democracy as a spectrum, the index is able to capture both big and small differences in the political systems of countries, and to record small changes within countries over time. This allows us to observe whether one country is more democratic than another, or whether a country has become more or less democratic over time.
The index also covers many countries and years. With the exception of microstates, it covers all countries in the world. Many countries are covered since 1900 — even while they were colonized by another country — and some of them as far back as 1789.
Finally, V-Dem takes many steps to make its assessments valid, precise, comparable across countries and time, and transparent. It relies on many country and subject experts answering detailed surveys to measure aspects of political systems that are often difficult to observe and acknowledges the remaining uncertainty in their assessments.
Whether V-Dem’s Electoral Democracy Index is a useful measure of democracy will depend on the questions we want to answer.
The index will not give us a satisfying answer if we are interested in non-electoral understandings of democracy (or different understandings of electoral democracy); if we are also interested in the political systems of microstates; and only interested in big differences in the political systems of countries.
In these cases, we will have to rely on other measures.
But if we value a sophisticated measure based on the knowledge of many country experts and are interested in big and small differences in electoral democracy, within and across countries, and far into the past, we can learn a lot from this data.
It is for these latter purposes we use the measure in some of our reporting on democracy.
I thank Edouard Mathieu, Hannah Ritchie, and Max Roser for their very helpful comments and ideas about how to improve this article.
","{""id"": ""wp-54245"", ""slug"": ""vdem-electoral-democracy-data"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Measuring the state of democracy across the world helps us understand the extent to which people have political rights and freedoms."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But measuring democracy comes with many challenges. People do not always agree on what characteristics define a democracy. These characteristics — such as whether an election was free and fair — are difficult to define and assess. The judgment of experts is to some degree subjective. They may disagree about a specific characteristic or how something as complex as a political system can be reduced into a single measure."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How do researchers address these challenges and measure democracy?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What is the Varieties of Democracy (V-Dem) project?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In some of our work on democracy, we rely on data published by the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.v-dem.net/vdemds.html"", ""children"": [{""text"": ""Varieties of Democracy (V-Dem) project"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Carl+Henrik+Knutsen%2C+Staffan+I.+Lindberg%2C+Jan+Teorell%2C+David+Altman%2C+Michael+Bernhard%2C+Agnes+Cornell%2C+M.+Steven+Fish%2C+Lisa+Gastaldi%2C+Haakon+Gjerl%C3%B8w%2C+Adam+Glynn%2C+Ana+Good+God%2C+Sandra+Grahn%2C+Allen+Hicken%2C+Katrin+Kinzelbach%2C+Joshua+Krusell%2C+Kyle+L.+Marquardt%2C+Kelly+McMann%2C+Valeriya+Mechkova%2C+Juraj+Medzihorsky%2C+Natalia+Natsika%2C+Anja+Neundorf%2C+Pamela+Paxton%2C+Daniel+Pemstein%2C+Josefine+Pernes%2C+Oskar+Ryd%C3%A9n%2C+Johannes+von+R%C3%B6mer%2C+Brigitte+Seim%2C+Rachel+Sigman%2C+Svend-Erik+Skaaning%2C+Jeffrey+Staton%2C+Aksel+Sundstr%C3%B6m%2C+Eitan+Tzelgov%2C+Yi-ting+Wang%2C+Tore+Wig%2C+Steven+Wilson+and+Daniel+Ziblatt.+2023.+V-Dem+%5BCountry-Year%2FCountry-Date%5D+Dataset+v13.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG="", ""children"": [{""text"": ""V-Dem [Country-Year/Country-Date] Dataset v13."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" Varieties of Democracy (V-Dem) Project."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""Pemstein, Daniel, Kyle L. Marquardt, Eitan Tzelgov, Yi-ting Wang, Juraj Medzihorsky, Joshua Krusell, Farhad Miri, and Johannes von Römer. 2023. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://v-dem.net/media/publications/Working_Paper_21_z5BldB1.pdf"", ""children"": [{""text"": ""The V-Dem Measurement Model: Latent Variable Analysis for Cross-National and Cross-Temporal Expert-Coded Data."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" V-Dem Working Paper No. 21. University of Gothenburg: Varieties of Democracy Institute.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.v-dem.net/about/v-dem-project/"", ""children"": [{""text"": ""run"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem characterize democracy?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""True to its name, the Varieties of Democracy project acknowledges that democracy can be characterized differently, and measures electoral, liberal, participatory, deliberative, and egalitarian characterizations of democracy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""At Our World in Data we primarily use V-Dem’s Electoral Democracy Index to measure democracy.{ref}The index is sometimes also called the Polyarchy Index.{/ref} The index is used in all of V-Dem’s other democracy indices because V-Dem considers there to be no democracy without elections. The other aspects can therefore be thought of as measuring the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""quality"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of a democracy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem characterizes electoral democracy as a political system in which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed. More specifically, this means:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Elected political leaders"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": broad elections choose the chief executive and legislature"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Comprehensive voting rights"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": all adult citizens have the legal right to vote in national elections"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Free and fair elections"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": no election violence, government intimidation, fraud, large irregularities, and vote buying"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Freedom of association"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": parties and civil society organizations can form and operate freely"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Freedom of expression"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "": people can voice their views and the media can present different political perspectives"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""You can find data on the other democracy indices, electoral democracy’s characteristics, and other derived measures in our "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Varieties+of+Democracy&Metric=Electoral+democracy&Sub-metric=Main+index"", ""children"": [{""text"": ""Democracy Data Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How is democracy scored?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Electoral Democracy Index scores each country on a spectrum, with some countries being more democratic than others."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The spectrum ranges from 0 (‘highly undemocratic’) to 1 (‘highly democratic’)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This scoring thereby differs from other approaches such as "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/regimes-of-the-world-data"", ""children"": [{""text"": ""‘Regimes of the World’"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracies-measurement"", ""children"": [{""text"": ""other projects"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", which classify countries as a binary: either they are a democracy or not."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/electoral-democracy"", ""type"": ""chart"", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-electoral-democracy-vdem"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What years and countries are covered?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As of version 13 of the dataset, V-Dem covers 202 countries, going back in time as far as 1789. Many countries have been covered since 1900, including before they became independent from their colonial powers."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How is democracy measured?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem work to make its assessments valid?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To actually measure what it wants to capture, V-Dem assesses the characteristics of democracy mostly through evaluations by experts.{ref} For more details, see: Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These anonymous experts are primarily academics and members of the media and civil society. They are also often nationals or residents of the country they assess, and therefore know its political system well and can evaluate aspects that are difficult to observe."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem’s own team of researchers supplements the expert evaluations. They code some easier-to-observe rules and laws of the political system, such as whether the legislature has a lower and upper house."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem work to make its assessments precise and reliable?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem uses several experts per country, year, and topic, to make its assessments less subjective. In total, around 3,500 country-experts fill surveys for V-Dem every year."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""While there are fewer experts for small countries and for the time before 1900, they rely typically on 25 experts per country and 5 experts per topic."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How does V-Dem work to make its assessments comparable?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem also works to make their coders’ assessments comparable across countries and time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The surveys ask the experts to answer very specific questions on completely explained scales about sub-characteristics of political systems — such as the presence or absence of election fraud — instead of making them rely on their broad impressions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The surveys are available in English, Arabic, French, Portuguese, Russian, and Spanish to reduce misunderstandings."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Experts further evaluate hypothetical countries, many coded several countries, and they denote their own uncertainty and personal demographic information."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem then uses this information to investigate expert biases, which they have found to be limited: they only find that experts from a country tend to be stricter in their assessments. {ref}“We have run extensive tests on how well such individual-level factors predict country-ratings but have found that the only factor consistently associated with country-ratings is country of origin (with “domestic” experts being harsher in their judgments).”"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How are the remaining differences in the data dealt with?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem uses a statistical model to address any remaining differences between coders.{ref}Specifically, it uses a Bayesian Item-Response Theory estimation strategy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Marquardt, Kyle, and Daniel Pemstein. 2018. IRT Models for Expert-Coded Panel Data. Political Analysis 26(4): 431-456.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The model combines the experts’ ratings of actual countries and hypothetical countries, as well as the experts’ stated uncertainties and personal demographics to produce best, upper-, and lower-bound estimates of many characteristics.{ref}Expressed precisely, V-Dem’s measurement model produces a probability distribution over the country-year scores. The best estimate is the distribution’s median, while the upper and lower bound estimates demarcate the interval in which the model places 68 percent of the probability mass.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem provides these different estimates for all of its main and supplementary indices, including the Electoral Democracy Index and the subindices for free and fair elections, freedom of association, and freedom of expression."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""With the different estimates, V-Dem explicitly acknowledges that its coders can be uncertain or make errors in their measurement."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The overall Electoral Democracy Index score is the result of weighing, multiplying, and adding up the subindices.{ref} The precise formula is:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""electoral democracy index = 0.5 * multiplicative polyarchy index + 0.5 * additive polyarchy index; with"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""multiplicative polyarchy index = elected officials * free and fair elections * comprehensive suffrage * freedom of association * freedom of expression; and "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""additive polyarchy index = 0.125 * elected officials + 0.25 * free and fair elections + 0.125 * comprehensive suffrage + 0.25 * freedom of association + 0.25 * freedom of expression{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The subindices are weighted because V-Dem considers some of them as more important than others: elected officials and voting rights are weighted less because they capture more formal requirements, as opposed to free and fair elections and the freedoms of association and expression that rely more on expert assessments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The subindices are partially multiplied and partially added up because V-Dem wants the subindices to partially compensate for one another, and partially for them to reinforce each other. An example of compensation is voting rights partially making up for a lack of rights to assemble and protest, whereas an example of reinforcement is voting rights mattering more if voters can also choose opposition candidates."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How is the data made accessible and transparent?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem releases "", ""spanType"": ""span-simple-text""}, {""url"": ""https://v-dem.net/data/the-v-dem-dataset/"", ""children"": [{""text"": ""its data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" publicly, and makes it straightforward to download and use."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It publishes the overall scores, the underlying subindices, and several hundred specific questions by country-year, country-date, and coder."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem also releases detailed descriptions of how they characterize democracy, the questions and coding procedures that guide the experts and researchers, as well as why it weighs, adds, and multiplies the scores for specific characteristics."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How do we change the data?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In our work, we expand the years covered by V-Dem further."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To expand the time coverage of today’s countries and include more of the period when they were still non-sovereign territories, we identified the historical entity they were a part of and used that regime’s data whenever available.{ref}For example, V-Dem only provides regime data since Bangladesh’s independence in 1971. There is, however, regime data for Pakistan and the colony of India, both of which the current territory of Bangladesh was a part. We, therefore, use the regime data of Pakistan for Bangladesh from 1947 to 1970, and the regime data of India from 1789 to 1946. We did so for all countries with a past or current population of more than one million.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We also calculated regional and global averages of the Electoral Democracy Index and its sub-indices, weighted and unweighted by population."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Our code and data are available "", ""spanType"": ""span-simple-text""}, {""url"": ""https://github.com/owid/notebooks/tree/main/BastianHerre/democracy"", ""children"": [{""text"": ""on GitHub"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and record our revisions in detail."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""How often and when is the data updated?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem releases a new version of the data each year in March."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We at Our World in Data aim to update our own data within a few weeks of the release."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What are the data’s shortcomings?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are shortcomings in the way the Electoral Democracy Index characterizes and measures democracy.{ref}This and the following section draw on several very helpful other articles summarizing and reviewing some of the leading democracy datasets:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Boese, Vanessa. 2019. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Boese%2C+Vanessa.+2019.+How+%28not%29+to+measure+democracy.+International+Area+Studies+Review+22%282%29%3A+95-127&btnG="", ""children"": [{""text"": ""How (not) to measure democracy"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". International Area Studies Review 22(2): 95-127."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Coppedge, Michael, John Gerring, Staffan I. Lindberg, Svend-Erik Skaaning, and Jan Teorell. 2017. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Staffan+I.+Lindberg%2C+Svend-Erik+Skaaning%2C+and+Jan+Teorell.+2017.+V-Dem+Comparisons+and+Contrasts+with+Other+Measurement+Projects.+V-Dem+Working+Paper+45.&btnG="", ""children"": [{""text"": ""V-Dem Comparisons and Contrasts with Other Measurement Projects"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". V-Dem Working Paper 45."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Møller, Jørgen and Svend-Erik Skaaning. 2021. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=M%C3%B8ller%2C+J%C3%B8rgen+and+Svend-Erik+Skaaning.+2021.+Varieties+of+Measurement%3A+A+Comparative+Assessment+of+Relatively+New+Democracy+Ratings+based+on+Original+Data.+V-Dem+Working+Paper+123.&btnG="", ""children"": [{""text"": ""Varieties of Measurement: A Comparative Assessment of Relatively New Democracy Ratings based on Original Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". V-Dem Working Paper 123.Skaaning, Svend-Erik. 2018. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Skaaning%2C+Svend-Erik.+2018.+Different+Types+of+Data+and+the+Validity+of+Democracy+Measures.+Politics+and+Governance+6%281%29%3A+105-116.&btnG="", ""children"": [{""text"": ""Different Types of Data and the Validity of Democracy Measures"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Politics and Governance 6(1): 105-116.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The index focuses on an electoral understanding of democracy and does not account for other characterizations, such as democracies as egalitarian political systems, in which political power is equally distributed to allow everyone to participate. This means that some of the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/economic-inequality-gini-index"", ""children"": [{""text"": ""most economically-unequal countries"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in the world, such as Brazil and South Africa, are classified as broadly democratic in recent years.{ref}True to its name, however, V-Dem provides several democracy indices in addition to the Electoral Democracy Index, and also measures liberal, participatory, deliberative, and egalitarian characterizations of democracy.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""V-Dem also does not cover some countries with very small populations."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Furthermore, the index is more difficult to interpret than other measures. Measures that group countries into democracies and autocracies, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/regimes-of-the-world-data"", ""children"": [{""text"": ""such as the Regimes of the World classification"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", make it possible to say which country was a democracy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The Electoral Democracy Index makes no clear assessment there, and only allows us to say whether a country is relatively democratic by comparing it to the range of the index, to other countries, or to the same country at another point in time. And when doing so, it is still difficult to say how large these differences are.{ref}This can be made easier by comparing how a score relates to the index’s overall distribution or its distribution for a specific year.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The assessment of the Electoral Democracy Index remains to some extent subjective. Its index is built on difficult evaluations by experts that rely less on easier-to-observe characteristics, such as whether regular elections are held."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Finally, the index’s aggregation remains to some extent arbitrary. It is unclear why these specific subindices were chosen; and why two subindices, elected officials and voting rights, are weighted less than the others."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What are the data’s strengths?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Despite these shortcomings, the index tells us a lot about how democratic the world was in the past and today."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Its characterization of democracy as an electoral political system, in which citizens get to participate in free and fair elections, is commonly recognized as the basic principle of democracy and shared by "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracies-measurement"", ""children"": [{""text"": ""all of the leading approaches of measuring democracy"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Because it treats democracy as a spectrum, the index is able to capture both big and small differences in the political systems of countries, and to record small changes within countries over time. This allows us to observe whether one country is more democratic than another, or whether a country has become more or less democratic over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The index also covers many countries and years. With the exception of microstates, it covers all countries in the world. Many countries are covered since 1900 — even while they were colonized by another country — and some of them as far back as 1789."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Finally, V-Dem takes many steps to make its assessments valid, precise, comparable across countries and time, and transparent. It relies on many country and subject experts answering detailed surveys to measure aspects of political systems that are often difficult to observe and acknowledges the remaining uncertainty in their assessments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""What is our summary assessment?"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Whether V-Dem’s Electoral Democracy Index is a useful measure of democracy will depend on the questions we want to answer."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The index will not give us a satisfying answer if we are interested in non-electoral understandings of democracy (or different understandings of electoral democracy); if we are also interested in the political systems of microstates; and only interested in big differences in the political systems of countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In these cases, we will have to rely on "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracies-measurement"", ""children"": [{""text"": ""other measures"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But if we value a sophisticated measure based on the knowledge of many country experts and are interested in big and small differences in electoral democracy, within and across countries, and far into the past, we can learn a lot from this data."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It is for these latter purposes we use the measure in some of our reporting on democracy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""Keep reading on "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""children"": [{""children"": [{""text"": ""Our World in Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/democracy"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""Acknowledgments"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I thank Edouard Mathieu, Hannah Ritchie, and Max Roser for their very helpful comments and ideas about how to improve this article."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""The 'Varieties of Democracy' data: how do researchers measure democracy?"", ""authors"": [""Bastian Herre""], ""excerpt"": ""There are many ways to measure democracy. Here is how the Varieties of Democracy project does it, one of the leading sources of global democracy data."", ""dateline"": ""November 30, 2022"", ""subtitle"": ""There are many ways to measure democracy. Here is how the Varieties of Democracy project does it, one of the leading sources of global democracy data."", ""sidebar-toc"": false, ""featured-image"": ""democratic_world.png""}, ""createdAt"": ""2022-11-01T09:38:34.000Z"", ""published"": false, ""updatedAt"": ""2023-06-08T09:56:26.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-11-30T12:54:36.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}, {""name"": ""columns block expects 2 children"", ""details"": ""Got 1 children instead""}], ""numBlocks"": 87, ""numErrors"": 2, ""wpTagCounts"": {""html"": 2, ""list"": 1, ""column"": 1, ""columns"": 1, ""heading"": 17, ""paragraph"": 66, ""owid/prominent-link"": 1}, ""htmlTagCounts"": {""p"": 66, ""h4"": 13, ""h5"": 4, ""ul"": 1, ""div"": 2, ""iframe"": 2}}",2022-11-30 12:54:36,2024-02-16 14:22:54,14zTYMg-mkPcMC68DqgQNb1eAe8VBwoD93yzyKGIeguw,"[""Bastian Herre""]","There are many ways to measure democracy. Here is how the Varieties of Democracy project does it, one of the leading sources of global democracy data.",2022-11-01 09:38:34,2023-06-08 09:56:26,https://ourworldindata.org/wp-content/uploads/2022/07/democratic_world.png,{},"Measuring the state of democracy across the world helps us understand the extent to which people have political rights and freedoms. But measuring democracy comes with many challenges. People do not always agree on what characteristics define a democracy. These characteristics — such as whether an election was free and fair — are difficult to define and assess. The judgment of experts is to some degree subjective. They may disagree about a specific characteristic or how something as complex as a political system can be reduced into a single measure. How do researchers address these challenges and measure democracy? ## **What is the Varieties of Democracy (V-Dem) project?** In some of our work on democracy, we rely on data published by the [Varieties of Democracy (V-Dem) project](https://www.v-dem.net/vdemds.html).{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. [V-Dem [Country-Year/Country-Date] Dataset v13.](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Carl+Henrik+Knutsen%2C+Staffan+I.+Lindberg%2C+Jan+Teorell%2C+David+Altman%2C+Michael+Bernhard%2C+Agnes+Cornell%2C+M.+Steven+Fish%2C+Lisa+Gastaldi%2C+Haakon+Gjerl%C3%B8w%2C+Adam+Glynn%2C+Ana+Good+God%2C+Sandra+Grahn%2C+Allen+Hicken%2C+Katrin+Kinzelbach%2C+Joshua+Krusell%2C+Kyle+L.+Marquardt%2C+Kelly+McMann%2C+Valeriya+Mechkova%2C+Juraj+Medzihorsky%2C+Natalia+Natsika%2C+Anja+Neundorf%2C+Pamela+Paxton%2C+Daniel+Pemstein%2C+Josefine+Pernes%2C+Oskar+Ryd%C3%A9n%2C+Johannes+von+R%C3%B6mer%2C+Brigitte+Seim%2C+Rachel+Sigman%2C+Svend-Erik+Skaaning%2C+Jeffrey+Staton%2C+Aksel+Sundstr%C3%B6m%2C+Eitan+Tzelgov%2C+Yi-ting+Wang%2C+Tore+Wig%2C+Steven+Wilson+and+Daniel+Ziblatt.+2023.+V-Dem+%5BCountry-Year%2FCountry-Date%5D+Dataset+v13.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG=) Varieties of Democracy (V-Dem) Project. Pemstein, Daniel, Kyle L. Marquardt, Eitan Tzelgov, Yi-ting Wang, Juraj Medzihorsky, Joshua Krusell, Farhad Miri, and Johannes von Römer. 2023. [The V-Dem Measurement Model: Latent Variable Analysis for Cross-National and Cross-Temporal Expert-Coded Data.](https://v-dem.net/media/publications/Working_Paper_21_z5BldB1.pdf) V-Dem Working Paper No. 21. University of Gothenburg: Varieties of Democracy Institute.{/ref} The project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is [run](https://www.v-dem.net/about/v-dem-project/) by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators. V-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation. ## **How does V-Dem characterize democracy?** True to its name, the Varieties of Democracy project acknowledges that democracy can be characterized differently, and measures electoral, liberal, participatory, deliberative, and egalitarian characterizations of democracy. At Our World in Data we primarily use V-Dem’s Electoral Democracy Index to measure democracy.{ref}The index is sometimes also called the Polyarchy Index.{/ref} The index is used in all of V-Dem’s other democracy indices because V-Dem considers there to be no democracy without elections. The other aspects can therefore be thought of as measuring the _quality_ of a democracy. V-Dem characterizes electoral democracy as a political system in which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed. More specifically, this means: * **Elected political leaders**: broad elections choose the chief executive and legislature * **Comprehensive voting rights**: all adult citizens have the legal right to vote in national elections * **Free and fair elections**: no election violence, government intimidation, fraud, large irregularities, and vote buying * **Freedom of association**: parties and civil society organizations can form and operate freely * **Freedom of expression**: people can voice their views and the media can present different political perspectives You can find data on the other democracy indices, electoral democracy’s characteristics, and other derived measures in our [Democracy Data Explorer](https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Varieties+of+Democracy&Metric=Electoral+democracy&Sub-metric=Main+index). ## **How is democracy scored?** The Electoral Democracy Index scores each country on a spectrum, with some countries being more democratic than others. The spectrum ranges from 0 (‘highly undemocratic’) to 1 (‘highly democratic’). This scoring thereby differs from other approaches such as [‘Regimes of the World’](https://ourworldindata.org/regimes-of-the-world-data) and [other projects](https://ourworldindata.org/democracies-measurement), which classify countries as a binary: either they are a democracy or not.Measuring the state of democracy across the world helps us understand the extent to which people have political rights and freedoms.
\n\n\n\nBut measuring democracy comes with many challenges. People do not always agree on what characteristics define a democracy. These characteristics — such as whether an election was free and fair — are difficult to define and assess. The judgment of experts is to some degree subjective. They may disagree about a specific characteristic or how something as complex as a political system can be reduced into a single measure.
\n\n\n\nHow do researchers address these challenges and measure democracy?
\n\n\n\nIn some of our work on democracy, we rely on data published by the Varieties of Democracy (V-Dem) project.{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. V-Dem [Country-Year/Country-Date] Dataset v13. Varieties of Democracy (V-Dem) Project.
Pemstein, Daniel, Kyle L. Marquardt, Eitan Tzelgov, Yi-ting Wang, Juraj Medzihorsky, Joshua Krusell, Farhad Miri, and Johannes von Römer. 2023. The V-Dem Measurement Model: Latent Variable Analysis for Cross-National and Cross-Temporal Expert-Coded Data. V-Dem Working Paper No. 21. University of Gothenburg: Varieties of Democracy Institute.{/ref}
The project is managed by the V-Dem Institute, based at the University of Gothenburg in Sweden. It spans seven more regional centers around the world and is run by five principal investigators, dozens of project and regional managers, and more than 100 country coordinators.
\n\n\n\nV-Dem is funded through grants and donations by government agencies and private foundations, such as the Swedish Research Council, the European Commission, and the Marcus and Marianne Wallenberg Foundation.
\n\n\n\nTrue to its name, the Varieties of Democracy project acknowledges that democracy can be characterized differently, and measures electoral, liberal, participatory, deliberative, and egalitarian characterizations of democracy.
\n\n\n\nAt Our World in Data we primarily use V-Dem’s Electoral Democracy Index to measure democracy.{ref}The index is sometimes also called the Polyarchy Index.{/ref} The index is used in all of V-Dem’s other democracy indices because V-Dem considers there to be no democracy without elections. The other aspects can therefore be thought of as measuring the quality of a democracy.
\n\n\n\nV-Dem characterizes electoral democracy as a political system in which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed. More specifically, this means:
\n\n\n\nYou can find data on the other democracy indices, electoral democracy’s characteristics, and other derived measures in our Democracy Data Explorer.
\n\n\n\nThe Electoral Democracy Index scores each country on a spectrum, with some countries being more democratic than others.
\n\n\n\nThe spectrum ranges from 0 (‘highly undemocratic’) to 1 (‘highly democratic’).
\n\n\n\nThis scoring thereby differs from other approaches such as ‘Regimes of the World’ and other projects, which classify countries as a binary: either they are a democracy or not.
\n\n\n\n\n\n\n\n\n\n\n\nAs of version 13 of the dataset, V-Dem covers 202 countries, going back in time as far as 1789. Many countries have been covered since 1900, including before they became independent from their colonial powers.
\n\n\n\nTo actually measure what it wants to capture, V-Dem assesses the characteristics of democracy mostly through evaluations by experts.{ref} For more details, see: Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
\n\n\n\nThese anonymous experts are primarily academics and members of the media and civil society. They are also often nationals or residents of the country they assess, and therefore know its political system well and can evaluate aspects that are difficult to observe.
\n\n\n\nV-Dem’s own team of researchers supplements the expert evaluations. They code some easier-to-observe rules and laws of the political system, such as whether the legislature has a lower and upper house.
\n\n\n\nV-Dem uses several experts per country, year, and topic, to make its assessments less subjective. In total, around 3,500 country-experts fill surveys for V-Dem every year.
\n\n\n\nWhile there are fewer experts for small countries and for the time before 1900, they rely typically on 25 experts per country and 5 experts per topic.
\n\n\n\nV-Dem also works to make their coders’ assessments comparable across countries and time.
\n\n\n\nThe surveys ask the experts to answer very specific questions on completely explained scales about sub-characteristics of political systems — such as the presence or absence of election fraud — instead of making them rely on their broad impressions.
\n\n\n\nThe surveys are available in English, Arabic, French, Portuguese, Russian, and Spanish to reduce misunderstandings.
\n\n\n\nExperts further evaluate hypothetical countries, many coded several countries, and they denote their own uncertainty and personal demographic information.
\n\n\n\nV-Dem then uses this information to investigate expert biases, which they have found to be limited: they only find that experts from a country tend to be stricter in their assessments. {ref}“We have run extensive tests on how well such individual-level factors predict country-ratings but have found that the only factor consistently associated with country-ratings is country of origin (with “domestic” experts being harsher in their judgments).”
\n\n\n\nCoppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan Lindberg, Jan Teorell, Kyle Marquardt, Juraj Medzihorsky, Daniel Pemstein, Nazifa Alizada, Lisa Gastaldi, Garry Hindle, Josefine Pernes, Johannes von Römer, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2021. V-Dem Methodology v11.1. Varieties of Democracy (V-Dem) Project: page 24.{/ref}
\n\n\n\nV-Dem uses a statistical model to address any remaining differences between coders.{ref}Specifically, it uses a Bayesian Item-Response Theory estimation strategy.
\n\n\n\nMarquardt, Kyle, and Daniel Pemstein. 2018. IRT Models for Expert-Coded Panel Data. Political Analysis 26(4): 431-456.{/ref}
\n\n\n\nThe model combines the experts’ ratings of actual countries and hypothetical countries, as well as the experts’ stated uncertainties and personal demographics to produce best, upper-, and lower-bound estimates of many characteristics.{ref}Expressed precisely, V-Dem’s measurement model produces a probability distribution over the country-year scores. The best estimate is the distribution’s median, while the upper and lower bound estimates demarcate the interval in which the model places 68 percent of the probability mass.{/ref}
\n\n\n\nV-Dem provides these different estimates for all of its main and supplementary indices, including the Electoral Democracy Index and the subindices for free and fair elections, freedom of association, and freedom of expression.
\n\n\n\nWith the different estimates, V-Dem explicitly acknowledges that its coders can be uncertain or make errors in their measurement.
\n\n\n\nThe overall Electoral Democracy Index score is the result of weighing, multiplying, and adding up the subindices.{ref} The precise formula is:
\n\n\n\nelectoral democracy index = 0.5 * multiplicative polyarchy index + 0.5 * additive polyarchy index; with
\n\n\n\nmultiplicative polyarchy index = elected officials * free and fair elections * comprehensive suffrage * freedom of association * freedom of expression; and
\n\n\n\nadditive polyarchy index = 0.125 * elected officials + 0.25 * free and fair elections + 0.125 * comprehensive suffrage + 0.25 * freedom of association + 0.25 * freedom of expression{/ref}
\n\n\n\nThe subindices are weighted because V-Dem considers some of them as more important than others: elected officials and voting rights are weighted less because they capture more formal requirements, as opposed to free and fair elections and the freedoms of association and expression that rely more on expert assessments.
\n\n\n\nThe subindices are partially multiplied and partially added up because V-Dem wants the subindices to partially compensate for one another, and partially for them to reinforce each other. An example of compensation is voting rights partially making up for a lack of rights to assemble and protest, whereas an example of reinforcement is voting rights mattering more if voters can also choose opposition candidates.
\n\n\n\nV-Dem releases its data publicly, and makes it straightforward to download and use.
\n\n\n\nIt publishes the overall scores, the underlying subindices, and several hundred specific questions by country-year, country-date, and coder.
\n\n\n\nV-Dem also releases detailed descriptions of how they characterize democracy, the questions and coding procedures that guide the experts and researchers, as well as why it weighs, adds, and multiplies the scores for specific characteristics.
\n\n\n\nIn our work, we expand the years covered by V-Dem further.
\n\n\n\nTo expand the time coverage of today’s countries and include more of the period when they were still non-sovereign territories, we identified the historical entity they were a part of and used that regime’s data whenever available.{ref}For example, V-Dem only provides regime data since Bangladesh’s independence in 1971. There is, however, regime data for Pakistan and the colony of India, both of which the current territory of Bangladesh was a part. We, therefore, use the regime data of Pakistan for Bangladesh from 1947 to 1970, and the regime data of India from 1789 to 1946. We did so for all countries with a past or current population of more than one million.{/ref}
\n\n\n\nWe also calculated regional and global averages of the Electoral Democracy Index and its sub-indices, weighted and unweighted by population.
\n\n\n\nOur code and data are available on GitHub and record our revisions in detail.
\n\n\n\nV-Dem releases a new version of the data each year in March.
\n\n\n\nWe at Our World in Data aim to update our own data within a few weeks of the release.
\n\n\n\nThere are shortcomings in the way the Electoral Democracy Index characterizes and measures democracy.{ref}This and the following section draw on several very helpful other articles summarizing and reviewing some of the leading democracy datasets:
\n\n\n\nBoese, Vanessa. 2019. How (not) to measure democracy. International Area Studies Review 22(2): 95-127.
\n\n\n\nCoppedge, Michael, John Gerring, Staffan I. Lindberg, Svend-Erik Skaaning, and Jan Teorell. 2017. V-Dem Comparisons and Contrasts with Other Measurement Projects. V-Dem Working Paper 45.
\n\n\n\nMøller, Jørgen and Svend-Erik Skaaning. 2021. Varieties of Measurement: A Comparative Assessment of Relatively New Democracy Ratings based on Original Data. V-Dem Working Paper 123.Skaaning, Svend-Erik. 2018. Different Types of Data and the Validity of Democracy Measures. Politics and Governance 6(1): 105-116.{/ref}
\n\n\n\nThe index focuses on an electoral understanding of democracy and does not account for other characterizations, such as democracies as egalitarian political systems, in which political power is equally distributed to allow everyone to participate. This means that some of the most economically-unequal countries in the world, such as Brazil and South Africa, are classified as broadly democratic in recent years.{ref}True to its name, however, V-Dem provides several democracy indices in addition to the Electoral Democracy Index, and also measures liberal, participatory, deliberative, and egalitarian characterizations of democracy.{/ref}
\n\n\n\nV-Dem also does not cover some countries with very small populations.
\n\n\n\nFurthermore, the index is more difficult to interpret than other measures. Measures that group countries into democracies and autocracies, such as the Regimes of the World classification, make it possible to say which country was a democracy.
\n\n\n\nThe Electoral Democracy Index makes no clear assessment there, and only allows us to say whether a country is relatively democratic by comparing it to the range of the index, to other countries, or to the same country at another point in time. And when doing so, it is still difficult to say how large these differences are.{ref}This can be made easier by comparing how a score relates to the index’s overall distribution or its distribution for a specific year.{/ref}
\n\n\n\nThe assessment of the Electoral Democracy Index remains to some extent subjective. Its index is built on difficult evaluations by experts that rely less on easier-to-observe characteristics, such as whether regular elections are held.
\n\n\n\nFinally, the index’s aggregation remains to some extent arbitrary. It is unclear why these specific subindices were chosen; and why two subindices, elected officials and voting rights, are weighted less than the others.
\n\n\n\nDespite these shortcomings, the index tells us a lot about how democratic the world was in the past and today.
\n\n\n\nIts characterization of democracy as an electoral political system, in which citizens get to participate in free and fair elections, is commonly recognized as the basic principle of democracy and shared by all of the leading approaches of measuring democracy.
\n\n\n\nBecause it treats democracy as a spectrum, the index is able to capture both big and small differences in the political systems of countries, and to record small changes within countries over time. This allows us to observe whether one country is more democratic than another, or whether a country has become more or less democratic over time.
\n\n\n\nThe index also covers many countries and years. With the exception of microstates, it covers all countries in the world. Many countries are covered since 1900 — even while they were colonized by another country — and some of them as far back as 1789.
\n\n\n\nFinally, V-Dem takes many steps to make its assessments valid, precise, comparable across countries and time, and transparent. It relies on many country and subject experts answering detailed surveys to measure aspects of political systems that are often difficult to observe and acknowledges the remaining uncertainty in their assessments.
\n\n\n\nWhether V-Dem’s Electoral Democracy Index is a useful measure of democracy will depend on the questions we want to answer.
\n\n\n\nThe index will not give us a satisfying answer if we are interested in non-electoral understandings of democracy (or different understandings of electoral democracy); if we are also interested in the political systems of microstates; and only interested in big differences in the political systems of countries.
\n\n\n\nIn these cases, we will have to rely on other measures.
\n\n\n\nBut if we value a sophisticated measure based on the knowledge of many country experts and are interested in big and small differences in electoral democracy, within and across countries, and far into the past, we can learn a lot from this data.
\n\n\n\nIt is for these latter purposes we use the measure in some of our reporting on democracy.
\nI thank Edouard Mathieu, Hannah Ritchie, and Max Roser for their very helpful comments and ideas about how to improve this article.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""There are many ways to measure democracy. Here is how the Varieties of Democracy project does it, one of the leading sources of global democracy data."", ""protected"": false}, ""date_gmt"": ""2022-11-30T12:54:36"", ""modified"": ""2023-06-08T10:56:26"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre""], ""modified_gmt"": ""2023-06-08T09:56:26"", ""comment_status"": ""closed"", ""featured_media"": 52089, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/07/democratic_world-e1657029841780-150x112.png"", ""medium_large"": ""/app/uploads/2022/07/democratic_world-e1657029841780-768x572.png""}}" 54076,We just published a data explorer on Global Health,global-health-explorer-launch,post,publish,,"{""id"": ""wp-54076"", ""slug"": ""global-health-explorer-launch"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We just published a data explorer on Global Health"", ""authors"": [""Fiona Spooner"", ""Edouard Mathieu"", ""Hannah Ritchie""], ""excerpt"": ""What do people die from? How do health outcomes and systems vary across the world? Explore data for all countries in our data explorer."", ""dateline"": ""November 2, 2022"", ""subtitle"": ""What do people die from? How do health outcomes and systems vary across the world? Explore data for all countries in our data explorer."", ""sidebar-toc"": false, ""featured-image"": ""Global-Health-Explorer.png""}, ""createdAt"": ""2022-10-25T16:35:11.000Z"", ""published"": false, ""updatedAt"": ""2022-11-02T11:28:15.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-11-02T10:30:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2022-11-02 10:30:00,2024-02-16 14:22:54,,"[""Fiona Spooner"", ""Edouard Mathieu"", ""Hannah Ritchie""]",What do people die from? How do health outcomes and systems vary across the world? Explore data for all countries in our data explorer.,2022-10-25 16:35:11,2022-11-02 11:28:15,https://ourworldindata.org/wp-content/uploads/2022/10/Global-Health-Explorer.png,{},,"{""id"": 54076, ""date"": ""2022-11-02T10:30:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=54076""}, ""link"": ""https://owid.cloud/global-health-explorer-launch"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""global-health-explorer-launch"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We just published a data explorer on Global Health""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54076""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/53"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=54076"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=54076"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=54076"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=54076""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54076/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/54241"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 54079, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/54076/revisions/54079""}]}, ""author"": 53, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""What do people die from? How do health outcomes and systems vary across the world? Explore data for all countries in our data explorer."", ""protected"": false}, ""date_gmt"": ""2022-11-02T10:30:00"", ""modified"": ""2022-11-02T11:28:15"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Fiona Spooner"", ""Edouard Mathieu"", ""Hannah Ritchie""], ""modified_gmt"": ""2022-11-02T11:28:15"", ""comment_status"": ""closed"", ""featured_media"": 54241, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/10/Global-Health-Explorer-150x79.png"", ""medium_large"": ""/app/uploads/2022/10/Global-Health-Explorer-768x403.png""}}" 53807,We just published our new data explorer on Poverty,poverty-explorer-launch,post,publish,,"{""id"": ""wp-53807"", ""slug"": ""poverty-explorer-launch"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We just published our new data explorer on Poverty"", ""authors"": [""Joe Hasell"", ""Pablo Arriagada""], ""excerpt"": ""Explore data on poverty across the world, and how this is changing over time."", ""dateline"": ""October 18, 2022"", ""subtitle"": ""Explore data on poverty across the world, and how this is changing over time."", ""sidebar-toc"": false, ""featured-image"": ""chart.png""}, ""createdAt"": ""2022-10-17T12:18:26.000Z"", ""published"": false, ""updatedAt"": ""2022-10-18T10:52:13.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-10-18T09:58:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2022-10-18 09:58:00,2024-02-16 14:22:54,,"[""Joe Hasell"", ""Pablo Arriagada""]","Explore data on poverty across the world, and how this is changing over time.",2022-10-17 12:18:26,2022-10-18 10:52:13,https://ourworldindata.org/wp-content/uploads/2022/10/chart.png,{},,"{""id"": 53807, ""date"": ""2022-10-18T10:58:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=53807""}, ""link"": ""https://owid.cloud/poverty-explorer-launch"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""poverty-explorer-launch"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We just published our new data explorer on Poverty""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/53807""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/14"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=53807"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=53807"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=53807"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=53807""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/53807/revisions"", ""count"": 3}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/53946"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 53811, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/53807/revisions/53811""}]}, ""author"": 14, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore data on poverty across the world, and how this is changing over time."", ""protected"": false}, ""date_gmt"": ""2022-10-18T09:58:00"", ""modified"": ""2022-10-18T11:52:13"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Joe Hasell"", ""Pablo Arriagada""], ""modified_gmt"": ""2022-10-18T10:52:13"", ""comment_status"": ""closed"", ""featured_media"": 53946, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/10/chart-150x79.png"", ""medium_large"": ""/app/uploads/2022/10/chart-768x402.png""}}" 53796,Introducing our updated work on Poverty: a new design for our content,owid-entry-redesign,post,publish,"This redesign project was built by Joe Hasell, Matthew Conlen, Marwa Boukarim, Matthieu Bergel, Ike Saunders, Daniel Bachler, Pablo Arriagada, and Max Roser.
At Our World in Data, we present data and research in a number of ways: as individual charts, data explorers that allow users to switch between metrics, technical articles that explain what particular metrics mean, and articles that explain what the data tells us about the world to make progress.
We have more than 3,000 interactive charts spanning hundreds of topics.
One of our biggest challenges has been bringing all of this data and research together in a way that users can navigate easily. People should be able to quickly find exactly what they’re looking for. And be able to find it again when they come back.
To help solve this challenge, over the past year, our product and engineering teams have been working on the design of a new format for collecting and presenting all our data, research, and writing on particular topics.
We’ve now published the first example of this redesign for one topic – Poverty.
We plan to progressively roll this design out across the other topics on our site in the next year.
At Our World in Data, we strive to make data both easy to access and easy to understand. And we want to make the information that we’re presenting useful to a wide range of people, from the government official who wants to make a policy decision informed by the best available data to a curious reader who wants to understand what is going on in the world around them.
One of Our World in Data’s biggest contributions is that we provide important context – of methodological challenges, data quality, and key insights – to the data and research that we present. This redesign aims to make sure that these important insights and technical details are easily discoverable without overwhelming readers through information overload.
Our new topic pages seek to achieve this by adopting the structure of Ben Schneiderman’s Visual Information-Seeking Mantra, a tried and true approach to designing interactive, information-dense user interfaces: overview first, zoom and filter, then details-on-demand.
At the top of the new topic pages are key insights, a collection of the main takeaways of a given topic curated by expert authors for those that want an overview.
These insights combine data visualizations with short explanations that allow readers to quickly understand the most important aspects of the subject.
Next, interactive data explorers allow readers to engage with and manipulate the data, for example, by quickly switching between metrics and comparing countries.
These explorers enable readers to discover their own insights by filtering and zooming charts to find relevant subsets of the data and create the exact visualizations that they are looking for.
Finally, the most interested readers can dig into additional details: methodological explanations of the definitions of key metrics explain what they mean, and identify important points about data quality; a collection of research and writing provides links to in-depth articles that look at global problems and how we make progress on solutions; a visual gallery of charts provides quick navigation to all of the visualizations that we’ve made on the topic.
To continue to improve our work, we really appreciate your feedback on this new design. Do you find it easy to navigate? What do you find confusing? If you have comments or suggestions please get in touch at info@ourworldindata.org.
This redesign project was built by Joe Hasell, Matthew Conlen, Marwa Boukarim, Matthieu Bergel, Ike Saunders, Daniel Bachler, Pablo Arriagada, and Max Roser.
\n\nAt Our World in Data, we present data and research in a number of ways: as individual charts, data explorers that allow users to switch between metrics, technical articles that explain what particular metrics mean, and articles that explain what the data tells us about the world to make progress.
\n\n\n\nWe have more than 3,000 interactive charts spanning hundreds of topics.
\n\n\n\nOne of our biggest challenges has been bringing all of this data and research together in a way that users can navigate easily. People should be able to quickly find exactly what they’re looking for. And be able to find it again when they come back.
\n\n\n\nTo help solve this challenge, over the past year, our product and engineering teams have been working on the design of a new format for collecting and presenting all our data, research, and writing on particular topics.
\n\n\n\nWe’ve now published the first example of this redesign for one topic – Poverty.
\n\n\n\nWe plan to progressively roll this design out across the other topics on our site in the next year.
\n\n\nAt Our World in Data, we strive to make data both easy to access and easy to understand. And we want to make the information that we’re presenting useful to a wide range of people, from the government official who wants to make a policy decision informed by the best available data to a curious reader who wants to understand what is going on in the world around them.
\n\n\n\nOne of Our World in Data’s biggest contributions is that we provide important context – of methodological challenges, data quality, and key insights – to the data and research that we present. This redesign aims to make sure that these important insights and technical details are easily discoverable without overwhelming readers through information overload.
\n\n\n\nOur new topic pages seek to achieve this by adopting the structure of Ben Schneiderman’s Visual Information-Seeking Mantra, a tried and true approach to designing interactive, information-dense user interfaces: overview first, zoom and filter, then details-on-demand.
\nAt the top of the new topic pages are key insights, a collection of the main takeaways of a given topic curated by expert authors for those that want an overview.
\n\n\n\nThese insights combine data visualizations with short explanations that allow readers to quickly understand the most important aspects of the subject.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/10/Key-insights-block.png\"")
Next, interactive data explorers allow readers to engage with and manipulate the data, for example, by quickly switching between metrics and comparing countries.
\n\n\n\nThese explorers enable readers to discover their own insights by filtering and zooming charts to find relevant subsets of the data and create the exact visualizations that they are looking for.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/10/Explorers-block.png\"")
Finally, the most interested readers can dig into additional details: methodological explanations of the definitions of key metrics explain what they mean, and identify important points about data quality; a collection of research and writing provides links to in-depth articles that look at global problems and how we make progress on solutions; a visual gallery of charts provides quick navigation to all of the visualizations that we’ve made on the topic.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/10/Research-and-writing-block.png\"")
To continue to improve our work, we really appreciate your feedback on this new design. Do you find it easy to navigate? What do you find confusing? If you have comments or suggestions please get in touch at info@ourworldindata.org.
\n\n\n\nThe data in this article uses a previous release of the World Bank's poverty and inequality data in which incomes are expressed in 2011 international-$.
The World Bank has since updated its methods, and now measures incomes in 2017 international-$. As part of this change, the International Poverty Line used to measure extreme poverty has also been updated: from $1.90 (in 2011 prices) to $2.15 (in 2017 prices).
This has had little effect on our overall understanding of poverty and inequality around the world. But because of the change of units, many of the figures mentioned in this article will differ from the latest World Bank figures.
Read more about the World Bank's updated methodology
","{""id"": ""wp-53780"", ""slug"": ""world-bank-pip-data-still-in-2011-ppps-warning"", ""content"": {""toc"": [], ""body"": [{""text"": [{""text"": ""The World Bank has updated its poverty and inequality data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The data in this article uses a previous release of the World Bank's poverty and inequality data in which incomes are expressed in 2011 international-$."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The World Bank has since updated its methods, and now measures incomes in 2017 international-$. As part of this change, the International Poverty Line used to measure extreme poverty has also been updated: from $1.90 (in 2011 prices) to $2.15 (in 2017 prices)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This has had little effect on our overall understanding of poverty and inequality around the world. But because of the change of units, many of the figures mentioned in this article will differ from the latest World Bank figures."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/from-1-90-to-2-15-a-day-the-updated-international-poverty-line"", ""type"": ""prominent-link"", ""title"": """", ""description"": ""Read more about the World Bank's updated methodology"", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/explorers/poverty-explorer"", ""type"": ""prominent-link"", ""title"": ""Explore the latest World Bank data on poverty and inequality"", ""description"": """", ""parseErrors"": []}], ""type"": ""article"", ""title"": ""World Bank PIP data still in 2011 PPPs warning"", ""authors"": [null], ""dateline"": ""October 17, 2022"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2022-10-17T10:36:56.000Z"", ""published"": false, ""updatedAt"": ""2022-10-17T10:36:56.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-10-17T10:36:56.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag owid/help""}], ""numBlocks"": 6, ""numErrors"": 1, ""wpTagCounts"": {""heading"": 1, ""owid/help"": 1, ""paragraph"": 5, ""owid/prominent-link"": 2}, ""htmlTagCounts"": {""p"": 5, ""h4"": 1}}",2022-10-17 10:36:56,2024-02-16 14:23:03,,[null],,,2022-10-17 10:36:56,,{},"## The World Bank has updated its poverty and inequality data The data in this article uses a previous release of the World Bank's poverty and inequality data in which incomes are expressed in 2011 international-$. The World Bank has since updated its methods, and now measures incomes in 2017 international-$. As part of this change, the International Poverty Line used to measure extreme poverty has also been updated: from $1.90 (in 2011 prices) to $2.15 (in 2017 prices). This has had little effect on our overall understanding of poverty and inequality around the world. But because of the change of units, many of the figures mentioned in this article will differ from the latest World Bank figures. ### Read more about the World Bank's updated methodology https://ourworldindata.org/from-1-90-to-2-15-a-day-the-updated-international-poverty-line ### Explore the latest World Bank data on poverty and inequality https://ourworldindata.org/explorers/poverty-explorer","{""data"": {""wpBlock"": {""content"": ""\tThe data in this article uses a previous release of the World Bank’s poverty and inequality data in which incomes are expressed in 2011 international-$.
\n\n\n\nThe World Bank has since updated its methods, and now measures incomes in 2017 international-$. As part of this change, the International Poverty Line used to measure extreme poverty has also been updated: from $1.90 (in 2011 prices) to $2.15 (in 2017 prices).
\n\n\n\nThis has had little effect on our overall understanding of poverty and inequality around the world. But because of the change of units, many of the figures mentioned in this article will differ from the latest World Bank figures.
\n\n\nRead more about the World Bank’s updated methodology
\n\nGlobally, seasonal influenza kills 400,000 people from respiratory disease each year on average. During large flu pandemics, when influenza strains evolved substantially, the death toll was even higher.
But the risk of dying from influenza has declined substantially over time from improvements in sanitation, healthcare, and vaccination.
People born in 1940 had around a third of the risk of dying from influenza as those born in 1900 – even when they reached the same age. This decline continued, and those born in 1980 have a risk of half that of those born in 1940.
Influenza still remains a large burden around the world, because of an aging population and a lack of access to healthcare and sanitation in many countries.
In this article, we look into these developments in detail: how many people die from seasonal influenza and how this has changed over time.
We will also look at which factors increase the risk of dying from the flu and understand why, in some years, influenza has led to large pandemics that caused millions of deaths. This knowledge can inform us about the risks of influenza in the future.
Although this is an important question, it is often difficult to answer.
Despite being a well-understood disease, it can be hard to count the number of deaths from influenza for several reasons.{ref}Gordon, A., & Reingold, A. (2018). The Burden of Influenza: A Complex Problem. Current Epidemiology Reports, 5(1), 1–9. https://doi.org/10.1007/s40471-018-0136-1{/ref}
One problem is that the symptoms of influenza look similar to other infections, such as respiratory syncytial virus and rhinovirus. In many countries, only a fraction of patients with an ""influenza-like illness"" are tested to confirm whether they were infected by the virus.{ref}Charbonneau, D. H., & James, L. N. (2019). FluView and FluNet: Tools for Influenza Activity and Surveillance. Medical Reference Services Quarterly, 38(4), 358–368. https://doi.org/10.1080/02763869.2019.1657734{/ref} This means we miss many – or, in some countries, most – infections.
Another problem is that influenza can lead to death in a number of indirect ways. It can cause death through respiratory complications such as pneumonia, but also from cardiovascular complications such as heart attacks and strokes, or other serious infections. This is especially true for the elderly and people who have chronic health conditions.{ref}Macias, A. E., McElhaney, J. E., Chaves, S. S., Nealon, J., Nunes, M. C., Samson, S. I., Seet, B. T., Weinke, T., & Yu, H. (2021). The disease burden of influenza beyond respiratory illness. Vaccine, 39, A6–A14. https://doi.org/10.1016/j.vaccine.2020.09.048{/ref} Without accounting for these deaths, we would underestimate the number of flu deaths.
To overcome this, researchers estimate the burden of influenza with other methods. They can estimate the number of excess deaths that occur during flu seasons, and use routine surveillance data and mortality records, to estimate how many of these are caused by the flu.
The annual mortality caused by seasonal influenza was estimated by the Global Pandemic Mortality Project II using data between 2002 and 2011. They estimated that, during this period, seasonal influenza caused between 294,000 and 518,000 deaths each year globally.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. https://jogh.org/documents/issue201902/jogh-09-020421.pdf
This shows the mean estimate of annual influenza mortality between 2002–2011, excluding the 2009 ""Swine flu"" pandemic influenza season. You can find estimated numbers for world regions in Table 2 of the paper. Rates for other age groups can be found here: https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf
Other global estimates of seasonal influenza mortality have been made by the Institute of Health Metrics and Evaluation (IHME) and the Centers for Disease Control and Prevention (CDC).
Estimates made by GLaMOR were comparable to those by the CDC, while estimates made by IHME were around 4-5 times lower. This may be because the IHME estimated influenza mortality by first estimating the number of deaths caused by lower respiratory diseases, then estimating the fraction of those that were primarily attributed to influenza in vital records, verbal autopsies, and other mortality data. This approach would have missed many deaths caused by complications of influenza and deaths that were not specified on records to be caused by influenza due to limited testing.
Both CDC and GLaMOR's models are also likely to be underestimating the total mortality burden from influenza, as they only use data from respiratory-associated deaths. While this would include deaths caused by influenza that had, for example, influenza listed as a secondary cause of death on death certificates, it would miss some that were caused by influenza but attributed to another cause like cardiovascular disease. Had these models used all-cause mortality to estimate deaths caused by influenza, they would have been more sensitive (captured more deaths caused by influenza) but also less specific (captured more deaths caused by other diseases that could not be distinguished easily).
In comparison to CDC estimates, GLaMOR used many country-specific indicators in order to extrapolate seasonal influenza mortality to countries that did not provide weekly or monthly influenza mortality records or influenza surveillance data, while the CDC extrapolated this using mainly the WHO Global Health Estimates of respiratory mortality.{/ref}
These estimates focus on deaths where people had respiratory disease. This means they miss some flu deaths, as some people may die from cardiovascular complications of the flu without having respiratory disease.{ref}The global number of people who die from other complications of the flu is unclear.
Paget et al. (the authors of the GLaMOR project) state in their paper that their estimate “does not cover cardiovascular deaths, something that could at least double the estimate of influenza-associated deaths.”
In recent meta-analyses, Behrouzi et al. found that influenza vaccination reduces the chances of major cardiovascular events (such as heart attacks and strokes) by around 34%, in clinical trials of the elderly.
This suggests the death toll from other complications could be large. However, global estimates have not been made of these types of deaths from flu.
Paget, J., Danielle Iuliano, A., Taylor, R. J., Simonsen, L., Viboud, C., & Spreeuwenberg, P. (2022). Estimates of mortality associated with seasonal influenza for the European Union from the GLaMOR project. Vaccine, 40(9), 1361–1369. https://doi.org/10.1016/j.vaccine.2021.11.080
Behrouzi, B., Bhatt, D. L., Cannon, C. P., Vardeny, O., Lee, D. S., Solomon, S. D., & Udell, J. A. (2022). Association of Influenza Vaccination With Cardiovascular Risk: A Meta-analysis. JAMA Network Open, 5(4), e228873. https://doi.org/10.1001/jamanetworkopen.2022.8873{/ref}
On the map, you can see the estimates of flu mortality shown as a rate per 100,000 people, among people aged over 65.
In Europe, the rate of deaths from the flu was 30.8 per 100,000 each year, among those aged over 65. This is more than three times the risk from traffic accidents, which kill 9 per 100,000, in the same age group.{ref}Eurostat. (2022). Causes of death—Standardised death rate. European Commission. https://ec.europa.eu/eurostat/databrowser/view/HLTH_CD_ASDR2__custom_3500876/default/table?lang=en
Transport accidents are counted under (V01–V99, Y85) in the ICD-10. Across 27 EU countries, these rates were 9.02, 9.15, and 8.77 per 100,000 people aged over 65 in 2015, 2016 and 2017 respectively.
Estimates of the rate of death from influenza are much lower in the ICD-10 since they only consider deaths where influenza is listed as the cause of death on death certificates, while the estimates we show above also include those that are caused by flu indirectly. This means ICD-10 death rates are likely to be highly underestimated for the flu. However, deaths caused by traffic accidents are more likely to be listed as the primary cause of death on death certificates and are much less underestimated by death certificate data.{/ref}
In low-income countries, these estimates tend to be less certain, due to lower levels of testing for influenza and limited mortality records.
But flu is estimated to be more deadly in countries in South America, Africa, and South Asia than in Europe and North America. For example, Indonesia has more than twice the death rate of Canada. These disparities are at least partly due to poverty, poorer underlying health, and lower access to healthcare.
Death rates from influenza are much lower than they were in the past.
We don’t have good long-term estimates of flu deaths across most countries. But, researchers have produced weekly estimates of influenza deaths over long periods in the United States. This allows us to see how modern rates compare to the past. You can see this data in the chart.{ref}Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y
From 1960 to 2015, the number of deaths from influenza was estimated using a Serfling model, which estimates the excess number of deaths during flu seasons using data from the rest of the year and accounting for changes that occur year by year. Since 1997, there has also been routine testing for ""influenza-like illnesses"" in hospitals to determine the share of them that are actually caused by influenza, rather than other diseases. Therefore, estimates from 1997 to 2015 were also calculated using a Serfling-surveillance model, which accounted for the share of tests that were positive for influenza. This also validates the estimates from the regular Serfling model. In addition, deaths among children aged under 5 are excluded in both models, as they would be likely to include deaths from respiratory syncytial virus.
Mortality in the US was slightly lower in during the 2009 Swine flu pandemic season than usual flu seasons, as severe disease shifted away from the elderly to young and middle-aged adults. However, the 2009 Swine flu pandemic led to more deaths than regular flu seasons in other countries such as Mexico.
Gagnon, A., Acosta, E., Hallman, S., Bourbeau, R., Dillon, L. Y., Ouellette, N., Earn, D. J. D., Herring, D. A., Inwood, K., Madrenas, J., & Miller, M. S. (2018). Pandemic Paradox: Early Life H2N2 Pandemic Influenza Infection Enhanced Susceptibility to Death during the 2009 H1N1 Pandemic. MBio, 9(1), e02091-17. https://doi.org/10.1128/mBio.02091-17{/ref}
Deaths from influenza fluctuate across the year, with large peaks in the winter.{ref}Influenza viruses are thought to transmit more efficiently during the winter due to lower temperatures and humidity. But in many tropical countries, flu epidemics coincide with warm rainy seasons, so the trends may have more causes. Other explanations include seasonal changes in human immunity or changes in human behavior, such as more indoor mixing and crowding. Petrova, V. N., & Russell, C. A. (2018). The evolution of seasonal influenza viruses. Nature Reviews Microbiology, 16(1), 47–60. https://doi.org/10.1038/nrmicro.2017.118{/ref} The total number of deaths from influenza has been roughly stable in the United States over the last 65 years. You can see this in the top panel of the chart.
However, a large part of this is due to the fact that the population has been growing and aging.
If we look at death rates within age groups, the rate of deaths from influenza has been falling. You can see this in the bottom panel, which accounts for changes in the size and age structure of the population.
This means the likelihood that someone dies from influenza at a given age has declined over time. But, because the population is getting larger and older, the total number of deaths has remained stable.
Influenza mortality has declined over several generations. We know this from historical data from the United States, which has been used to estimate ""cohort effects"". This tells us whether people who were born more recently have lower risks of death, after accounting for their younger age.
Since 1900, there has been a long-term decline in the risk of dying from the flu.{ref} Between 1860 and 1900, there was a slight increase in the risk of death from influenza, which may have been due to worsening health conditions as more people moved into crowded urban areas.
Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y{/ref} There are several reasons for this.
One is that there were large projects to improve sanitation in cities across the United States in the early 1900s.{ref}Cutler, D., & Miller, G. (2005). The role of public health improvements in health advances: The twentieth-century United States. Demography, 42(1), 1–22. https://doi.org/10.1353/dem.2005.0002{/ref} Over the twentieth century, there were also improvements in neonatal healthcare and increases in the rate of childhood vaccinations. All of these factors had benefits that carried forward as people aged: they protected people from developing comorbidities that increased the risk of dying from influenza.
There has also been an increase in the rate of flu vaccinations. Influenza vaccines were developed for the first time in the 1930s and 1940s. In 1952, the World Health Organization began a surveillance system to monitor which flu strains were circulating worldwide. This helped researchers develop new vaccines each year that matched those strains.{ref}Barberis, I., Myles, P., Ault, S. K., Bragazzi, N. L., & Martini, M. (2016). History and evolution of influenza control through vaccination: From the first monovalent vaccine to universal vaccines. Journal of Preventive Medicine and Hygiene, 57(3), E115–E120. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5139605/ {/ref} Over the following decades, the rate of influenza vaccinations among the elderly began to grow.{ref}Centers for Disease Control and Prevention, & National Center for Immunization and Respiratory Diseases. (2021). Historical Reference of Seasonal Influenza Vaccine Doses Distributed. https://www.cdc.gov/flu/prevent/vaccine-supply-historical.htm{/ref}
The historical decline in influenza mortality has been substantial, as you can see in the chart.
Even when they reached the same age, people born in 1940 had around a third of the risk of dying from influenza as those born in 1900. This decline continued, and those born in 1980 had a risk of half that of those born in 1940.{ref}These estimates are more variable for young age groups, particularly those born after 1990 because the number of influenza deaths to make comparisons were low, especially as deaths in children under 5 were excluded from the models, to avoid counting deaths from respiratory syncytial virus.
Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y{/ref}
Age is a major risk factor of dying from the flu. As you can see in the chart, infants and the elderly tend to have a much higher risk of death from a range of respiratory diseases, including influenza, compared to young adults. For example, 60-year-olds have a ten times greater risk of death from influenza than 20-year-olds.{ref}Metcalf, C. J. E., Paireau, J., O’Driscoll, M., Pivette, M., Hubert, B., Pontais, I., Nickbakhsh, S., Cummings, D. A. T., Cauchemez, S., & Salje, H. (2022). Comparing the age and sex trajectories of SARS-CoV-2 morbidity and mortality with other respiratory pathogens. Royal Society Open Science, 9(6), 211498. https://doi.org/10.1098/rsos.211498
In this post, we show the relative risks of death. This is because it's more difficult to give an absolute risk of death from influenza at different ages, because mortality rates vary widely over time and between countries, as we saw earlier. Absolute risks depend on how many people are infected during a flu season, the availability of healthcare, the rates of vaccination, and so on. However, the relative risk of death – ratio between the risk of death in one age group versus another – tends to be more consistent.{/ref}
Once someone reaches their twenties, their mortality risk from the flu increases exponentially. This shape follows the risk of death from all causes.{ref}The shape of this age–mortality curve is often described by the Gompertz function. Olshansky, S. J., & Carnes, B. A. (1997). Ever since gompertz. Demography, 34(1), 1-15. https://link.springer.com/content/pdf/10.2307/2061656.pdf {/ref}
The risk of dying from influenza also depends on other factors such as the quality of healthcare, the strain of influenza, and whether the person received the flu vaccine.{ref}Belongia, E. A., Simpson, M. D., King, J. P., Sundaram, M. E., Kelley, N. S., Osterholm, M. T., & McLean, H. Q. (2016). Variable influenza vaccine effectiveness by subtype: A systematic review and meta-analysis of test-negative design studies. The Lancet Infectious Diseases, 16(8), 942–951. https://doi.org/10.1016/S1473-3099(16)00129-8{/ref}
Every year, influenza vaccines are reformulated to match the strains of flu that are expected to dominate during the winter. When there is a mismatch between the strains in the vaccine and the flu strains that are circulating, the vaccines tend to have lower efficacy, and flu seasons tend to be more severe.{ref}Tricco, A. C., Chit, A., Soobiah, C., Hallett, D., Meier, G., Chen, M. H., Tashkandi, M., Bauch, C. T., & Loeb, M. (2013). Comparing influenza vaccine efficacy against mismatched and matched strains: A systematic review and meta-analysis. BMC Medicine, 11(1), 153. https://doi.org/10.1186/1741-7015-11-153{/ref}
Some seasons are far more severe than usual seasonal influenza. This tends to occur when new influenza strains arise and cause influenza pandemics.
Over time, influenza viruses that are circulating in the population tend to mutate through a process called ""antigenic drift"". This gives them the ability to evade people's immunity.
But, influenza viruses can also evolve with large and sudden changes. This happens in a process called “antigenic shift”, when parts of different strains combine with each other. These new combinations can be more infectious and lethal than previous strains, leading to deadlier pandemics.
For example, the Spanish flu evolved from a combination of human influenza and another animal influenza, which formed a new H1N1 influenza virus. As you can see in the chart, it caused the largest influenza pandemic in history: research by Spreeuwenberg et al. (2018) suggests that the Spanish flu killed around 17.4 milion people. Other estimates are even higher: Johnson and Mueller (2002) suggest that the Spanish flu killed between 50 to 100 million people.{ref}“Reassessing the Global Mortality Burden of the 1918 Influenza Pandemic”. American Journal of Epidemiology. 187 (12): 2561–2567. doi:10.1093/aje/kwy191. PMID 30202996. Online here.
Johnson, N. P., & Mueller, J. (2002). Updating the accounts: Global mortality of the 1918-1920"" Spanish"" influenza pandemic. Bulletin of the History of Medicine, 105–115.{/ref}
This death toll massively exceeds the number who die in a typical year from the flu – it is between 30 to 340 times higher than the estimate of 294,000 to 518,000 deaths that are caused by seasonal influenza each year, even though the global population was much smaller at the time.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. https://pubmed.ncbi.nlm.nih.gov/31673337/ {/ref}
Mortality during the Spanish flu rose sharply in young adults, compared to previous seasons. Research suggests that this is because they lacked immunity to H1 influenza viruses because they had been exposed to different influenza strains in their childhood. In contrast, older generations had been exposed to similar H1 influenza viruses decades before the pandemic began, which gave them some protection from the H1N1 pandemic strain.{ref}Worobey, M., Han, G.-Z., & Rambaut, A. (2014). Genesis and pathogenesis of the 1918 pandemic H1N1 influenza A virus. Proceedings of the National Academy of Sciences, 111(22), 8107–8112. https://doi.org/10.1073/pnas.1324197111
Gagnon, A., Miller, M. S., Hallman, S. A., Bourbeau, R., Herring, D. A., Earn, D. J. D., & Madrenas, J. (2013). Age-specific mortality during the 1918 influenza pandemic: Unravelling the mystery of high young adult mortality. PloS One, 8(8), e69586. https://doi.org/10.1371/journal.pone.0069586
Luk, J., Gross, P., & Thompson, W. W. (2001). Observations on Mortality during the 1918 Influenza Pandemic. Clinical Infectious Diseases, 33(8), 1375–1378. https://doi.org/10.1086/322662
Ma, J., Dushoff, J., & Earn, D. J. D. (2011). Age-specific mortality risk from pandemic influenza. Journal of Theoretical Biology, 288, 29–34. https://doi.org/10.1016/j.jtbi.2011.08.003{/ref}
Seasonal flu causes 400,000 respiratory deaths each year on average. But the burden is far lower than it was in the past, due to improvements in sanitation, healthcare, and vaccination.
The flu also remains a large burden around the world for two major reasons. One is that many people around the world still lack access to healthcare and have low rates of influenza vaccination, which increases the risk of death.
Another reason is that the populations of many countries have been aging rapidly. In lower-income countries, the flu could become a larger burden as they face aging populations in the future.
To tackle this risk, the world can take lessons from how the burden has been reduced in the past. One way is to increase the rates of influenza vaccination, as well as other routine vaccinations, which also reduce the risk that flu is severe. Another is to improve sanitation and access to healthcare around the world.
We've already seen a huge decline in the burden of the flu over many decades, and with greater efforts, we could see that burden decline even further.
Keep reading on Our World in Data:
Acknowledgments: Hannah Ritchie, Max Roser and Edouard Mathieu provided very helpful guidance and comments that helped improve this post.
","{""id"": ""wp-53550"", ""slug"": ""influenza-deaths"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Globally, seasonal influenza kills 400,000 people from respiratory disease each year on average. During large flu pandemics, when influenza strains evolved substantially, the death toll was even higher."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But the risk of dying from influenza has declined substantially over time from improvements in sanitation, healthcare, and vaccination. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""People born in 1940 had around a third of the risk of dying from influenza as those born in 1900 – even when they reached the same age. This decline continued, and those born in 1980 have a risk of half that of those born in 1940."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Influenza still remains a large burden around the world, because of an aging population and a lack of access to healthcare and sanitation in many countries. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this article, we look into these developments in detail: how many people die from seasonal influenza and how this has changed over time. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We will also look at which factors increase the risk of dying from the flu and understand why, in some years, influenza has led to large pandemics that caused millions of deaths. This knowledge can inform us about the risks of influenza in the future."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How many people die from seasonal influenza?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Although this is an important question, it is often difficult to answer."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Despite being a well-understood disease, it can be hard to count the number of deaths from influenza for several reasons.{ref}Gordon, A., & Reingold, A. (2018). The Burden of Influenza: A Complex Problem. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Current Epidemiology Reports"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""5"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(1), 1–9."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1007/s40471-018-0136-1"", ""children"": [{""text"": "" https://doi.org/10.1007/s40471-018-0136-1"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One problem is that the symptoms of influenza look similar to other infections, such as respiratory syncytial virus and rhinovirus. In many countries, only a fraction of patients with an \""influenza-like illness\"" are tested to confirm whether they were infected by the virus.{ref}Charbonneau, D. H., & James, L. N. (2019). FluView and FluNet: Tools for Influenza Activity and Surveillance. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Medical Reference Services Quarterly"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""38"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(4), 358–368."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1080/02763869.2019.1657734"", ""children"": [{""text"": "" https://doi.org/10.1080/02763869.2019.1657734"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} This means we miss many – or, in some countries, most – infections."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Another problem is that influenza can lead to death in a number of indirect ways. It can cause death through respiratory complications such as pneumonia, but also from cardiovascular complications such as heart attacks and strokes, or other serious infections. This is especially true for the elderly and people who have chronic health conditions.{ref}Macias, A. E., McElhaney, J. E., Chaves, S. S., Nealon, J., Nunes, M. C., Samson, S. I., Seet, B. T., Weinke, T., & Yu, H. (2021). The disease burden of influenza beyond respiratory illness. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Vaccine"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""39"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", A6–A14."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/j.vaccine.2020.09.048"", ""children"": [{""text"": "" "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://doi.org/10.1016/j.vaccine.2020.09.048%7B/ref"", ""children"": [{""text"": ""https://doi.org/10.1016/j.vaccine.2020.09.048"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} Without accounting for these deaths, we would underestimate the number of flu deaths."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To overcome this, researchers estimate the burden of influenza with other methods. They can estimate the number of "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""excess"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" deaths that occur during flu seasons, and use routine surveillance data and mortality records, to estimate how many of these are caused by the flu."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The annual mortality caused by seasonal influenza was estimated by the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nivel.nl/en/project/glamor-ii-global-influenza-mortality-research"", ""children"": [{""text"": ""Global Pandemic Mortality Project II"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" using data between 2002 and 2011. They estimated that, during this period, seasonal influenza caused between 294,000 and 518,000 deaths each year globally.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://jogh.org/documents/issue201902/jogh-09-020421.pdf"", ""children"": [{""text"": ""https://jogh.org/documents/issue201902/jogh-09-020421.pdf"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""This shows the mean estimate of annual influenza mortality between 2002–2011, excluding the 2009 \""Swine flu\"" pandemic influenza season. You can find estimated numbers for world regions in Table 2 of the paper. Rates for other age groups can be found here: "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf"", ""children"": [{""text"": ""https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""Other global estimates of seasonal influenza mortality have been made by the Institute of Health Metrics and Evaluation (IHME) and the Centers for Disease Control and Prevention (CDC). "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Estimates made by GLaMOR were comparable to those by the CDC, while estimates made by IHME were around 4-5 times lower. This may be because the IHME estimated influenza mortality by first estimating the number of deaths caused by lower respiratory diseases, then estimating the fraction of those that were primarily attributed to influenza in vital records, verbal autopsies, and other mortality data. This approach would have missed many deaths caused by complications of influenza and deaths that were not specified on records to be caused by influenza due to limited testing."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Both CDC and GLaMOR's models are also likely to be underestimating the total mortality burden from influenza, as they only use data from respiratory-associated deaths. While this "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""would"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" include deaths caused by influenza that had, for example, influenza listed as a secondary cause of death on death certificates, it "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""would"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""children"": [{""text"": ""miss"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" some that were caused by influenza but attributed to another cause like cardiovascular disease. Had these models used all-cause mortality to estimate deaths caused by influenza, they would have been more sensitive (captured more deaths caused by influenza) but also less specific (captured more deaths caused by other diseases that could not be distinguished easily)."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""In comparison to CDC estimates, GLaMOR used many country-specific indicators in order to extrapolate seasonal influenza mortality to countries that did not provide weekly or monthly influenza mortality records or influenza surveillance data, while the CDC extrapolated this using mainly the WHO Global Health Estimates of respiratory mortality.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""These estimates focus on deaths where people had respiratory disease. This means they miss some flu deaths, as some people may die from cardiovascular complications of the flu without having respiratory disease.{ref}The global number of people who die from other complications of the flu is unclear. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Paget et al. (the authors of the GLaMOR project) state in their paper that their estimate “does not cover cardiovascular deaths, something that could at least double the estimate of influenza-associated deaths.”"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In recent meta-analyses, Behrouzi et al. found that influenza vaccination reduces the chances of major cardiovascular events (such as heart attacks and strokes) by around 34%, in clinical trials of the elderly."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This suggests the death toll from other complications could be large. However, global estimates have not been made of these types of deaths from flu."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Paget, J., Danielle Iuliano, A., Taylor, R. J., Simonsen, L., Viboud, C., & Spreeuwenberg, P. (2022). Estimates of mortality associated with seasonal influenza for the European Union from the GLaMOR project. Vaccine, 40(9), 1361–1369. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/j.vaccine.2021.11.080"", ""children"": [{""text"": ""https://doi.org/10.1016/j.vaccine.2021.11.080"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Behrouzi, B., Bhatt, D. L., Cannon, C. P., Vardeny, O., Lee, D. S., Solomon, S. D., & Udell, J. A. (2022). Association of Influenza Vaccination With Cardiovascular Risk: A Meta-analysis. JAMA Network Open, 5(4), e228873. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1001/jamanetworkopen.2022.8873"", ""children"": [{""text"": ""https://doi.org/10.1001/jamanetworkopen.2022.8873"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""On the map, you can see the estimates of flu mortality shown as a rate per 100,000 people, among people aged over 65. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In Europe, the rate of deaths from the flu was 30.8 per 100,000 each year, among those aged over 65. This is more than three times the risk from traffic accidents, which kill 9 per 100,000, in the same age group.{ref}Eurostat. (2022). "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Causes of death—Standardised death rate"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". European Commission."", ""spanType"": ""span-simple-text""}, {""url"": ""https://ec.europa.eu/eurostat/databrowser/view/HLTH_CD_ASDR2__custom_3500876/default/table?lang=en"", ""children"": [{""text"": "" https://ec.europa.eu/eurostat/databrowser/view/HLTH_CD_ASDR2__custom_3500876/default/table?lang=en"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""text"": ""Transport accidents are counted under (V01–V99, Y85) in the ICD-10. Across 27 EU countries, these rates were 9.02, 9.15, and 8.77 per 100,000 people aged over 65 in 2015, 2016 and 2017 respectively."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Estimates of the rate of death from influenza are much lower in the ICD-10 since they only consider deaths where influenza is listed as the cause of death on death certificates, while the estimates we show above also include those that are caused by flu indirectly. This means ICD-10 death rates are likely to be highly underestimated for the flu. However, deaths caused by traffic accidents are more likely to be listed as the primary cause of death on death certificates and are much less underestimated by death certificate data.{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In low-income countries, these estimates tend to be less certain, due to lower levels of testing for influenza and limited mortality records. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But flu is estimated to be more deadly in countries in South America, Africa, and South Asia than in Europe and North America. For example, Indonesia has more than twice the death rate of Canada. These disparities are at least partly due to poverty, poorer underlying health, and lower access to healthcare."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/annual-mortality-rate-from-seasonal-influenza-ages-65?tab=map&country=South-East+Asia~OWID_WRL~Western+Pacific~Eastern+Mediterranean~Africa~Europe~Americas"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""What did influenza mortality look like in the past?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Death rates from influenza are much lower than they were in the past."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We don’t have good long-term estimates of flu deaths across most countries. But, researchers have produced weekly estimates of influenza deaths over long periods in the United States. This allows us to see how modern rates compare to the past. You can see this data in the chart.{ref}Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Demography"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""56"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(5), 1723–1746."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1007/s13524-019-00809-y"", ""children"": [{""text"": "" https://doi.org/10.1007/s13524-019-00809-y"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""From 1960 to 2015, the number of deaths from influenza was estimated using a Serfling model, which estimates the excess number of deaths during flu seasons using data from the rest of the year and accounting for changes that occur year by year. Since 1997, there has also been routine testing for \""influenza-like illnesses\"" in hospitals to determine the share of them that are actually caused by influenza, rather than other diseases. Therefore, estimates from 1997 to 2015 were also calculated using a Serfling-surveillance model, which accounted for the share of tests that were positive for influenza. This also validates the estimates from the regular Serfling model. In addition, deaths among children aged under 5 are excluded in both models, as they would be likely to include deaths from respiratory syncytial virus."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Mortality in the US was slightly lower in during the 2009 Swine flu pandemic season than usual flu seasons, as severe disease shifted away from the elderly to young and middle-aged adults. However, the 2009 Swine flu pandemic led to more deaths than regular flu seasons in other countries such as Mexico. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Gagnon, A., Acosta, E., Hallman, S., Bourbeau, R., Dillon, L. Y., Ouellette, N., Earn, D. J. D., Herring, D. A., Inwood, K., Madrenas, J., & Miller, M. S. (2018). Pandemic Paradox: Early Life H2N2 Pandemic Influenza Infection Enhanced Susceptibility to Death during the 2009 H1N1 Pandemic. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""MBio"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""9"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(1), e02091-17. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1128/mBio.02091-17"", ""children"": [{""text"": ""https://doi.org/10.1128/mBio.02091-17"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Deaths from influenza fluctuate across the year, with large peaks in the winter.{ref}Influenza viruses are thought to transmit more efficiently during the winter due to lower temperatures and humidity. But in many tropical countries, flu epidemics coincide with warm rainy seasons, so the trends may have more causes. Other explanations include seasonal changes in human immunity or changes in human behavior, such as more indoor mixing and crowding. Petrova, V. N., & Russell, C. A. (2018). The evolution of seasonal influenza viruses. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Nature Reviews Microbiology"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""16"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(1), 47–60."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1038/nrmicro.2017.118"", ""children"": [{""text"": "" https://doi.org/10.1038/nrmicro.2017.118"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} The "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""total"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" number of deaths from influenza has been roughly stable in the United States over the last 65 years. You can see this in the top panel of the chart."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""However, a large part of this is due to the fact that the population has been growing and aging."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If we look at death rates "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""within"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" age groups, the rate of deaths from influenza has been falling. You can see this in the bottom panel, which accounts for changes in the size and age structure of the population. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This means the likelihood that someone dies from influenza "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""at a given age"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" has declined over time. But, because the population is getting larger and older, the total number of deaths has remained stable."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Historical-influenza-deaths-USA-3-panels-1.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Why has influenza mortality declined over time?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Influenza mortality has declined over several generations. We know this from historical data from the United States, which has been used to estimate \""cohort effects\"". This tells us whether people who were born more recently have lower risks of death, after accounting for their younger age."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Since 1900, there has been a long-term decline in the risk of dying from the flu.{ref} Between 1860 and 1900, there was a slight increase in the risk of death from influenza, which may have been due to worsening health conditions as more people moved into crowded urban areas."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1007/s13524-019-00809-y"", ""children"": [{""text"": ""https://doi.org/10.1007/s13524-019-00809-y"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} There are several reasons for this. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One is that there were large projects to improve sanitation in cities across the United States in the early 1900s.{ref}Cutler, D., & Miller, G. (2005). The role of public health improvements in health advances: The twentieth-century United States. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Demography"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""42"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(1), 1–22."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1353/dem.2005.0002"", ""children"": [{""text"": "" "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://doi.org/10.1353/dem.2005.0002%7B/ref"", ""children"": [{""text"": ""https://doi.org/10.1353/dem.2005.0002"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} Over the twentieth century, there were also improvements in neonatal healthcare and increases in the rate of "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/vaccination"", ""children"": [{""text"": ""childhood vaccinations"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". All of these factors had benefits that carried forward as people aged: they protected people from developing comorbidities that increased the risk of dying from influenza."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There has also been an increase in the rate of flu vaccinations. Influenza vaccines were developed for the first time in the 1930s and 1940s. In 1952, the World Health Organization began a surveillance system to monitor which flu strains were circulating worldwide. This helped researchers develop new vaccines each year that matched those strains.{ref}Barberis, I., Myles, P., Ault, S. K., Bragazzi, N. L., & Martini, M. (2016). History and evolution of influenza control through vaccination: From the first monovalent vaccine to universal vaccines. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Journal of Preventive Medicine and Hygiene"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""57"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(3), E115–E120. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5139605/"", ""children"": [{""text"": ""https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5139605/"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref} Over the following decades, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/influenza-vaccination-rate?country=NZL~GRC~ISR~TUR~LVA~POL~EST~SVK~AUT~HUN~CZE~LTU~SVN~LUX~DEU~FIN~ISL~BEL~FRA~SWE~CRI~AUS~CAN~CHL~NLD~MEX~GBR~USA~CHE~KOR~ESP~IRL~ITA~JPN~NOR~PRT~DNK"", ""children"": [{""text"": ""the rate of influenza vaccinations among the elderly"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" began to grow.{ref}Centers for Disease Control and Prevention, & National Center for Immunization and Respiratory Diseases. (2021). "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Historical Reference of Seasonal Influenza Vaccine Doses Distributed"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""."", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.cdc.gov/flu/prevent/vaccine-supply-historical.htm"", ""children"": [{""text"": "" https://www.cdc.gov/flu/prevent/vaccine-supply-historical.htm"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The historical decline in influenza mortality has been substantial, as you can see in the chart."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Even when they reached the same age, people born in 1940 had around a third of the risk of dying from influenza as those born in 1900. This decline continued, and those born in 1980 had a risk of half that of those born in 1940.{ref}These estimates are more variable for young age groups, particularly those born after 1990 because the number of influenza deaths to make comparisons were low, especially as deaths in children under 5 were excluded from the models, to avoid counting deaths from respiratory syncytial virus."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Demography"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""56"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(5), 1723–1746."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1007/s13524-019-00809-y"", ""children"": [{""text"": "" https://doi.org/10.1007/s13524-019-00809-y"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": ""People born more recently have a lower risk of dying from influenza. Even when they reached the same age, people born in 1940 had a third of the risk of dying from flu than those born in 1900. This risk halved further for those born in 1980."", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Historical-decline-in-influenza-deaths-by-birth-cohort.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Which factors affect the number of deaths from the flu?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Age is a major risk factor of dying from the flu. As you can see in the chart, infants and the elderly tend to have a much higher risk of death from a range of respiratory diseases, including influenza, compared to young adults. For example, 60-year-olds have a ten times greater risk of death from influenza than 20-year-olds.{ref}Metcalf, C. J. E., Paireau, J., O’Driscoll, M., Pivette, M., Hubert, B., Pontais, I., Nickbakhsh, S., Cummings, D. A. T., Cauchemez, S., & Salje, H. (2022). Comparing the age and sex trajectories of SARS-CoV-2 morbidity and mortality with other respiratory pathogens. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Royal Society Open Science"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""9"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(6), 211498."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1098/rsos.211498"", ""children"": [{""text"": "" https://doi.org/10.1098/rsos.211498"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this post, we show the relative risks of death. This is because it's more difficult to give an absolute risk of death from influenza at different ages, because mortality rates vary widely over time and between countries, as we saw earlier. Absolute risks depend on how many people are infected during a flu season, the availability of healthcare, the rates of vaccination, and so on. However, the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""relative"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" risk of death – ratio between the risk of death in one age group versus another – tends to be more consistent.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Once someone reaches their twenties, their mortality risk from the flu increases exponentially. This shape follows the risk of death from all causes.{ref}The shape of this age–mortality curve is often described by the Gompertz function. Olshansky, S. J., & Carnes, B. A. (1997). Ever since gompertz. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Demography"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""34"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(1), 1-15. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://link.springer.com/content/pdf/10.2307/2061656.pdf"", ""children"": [{""text"": ""https://link.springer.com/content/pdf/10.2307/2061656.pdf"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The risk of dying from influenza also depends on other factors such as the quality of healthcare, the strain of influenza, and whether the person received the flu vaccine.{ref}Belongia, E. A., Simpson, M. D., King, J. P., Sundaram, M. E., Kelley, N. S., Osterholm, M. T., & McLean, H. Q. (2016). Variable influenza vaccine effectiveness by subtype: A systematic review and meta-analysis of test-negative design studies. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""The Lancet Infectious Diseases"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""16"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(8), 942–951."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/S1473-3099(16)00129-8"", ""children"": [{""text"": "" https://doi.org/10.1016/S1473-3099(16)00129-8"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Every year, influenza vaccines are reformulated to match the strains of flu that are expected to dominate during the winter. When there is a mismatch between the strains in the vaccine and the flu strains that are circulating, the vaccines tend to have lower efficacy, and flu seasons tend to be more severe.{ref}Tricco, A. C., Chit, A., Soobiah, C., Hallett, D., Meier, G., Chen, M. H., Tashkandi, M., Bauch, C. T., & Loeb, M. (2013). Comparing influenza vaccine efficacy against mismatched and matched strains: A systematic review and meta-analysis. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""BMC Medicine"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""11"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(1), 153."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1186/1741-7015-11-153"", ""children"": [{""text"": "" https://doi.org/10.1186/1741-7015-11-153"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": ""The risk of death from respiratory diseases increases exponentially with age. This is shown for influenza, MERS, Covid-19 and SARS. On average, people aged 20 have a tenth of the risk of death from influenza as people aged 60."", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Risk-of-death-respiratory-diseases-by-age.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Why are some flu seasons so severe?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Some seasons are far more severe than usual seasonal influenza. This tends to occur when new influenza strains arise and cause influenza pandemics. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Over time, influenza viruses that are circulating in the population tend to mutate through a process called \""antigenic drift\"". This gives them the ability to evade people's immunity."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But, influenza viruses can also evolve with large and sudden changes. This happens in a process called “antigenic shift”, when parts of different strains combine with each other. These new combinations can be more infectious and lethal than previous strains, leading to deadlier pandemics."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""For example, the Spanish flu evolved from a combination of human influenza and another animal influenza, which formed a new H1N1 influenza virus. As you can see in the chart, it caused "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/spanish-flu-largest-influenza-pandemic-in-history"", ""children"": [{""text"": ""the largest influenza pandemic in history"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "": research by Spreeuwenberg et al. (2018) suggests that the Spanish flu killed around 17.4 milion people. Other estimates are even higher: Johnson and Mueller (2002) suggest that the Spanish flu killed between 50 to 100 million people.{ref}“Reassessing the Global Mortality Burden of the 1918 Influenza Pandemic”. American Journal of Epidemiology. 187 (12): 2561–2567. doi:10.1093/aje/kwy191. PMID 30202996. Online "", ""spanType"": ""span-simple-text""}, {""url"": ""https://academic.oup.com/aje/article/187/12/2561/5092383"", ""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""Johnson, N. P., & Mueller, J. (2002). Updating the accounts: Global mortality of the 1918-1920\"" Spanish\"" influenza pandemic. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Bulletin of the History of Medicine"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 105–115.{/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This death toll massively exceeds the number who die in a typical year from the flu – it is between 30 to 340 times higher than the estimate of 294,000 to 518,000 deaths that are caused by seasonal influenza each year, even though the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/world-population-growth"", ""children"": [{""text"": ""global population was much smaller at the time"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://pubmed.ncbi.nlm.nih.gov/31673337/"", ""children"": [{""text"": ""https://pubmed.ncbi.nlm.nih.gov/31673337/"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" {/ref} "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Mortality during the Spanish flu rose sharply in young adults, compared to previous seasons. Research suggests that this is because they lacked immunity to H1 influenza viruses because they had been exposed to different influenza strains in their childhood. In contrast, older generations had been exposed to similar H1 influenza viruses decades before the pandemic began, which gave them some protection from the H1N1 pandemic strain.{ref}Worobey, M., Han, G.-Z., & Rambaut, A. (2014). Genesis and pathogenesis of the 1918 pandemic H1N1 influenza A virus. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Proceedings of the National Academy of Sciences"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""111"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(22), 8107–8112."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1073/pnas.1324197111"", ""children"": [{""text"": "" https://doi.org/10.1073/pnas.1324197111"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}], ""spanType"": ""span-link""}, {""text"": ""Gagnon, A., Miller, M. S., Hallman, S. A., Bourbeau, R., Herring, D. A., Earn, D. J. D., & Madrenas, J. (2013). Age-specific mortality during the 1918 influenza pandemic: Unravelling the mystery of high young adult mortality. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""PloS One"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""8"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(8), e69586."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1371/journal.pone.0069586"", ""children"": [{""text"": "" https://doi.org/10.1371/journal.pone.0069586"", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}], ""spanType"": ""span-link""}, {""text"": ""Luk, J., Gross, P., & Thompson, W. W. (2001). Observations on Mortality during the 1918 Influenza Pandemic. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Clinical Infectious Diseases"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""33"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""(8), 1375–1378."", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1086/322662"", ""children"": [{""text"": "" https://doi.org/10.1086/322662"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""spanType"": ""span-newline""}, {""text"": ""Ma, J., Dushoff, J., & Earn, D. J. D. (2011). Age-specific mortality risk from pandemic influenza. "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Journal of Theoretical Biology"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""288"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 29–34. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://doi.org/10.1016/j.jtbi.2011.08.003"", ""children"": [{""text"": ""https://doi.org/10.1016/j.jtbi.2011.08.003"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Influenza-pandemics-in-comparison-1.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Conclusion"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Seasonal flu causes 400,000 respiratory deaths each year on average. But the burden is far lower than it was in the past, due to improvements in sanitation, healthcare, and vaccination."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The flu also remains a large burden around the world for two major reasons. One is that many people around the world still lack access to healthcare and have low rates of influenza vaccination, which increases the risk of death."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Another reason is that the populations of many countries have been aging rapidly. In lower-income countries, the flu could become a larger burden as they face "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/historic-and-un-pop-projections-by-age"", ""children"": [{""text"": ""aging populations in the future"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To tackle this risk, the world can take lessons from how the burden has been reduced in the past. One way is to increase the rates of influenza vaccination, as well as other routine vaccinations, which also reduce the risk that flu is severe. Another is to improve sanitation and access to healthcare around the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We've already seen a huge decline in the burden of the flu over many decades, and with greater efforts, we could see that burden decline even further."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""children"": [{""text"": ""Keep reading on Our World in Data:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""spanType"": ""span-italic""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Acknowledgments:"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "" Hannah Ritchie, Max Roser and Edouard Mathieu provided very helpful guidance and comments that helped improve this post."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""How many people die from the flu?"", ""authors"": [""Saloni Dattani"", ""Fiona Spooner""], ""excerpt"": ""The risk of death from influenza has declined over time, but globally, hundreds of thousands of people still die from the disease each year."", ""dateline"": ""October 20, 2022"", ""subtitle"": ""The risk of death from influenza has declined over time, but globally, hundreds of thousands of people still die from the disease each year."", ""sidebar-toc"": false, ""featured-image"": ""Flu-deaths-thumbnail.png""}, ""createdAt"": ""2022-10-17T11:22:54.000Z"", ""published"": false, ""updatedAt"": ""2023-05-17T21:17:28.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-10-20T10:45:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}, {""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}], ""numBlocks"": 24, ""numErrors"": 9, ""wpTagCounts"": {""html"": 1, ""image"": 4, ""column"": 26, ""columns"": 13, ""heading"": 6, ""paragraph"": 59, ""separator"": 1, ""owid/prominent-link"": 3}, ""htmlTagCounts"": {""p"": 59, ""h4"": 6, ""hr"": 1, ""div"": 39, ""figure"": 4, ""iframe"": 1}}",2022-10-20 10:45:00,2024-02-16 14:22:54,1GhlZrbyAReOwQhdFG9eTsjLXTS_R8YhEK5WvxnM2O9Y,"[""Saloni Dattani"", ""Fiona Spooner""]","The risk of death from influenza has declined over time, but globally, hundreds of thousands of people still die from the disease each year.",2022-10-17 11:22:54,2023-05-17 21:17:28,https://ourworldindata.org/wp-content/uploads/2022/10/Flu-deaths-thumbnail.png,{},"Globally, seasonal influenza kills 400,000 people from respiratory disease each year on average. During large flu pandemics, when influenza strains evolved substantially, the death toll was even higher. But the risk of dying from influenza has declined substantially over time from improvements in sanitation, healthcare, and vaccination. People born in 1940 had around a third of the risk of dying from influenza as those born in 1900 – even when they reached the same age. This decline continued, and those born in 1980 have a risk of half that of those born in 1940. Influenza still remains a large burden around the world, because of an aging population and a lack of access to healthcare and sanitation in many countries. In this article, we look into these developments in detail: how many people die from seasonal influenza and how this has changed over time. We will also look at which factors increase the risk of dying from the flu and understand why, in some years, influenza has led to large pandemics that caused millions of deaths. This knowledge can inform us about the risks of influenza in the future. ## How many people die from seasonal influenza? Although this is an important question, it is often difficult to answer. Despite being a well-understood disease, it can be hard to count the number of deaths from influenza for several reasons.{ref}Gordon, A., & Reingold, A. (2018). The Burden of Influenza: A Complex Problem. _Current Epidemiology Reports_, _5_(1), 1–9.[ https://doi.org/10.1007/s40471-018-0136-1](https://doi.org/10.1007/s40471-018-0136-1){/ref} One problem is that the symptoms of influenza look similar to other infections, such as respiratory syncytial virus and rhinovirus. In many countries, only a fraction of patients with an ""influenza-like illness"" are tested to confirm whether they were infected by the virus.{ref}Charbonneau, D. H., & James, L. N. (2019). FluView and FluNet: Tools for Influenza Activity and Surveillance. _Medical Reference Services Quarterly_, _38_(4), 358–368.[ https://doi.org/10.1080/02763869.2019.1657734](https://doi.org/10.1080/02763869.2019.1657734){/ref} This means we miss many – or, in some countries, most – infections. Another problem is that influenza can lead to death in a number of indirect ways. It can cause death through respiratory complications such as pneumonia, but also from cardiovascular complications such as heart attacks and strokes, or other serious infections. This is especially true for the elderly and people who have chronic health conditions.{ref}Macias, A. E., McElhaney, J. E., Chaves, S. S., Nealon, J., Nunes, M. C., Samson, S. I., Seet, B. T., Weinke, T., & Yu, H. (2021). The disease burden of influenza beyond respiratory illness. _Vaccine_, _39_, A6–A14.[ ](https://doi.org/10.1016/j.vaccine.2020.09.048)[https://doi.org/10.1016/j.vaccine.2020.09.048](https://doi.org/10.1016/j.vaccine.2020.09.048%7B/ref){/ref} Without accounting for these deaths, we would underestimate the number of flu deaths. To overcome this, researchers estimate the burden of influenza with other methods. They can estimate the number of _excess_ deaths that occur during flu seasons, and use routine surveillance data and mortality records, to estimate how many of these are caused by the flu. The annual mortality caused by seasonal influenza was estimated by the [Global Pandemic Mortality Project II](https://www.nivel.nl/en/project/glamor-ii-global-influenza-mortality-research) using data between 2002 and 2011. They estimated that, during this period, seasonal influenza caused between 294,000 and 518,000 deaths each year globally.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. [https://jogh.org/documents/issue201902/jogh-09-020421.pdf](https://jogh.org/documents/issue201902/jogh-09-020421.pdf) This shows the mean estimate of annual influenza mortality between 2002–2011, excluding the 2009 ""Swine flu"" pandemic influenza season. You can find estimated numbers for world regions in Table 2 of the paper. Rates for other age groups can be found here: [https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf](https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf) Other global estimates of seasonal influenza mortality have been made by the Institute of Health Metrics and Evaluation (IHME) and the Centers for Disease Control and Prevention (CDC). Estimates made by GLaMOR were comparable to those by the CDC, while estimates made by IHME were around 4-5 times lower. This may be because the IHME estimated influenza mortality by first estimating the number of deaths caused by lower respiratory diseases, then estimating the fraction of those that were primarily attributed to influenza in vital records, verbal autopsies, and other mortality data. This approach would have missed many deaths caused by complications of influenza and deaths that were not specified on records to be caused by influenza due to limited testing. Both CDC and GLaMOR's models are also likely to be underestimating the total mortality burden from influenza, as they only use data from respiratory-associated deaths. While this _would_ include deaths caused by influenza that had, for example, influenza listed as a secondary cause of death on death certificates, it _would__miss_ some that were caused by influenza but attributed to another cause like cardiovascular disease. Had these models used all-cause mortality to estimate deaths caused by influenza, they would have been more sensitive (captured more deaths caused by influenza) but also less specific (captured more deaths caused by other diseases that could not be distinguished easily). In comparison to CDC estimates, GLaMOR used many country-specific indicators in order to extrapolate seasonal influenza mortality to countries that did not provide weekly or monthly influenza mortality records or influenza surveillance data, while the CDC extrapolated this using mainly the WHO Global Health Estimates of respiratory mortality.{/ref} These estimates focus on deaths where people had respiratory disease. This means they miss some flu deaths, as some people may die from cardiovascular complications of the flu without having respiratory disease.{ref}The global number of people who die from other complications of the flu is unclear. Paget et al. (the authors of the GLaMOR project) state in their paper that their estimate “does not cover cardiovascular deaths, something that could at least double the estimate of influenza-associated deaths.” In recent meta-analyses, Behrouzi et al. found that influenza vaccination reduces the chances of major cardiovascular events (such as heart attacks and strokes) by around 34%, in clinical trials of the elderly. This suggests the death toll from other complications could be large. However, global estimates have not been made of these types of deaths from flu. Paget, J., Danielle Iuliano, A., Taylor, R. J., Simonsen, L., Viboud, C., & Spreeuwenberg, P. (2022). Estimates of mortality associated with seasonal influenza for the European Union from the GLaMOR project. Vaccine, 40(9), 1361–1369. [https://doi.org/10.1016/j.vaccine.2021.11.080](https://doi.org/10.1016/j.vaccine.2021.11.080) Behrouzi, B., Bhatt, D. L., Cannon, C. P., Vardeny, O., Lee, D. S., Solomon, S. D., & Udell, J. A. (2022). Association of Influenza Vaccination With Cardiovascular Risk: A Meta-analysis. JAMA Network Open, 5(4), e228873. [https://doi.org/10.1001/jamanetworkopen.2022.8873](https://doi.org/10.1001/jamanetworkopen.2022.8873){/ref} On the map, you can see the estimates of flu mortality shown as a rate per 100,000 people, among people aged over 65. In Europe, the rate of deaths from the flu was 30.8 per 100,000 each year, among those aged over 65. This is more than three times the risk from traffic accidents, which kill 9 per 100,000, in the same age group.{ref}Eurostat. (2022). _Causes of death—Standardised death rate_. European Commission.[ https://ec.europa.eu/eurostat/databrowser/view/HLTH_CD_ASDR2__custom_3500876/default/table?lang=en](https://ec.europa.eu/eurostat/databrowser/view/HLTH_CD_ASDR2__custom_3500876/default/table?lang=en) Transport accidents are counted under (V01–V99, Y85) in the ICD-10. Across 27 EU countries, these rates were 9.02, 9.15, and 8.77 per 100,000 people aged over 65 in 2015, 2016 and 2017 respectively. Estimates of the rate of death from influenza are much lower in the ICD-10 since they only consider deaths where influenza is listed as the cause of death on death certificates, while the estimates we show above also include those that are caused by flu indirectly. This means ICD-10 death rates are likely to be highly underestimated for the flu. However, deaths caused by traffic accidents are more likely to be listed as the primary cause of death on death certificates and are much less underestimated by death certificate data.{/ref} In low-income countries, these estimates tend to be less certain, due to lower levels of testing for influenza and limited mortality records. But flu is estimated to be more deadly in countries in South America, Africa, and South Asia than in Europe and North America. For example, Indonesia has more than twice the death rate of Canada. These disparities are at least partly due to poverty, poorer underlying health, and lower access to healthcare.Globally, seasonal influenza kills 400,000 people from respiratory disease each year on average. During large flu pandemics, when influenza strains evolved substantially, the death toll was even higher.
\n\n\n\nBut the risk of dying from influenza has declined substantially over time from improvements in sanitation, healthcare, and vaccination.
\n\n\n\nPeople born in 1940 had around a third of the risk of dying from influenza as those born in 1900 – even when they reached the same age. This decline continued, and those born in 1980 have a risk of half that of those born in 1940.
\n\n\n\nInfluenza still remains a large burden around the world, because of an aging population and a lack of access to healthcare and sanitation in many countries.
\n\n\n\nIn this article, we look into these developments in detail: how many people die from seasonal influenza and how this has changed over time.
\n\n\n\nWe will also look at which factors increase the risk of dying from the flu and understand why, in some years, influenza has led to large pandemics that caused millions of deaths. This knowledge can inform us about the risks of influenza in the future.
\nAlthough this is an important question, it is often difficult to answer.
\n\n\n\nDespite being a well-understood disease, it can be hard to count the number of deaths from influenza for several reasons.{ref}Gordon, A., & Reingold, A. (2018). The Burden of Influenza: A Complex Problem. Current Epidemiology Reports, 5(1), 1–9. https://doi.org/10.1007/s40471-018-0136-1{/ref}
\n\n\n\nOne problem is that the symptoms of influenza look similar to other infections, such as respiratory syncytial virus and rhinovirus. In many countries, only a fraction of patients with an “influenza-like illness” are tested to confirm whether they were infected by the virus.{ref}Charbonneau, D. H., & James, L. N. (2019). FluView and FluNet: Tools for Influenza Activity and Surveillance. Medical Reference Services Quarterly, 38(4), 358–368. https://doi.org/10.1080/02763869.2019.1657734{/ref} This means we miss many – or, in some countries, most – infections.
\n\n\n\nAnother problem is that influenza can lead to death in a number of indirect ways. It can cause death through respiratory complications such as pneumonia, but also from cardiovascular complications such as heart attacks and strokes, or other serious infections. This is especially true for the elderly and people who have chronic health conditions.{ref}Macias, A. E., McElhaney, J. E., Chaves, S. S., Nealon, J., Nunes, M. C., Samson, S. I., Seet, B. T., Weinke, T., & Yu, H. (2021). The disease burden of influenza beyond respiratory illness. Vaccine, 39, A6–A14. https://doi.org/10.1016/j.vaccine.2020.09.048{/ref} Without accounting for these deaths, we would underestimate the number of flu deaths.
\n\n\n\nTo overcome this, researchers estimate the burden of influenza with other methods. They can estimate the number of excess deaths that occur during flu seasons, and use routine surveillance data and mortality records, to estimate how many of these are caused by the flu.
\n\n\n\nThe annual mortality caused by seasonal influenza was estimated by the Global Pandemic Mortality Project II using data between 2002 and 2011. They estimated that, during this period, seasonal influenza caused between 294,000 and 518,000 deaths each year globally.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. https://jogh.org/documents/issue201902/jogh-09-020421.pdf
This shows the mean estimate of annual influenza mortality between 2002–2011, excluding the 2009 “Swine flu” pandemic influenza season. You can find estimated numbers for world regions in Table 2 of the paper. Rates for other age groups can be found here: https://www.nivel.nl/sites/default/files/influenza-nieuwsbrief/GLaMOR%20project_seasonal%20estimates.pdf
Other global estimates of seasonal influenza mortality have been made by the Institute of Health Metrics and Evaluation (IHME) and the Centers for Disease Control and Prevention (CDC).
Estimates made by GLaMOR were comparable to those by the CDC, while estimates made by IHME were around 4-5 times lower. This may be because the IHME estimated influenza mortality by first estimating the number of deaths caused by lower respiratory diseases, then estimating the fraction of those that were primarily attributed to influenza in vital records, verbal autopsies, and other mortality data. This approach would have missed many deaths caused by complications of influenza and deaths that were not specified on records to be caused by influenza due to limited testing.
\n\n\n\nBoth CDC and GLaMOR’s models are also likely to be underestimating the total mortality burden from influenza, as they only use data from respiratory-associated deaths. While this would include deaths caused by influenza that had, for example, influenza listed as a secondary cause of death on death certificates, it would miss some that were caused by influenza but attributed to another cause like cardiovascular disease. Had these models used all-cause mortality to estimate deaths caused by influenza, they would have been more sensitive (captured more deaths caused by influenza) but also less specific (captured more deaths caused by other diseases that could not be distinguished easily).
In comparison to CDC estimates, GLaMOR used many country-specific indicators in order to extrapolate seasonal influenza mortality to countries that did not provide weekly or monthly influenza mortality records or influenza surveillance data, while the CDC extrapolated this using mainly the WHO Global Health Estimates of respiratory mortality.{/ref}
These estimates focus on deaths where people had respiratory disease. This means they miss some flu deaths, as some people may die from cardiovascular complications of the flu without having respiratory disease.{ref}The global number of people who die from other complications of the flu is unclear.
\n\n\n\nPaget et al. (the authors of the GLaMOR project) state in their paper that their estimate “does not cover cardiovascular deaths, something that could at least double the estimate of influenza-associated deaths.”
\n\n\n\nIn recent meta-analyses, Behrouzi et al. found that influenza vaccination reduces the chances of major cardiovascular events (such as heart attacks and strokes) by around 34%, in clinical trials of the elderly.
\n\n\n\nThis suggests the death toll from other complications could be large. However, global estimates have not been made of these types of deaths from flu.
\n\n\n\nPaget, J., Danielle Iuliano, A., Taylor, R. J., Simonsen, L., Viboud, C., & Spreeuwenberg, P. (2022). Estimates of mortality associated with seasonal influenza for the European Union from the GLaMOR project. Vaccine, 40(9), 1361–1369. https://doi.org/10.1016/j.vaccine.2021.11.080
\n\n\n\nBehrouzi, B., Bhatt, D. L., Cannon, C. P., Vardeny, O., Lee, D. S., Solomon, S. D., & Udell, J. A. (2022). Association of Influenza Vaccination With Cardiovascular Risk: A Meta-analysis. JAMA Network Open, 5(4), e228873. https://doi.org/10.1001/jamanetworkopen.2022.8873{/ref}
\n\n\n\nOn the map, you can see the estimates of flu mortality shown as a rate per 100,000 people, among people aged over 65.
\n\n\n\nIn Europe, the rate of deaths from the flu was 30.8 per 100,000 each year, among those aged over 65. This is more than three times the risk from traffic accidents, which kill 9 per 100,000, in the same age group.{ref}Eurostat. (2022). Causes of death—Standardised death rate. European Commission. https://ec.europa.eu/eurostat/databrowser/view/HLTH_CD_ASDR2__custom_3500876/default/table?lang=en
Transport accidents are counted under (V01–V99, Y85) in the ICD-10. Across 27 EU countries, these rates were 9.02, 9.15, and 8.77 per 100,000 people aged over 65 in 2015, 2016 and 2017 respectively.
Estimates of the rate of death from influenza are much lower in the ICD-10 since they only consider deaths where influenza is listed as the cause of death on death certificates, while the estimates we show above also include those that are caused by flu indirectly. This means ICD-10 death rates are likely to be highly underestimated for the flu. However, deaths caused by traffic accidents are more likely to be listed as the primary cause of death on death certificates and are much less underestimated by death certificate data.{/ref}
\n\n\n\nIn low-income countries, these estimates tend to be less certain, due to lower levels of testing for influenza and limited mortality records.
\n\n\n\nBut flu is estimated to be more deadly in countries in South America, Africa, and South Asia than in Europe and North America. For example, Indonesia has more than twice the death rate of Canada. These disparities are at least partly due to poverty, poorer underlying health, and lower access to healthcare.
\nDeath rates from influenza are much lower than they were in the past.
\n\n\n\nWe don’t have good long-term estimates of flu deaths across most countries. But, researchers have produced weekly estimates of influenza deaths over long periods in the United States. This allows us to see how modern rates compare to the past. You can see this data in the chart.{ref}Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y
\n\n\n\nFrom 1960 to 2015, the number of deaths from influenza was estimated using a Serfling model, which estimates the excess number of deaths during flu seasons using data from the rest of the year and accounting for changes that occur year by year. Since 1997, there has also been routine testing for “influenza-like illnesses” in hospitals to determine the share of them that are actually caused by influenza, rather than other diseases. Therefore, estimates from 1997 to 2015 were also calculated using a Serfling-surveillance model, which accounted for the share of tests that were positive for influenza. This also validates the estimates from the regular Serfling model. In addition, deaths among children aged under 5 are excluded in both models, as they would be likely to include deaths from respiratory syncytial virus.
\n\n\n\nMortality in the US was slightly lower in during the 2009 Swine flu pandemic season than usual flu seasons, as severe disease shifted away from the elderly to young and middle-aged adults. However, the 2009 Swine flu pandemic led to more deaths than regular flu seasons in other countries such as Mexico.
\n\n\n\nGagnon, A., Acosta, E., Hallman, S., Bourbeau, R., Dillon, L. Y., Ouellette, N., Earn, D. J. D., Herring, D. A., Inwood, K., Madrenas, J., & Miller, M. S. (2018). Pandemic Paradox: Early Life H2N2 Pandemic Influenza Infection Enhanced Susceptibility to Death during the 2009 H1N1 Pandemic. MBio, 9(1), e02091-17. https://doi.org/10.1128/mBio.02091-17{/ref}
\n\n\n\nDeaths from influenza fluctuate across the year, with large peaks in the winter.{ref}Influenza viruses are thought to transmit more efficiently during the winter due to lower temperatures and humidity. But in many tropical countries, flu epidemics coincide with warm rainy seasons, so the trends may have more causes. Other explanations include seasonal changes in human immunity or changes in human behavior, such as more indoor mixing and crowding. Petrova, V. N., & Russell, C. A. (2018). The evolution of seasonal influenza viruses. Nature Reviews Microbiology, 16(1), 47–60. https://doi.org/10.1038/nrmicro.2017.118{/ref} The total number of deaths from influenza has been roughly stable in the United States over the last 65 years. You can see this in the top panel of the chart.
\n\n\n\nHowever, a large part of this is due to the fact that the population has been growing and aging.
\n\n\n\nIf we look at death rates within age groups, the rate of deaths from influenza has been falling. You can see this in the bottom panel, which accounts for changes in the size and age structure of the population.
\n\n\n\nThis means the likelihood that someone dies from influenza at a given age has declined over time. But, because the population is getting larger and older, the total number of deaths has remained stable.
\n![\""\""](\""https://owid.cloud/app/uploads/2022/10/Historical-influenza-deaths-USA-3-panels-1-224x400.png\"")
Influenza mortality has declined over several generations. We know this from historical data from the United States, which has been used to estimate “cohort effects”. This tells us whether people who were born more recently have lower risks of death, after accounting for their younger age.
\n\n\n\nSince 1900, there has been a long-term decline in the risk of dying from the flu.{ref} Between 1860 and 1900, there was a slight increase in the risk of death from influenza, which may have been due to worsening health conditions as more people moved into crowded urban areas.
\n\n\n\nAcosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y{/ref} There are several reasons for this.
\n\n\n\nOne is that there were large projects to improve sanitation in cities across the United States in the early 1900s.{ref}Cutler, D., & Miller, G. (2005). The role of public health improvements in health advances: The twentieth-century United States. Demography, 42(1), 1–22. https://doi.org/10.1353/dem.2005.0002{/ref} Over the twentieth century, there were also improvements in neonatal healthcare and increases in the rate of childhood vaccinations. All of these factors had benefits that carried forward as people aged: they protected people from developing comorbidities that increased the risk of dying from influenza.
\n\n\n\nThere has also been an increase in the rate of flu vaccinations. Influenza vaccines were developed for the first time in the 1930s and 1940s. In 1952, the World Health Organization began a surveillance system to monitor which flu strains were circulating worldwide. This helped researchers develop new vaccines each year that matched those strains.{ref}Barberis, I., Myles, P., Ault, S. K., Bragazzi, N. L., & Martini, M. (2016). History and evolution of influenza control through vaccination: From the first monovalent vaccine to universal vaccines. Journal of Preventive Medicine and Hygiene, 57(3), E115–E120. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5139605/ {/ref} Over the following decades, the rate of influenza vaccinations among the elderly began to grow.{ref}Centers for Disease Control and Prevention, & National Center for Immunization and Respiratory Diseases. (2021). Historical Reference of Seasonal Influenza Vaccine Doses Distributed. https://www.cdc.gov/flu/prevent/vaccine-supply-historical.htm{/ref}
\n\n\n\nThe historical decline in influenza mortality has been substantial, as you can see in the chart.
\n\n\n\nEven when they reached the same age, people born in 1940 had around a third of the risk of dying from influenza as those born in 1900. This decline continued, and those born in 1980 had a risk of half that of those born in 1940.{ref}These estimates are more variable for young age groups, particularly those born after 1990 because the number of influenza deaths to make comparisons were low, especially as deaths in children under 5 were excluded from the models, to avoid counting deaths from respiratory syncytial virus.
Acosta, E., Hallman, S. A., Dillon, L. Y., Ouellette, N., Bourbeau, R., Herring, D. A., Inwood, K., Earn, D. J. D., Madrenas, J., Miller, M. S., & Gagnon, A. (2019). Determinants of Influenza Mortality Trends: Age-Period-Cohort Analysis of Influenza Mortality in the United States, 1959–2016. Demography, 56(5), 1723–1746. https://doi.org/10.1007/s13524-019-00809-y{/ref}
Age is a major risk factor of dying from the flu. As you can see in the chart, infants and the elderly tend to have a much higher risk of death from a range of respiratory diseases, including influenza, compared to young adults. For example, 60-year-olds have a ten times greater risk of death from influenza than 20-year-olds.{ref}Metcalf, C. J. E., Paireau, J., O’Driscoll, M., Pivette, M., Hubert, B., Pontais, I., Nickbakhsh, S., Cummings, D. A. T., Cauchemez, S., & Salje, H. (2022). Comparing the age and sex trajectories of SARS-CoV-2 morbidity and mortality with other respiratory pathogens. Royal Society Open Science, 9(6), 211498. https://doi.org/10.1098/rsos.211498
\n\n\n\nIn this post, we show the relative risks of death. This is because it’s more difficult to give an absolute risk of death from influenza at different ages, because mortality rates vary widely over time and between countries, as we saw earlier. Absolute risks depend on how many people are infected during a flu season, the availability of healthcare, the rates of vaccination, and so on. However, the relative risk of death – ratio between the risk of death in one age group versus another – tends to be more consistent.{/ref}
\n\n\n\nOnce someone reaches their twenties, their mortality risk from the flu increases exponentially. This shape follows the risk of death from all causes.{ref}The shape of this age–mortality curve is often described by the Gompertz function. Olshansky, S. J., & Carnes, B. A. (1997). Ever since gompertz. Demography, 34(1), 1-15. https://link.springer.com/content/pdf/10.2307/2061656.pdf {/ref}
\n\n\n\nThe risk of dying from influenza also depends on other factors such as the quality of healthcare, the strain of influenza, and whether the person received the flu vaccine.{ref}Belongia, E. A., Simpson, M. D., King, J. P., Sundaram, M. E., Kelley, N. S., Osterholm, M. T., & McLean, H. Q. (2016). Variable influenza vaccine effectiveness by subtype: A systematic review and meta-analysis of test-negative design studies. The Lancet Infectious Diseases, 16(8), 942–951. https://doi.org/10.1016/S1473-3099(16)00129-8{/ref}
\n\n\n\nEvery year, influenza vaccines are reformulated to match the strains of flu that are expected to dominate during the winter. When there is a mismatch between the strains in the vaccine and the flu strains that are circulating, the vaccines tend to have lower efficacy, and flu seasons tend to be more severe.{ref}Tricco, A. C., Chit, A., Soobiah, C., Hallett, D., Meier, G., Chen, M. H., Tashkandi, M., Bauch, C. T., & Loeb, M. (2013). Comparing influenza vaccine efficacy against mismatched and matched strains: A systematic review and meta-analysis. BMC Medicine, 11(1), 153. https://doi.org/10.1186/1741-7015-11-153{/ref}
\n![\""The](\""https://owid.cloud/app/uploads/2022/10/Risk-of-death-respiratory-diseases-by-age-693x550.png\"")
Some seasons are far more severe than usual seasonal influenza. This tends to occur when new influenza strains arise and cause influenza pandemics.
\n\n\n\nOver time, influenza viruses that are circulating in the population tend to mutate through a process called “antigenic drift”. This gives them the ability to evade people’s immunity.
\n\n\n\nBut, influenza viruses can also evolve with large and sudden changes. This happens in a process called “antigenic shift”, when parts of different strains combine with each other. These new combinations can be more infectious and lethal than previous strains, leading to deadlier pandemics.
\n\n\n\nFor example, the Spanish flu evolved from a combination of human influenza and another animal influenza, which formed a new H1N1 influenza virus. As you can see in the chart, it caused the largest influenza pandemic in history: research by Spreeuwenberg et al. (2018) suggests that the Spanish flu killed around 17.4 milion people. Other estimates are even higher: Johnson and Mueller (2002) suggest that the Spanish flu killed between 50 to 100 million people.{ref}“Reassessing the Global Mortality Burden of the 1918 Influenza Pandemic”. American Journal of Epidemiology. 187 (12): 2561–2567. doi:10.1093/aje/kwy191. PMID 30202996. Online here.
Johnson, N. P., & Mueller, J. (2002). Updating the accounts: Global mortality of the 1918-1920″ Spanish” influenza pandemic. Bulletin of the History of Medicine, 105–115.{/ref}
This death toll massively exceeds the number who die in a typical year from the flu – it is between 30 to 340 times higher than the estimate of 294,000 to 518,000 deaths that are caused by seasonal influenza each year, even though the global population was much smaller at the time.{ref}Paget J, Spreeuwenberg P, Charu V, Taylor RJ, Iuliano AD, Bresee J, Simonsen L, Viboud C; Global Seasonal Influenza-associated Mortality Collaborator Network and GLaMOR Collaborating Teams*. Global mortality associated with seasonal influenza epidemics: New burden estimates and predictors from the GLaMOR Project. J Glob Health. 2019 Dec;9(2):020421. doi: 10.7189/jogh.09.020421. https://pubmed.ncbi.nlm.nih.gov/31673337/ {/ref}
\n\n\n\nMortality during the Spanish flu rose sharply in young adults, compared to previous seasons. Research suggests that this is because they lacked immunity to H1 influenza viruses because they had been exposed to different influenza strains in their childhood. In contrast, older generations had been exposed to similar H1 influenza viruses decades before the pandemic began, which gave them some protection from the H1N1 pandemic strain.{ref}Worobey, M., Han, G.-Z., & Rambaut, A. (2014). Genesis and pathogenesis of the 1918 pandemic H1N1 influenza A virus. Proceedings of the National Academy of Sciences, 111(22), 8107–8112. https://doi.org/10.1073/pnas.1324197111
Gagnon, A., Miller, M. S., Hallman, S. A., Bourbeau, R., Herring, D. A., Earn, D. J. D., & Madrenas, J. (2013). Age-specific mortality during the 1918 influenza pandemic: Unravelling the mystery of high young adult mortality. PloS One, 8(8), e69586. https://doi.org/10.1371/journal.pone.0069586
Luk, J., Gross, P., & Thompson, W. W. (2001). Observations on Mortality during the 1918 Influenza Pandemic. Clinical Infectious Diseases, 33(8), 1375–1378. https://doi.org/10.1086/322662
Ma, J., Dushoff, J., & Earn, D. J. D. (2011). Age-specific mortality risk from pandemic influenza. Journal of Theoretical Biology, 288, 29–34. https://doi.org/10.1016/j.jtbi.2011.08.003{/ref}
Seasonal flu causes 400,000 respiratory deaths each year on average. But the burden is far lower than it was in the past, due to improvements in sanitation, healthcare, and vaccination.
\n\n\n\nThe flu also remains a large burden around the world for two major reasons. One is that many people around the world still lack access to healthcare and have low rates of influenza vaccination, which increases the risk of death.
\n\n\n\nAnother reason is that the populations of many countries have been aging rapidly. In lower-income countries, the flu could become a larger burden as they face aging populations in the future.
\n\n\n\nTo tackle this risk, the world can take lessons from how the burden has been reduced in the past. One way is to increase the rates of influenza vaccination, as well as other routine vaccinations, which also reduce the risk that flu is severe. Another is to improve sanitation and access to healthcare around the world.
\n\n\n\nWe’ve already seen a huge decline in the burden of the flu over many decades, and with greater efforts, we could see that burden decline even further.
\nKeep reading on Our World in Data:
\n\n\nAcknowledgments: Hannah Ritchie, Max Roser and Edouard Mathieu provided very helpful guidance and comments that helped improve this post.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""The risk of death from influenza has declined over time, but globally, hundreds of thousands of people still die from the disease each year."", ""protected"": false}, ""date_gmt"": ""2022-10-20T10:45:00"", ""modified"": ""2023-05-17T22:17:28"", ""template"": """", ""categories"": [46, 171], ""ping_status"": ""closed"", ""authors_name"": [""Saloni Dattani"", ""Fiona Spooner""], ""modified_gmt"": ""2023-05-17T21:17:28"", ""comment_status"": ""closed"", ""featured_media"": 53551, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/10/Flu-deaths-thumbnail-150x59.png"", ""medium_large"": ""/app/uploads/2022/10/Flu-deaths-thumbnail-768x303.png""}}" 53526,Which countries have put a price on carbon?,carbon-pricing,post,publish,"The true cost of burning fossil fuels is not reflected in their market price. People often compare the monetary price of fossil fuels to low-carbon alternatives such as renewables or nuclear energy. But these comparisons do not capture each option’s social impact, even in a purely economic sense.
Burning fossil fuels drives climate change. The impacts of climate change come at a cost – these impacts are already happening.
They also cause local air pollution, which kills millions every year and has negative health impacts for many more.
One way to address this is to put a price on carbon. The purpose of setting a carbon price is to capture some of these external costs in the market.
My colleague, Max Roser, wrote a detailed case for a carbon price here.
Putting a price on carbon achieves a couple of things. First, it makes more polluting fuels, products, and services more expensive. Burning coal becomes much more expensive than using solar energy. Beef gets more expensive relative to tofu or meat alternatives. It makes the ‘cleaner’ option less expensive.
Second, it means it’s those who emit greenhouse gases that pay for it.
There are a few policies through which countries can put a price on carbon:
Many countries have adopted carbon pricing instruments.
In this article, we provide an overview of which countries have them, and how the price of carbon is changing over time. The underlying data – from the World Carbon Pricing Dataset – is updated annually.{ref}Dolphin, G., & Xiahou, Q. (2022). World Carbon Pricing Database: Sources and Methods. Nature Scientific Data.{/ref} It is assembled by Geoffroy Dolphin. We will keep the data in this article updated.
The first chart shows which countries have implemented a carbon tax.
It’s often the case that only specific sectors, or specific fuels, in a given country are subject to a carbon tax. For example, heavy industry or household electricity might have a carbon tax, but road transport might not.
A country is coded as having a carbon tax if any of its sectors have one. This means the country does not necessarily have an economy-wide tax in place.
One final note: the World Carbon Pricing Database only looks at taxes applied to carbon dioxide (CO2) emissions. It does not consider taxes on non-CO2 greenhouse gases, such as methane or nitrous oxide.
Having a carbon tax does not guarantee that it will be effective. The level of carbon tax rates matters. It needs to be sufficiently expensive to de-incentivize carbon-emitting activities, and incentivize low-carbon ones.
In the chart, we see the carbon price in countries that have implemented a tax.
Since carbon prices might vary from sector to sector, or some sectors will have a carbon price while others won’t, the authors calculate an ‘emissions-weighted’ carbon price. This price is weighted according to each sector’s contribution to a country’s CO2 emissions.
The alternative to a carbon tax is a ‘cap and trade’ system. This is where the number of emission allowances/licenses, or the amount that a country or sector can emit, is fixed. Companies and sectors within the economy can then buy and sell carbon credits. This supply and demand determine the price of carbon in the marketplace.
The most well-known ‘cap and trade’ system is the European Union’s Emissions Trading System (EU ETS). It was first introduced in 2005.
But a number of other countries have implemented one at the national or sub-national level.
In the chart, we see which countries have implemented an emissions trading system. Again, a country is coded as having a system if at least one sector is covered by one.
The supply and demand for carbon credits within an economy determines the carbon price. If the amount of available credits is low compared to demand, then the price will be higher. If there are many available credits then the price will be low, and companies will not be incentivized to reduce their emissions. This means the carbon price changes over time.
In the map, we see the carbon price in emissions trading systems across the world. Again, if the price varies by sector, these prices are weighted by each sector’s contribution to the country’s CO2 emissions.
If we want the climate cost of fuels and products to be reflected in their market price we would want a carbon pricing mechanism everywhere.
In 2020, around 12% of emissions were in countries or sectors that had a carbon tax. Just 6% were covered by a trading system. This means that, in total, a carbon price had to be paid on 18% of global emissions.
We see the share of global CO2 emissions that are covered by each in the chart. The map also shows the share of emissions in each country that are covered by either a carbon tax or emissions trading system.
This data is currently only shown to 2020. As of 2021, this coverage expanded dramatically – to 25% of global CO2 emissions – because China started its national emissions trading system. This will be reflected in the next update of this dataset.
Many thanks to Geoffroy Dolphin and Felix Pretis for the provision of data from the World Carbon Pricing Database, and feedback on this work. Thanks also to Max Roser, Edouard Mathieu and Bastian Herre for editorial feedback on this article.
","{""id"": ""wp-53526"", ""slug"": ""carbon-pricing"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""The true cost of burning fossil fuels is not reflected in their market price. People "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/cheap-renewables-growth"", ""children"": [{""text"": ""often compare"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""monetary price"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of fossil fuels to low-carbon alternatives such as renewables or nuclear energy. But these comparisons do not capture each option’s social impact, even in a purely economic sense."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Burning fossil fuels drives "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/climate-change"", ""children"": [{""text"": ""climate change"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". The impacts of climate change come at a cost – these impacts are already happening."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""They also cause local air pollution, which "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/data-review-air-pollution-deaths"", ""children"": [{""text"": ""kills millions"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" every year and has negative health impacts for many more."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One way to address this is to put a price on carbon. The purpose of setting a carbon price is to capture some of these external costs in the market."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""My colleague, Max Roser, wrote a detailed case for a carbon price "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/carbon-price"", ""children"": [{""children"": [{""text"": ""here"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Putting a price on carbon achieves a couple of things. First, it makes more polluting fuels, products, and services more expensive. Burning coal becomes "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/safest-sources-of-energy"", ""children"": [{""text"": ""much more expensive"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" than using solar energy. Beef gets "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/food-footprints"", ""children"": [{""text"": ""more expensive"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" relative to tofu or meat alternatives. It makes the ‘cleaner’ option less expensive."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Second, it means it’s those who emit greenhouse gases that pay for it."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are a few policies through which countries can put a price on carbon:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""list"", ""items"": [{""type"": ""text"", ""value"": [{""children"": [{""text"": ""Carbon tax: "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""this is a levy that is applied to the production of greenhouse gas emissions directly or fuels that emit these gases when they’re burned. This means goods and services which emit more greenhouse gases in their production will have a higher tax."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Emissions trading system: "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": ""this is sometimes called a ‘cap and trade’ system. Here, the carbon price changes over time. A maximum level of pollution (a ‘cap’) is defined and manufacturers need licenses to emit greenhouse gases. How expensive these licenses are is determined by a trading system. The price of a license increases as emissions approach the cap."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many countries have adopted carbon pricing instruments."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this article, we provide an overview of which countries have them, and how the price of carbon is changing over time. The underlying data – from the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.rff.org/publications/working-papers/world-carbon-pricing-database-sources-and-methods/"", ""children"": [{""text"": ""World Carbon Pricing Dataset"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" – is updated annually.{ref}Dolphin, G., & Xiahou, Q. (2022). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/s41597-022-01659-x"", ""children"": [{""text"": ""World Carbon Pricing Database: Sources and Methods"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Nature Scientific Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "".{/ref} It is assembled by Geoffroy Dolphin. We will keep the data in this article updated. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Which countries have a carbon tax?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The first chart shows which countries have implemented a carbon tax."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""It’s often the case that only specific sectors, or specific fuels, in a given country are subject to a carbon tax. For example, heavy industry or household electricity might have a carbon tax, but road transport might not."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A country is coded as having a carbon tax if "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""any"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of its sectors have one. This means the country does not necessarily have an economy-wide tax in place. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One final note: the World Carbon Pricing Database only looks at taxes applied to carbon dioxide (CO"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-subscript""}, {""text"": "") emissions. It does not consider taxes on non-CO"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2 "", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-subscript""}, {""text"": ""greenhouse gases, such as methane or nitrous oxide."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/carbon-tax-instruments"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""What is the level of carbon taxes?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Having a carbon tax does not guarantee that it will be effective. The level of carbon tax rates matters. It needs to be sufficiently expensive to de-incentivize carbon-emitting activities, and incentivize low-carbon ones."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart, we see the carbon price in countries that have implemented a tax."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Since carbon prices might vary from sector to sector, or some sectors will have a carbon price while others won’t, the authors calculate an ‘emissions-weighted’ carbon price. This price is weighted according to each sector’s contribution to a country’s CO"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-subscript""}, {""text"": "" emissions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/emissions-weighted-carbon-price?tab=map"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Which countries have a carbon emissions trading system?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The alternative to a carbon tax is a ‘cap and trade’ system. This is where the number of emission allowances/licenses, or the amount that a country or sector can emit, is fixed. Companies and sectors within the economy can then buy and sell carbon credits. This supply and demand determine the price of carbon in the marketplace."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The most well-known ‘cap and trade’ system is the European Union’s "", ""spanType"": ""span-simple-text""}, {""url"": ""https://en.wikipedia.org/wiki/European_Union_Emissions_Trading_System"", ""children"": [{""text"": ""Emissions Trading System"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (EU ETS). It was first introduced in 2005."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But a number of other countries have implemented one at the national or sub-national level."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart, we see which countries have implemented an emissions trading system. Again, a country is coded as having a system if at least one sector is covered by one."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/carbon-emissions-trading-system"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""What is the carbon price in emissions trading systems?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""The supply and demand for carbon credits within an economy determines the carbon price. If the amount of available credits is low compared to demand, then the price will be higher. If there are many available credits then the price will be low, and companies will not be incentivized to reduce their emissions. This means the carbon price changes over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the map, we see the carbon price in emissions trading systems across the world. Again, if the price varies by sector, these prices are weighted by each sector’s contribution to the country’s CO"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-subscript""}, {""text"": "" emissions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/weighted-carbon-price-ets"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""How much of the world’s CO"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-subscript""}, {""text"": "" emissions are covered by a carbon tax or emissions trading system?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If we want the climate cost of fuels and products to be reflected in their market price we would want a carbon pricing mechanism everywhere."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": "" In 2020, around 12% of emissions were in countries or sectors that had a carbon tax. Just 6% were covered by a trading system. This means that, in total, a carbon price had to be paid on 18% of global emissions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We see the share of global CO"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-subscript""}, {""text"": "" emissions that are covered by each in the chart. The map also shows the share of emissions in each country that are covered by either a carbon tax or emissions trading system."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data is currently only shown to 2020. As of 2021, this coverage expanded dramatically – to 25% of global CO"", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""2"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-subscript""}, {""text"": "" emissions – because China started its national emissions trading system. This will be reflected in the next update of this dataset."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/grapher/carbon-tax-trading-coverage?tab=chart&country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/grapher/carbon-tax-trading-coverage?tab=map&country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Keep reading at "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Our World in Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""type"": ""heading"", ""level"": 4, ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/carbon-price"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/cheap-renewables-growth"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/data-review-air-pollution-deaths"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""text"": [{""text"": ""Acknowledgments"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 4, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many thanks to Geoffroy Dolphin and Felix Pretis for the provision of data from the World Carbon Pricing Database, and feedback on this work. Thanks also to Max Roser, Edouard Mathieu and Bastian Herre for editorial feedback on this article."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Which countries have put a price on carbon?"", ""authors"": [""Hannah Ritchie"", ""Pablo Rosado""], ""excerpt"": ""Putting a price on carbon helps us account for the real costs of fossil fuels in the market. Which countries have a carbon tax or trading system?"", ""dateline"": ""October 14, 2022"", ""subtitle"": ""Putting a price on carbon helps us account for the real costs of fossil fuels in the market. Which countries have a carbon tax or trading system?"", ""sidebar-toc"": false, ""featured-image"": ""Carbon-taxes-featured-image.png""}, ""createdAt"": ""2022-10-11T10:22:07.000Z"", ""published"": false, ""updatedAt"": ""2022-10-14T11:06:38.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-10-14T10:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag list""}], ""numBlocks"": 31, ""numErrors"": 1, ""wpTagCounts"": {""html"": 6, ""list"": 1, ""column"": 12, ""columns"": 6, ""heading"": 7, ""paragraph"": 28, ""owid/prominent-link"": 3}, ""htmlTagCounts"": {""p"": 28, ""h3"": 3, ""h4"": 2, ""h5"": 2, ""ul"": 1, ""div"": 18, ""iframe"": 6}}",2022-10-14 10:00:00,2024-02-16 14:22:54,1a7detrKx0jOujqXN9k0pj0wPiE-NnbC4r7xAUeimgiA,"[""Hannah Ritchie"", ""Pablo Rosado""]",Putting a price on carbon helps us account for the real costs of fossil fuels in the market. Which countries have a carbon tax or trading system?,2022-10-11 10:22:07,2022-10-14 11:06:38,https://ourworldindata.org/wp-content/uploads/2022/10/Carbon-taxes-featured-image.png,{},"The true cost of burning fossil fuels is not reflected in their market price. People [often compare](https://ourworldindata.org/cheap-renewables-growth) the _monetary price_ of fossil fuels to low-carbon alternatives such as renewables or nuclear energy. But these comparisons do not capture each option’s social impact, even in a purely economic sense. Burning fossil fuels drives [climate change](http://ourworldindata.org/climate-change). The impacts of climate change come at a cost – these impacts are already happening. They also cause local air pollution, which [kills millions](https://ourworldindata.org/data-review-air-pollution-deaths) every year and has negative health impacts for many more. One way to address this is to put a price on carbon. The purpose of setting a carbon price is to capture some of these external costs in the market. My colleague, Max Roser, wrote a detailed case for a carbon price [**here**](https://ourworldindata.org/carbon-price). Putting a price on carbon achieves a couple of things. First, it makes more polluting fuels, products, and services more expensive. Burning coal becomes [much more expensive](https://ourworldindata.org/safest-sources-of-energy) than using solar energy. Beef gets [more expensive](https://ourworldindata.org/explorers/food-footprints) relative to tofu or meat alternatives. It makes the ‘cleaner’ option less expensive. Second, it means it’s those who emit greenhouse gases that pay for it. There are a few policies through which countries can put a price on carbon: * **Carbon tax: **this is a levy that is applied to the production of greenhouse gas emissions directly or fuels that emit these gases when they’re burned. This means goods and services which emit more greenhouse gases in their production will have a higher tax. * **Emissions trading system: **this is sometimes called a ‘cap and trade’ system. Here, the carbon price changes over time. A maximum level of pollution (a ‘cap’) is defined and manufacturers need licenses to emit greenhouse gases. How expensive these licenses are is determined by a trading system. The price of a license increases as emissions approach the cap. Many countries have adopted carbon pricing instruments. In this article, we provide an overview of which countries have them, and how the price of carbon is changing over time. The underlying data – from the [World Carbon Pricing Dataset](https://www.rff.org/publications/working-papers/world-carbon-pricing-database-sources-and-methods/) – is updated annually.{ref}Dolphin, G., & Xiahou, Q. (2022). [World Carbon Pricing Database: Sources and Methods](https://www.nature.com/articles/s41597-022-01659-x). _Nature Scientific Data_.{/ref} It is assembled by Geoffroy Dolphin. We will keep the data in this article updated. ## Which countries have a carbon tax? The first chart shows which countries have implemented a carbon tax. It’s often the case that only specific sectors, or specific fuels, in a given country are subject to a carbon tax. For example, heavy industry or household electricity might have a carbon tax, but road transport might not. A country is coded as having a carbon tax if _any_ of its sectors have one. This means the country does not necessarily have an economy-wide tax in place. One final note: the World Carbon Pricing Database only looks at taxes applied to carbon dioxide (CO2) emissions. It does not consider taxes on non-CO2 greenhouse gases, such as methane or nitrous oxide.The true cost of burning fossil fuels is not reflected in their market price. People often compare the monetary price of fossil fuels to low-carbon alternatives such as renewables or nuclear energy. But these comparisons do not capture each option’s social impact, even in a purely economic sense.
\n\n\n\nBurning fossil fuels drives climate change. The impacts of climate change come at a cost – these impacts are already happening.
\n\n\n\nThey also cause local air pollution, which kills millions every year and has negative health impacts for many more.
\n\n\n\nOne way to address this is to put a price on carbon. The purpose of setting a carbon price is to capture some of these external costs in the market.
\n\n\n\nMy colleague, Max Roser, wrote a detailed case for a carbon price here.
\n\n\n\nPutting a price on carbon achieves a couple of things. First, it makes more polluting fuels, products, and services more expensive. Burning coal becomes much more expensive than using solar energy. Beef gets more expensive relative to tofu or meat alternatives. It makes the ‘cleaner’ option less expensive.
\n\n\n\nSecond, it means it’s those who emit greenhouse gases that pay for it.
\n\n\n\nThere are a few policies through which countries can put a price on carbon:
\n\n\n\nMany countries have adopted carbon pricing instruments.
\n\n\n\nIn this article, we provide an overview of which countries have them, and how the price of carbon is changing over time. The underlying data – from the World Carbon Pricing Dataset – is updated annually.{ref}Dolphin, G., & Xiahou, Q. (2022). World Carbon Pricing Database: Sources and Methods. Nature Scientific Data.{/ref} It is assembled by Geoffroy Dolphin. We will keep the data in this article updated.
\n\n\n\nThe first chart shows which countries have implemented a carbon tax.
\n\n\n\nIt’s often the case that only specific sectors, or specific fuels, in a given country are subject to a carbon tax. For example, heavy industry or household electricity might have a carbon tax, but road transport might not.
\n\n\n\nA country is coded as having a carbon tax if any of its sectors have one. This means the country does not necessarily have an economy-wide tax in place.
\n\n\n\nOne final note: the World Carbon Pricing Database only looks at taxes applied to carbon dioxide (CO2) emissions. It does not consider taxes on non-CO2 greenhouse gases, such as methane or nitrous oxide.
\nHaving a carbon tax does not guarantee that it will be effective. The level of carbon tax rates matters. It needs to be sufficiently expensive to de-incentivize carbon-emitting activities, and incentivize low-carbon ones.
\n\n\n\nIn the chart, we see the carbon price in countries that have implemented a tax.
\n\n\n\nSince carbon prices might vary from sector to sector, or some sectors will have a carbon price while others won’t, the authors calculate an ‘emissions-weighted’ carbon price. This price is weighted according to each sector’s contribution to a country’s CO2 emissions.
\nThe alternative to a carbon tax is a ‘cap and trade’ system. This is where the number of emission allowances/licenses, or the amount that a country or sector can emit, is fixed. Companies and sectors within the economy can then buy and sell carbon credits. This supply and demand determine the price of carbon in the marketplace.
\n\n\n\nThe most well-known ‘cap and trade’ system is the European Union’s Emissions Trading System (EU ETS). It was first introduced in 2005.
\n\n\n\nBut a number of other countries have implemented one at the national or sub-national level.
\n\n\n\nIn the chart, we see which countries have implemented an emissions trading system. Again, a country is coded as having a system if at least one sector is covered by one.
\nThe supply and demand for carbon credits within an economy determines the carbon price. If the amount of available credits is low compared to demand, then the price will be higher. If there are many available credits then the price will be low, and companies will not be incentivized to reduce their emissions. This means the carbon price changes over time.
\n\n\n\nIn the map, we see the carbon price in emissions trading systems across the world. Again, if the price varies by sector, these prices are weighted by each sector’s contribution to the country’s CO2 emissions.
\nIf we want the climate cost of fuels and products to be reflected in their market price we would want a carbon pricing mechanism everywhere.
\n\n\n\nIn 2020, around 12% of emissions were in countries or sectors that had a carbon tax. Just 6% were covered by a trading system. This means that, in total, a carbon price had to be paid on 18% of global emissions.
\n\n\n\nWe see the share of global CO2 emissions that are covered by each in the chart. The map also shows the share of emissions in each country that are covered by either a carbon tax or emissions trading system.
\n\n\n\nThis data is currently only shown to 2020. As of 2021, this coverage expanded dramatically – to 25% of global CO2 emissions – because China started its national emissions trading system. This will be reflected in the next update of this dataset.
\nMany thanks to Geoffroy Dolphin and Felix Pretis for the provision of data from the World Carbon Pricing Database, and feedback on this work. Thanks also to Max Roser, Edouard Mathieu and Bastian Herre for editorial feedback on this article.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Putting a price on carbon helps us account for the real costs of fossil fuels in the market. Which countries have a carbon tax or trading system?"", ""protected"": false}, ""date_gmt"": ""2022-10-14T10:00:00"", ""modified"": ""2022-10-14T12:06:38"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie"", ""Pablo Rosado""], ""modified_gmt"": ""2022-10-14T11:06:38"", ""comment_status"": ""closed"", ""featured_media"": 53529, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/10/Carbon-taxes-featured-image-150x79.png"", ""medium_large"": ""/app/uploads/2022/10/Carbon-taxes-featured-image-768x402.png""}}" 53401,Ocean plastics: How much do rich countries contribute by shipping their waste overseas?,plastic-waste-trade,post,publish,"Most of the plastic that enters the oceans from land comes from rivers in Asia.{ref}It’s estimated that around 70% to 80% of the plastic in the ocean comes from land. The other 20% to 30% comes from marine sources, such as fishing nets and lines.
In some parts of the ocean, the contribution of marine sources is higher. For example, it’s estimated that up to 52% of the plastic mass in the ‘Great Pacific Garbage Patch’ is plastic lines, ropes, and fishing nets.
Li, W. C., Tse, H. F., & Fok, L. (2016). Plastic waste in the marine environment: A review of sources, occurrence and effects. Science of the Total Environment, 566, 333-349.
Lebreton, L., Slat, B., Ferrari, F., Sainte-Rose, B., Aitken, J., Marthouse, R., … & Noble, K. (2018). Evidence that the Great Pacific Garbage Patch is rapidly accumulating plastic. Scientific Reports, 8(1), 4666.{/ref} More than 80% of it.{ref}One of the most recent estimates from Meijer et al. (2021) estimates that 81% of plastic waste that is emitted to the ocean comes from Asia.
An earlier study, by Lebreton et al. (2017), estimated a similar figure of 86% coming from Asian rivers.
Meijer, J.J.L, Emmerik, T., Ent, R., Schmidt, C., Lebreton, L. (2021). More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean. Science Advances.
Lebreton, L. C., Van der Zwet, J., Damsteeg, J. W., Slat, B., Andrady, A., & Reisser, J. (2017). River plastic emissions to the world’s oceans. Nature Communications, 8, 15611.{/ref}
Only a small amount comes from rivers across Europe and North America. Together, rivers in these regions contribute just 5% of the global total. This would suggest that the world’s richest countries don’t contribute much to the problem of plastic pollution.
But, these numbers only look at the plastic that is emitted domestically. They don’t consider the fact that many countries export plastic waste overseas. If it was the case that the UK exported a lot of its plastic waste to countries where waste management systems are poor, and lots of plastic leaks into the environment, the UK would have a large indirect impact on ocean pollution.
Here I use global data to understand the scale of plastic waste trade. I look at who the biggest exporters and importers are, and where this waste ends up.
I estimate that a few percent – possibly up to 5% – of the world’s ocean plastics could come from rich countries exporting their waste overseas. That could bring the total up to 10%: 5% directly from rivers in these regions, plus a further 5% from trade.
Importing plastics can often bring economic benefits. Recycled plastics can be repurposed into other goods, and fed into manufacturing industries. This is often cheaper than buying or making virgin plastics from scratch.
In 2020, around 5 million tonnes of plastic waste was traded globally.{ref}Here I’m using data on plastic trade from the UN Comtrade Database. Data is sourced from the commodity category: ‘3915 – Waste, parings, and scrap, of plastics’.{/ref} We might imagine that the pandemic forced a large reduction in plastic trade, but this doesn’t seem to be the case. In 2019, rates were only slightly higher, at around 6 million tonnes.
Let’s put those 5 million tonnes into context.
The world generates around 350 million tonnes of plastic waste per year. That means that around 2% of waste is traded.{ref}5 million is 2% of 350 million.{/ref}
The remaining 98% is handled domestically. It’s sent to a landfill, recycled, or incinerated in the country where the waste was generated. The idea that most of the world’s plastic waste is shipped overseas is incorrect. One reason why this figure is so low is that it’s mostly recycled waste that’s traded, and only 20% of the world’s plastic is recycled.{ref}Geyer, R., Jambeck, J. R., & Law, K. L. (2017). Production, use, and fate of all plastics ever made. Science Advances, 3(7), e1700782.{/ref}
Over the last decade, we’ve seen a large decline in the amount of plastic waste traded. Rates have fallen by two-thirds since 2010.
Policies in China toward plastic trade have had a large impact on the global change shown in the previous chart.
In 2016, China was importing more than half of the world’s traded plastic waste. By 2018, this had plummeted to less than 1%. We see this in the chart.
The reason for this dramatic decline was the Chinese government banning the import of most types of plastic waste in 2017. This was part of a broader policy decision to stop the import of 24 different types of solid waste including paper, textiles, and plastics. These bans were implemented as a result of environmental and health concerns from contaminated waste streams.
This ban had two major impacts.{ref}Wen, Z., Xie, Y., Chen, M., & Dinga, C. D. (2021). China’s plastic import ban increases prospects of environmental impact mitigation of plastic waste trade flow worldwide. Nature Communications, 12(1), 1-9.
Brooks, A. L., Wang, S., & Jambeck, J. R. (2018). The Chinese import ban and its impact on global plastic waste trade. Science Advances, 4(6).{/ref} The first was that the total volume of plastic trade globally dropped significantly – we saw earlier that global trade has halved since 2017. The second was that other countries emerged to take China’s place as major importers. Most of them are also countries in Asia – Malaysia, Vietnam, Indonesia, the Philippines, and Turkey, started to import much more plastic than in previous years.
Where does plastic waste flow across the world?
Europe is the region that exports the most plastic, but it's also the region that imports the most.
This is true more generally. Most plastic is traded within a given region. European countries export most plastic to other European countries. Asian countries export most to other Asian countries.
In the visualization, we see the flow of plastic across the world.{ref}This is based on data from the OECD report:
Brown, A., F. Laubinger and P. Börkey (2022), ""Monitoring trade in plastic waste and scrap"", OECD Environment Working Papers, No. 194, OECD Publishing, Paris.{/ref} On the left we have the exporters of plastic waste; on the right, we see where that plastic ends up. The height of each bar is proportional to the amount of plastic that is traded.
Europe is the biggest exporter of plastic. But, it’s also the biggest importer. Many countries across Europe trade with one another. At the national level, Germany is the biggest exporter and importer – it trades different plastics with its neighbors including the Netherlands, Poland, Austria, and Switzerland.{ref}Brown, A., F. Laubinger and P. Börkey (2022), ""Monitoring trade in plastic waste and scrap"", OECD Environment Working Papers, No. 194, OECD Publishing, Paris.{/ref}
This is also true of Asia, where Japan is the biggest exporter to other countries in Asia.
Many people think that rich countries ship most of their plastic waste overseas. But is this really true?
The short answer is no: many countries export some of their waste, but they still handle most of it domestically.
Let’s take the example of the UK. In 2010, it generated an estimated 4.93 million tonnes of plastic waste.{ref}Jambeck, J. R., Geyer, R., Wilcox, C., Siegler, T. R., Perryman, M., Andrady, A., ... & Law, K. L. (2015). Plastic waste inputs from land into the ocean. Science, 347(6223), 768-771.{/ref} It exported 838,000 tonnes overseas.
That means it exported about 17% of its plastic waste. That’s a substantial fraction – nearly one-fifth of it.
This data is for 2010, a year with good high-quality estimates of plastic waste generation. It’s still likely to be a reasonable estimate today. If anything, this share might have declined slightly, because waste exports have not increased, and waste generation probably has.
When it comes to the fraction of plastic waste that is exported, the UK is one of the largest exporters. For context, the US exported about 5% of its plastic waste in 2010. France exported 11%, and the Netherlands exported 14%.
Most rich countries are net exporters of plastic waste. And this can be a significant fraction of their waste. But it’s not the case that they handle most of it by offshoring it to other countries.
This is the crucial question. While we don’t have an exact answer, we can give a plausible range.
To give an exact answer we would need to trace each piece of plastic pollution back to its original source. But we can do some calculations to estimate how much plastic is at higher risk of entering the ocean because of this trade.
In 2020, low-to-middle-income countries – where plastic waste was at a ‘higher risk’ of entering the ocean (because of poorer waste management systems) – imported around 1.6 million tonnes of plastic waste from rich countries. Here ‘rich countries’ include all countries in Europe and North America, plus Japan, Hong Kong, and OECD countries from other regions.{ref}A report by the OECD provides a summary of trade flows of plastic waste across the world. We saw this in the Sankey diagram, earlier in the post. The authors use data from the UN Comtrade database to calculate these figures.
From these figures, we see that in 2020, non-OECD countries in Asia imported around 1.9 million tonnes of plastic waste. ‘Rest of the World’, which in this case is mostly lower-income countries across Africa and South America imported 0.12 million tonnes, and China also imported 0.12 million tonnes. Combined, these low-to-middle-income countries imported 2.14 million tonnes.
0.57 million tonnes of this came from countries within this group i.e. low-to-middle-income countries.
That leaves around 1.6 million tonnes that come from richer countries, which is the sum of Europe, North America, Hong Kong, and Japan.{/ref}
How much of this plastic ends up in the ocean?
Again, we don’t know for sure. But we can run through a worst and best-case scenario.
Here we will assume that all of this traded waste was ‘mismanaged’, meaning it was not formally managed and was either littered or dumped in open landfills. In reality, not all of it will be mismanaged, but let’s be conservative here.
The probability that mismanaged waste ends up in the ocean varies a lot by country.
The country where the probability is highest is the Phillippines – an estimated 7% ends up in the ocean.{ref}This estimate comes from the work of Meijer et al. (2021), published in Science Advances.
Meijer, L. J., van Emmerik, T., van der Ent, R., Schmidt, C., & Lebreton, L. (2021). More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean. Science Advances, 7(18), eaaz5803.{/ref} We could imagine this being our ‘worst-case’ scenario: if rich countries exported all of their plastic trade to the Philippines, 7% of it might end up in the ocean. That would be 112,000 tonnes.{ref}We can calculate this as 7% of 1.6 million tonnes, which is around 112,000 tonnes.{/ref}
In a ‘best-case’ scenario, only around 1% of mismanaged waste would end up in the ocean. Most countries across the world have a risk of just under 1%. In Asia, this would be typical of countries such as Thailand and Cambodia. In this ‘best-case’ scenario, around 16,000 tonnes of ocean plastics each year would enter the ocean from trade.{ref}We can calculate this figure as 1% of 1.6 million tonnes, which gives us 16,000 tonnes.{/ref}
This gives us an upper and lower bound for the contribution of trade from rich countries. Since around one million tonnes of plastic enters the ocean each year, rich countries would contribute between 1.6% (in the best case) and 11% (in the worst case) of ocean plastics through shipping waste overseas.{ref}Meijer et al. (2021) estimate that around 1 million tonnes of plastic are emitted into the oceans each year. They put the uncertainty range on this between 0.8 and 2.6 million tonnes.
Meijer, J.J.L, Emmerik, T., Ent, R., Schmidt, C., Lebreton, L. (2021). More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean. Science Advances.
We get these figures by calculating 16,000 and 112,000 tonnes as a share of 1 million. That comes to 1.6% and 11%.{/ref}
The true figure probably falls somewhere in between. A reasonable estimate might be around 5% of ocean plastics. In reality, it might be a bit lower because a tonne of waste that is bought and traded is more likely to be managed well than the average tonne of waste in a country.
I estimate that a few percent of ocean plastics could result from trade from rich countries. A figure as high as 5% would not be unreasonable.
Stopping exports of plastic waste to countries with poor waste management would help to tackle ocean pollution.
If rich countries banned the export of plastic waste to these countries, we might reduce plastic pollution a bit: perhaps up to 5%.
But, an end to trade won’t stop plastic pollution. Only a small fraction of the world’s plastic waste is traded – under 2%. And most – two-thirds of it – ends up in richer countries, where it’s very unlikely to end up in the ocean.
There are obvious reasons to reduce these exports beyond the plastic pollution problem. Countries have been guilty of exporting contaminated recycling plastic packages – one of the drivers for countries to ban plastic imports. This is unacceptable: poorer countries are not a dumping ground for the rich.
Most of the world’s waste is handled domestically and most of the waste that enters the oceans stems from these countries. To really tackle the problem we need to do two things: scale waste management systems in rich countries; the fact that they are exporting waste overseas suggests they have under-invested in practices at home; and, importantly, improve waste management infrastructure and practices in low-to-middle-income countries, as this is where most plastic pollution originates.
Many thanks to Max Roser, Edouard Mathieu, and Bastian Herre for feedback and suggestions on this article.
","{""id"": ""wp-53401"", ""slug"": ""plastic-waste-trade"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Most of the plastic that enters the oceans from land "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ocean-plastics"", ""children"": [{""text"": ""comes from"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" rivers in Asia.{ref}It’s estimated that around 70% to 80% of the plastic in the ocean comes from land. The other 20% to 30% comes from marine sources, such as fishing nets and lines."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""In some parts of the ocean, the contribution of marine sources is higher. For example, it’s estimated that up to 52% of the plastic mass in the ‘Great Pacific Garbage Patch’ is plastic lines, ropes, and fishing nets."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""Li, W. C., Tse, H. F., & Fok, L. (2016). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.sciencedirect.com/science/article/pii/S0048969716310154"", ""children"": [{""text"": ""Plastic waste in the marine environment: A review of sources, occurrence and effects"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science of the Total Environment"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 566, 333-349."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Lebreton, L., Slat, B., Ferrari, F., Sainte-Rose, B., Aitken, J., Marthouse, R., … & Noble, K. (2018). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/s41598-018-22939-w"", ""children"": [{""text"": ""Evidence that the Great Pacific Garbage Patch is rapidly accumulating plastic"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Scientific Reports"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 8(1), 4666.{/ref} More than 80% of it.{ref}One of the most recent estimates from Meijer et al. (2021) estimates that 81% of plastic waste that is emitted to the ocean comes from Asia."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""An earlier study, by Lebreton et al. (2017), estimated a similar figure of 86% coming from Asian rivers."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Meijer, J.J.L, Emmerik, T., Ent, R., Schmidt, C., Lebreton, L. (2021). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://advances.sciencemag.org/content/7/18/eaaz5803"", ""children"": [{""text"": ""More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science Advances"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Lebreton, L. C., Van der Zwet, J., Damsteeg, J. W., Slat, B., Andrady, A., & Reisser, J. (2017). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/ncomms15611"", ""children"": [{""text"": ""River plastic emissions to the world’s oceans"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Nature Communications"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 8, 15611.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Only a small amount comes from rivers across Europe and North America. Together, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/share-of-global-plastic-waste-emitted-to-the-ocean?tab=chart&country=Africa~Asia~Europe~South+America~North+America~Oceania"", ""children"": [{""text"": ""rivers in these regions contribute"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" just 5% of the global total. This would suggest that the world’s richest countries don’t contribute much to the problem of plastic pollution."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But, these numbers only look at the plastic that is emitted "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""domestically"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "". They don’t consider the fact that many countries export plastic waste overseas. If it was the case that the UK exported a lot of its plastic waste to countries where waste management systems are poor, and lots of plastic leaks into the environment, the UK would have a large indirect impact on ocean pollution."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Here I use global data to understand the scale of plastic waste trade. I look at who the biggest exporters and importers are, and where this waste ends up."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I estimate that a few percent – possibly up to 5% – of the world’s ocean plastics could come from rich countries exporting their waste overseas. That could bring the total up to 10%: 5% directly from rivers in these regions, plus a further 5% from trade."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How much of the world’s plastic waste is traded?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Importing plastics can often bring economic benefits. Recycled plastics can be repurposed into other goods, and fed into manufacturing industries. This is often cheaper than buying or making virgin plastics from scratch."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In 2020, around 5 million tonnes of plastic waste was traded globally.{ref}Here I’m using data on plastic trade from the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://comtrade.un.org/data/"", ""children"": [{""text"": ""UN Comtrade Database"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Data is sourced from the commodity category: ‘3915 – Waste, parings, and scrap, of plastics’.{/ref} We might imagine that the pandemic forced a large reduction in plastic trade, but this doesn’t seem to be the case. In 2019, rates were only slightly higher, at around 6 million tonnes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Let’s put those 5 million tonnes into context."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The world generates around 350 million "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/plastic-waste-by-sector"", ""children"": [{""text"": ""tonnes of plastic waste"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" per year. That means that around 2% of waste is traded.{ref}5 million is 2% of 350 million.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The remaining 98% is handled domestically. It’s sent to a landfill, recycled, or incinerated in the country where the waste was generated. The idea that "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""most"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of the world’s plastic waste is shipped overseas is incorrect. One reason why this figure is so low is that it’s mostly recycled waste that’s traded, and only 20% of the world’s plastic is recycled.{ref}Geyer, R., Jambeck, J. R., & Law, K. L. (2017). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.science.org/doi/10.1126/sciadv.1700782"", ""children"": [{""text"": ""Production, use, and fate of all plastics ever made"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science Advances"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 3(7), e1700782.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Over the last decade, we’ve seen a large decline in the amount of plastic waste traded. Rates have fallen by two-thirds since 2010."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/explorers/plastic-pollution?time=2007..latest&facet=none&country=~OWID_WRL&Metric=Waste+trade&Per+capita=false&Share+of+world+total=false&Sub-metric=Exports&hideControls=true"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""What was the impact of the Chinese ban on plastic trade?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Policies in China toward plastic trade have had a large impact on the global change shown in the previous chart."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In 2016, China was importing more than half of the world’s traded plastic waste. By 2018, this had plummeted to less than 1%. We see this in the chart."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The reason for this dramatic decline was the Chinese government banning the import of most types of plastic waste in 2017. This was part of a broader policy decision to stop the import of 24 different types of solid waste including paper, textiles, and plastics. These bans were implemented as a result of environmental and health concerns from contaminated waste streams."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This ban had two major impacts.{ref}Wen, Z., Xie, Y., Chen, M., & Dinga, C. D. (2021). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.nature.com/articles/s41467-020-20741-9"", ""children"": [{""text"": ""China’s plastic import ban increases prospects of environmental impact mitigation of plastic waste trade flow worldwide"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Nature Communications"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 12(1), 1-9."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Brooks, A. L., Wang, S., & Jambeck, J. R. (2018). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.science.org/doi/10.1126/sciadv.aat0131"", ""children"": [{""text"": ""The Chinese import ban and its impact on global plastic waste trade"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science Advances"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "", 4(6).{/ref} The first was that the total volume of plastic trade globally dropped significantly – we saw earlier that global trade has halved since 2017. The second was that other countries emerged to take China’s place as major importers. Most of them are also countries in Asia – Malaysia, Vietnam, Indonesia, the Philippines, and Turkey, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://owid.cloud/admin/explorers/preview/plastic-pollution?time=earliest..2018&facet=none&country=MYS~IDN~TUR~VNM&Metric=Waste+trade&Sub-metric=Imports&Per+capita=false&Share+of+world+total=false"", ""children"": [{""text"": ""started to import"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" much more plastic than in previous years."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""left"": [{""url"": ""https://ourworldindata.org/explorers/plastic-pollution?time=2007..latest&facet=none&country=~CHN&Metric=Waste+trade&Per+capita=false&Share+of+world+total=false&Sub-metric=Imports&hideControls=true"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/explorers/plastic-pollution?time=2007..2018&facet=none&country=MYS~TUR~VNM~IDN&Metric=Waste+trade&Per+capita=false&Share+of+world+total=false&Sub-metric=Imports&hideControls=true"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Most plastic waste is traded within world regions, rather than between them"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Where does plastic waste flow across the world?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Europe is the region that exports the most plastic, but it's also the region that imports the most."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is true more generally. Most plastic is traded "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""within"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" a given region. European countries export most plastic to other European countries. Asian countries export most to other Asian countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the visualization, we see the flow of plastic across the world.{ref}This is based on data from the OECD report:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Brown, A., F. Laubinger and P. Börkey (2022), \"""", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.oecd-ilibrary.org/environment/monitoring-trade-in-plastic-waste-and-scrap_8f3e9c56-en"", ""children"": [{""text"": ""Monitoring trade in plastic waste and scrap"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""\"", OECD Environment Working Papers, No. 194, OECD Publishing, Paris.{/ref} On the left we have the exporters of plastic waste; on the right, we see where that plastic ends up. The height of each bar is proportional to the amount of plastic that is traded."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Europe is the biggest exporter of plastic. But, it’s also the biggest importer. Many countries across Europe trade with one another. At the national level, Germany is the biggest exporter "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""and"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" importer – it trades different plastics with its neighbors including the Netherlands, Poland, Austria, and Switzerland.{ref}Brown, A., F. Laubinger and P. Börkey (2022), \"""", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.oecd-ilibrary.org/environment/monitoring-trade-in-plastic-waste-and-scrap_8f3e9c56-en"", ""children"": [{""text"": ""Monitoring trade in plastic waste and scrap"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""\"", OECD Environment Working Papers, No. 194, OECD Publishing, Paris.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is also true of Asia, where Japan is the biggest exporter to other countries in Asia."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""Plastic-waste-trade-sankey.png"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Do rich countries export most of their plastic waste overseas?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many people think that rich countries ship "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""most"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of their plastic waste overseas. But is this really true?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The short answer is no: many countries export some of their waste, but they still handle most of it domestically."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Let’s take the example of the UK. In 2010, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/plastic-waste-generation-total"", ""children"": [{""text"": ""it generated"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" an estimated 4.93 million tonnes of plastic waste.{ref}Jambeck, J. R., Geyer, R., Wilcox, C., Siegler, T. R., Perryman, M., Andrady, A., ... & Law, K. L. (2015). Plastic waste inputs from land into the ocean. Science, 347(6223), 768-771.{/ref} It exported 838,000 tonnes overseas. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That means it exported about 17% of its plastic waste. That’s a substantial fraction – nearly one-fifth of it."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data is for 2010, a year with good high-quality estimates of plastic waste generation. It’s still likely to be a reasonable estimate today. If anything, this share might have declined slightly, because waste exports have not increased, and waste generation probably has."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""When it comes to the fraction of plastic waste that is exported, the UK is one of the largest exporters. For context, the US exported about 5% of its plastic waste in 2010. France exported 11%, and the Netherlands exported 14%. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Most rich countries "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""are"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" net exporters of plastic waste. And this can be a significant fraction of their waste. But it’s not the case that they handle "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""most"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of it by offshoring it to other countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How much do rich countries contribute to plastic pollution through their exported waste?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is the crucial question. While we don’t have an exact answer, we can give a plausible range."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To give an exact answer we would need to trace each piece of plastic pollution back to its original source. But we can do some calculations to estimate how much plastic is at "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""higher risk"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of entering the ocean because of this trade. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In 2020, low-to-middle-income countries – where plastic waste was at a ‘higher risk’ of entering the ocean (because of poorer waste management systems) – imported around 1.6 million tonnes of plastic waste from rich countries. Here ‘rich countries’ include all countries in Europe and North America, plus Japan, Hong Kong, and OECD countries from other regions.{ref}A "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.oecd-ilibrary.org/environment/monitoring-trade-in-plastic-waste-and-scrap_8f3e9c56-en"", ""children"": [{""text"": ""report by the OECD"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" provides a summary of trade flows of plastic waste across the world. We saw this in the Sankey diagram, earlier in the post. The authors use data from the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://comtradeplus.un.org/"", ""children"": [{""text"": ""UN Comtrade database"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to calculate these figures."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""From these figures, we see that in 2020, non-OECD countries in Asia imported around 1.9 million tonnes of plastic waste. ‘Rest of the World’, which in this case is mostly lower-income countries across Africa and South America imported 0.12 million tonnes, and China also imported 0.12 million tonnes. Combined, these low-to-middle-income countries imported 2.14 million tonnes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""0.57 million tonnes of this came from countries "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""within"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" this group i.e. low-to-middle-income countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""That leaves around 1.6 million tonnes that come from richer countries, which is the sum of Europe, North America, Hong Kong, and Japan.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How much of this plastic ends up in the ocean?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Again, we don’t know for sure. But we can run through a worst and best-case scenario."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Here we will assume that all of this traded waste was ‘mismanaged’, meaning it was not formally managed and was either littered or dumped in open landfills. In reality, not all of it will be mismanaged, but let’s be conservative here."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/probability-mismanaged-plastic-ocean"", ""children"": [{""text"": ""probability that"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" mismanaged waste ends up in the ocean varies a lot by country. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The country where the probability is highest is the Phillippines – an estimated 7% ends up in the ocean.{ref}This estimate comes from the work of Meijer et al. (2021), published in "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science Advances"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Meijer, L. J., van Emmerik, T., van der Ent, R., Schmidt, C., & Lebreton, L. (2021). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.science.org/doi/10.1126/sciadv.aaz5803"", ""children"": [{""text"": ""More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Science Advances, 7(18), eaaz5803.{/ref} We could imagine this being our ‘worst-case’ scenario: if rich countries exported "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""all"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of their plastic trade to the Philippines, 7% of it might end up in the ocean. That would be 112,000 tonnes.{ref}We can calculate this as 7% of 1.6 million tonnes, which is around 112,000 tonnes.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In a ‘best-case’ scenario, only around 1% of mismanaged waste would end up in the ocean. Most countries across the world have a risk of just under 1%. In Asia, this would be typical of countries such as Thailand and Cambodia. In this ‘best-case’ scenario, around 16,000 tonnes of ocean plastics each year would enter the ocean from trade.{ref}We can calculate this figure as 1% of 1.6 million tonnes, which gives us 16,000 tonnes.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This gives us an upper and lower bound for the contribution of trade from rich countries. Since around "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/ocean-plastics"", ""children"": [{""text"": ""one million tonnes"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of plastic enters the ocean each year, rich countries would contribute between 1.6% (in the best case) and 11% (in the worst case) of ocean plastics through shipping waste overseas.{ref}Meijer et al. (2021) estimate that around 1 million tonnes of plastic are emitted into the oceans each year. They put the uncertainty range on this between 0.8 and 2.6 million tonnes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Meijer, J.J.L, Emmerik, T., Ent, R., Schmidt, C., Lebreton, L. (2021). "", ""spanType"": ""span-simple-text""}, {""url"": ""https://advances.sciencemag.org/content/7/18/eaaz5803"", ""children"": [{""text"": ""More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Science Advances"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We get these figures by calculating 16,000 and 112,000 tonnes as a share of 1 million. That comes to 1.6% and 11%.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The true figure probably falls somewhere in between. A reasonable estimate might be around 5% of ocean plastics. In reality, it might be a bit lower because a tonne of waste that is bought and traded is more likely to be managed well than the average tonne of waste in a country."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""I estimate that a few percent of ocean plastics could result from trade from rich countries. A figure as high as 5% would not be unreasonable."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""text"": [{""text"": ""Ending plastic trade would only do a bit to reduce plastic pollution – what is needed are better waste-management systems"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""left"": [{""type"": ""text"", ""value"": [{""text"": ""Stopping exports of plastic waste to countries with poor waste management would help to tackle ocean pollution. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""If rich countries banned the export of plastic waste to these countries, we might reduce plastic pollution a bit: perhaps up to 5%."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But, an end to trade won’t stop plastic pollution. Only a small fraction of the world’s plastic waste is traded – under 2%. And most – two-thirds of it – ends up in richer countries, where it’s very unlikely to end up in the ocean."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are obvious reasons to reduce these exports beyond the plastic pollution problem. Countries "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.bbc.co.uk/news/world-48444874"", ""children"": [{""text"": ""have been guilty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of exporting contaminated recycling plastic packages – one of the drivers for countries to ban plastic imports. This is unacceptable: poorer countries are not a dumping ground for the rich."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Most of the world’s waste is handled domestically and most of the waste that enters the oceans stems from these countries. To really tackle the problem we need to do two things: scale waste management systems in rich countries; the fact that they are exporting waste overseas suggests they have under-invested in practices at home; and, importantly, improve waste management infrastructure and practices in low-to-middle-income countries, as this is where most plastic pollution originates."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""sticky-right"", ""right"": [{""url"": ""https://ourworldindata.org/explorers/plastic-pollution?time=latest&facet=none&Metric=Waste+trade&Per+capita=false&Share+of+world+total=true&Sub-metric=Imports&country=Low-income+countries~High-income+countries~Lower-middle-income+countries~Upper-middle-income+countries&hideControls=true"", ""type"": ""chart"", ""parseErrors"": []}], ""parseErrors"": []}, {""text"": [{""text"": ""Keep reading at "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Our World in Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""..."", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 4, ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/plastic-pollution"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/ocean-plastics"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/global-plastics-production"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""text"": [{""text"": ""Acknowledgments"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 4, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many thanks to Max Roser, Edouard Mathieu, and Bastian Herre for feedback and suggestions on this article."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Ocean plastics: How much do rich countries contribute by shipping their waste overseas?"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""Many countries ship plastic waste overseas. How much of the world’s waste is traded, and how big is its role in the pollution of our oceans?"", ""dateline"": ""October 11, 2022"", ""subtitle"": ""Many countries ship plastic waste overseas. How much of the world’s waste is traded, and how big is its role in the pollution of our oceans?"", ""sidebar-toc"": false, ""featured-image"": ""Plastic-waste-trade-featured-image.png""}, ""createdAt"": ""2022-10-06T10:50:21.000Z"", ""published"": false, ""updatedAt"": ""2022-10-25T12:06:49.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-10-11T09:50:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}, {""name"": ""unhandled html tag found"", ""details"": ""Encountered the unhandled tag hr""}, {""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag separator""}], ""numBlocks"": 55, ""numErrors"": 3, ""wpTagCounts"": {""html"": 4, ""image"": 1, ""column"": 14, ""columns"": 7, ""heading"": 8, ""paragraph"": 59, ""separator"": 1, ""owid/prominent-link"": 3}, ""htmlTagCounts"": {""p"": 59, ""h3"": 5, ""h4"": 1, ""h5"": 2, ""hr"": 1, ""div"": 21, ""figure"": 1, ""iframe"": 4}}",2022-10-11 09:50:00,2024-02-16 14:22:54,1Zz6fB2VxqBBtsSSDN8grd8AhfLRjNHfhU43ksS2Qk-0,"[""Hannah Ritchie""]","Many countries ship plastic waste overseas. How much of the world’s waste is traded, and how big is its role in the pollution of our oceans?",2022-10-06 10:50:21,2022-10-25 12:06:49,https://ourworldindata.org/wp-content/uploads/2022/10/Plastic-waste-trade-featured-image.png,{},"Most of the plastic that enters the oceans from land [comes from](https://ourworldindata.org/ocean-plastics) rivers in Asia.{ref}It’s estimated that around 70% to 80% of the plastic in the ocean comes from land. The other 20% to 30% comes from marine sources, such as fishing nets and lines. In some parts of the ocean, the contribution of marine sources is higher. For example, it’s estimated that up to 52% of the plastic mass in the ‘Great Pacific Garbage Patch’ is plastic lines, ropes, and fishing nets. Li, W. C., Tse, H. F., & Fok, L. (2016). [Plastic waste in the marine environment: A review of sources, occurrence and effects](https://www.sciencedirect.com/science/article/pii/S0048969716310154). _Science of the Total Environment_, 566, 333-349. Lebreton, L., Slat, B., Ferrari, F., Sainte-Rose, B., Aitken, J., Marthouse, R., … & Noble, K. (2018). [Evidence that the Great Pacific Garbage Patch is rapidly accumulating plastic](https://www.nature.com/articles/s41598-018-22939-w). _Scientific Reports_, 8(1), 4666.{/ref} More than 80% of it.{ref}One of the most recent estimates from Meijer et al. (2021) estimates that 81% of plastic waste that is emitted to the ocean comes from Asia. An earlier study, by Lebreton et al. (2017), estimated a similar figure of 86% coming from Asian rivers. Meijer, J.J.L, Emmerik, T., Ent, R., Schmidt, C., Lebreton, L. (2021). [More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean](https://advances.sciencemag.org/content/7/18/eaaz5803). _Science Advances_. Lebreton, L. C., Van der Zwet, J., Damsteeg, J. W., Slat, B., Andrady, A., & Reisser, J. (2017). [River plastic emissions to the world’s oceans](https://www.nature.com/articles/ncomms15611). _Nature Communications_, 8, 15611.{/ref} Only a small amount comes from rivers across Europe and North America. Together, [rivers in these regions contribute](https://ourworldindata.org/grapher/share-of-global-plastic-waste-emitted-to-the-ocean?tab=chart&country=Africa~Asia~Europe~South+America~North+America~Oceania) just 5% of the global total. This would suggest that the world’s richest countries don’t contribute much to the problem of plastic pollution. But, these numbers only look at the plastic that is emitted _domestically_. They don’t consider the fact that many countries export plastic waste overseas. If it was the case that the UK exported a lot of its plastic waste to countries where waste management systems are poor, and lots of plastic leaks into the environment, the UK would have a large indirect impact on ocean pollution. Here I use global data to understand the scale of plastic waste trade. I look at who the biggest exporters and importers are, and where this waste ends up. I estimate that a few percent – possibly up to 5% – of the world’s ocean plastics could come from rich countries exporting their waste overseas. That could bring the total up to 10%: 5% directly from rivers in these regions, plus a further 5% from trade. ## How much of the world’s plastic waste is traded? Importing plastics can often bring economic benefits. Recycled plastics can be repurposed into other goods, and fed into manufacturing industries. This is often cheaper than buying or making virgin plastics from scratch. In 2020, around 5 million tonnes of plastic waste was traded globally.{ref}Here I’m using data on plastic trade from the [UN Comtrade Database](https://comtrade.un.org/data/). Data is sourced from the commodity category: ‘3915 – Waste, parings, and scrap, of plastics’.{/ref} We might imagine that the pandemic forced a large reduction in plastic trade, but this doesn’t seem to be the case. In 2019, rates were only slightly higher, at around 6 million tonnes. Let’s put those 5 million tonnes into context. The world generates around 350 million [tonnes of plastic waste](https://ourworldindata.org/grapher/plastic-waste-by-sector) per year. That means that around 2% of waste is traded.{ref}5 million is 2% of 350 million.{/ref} The remaining 98% is handled domestically. It’s sent to a landfill, recycled, or incinerated in the country where the waste was generated. The idea that _most_ of the world’s plastic waste is shipped overseas is incorrect. One reason why this figure is so low is that it’s mostly recycled waste that’s traded, and only 20% of the world’s plastic is recycled.{ref}Geyer, R., Jambeck, J. R., & Law, K. L. (2017). [Production, use, and fate of all plastics ever made](https://www.science.org/doi/10.1126/sciadv.1700782). _Science Advances_, 3(7), e1700782.{/ref} Over the last decade, we’ve seen a large decline in the amount of plastic waste traded. Rates have fallen by two-thirds since 2010.Most of the plastic that enters the oceans from land comes from rivers in Asia.{ref}It’s estimated that around 70% to 80% of the plastic in the ocean comes from land. The other 20% to 30% comes from marine sources, such as fishing nets and lines.
In some parts of the ocean, the contribution of marine sources is higher. For example, it’s estimated that up to 52% of the plastic mass in the ‘Great Pacific Garbage Patch’ is plastic lines, ropes, and fishing nets.
Li, W. C., Tse, H. F., & Fok, L. (2016). Plastic waste in the marine environment: A review of sources, occurrence and effects. Science of the Total Environment, 566, 333-349.
Lebreton, L., Slat, B., Ferrari, F., Sainte-Rose, B., Aitken, J., Marthouse, R., … & Noble, K. (2018). Evidence that the Great Pacific Garbage Patch is rapidly accumulating plastic. Scientific Reports, 8(1), 4666.{/ref} More than 80% of it.{ref}One of the most recent estimates from Meijer et al. (2021) estimates that 81% of plastic waste that is emitted to the ocean comes from Asia.
\n\n\n\nAn earlier study, by Lebreton et al. (2017), estimated a similar figure of 86% coming from Asian rivers.
\n\n\n\nMeijer, J.J.L, Emmerik, T., Ent, R., Schmidt, C., Lebreton, L. (2021). More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean. Science Advances.
\n\n\n\nLebreton, L. C., Van der Zwet, J., Damsteeg, J. W., Slat, B., Andrady, A., & Reisser, J. (2017). River plastic emissions to the world’s oceans. Nature Communications, 8, 15611.{/ref}
\n\n\n\nOnly a small amount comes from rivers across Europe and North America. Together, rivers in these regions contribute just 5% of the global total. This would suggest that the world’s richest countries don’t contribute much to the problem of plastic pollution.
\n\n\n\nBut, these numbers only look at the plastic that is emitted domestically. They don’t consider the fact that many countries export plastic waste overseas. If it was the case that the UK exported a lot of its plastic waste to countries where waste management systems are poor, and lots of plastic leaks into the environment, the UK would have a large indirect impact on ocean pollution.
\n\n\n\nHere I use global data to understand the scale of plastic waste trade. I look at who the biggest exporters and importers are, and where this waste ends up.
\n\n\n\nI estimate that a few percent – possibly up to 5% – of the world’s ocean plastics could come from rich countries exporting their waste overseas. That could bring the total up to 10%: 5% directly from rivers in these regions, plus a further 5% from trade.
\n\n\n\nImporting plastics can often bring economic benefits. Recycled plastics can be repurposed into other goods, and fed into manufacturing industries. This is often cheaper than buying or making virgin plastics from scratch.
\n\n\n\nIn 2020, around 5 million tonnes of plastic waste was traded globally.{ref}Here I’m using data on plastic trade from the UN Comtrade Database. Data is sourced from the commodity category: ‘3915 – Waste, parings, and scrap, of plastics’.{/ref} We might imagine that the pandemic forced a large reduction in plastic trade, but this doesn’t seem to be the case. In 2019, rates were only slightly higher, at around 6 million tonnes.
\n\n\n\nLet’s put those 5 million tonnes into context.
\n\n\n\nThe world generates around 350 million tonnes of plastic waste per year. That means that around 2% of waste is traded.{ref}5 million is 2% of 350 million.{/ref}
\n\n\n\nThe remaining 98% is handled domestically. It’s sent to a landfill, recycled, or incinerated in the country where the waste was generated. The idea that most of the world’s plastic waste is shipped overseas is incorrect. One reason why this figure is so low is that it’s mostly recycled waste that’s traded, and only 20% of the world’s plastic is recycled.{ref}Geyer, R., Jambeck, J. R., & Law, K. L. (2017). Production, use, and fate of all plastics ever made. Science Advances, 3(7), e1700782.{/ref}
\n\n\n\nOver the last decade, we’ve seen a large decline in the amount of plastic waste traded. Rates have fallen by two-thirds since 2010.
\nPolicies in China toward plastic trade have had a large impact on the global change shown in the previous chart.
\n\n\n\nIn 2016, China was importing more than half of the world’s traded plastic waste. By 2018, this had plummeted to less than 1%. We see this in the chart.
\n\n\n\nThe reason for this dramatic decline was the Chinese government banning the import of most types of plastic waste in 2017. This was part of a broader policy decision to stop the import of 24 different types of solid waste including paper, textiles, and plastics. These bans were implemented as a result of environmental and health concerns from contaminated waste streams.
\n\n\n\nThis ban had two major impacts.{ref}Wen, Z., Xie, Y., Chen, M., & Dinga, C. D. (2021). China’s plastic import ban increases prospects of environmental impact mitigation of plastic waste trade flow worldwide. Nature Communications, 12(1), 1-9.
\n\n\n\nBrooks, A. L., Wang, S., & Jambeck, J. R. (2018). The Chinese import ban and its impact on global plastic waste trade. Science Advances, 4(6).{/ref} The first was that the total volume of plastic trade globally dropped significantly – we saw earlier that global trade has halved since 2017. The second was that other countries emerged to take China’s place as major importers. Most of them are also countries in Asia – Malaysia, Vietnam, Indonesia, the Philippines, and Turkey, started to import much more plastic than in previous years.
\nWhere does plastic waste flow across the world?
\n\n\n\nEurope is the region that exports the most plastic, but it’s also the region that imports the most.
\n\n\n\nThis is true more generally. Most plastic is traded within a given region. European countries export most plastic to other European countries. Asian countries export most to other Asian countries.
\n\n\n\nIn the visualization, we see the flow of plastic across the world.{ref}This is based on data from the OECD report:
\n\n\n\nBrown, A., F. Laubinger and P. Börkey (2022), “Monitoring trade in plastic waste and scrap“, OECD Environment Working Papers, No. 194, OECD Publishing, Paris.{/ref} On the left we have the exporters of plastic waste; on the right, we see where that plastic ends up. The height of each bar is proportional to the amount of plastic that is traded.
\n\n\n\nEurope is the biggest exporter of plastic. But, it’s also the biggest importer. Many countries across Europe trade with one another. At the national level, Germany is the biggest exporter and importer – it trades different plastics with its neighbors including the Netherlands, Poland, Austria, and Switzerland.{ref}Brown, A., F. Laubinger and P. Börkey (2022), “Monitoring trade in plastic waste and scrap“, OECD Environment Working Papers, No. 194, OECD Publishing, Paris.{/ref}
\n\n\n\nThis is also true of Asia, where Japan is the biggest exporter to other countries in Asia.
\n\n\n\n\n![\""\""](\""https://owid.cloud/app/uploads/2022/10/Plastic-waste-trade-sankey.png\"")
Many people think that rich countries ship most of their plastic waste overseas. But is this really true?
\n\n\n\nThe short answer is no: many countries export some of their waste, but they still handle most of it domestically.
\n\n\n\nLet’s take the example of the UK. In 2010, it generated an estimated 4.93 million tonnes of plastic waste.{ref}Jambeck, J. R., Geyer, R., Wilcox, C., Siegler, T. R., Perryman, M., Andrady, A., … & Law, K. L. (2015). Plastic waste inputs from land into the ocean. Science, 347(6223), 768-771.{/ref} It exported 838,000 tonnes overseas.
\n\n\n\nThat means it exported about 17% of its plastic waste. That’s a substantial fraction – nearly one-fifth of it.
\n\n\n\nThis data is for 2010, a year with good high-quality estimates of plastic waste generation. It’s still likely to be a reasonable estimate today. If anything, this share might have declined slightly, because waste exports have not increased, and waste generation probably has.
\n\n\n\nWhen it comes to the fraction of plastic waste that is exported, the UK is one of the largest exporters. For context, the US exported about 5% of its plastic waste in 2010. France exported 11%, and the Netherlands exported 14%.
\n\n\n\nMost rich countries are net exporters of plastic waste. And this can be a significant fraction of their waste. But it’s not the case that they handle most of it by offshoring it to other countries.
\nThis is the crucial question. While we don’t have an exact answer, we can give a plausible range.
\n\n\n\nTo give an exact answer we would need to trace each piece of plastic pollution back to its original source. But we can do some calculations to estimate how much plastic is at higher risk of entering the ocean because of this trade.
\n\n\n\nIn 2020, low-to-middle-income countries – where plastic waste was at a ‘higher risk’ of entering the ocean (because of poorer waste management systems) – imported around 1.6 million tonnes of plastic waste from rich countries. Here ‘rich countries’ include all countries in Europe and North America, plus Japan, Hong Kong, and OECD countries from other regions.{ref}A report by the OECD provides a summary of trade flows of plastic waste across the world. We saw this in the Sankey diagram, earlier in the post. The authors use data from the UN Comtrade database to calculate these figures.
\n\n\n\nFrom these figures, we see that in 2020, non-OECD countries in Asia imported around 1.9 million tonnes of plastic waste. ‘Rest of the World’, which in this case is mostly lower-income countries across Africa and South America imported 0.12 million tonnes, and China also imported 0.12 million tonnes. Combined, these low-to-middle-income countries imported 2.14 million tonnes.
\n\n\n\n0.57 million tonnes of this came from countries within this group i.e. low-to-middle-income countries.
\n\n\n\nThat leaves around 1.6 million tonnes that come from richer countries, which is the sum of Europe, North America, Hong Kong, and Japan.{/ref}
\n\n\n\nHow much of this plastic ends up in the ocean?
\n\n\n\nAgain, we don’t know for sure. But we can run through a worst and best-case scenario.
\n\n\n\nHere we will assume that all of this traded waste was ‘mismanaged’, meaning it was not formally managed and was either littered or dumped in open landfills. In reality, not all of it will be mismanaged, but let’s be conservative here.
\n\n\n\nThe probability that mismanaged waste ends up in the ocean varies a lot by country.
\n\n\n\nThe country where the probability is highest is the Phillippines – an estimated 7% ends up in the ocean.{ref}This estimate comes from the work of Meijer et al. (2021), published in Science Advances.
\n\n\n\nMeijer, L. J., van Emmerik, T., van der Ent, R., Schmidt, C., & Lebreton, L. (2021). More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean. Science Advances, 7(18), eaaz5803.{/ref} We could imagine this being our ‘worst-case’ scenario: if rich countries exported all of their plastic trade to the Philippines, 7% of it might end up in the ocean. That would be 112,000 tonnes.{ref}We can calculate this as 7% of 1.6 million tonnes, which is around 112,000 tonnes.{/ref}
\n\n\n\nIn a ‘best-case’ scenario, only around 1% of mismanaged waste would end up in the ocean. Most countries across the world have a risk of just under 1%. In Asia, this would be typical of countries such as Thailand and Cambodia. In this ‘best-case’ scenario, around 16,000 tonnes of ocean plastics each year would enter the ocean from trade.{ref}We can calculate this figure as 1% of 1.6 million tonnes, which gives us 16,000 tonnes.{/ref}
\n\n\n\nThis gives us an upper and lower bound for the contribution of trade from rich countries. Since around one million tonnes of plastic enters the ocean each year, rich countries would contribute between 1.6% (in the best case) and 11% (in the worst case) of ocean plastics through shipping waste overseas.{ref}Meijer et al. (2021) estimate that around 1 million tonnes of plastic are emitted into the oceans each year. They put the uncertainty range on this between 0.8 and 2.6 million tonnes.
\n\n\n\nMeijer, J.J.L, Emmerik, T., Ent, R., Schmidt, C., Lebreton, L. (2021). More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean. Science Advances.
\n\n\n\nWe get these figures by calculating 16,000 and 112,000 tonnes as a share of 1 million. That comes to 1.6% and 11%.{/ref}
\n\n\n\nThe true figure probably falls somewhere in between. A reasonable estimate might be around 5% of ocean plastics. In reality, it might be a bit lower because a tonne of waste that is bought and traded is more likely to be managed well than the average tonne of waste in a country.
\n\n\n\nI estimate that a few percent of ocean plastics could result from trade from rich countries. A figure as high as 5% would not be unreasonable.
\nStopping exports of plastic waste to countries with poor waste management would help to tackle ocean pollution.
\n\n\n\nIf rich countries banned the export of plastic waste to these countries, we might reduce plastic pollution a bit: perhaps up to 5%.
\n\n\n\nBut, an end to trade won’t stop plastic pollution. Only a small fraction of the world’s plastic waste is traded – under 2%. And most – two-thirds of it – ends up in richer countries, where it’s very unlikely to end up in the ocean.
\n\n\n\nThere are obvious reasons to reduce these exports beyond the plastic pollution problem. Countries have been guilty of exporting contaminated recycling plastic packages – one of the drivers for countries to ban plastic imports. This is unacceptable: poorer countries are not a dumping ground for the rich.
\n\n\n\nMost of the world’s waste is handled domestically and most of the waste that enters the oceans stems from these countries. To really tackle the problem we need to do two things: scale waste management systems in rich countries; the fact that they are exporting waste overseas suggests they have under-invested in practices at home; and, importantly, improve waste management infrastructure and practices in low-to-middle-income countries, as this is where most plastic pollution originates.
\nMany thanks to Max Roser, Edouard Mathieu, and Bastian Herre for feedback and suggestions on this article.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Many countries ship plastic waste overseas. How much of the world’s waste is traded, and how big is its role in the pollution of our oceans?"", ""protected"": false}, ""date_gmt"": ""2022-10-11T09:50:00"", ""modified"": ""2022-10-25T13:06:49"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2022-10-25T12:06:49"", ""comment_status"": ""closed"", ""featured_media"": 53420, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/10/Plastic-waste-trade-featured-image-150x79.png"", ""medium_large"": ""/app/uploads/2022/10/Plastic-waste-trade-featured-image-768x402.png""}}" 53245,From $1.90 to $2.15 a day: the updated International Poverty Line,from-1-90-to-2-15-a-day-the-updated-international-poverty-line,post,publish,"To track progress towards its goal of eradicating extreme poverty by 2030, the UN relies on World Bank estimates of the share of the world population that fall below the International Poverty Line.
In September 2022, the figure at which this poverty line is set shifted from $1.90 to $2.15. This reflects a change in the units in which the World Bank expresses its poverty and inequality data – from international dollars given in 2011 prices to international dollars given in 2017 prices.
In this article we consider what the World Bank’s revised methodology means for our understanding of global poverty. We then also explain the updated methodology: what international dollars are, what the change to 2017 prices mean, and how this change affects the International Poverty Line and the World Bank’s poverty estimates.
The crucial fact to know about the World Bank’s updated International Poverty Line is that, whilst the number has risen – from $1.90 to $2.15 per day – in terms of the goods and services that it affords, the value is broadly the same.
This is because the units in which the old and new figures are counted have changed: the $1.90 figure was expressed in 2011 international-$, and the new figure of $2.15 is expressed in 2017 international-$.
Below we explain in much more detail what international dollars are and what the change in base year means. But broadly speaking, the difference in the two units reflects inflation over time: Because prices have gone up, one 2017 international-$ buys a smaller quantity of goods and services than one 2011 international-$. This leaves the real value of the revised International Poverty Line broadly equal to what it was before.
This continuity is reflected in the data: Overall the update only very modestly changes our understanding of the extent of extreme poverty globally. We see this in the two charts below that show the share and number of people in extreme poverty using the old and updated methodologies.
Estimates of the share of people living in extreme poverty globally for 2019 – the latest available year – are slightly lower using the updated methodology: 8.4% as compared to 8.7%. That translates to 20 million fewer people living in extreme poverty.
We also see a slightly faster decline in poverty over recent decades using the updated methodology. Looking back to the 1990s, estimates of the share in extreme poverty using the updated methodology are around 1.5% points higher – or around 80 million people.
There are only small changes to the World Bank’s extreme poverty estimates at the global level. But the updated methodology has resulted in more substantial revisions to poverty estimates for some countries. These changes reflect a revision to the adjustments made to account for differences in the cost of living across countries, as explained in more detail below.
These revisions go in both directions: some countries see higher poverty estimates and others see lower estimates. These opposing differences are cancelling each other out when we look at the global figures.
Nigeria is one country that has seen a significant downward revision, with 16 million fewer people estimated to be living in extreme poverty in 2018. You can read more about this particular change in this World Bank blogpost.
In Indonesia, the revision has gone in the other direction – the latest available estimates for 2021 show an additional 4 million people living in extreme poverty when calculated with the new methodology.
Despite significant differences in some individual countries, our overall understanding of the global distribution of extreme poverty has changed very little though. This can be seen from a comparison of the two map charts below which show the share of population in extreme poverty according to the old and updated World Bank methodologies.
The International Poverty Line is set by the World Bank to be representative of national definitions of poverty adopted in the world’s poorest countries.
In addition to this very low poverty line the World Bank also sets two higher global poverty lines for measuring poverty: one that reflects the definitions of poverty adopted in lower-middle income countries, and one that reflects the definitions adopted in upper-middle income countries. Within the updated methodology, these lines are set at $3.65 and $6.85 in 2017 international-$, replacing the previous $3.20 and $5.50 lines expressed in 2011 international-$.
How have poverty estimates around the world changed with the revision of these higher lines?
As with the International Poverty Line, the real value of the lower-middle income poverty line is roughly the same across the two methodologies. Accordingly, global poverty estimates are again very similar across the old and updated methodologies – as we see in the first chart below.
In contrast, the update of the upper-middle income poverty line represents a real increase in the value of the line: This revised higher line buys more goods and services than the previously used figure. This real-terms increase is due to a rise in the national poverty lines to which the World Bank anchors this line, as discussed in more detail below.
Since the real value of the poverty line is significantly higher, unsurprisingly the prevalence of poverty when measured with the updated methodology is higher – as shown in the second chart.
In this data explorer, you can compare the World Bank’s 2011 and 2017 international-$ estimates for a range of key poverty metrics.
At the heart of the update is a change in the units that the World Bank uses to count and compare the incomes of households in different countries at different times.
These units are called international dollars – a hypothetical currency that adjusts for inflation and differences in the cost of living between countries.
To understand the update to the World Bank’s methodology we need a good understanding of what these units mean and how they are calculated. In the box below we give a summary.
Much of the economic data we use to understand the world – for instance on the goods and services bought or produced by households, firms and governments, or the incomes they receive – is initially recorded in terms of the units in which these transactions took place. That means this data starts out being expressed in a variety of local currencies – as so many rupees, US dollars, or yuan, etc. – and without adjusting for inflation over time. This is known as being in ‘current prices’, or in ‘nominal’ terms.
Before these figures can be meaningfully compared, they need to be converted into common units.
International dollars (int.-$) are a hypothetical currency that is used for this. It is the result of adjusting both for inflation within countries over time and for differences in the cost of living between countries.
The goal of international-$ is to provide a unit whose purchasing power is held fixed over time and across countries, such that one int.-$ can buy the same quantity and quality of goods and services no matter where or when it is spent.
The price level in the US is used as the benchmark – or ‘numeraire’ – so that one 2017 int.-$ is defined as the value of goods and services that one US dollar would buy in the US in 2017.
The year 2017 here indicates two things, related to the two adjustments mentioned. Firstly, it tells us the base year used for the inflation adjustment within countries. This is the year whose prices are chosen to be the benchmark. If prices are higher than this benchmark year, nominal data will be adjusted downwards. If prices are lower, nominal data will be adjusted upwards. In the base year itself, the nominal and inflation-adjusted figures are the same by definition.
Secondly, 2017 indicates the year in which the differences in the cost of living between countries was assessed.
Converting data in local currencies to international-$ means dividing the figures by a set of ‘exchange’ rates, known as Purchasing Power Parity (PPP) rates. Unlike the exchange rates between currencies you would see at the foreign exchange counter, these account for differences in the cost of living between countries.
If you have ever shopped or eaten in a restaurant abroad, you may have noticed a country as being a particularly expensive or particularly cheap place to live. A given amount of your own currency, when exchanged for another country’s currency, may buy you considerably more or less there than it would have done at home.
The goal of PPP rates is to account for these price differences. They express, for each country, the amount of local currency that is needed to buy the same goods and services there as 1 US dollar buys in the US.
You can read more about this in our article What are PPP adjustments and why do we need them?
The calculation of PPP rates is the task of the International Comparison Program (ICP), which gathers data on the prices of thousands of goods and services in each country in a particular year.
The ICP does not calculate PPP rates every year, but rather conducts its work in ‘rounds’ that are several years apart. The most recent round was conducted in 2017 and the previous round was conducted in 2011.
In converting economic data to international-$, which round of PPPs are used to adjust for cost-of-living differences between countries is, in principle, a separate issue to the base year used to adjust for inflation over time. By convention, however, the same year tends to be chosen for both. When converted to 2017 international-$, nominal local currencies are first adjusted for inflation to local 2017 prices, and are then adjusted to US prices using the PPPs calculated in the ICP’s 2017 round. Likewise, 2011 international-$ adjust for inflation using 2011 local prices, and then use the 2011 PPPs to adjust for cost-of-living differences.
The World Bank’s shift from measuring household incomes in 2011 to 2017 international-$ has affected the data in two ways, relating to the two different adjustments made in the calculation of international-$.
Firstly, it reflects the impact of inflation between 2011 and 2017. Secondly, it reflects a revised assessment of how the cost of living compares in different countries, as presented in the 2017 Purchasing Power Parity rates published by the International Comparison Program.
As explained in more detail in the box above:
The amount of goods and services that one US dollar bought in the US changed over these years. There was inflation of around 9%, meaning that one US dollar bought 9% less in the US in 2017 than it did in 2011.{ref}According to World Bank data on the Consumer Price Index, the price level (relative to 2010) in 2011 was 103.2% and the price level in 2017 was 112.4%. The increase in prices was therefore 112.4/103.2 = 1.089, or approximately 9%. These are the same figures that the World Bank uses to adjust its data on household incomes for inflation within each country.{/ref}
This also means that one 2017 int.-$ buys around 9% less than one 2011 int.-$, and this is one key way in which the two units differ. Expressing incomes in 2017 international-$ generally results in higher numbers than when expressed in 2011 international-$.
You can see this general increase in the chart here which compares the average income or consumption per person, as expressed in 2011 international-$ (shown on the x-axis) and in 2017 international dollars (on the y-axis).
The 45 degree line indicates the point where the two numbers would be equal. You see that in almost all countries, average incomes are ‘higher’ when expressed in 2017 international-$ than in 2011 international-$.
But we also see in the chart above that the increase is not uniform across countries. In fact, some countries fall below the 45 degree line: their incomes count as fewer 2017 international-$ than they do 2011 international-$.
You can see these differences in more detail in this chart, which plots the average income or consumption expenditure of countries over time – as measured in both 2011 and 2017 international-$.
The incomes are plotted on a log scale to show proportional differences more clearly. Again we see that the switch to 2017 international-$ generally shifts the incomes data up by a certain proportion – but the proportion varies for each country. In the chart we have selected particular countries to demonstrate the range of differences we see in the data, but you can change which countries or regions are shown using the 'Add country' button.
At one end of the spectrum are countries such as Angola whose figures shift up by a large amount. The figure for the average level of consumption in Angola is 80% higher when expressed in 2017 international-$ than in 2011 international-$. At the opposite end of the spectrum is Ghana, whose figures fall by around 25%.
In the US – the ‘numeraire’ country for the purposes of the PPP adjustment – the rise is 9%, in-line with US inflation as discussed above.
The different shifts we see across countries means that, with the new 2017 international-$, we have an updated understanding of how incomes compare across countries. Those countries that see a bigger jump in their income figures when measured in the new units now appear relatively richer than countries with a smaller, or even negative jump.
But what explains these such very different jumps, and how are we to interpret them? Is it that our previous understanding, based on the old 2011 PPPs, was wrong?
Part of the complication of interpreting the switch to 2017 international-$ is that two factors are operating at the same time.
Firstly, the change in PPPs reflects changes in the methods and underlying price data used by the ICP. Producing these estimates, in as comparable a way as possible, is a huge and very complex task and not surprisingly the ICP adapts and refines the methods they use in different rounds. Moreover, the quality and coverage of the country price data that go into these calculations is also generally improving over time. It is not necessarily the case that the quality of the estimates improve in every aspect or for every country, but in general the new PPPs do indeed give us a better, truer picture of how incomes compare around the world.{ref}You can read about these changes and a summary of the ICP’s current methods in its 2017 report. Dean Joliffe and others (2022) provide a helpful comparison of the ICPs methods in the 2011 and 2017 rounds in the context of global poverty measurement.{/ref}
Secondly, the change in PPPs reflect actual changes in relative price levels across countries. Inflation happens at different rates in different countries. In part, such changes are accounted for by the inflation adjustment made when converting data to international-$, as discussed above. But the way that PPP rates are calculated means that the change in PPPs need not exactly track these various national inflation rates. When a new cross-country assessment of price levels is made, there can be a gap with what would have been expected based on the previous set of PPPs extended forward with national inflation data.{ref}A helpful – although quite detailed and technical – explanation of why this gap can exist can be found in this methodological note from the team behind the Maddison Project Database – another cross-country dataset of incomes that uses international-$ adjustments.
At the heart of the problem is that, when comparing the overall price level in two places or two years you need to account not just for the prices of each product but also the quantity of that product bought or produced in each country – the weights attached to each product when calculating the average price level. Both the prices and the weights differ between countries. This means that straightforward comparisons of the price level between different pairs of countries needn’t be consistent with each other – the difference in prices between country A and C, need not be the same as the combined differences between country A and country B and country B and country C (in technical terms, they are not transitive). For the purposes of making international comparisons this is clearly an unhelpful property. In calculating PPPs, the IPC use a method for comparing price levels that yields transitive rates. A consequence of this is that there can be inconsistency between the change in price levels found across various rounds of PPPs and the inflation rates recorded in individual countries.{/ref}
Because the change in PPPs in part pick up ‘real’ changes in the relative price levels between countries, some data providers expressing income data in international-$ - such as the Penn World Tables - use PPPs from multiple ICP rounds and adopt techniques to interpolate, or smooth out the jumps between the rounds. An important downside of this approach, however, is that the trends in income calculated with this approach do not then necessarily match up with the trends observed in the original national data, prior to conversion to international-$.
This ‘multiple benchmarks’ approach is not the route that the World Bank takes. It uses just one set of PPPs – from the latest available ICP round – applied back over the whole time span of its data. This makes the trends over time consistent with the original data, but it means that the adoption of a new set of PPPs can generate significant revisions to our understanding of income levels in different countries – both now and in the past.
A key aspect of the World Bank’s updated methodology is that, along with the change in the units by which it measures household incomes, it has also updated the International Poverty Line against which it measures extreme poverty.
To help explain this change, in the box below we describe what the International Poverty Line is and how it is calculated, along with the higher poverty lines set by the World Bank.
There is no single definition of poverty. Our understanding of the extent of poverty and how it is changing depends on which definition we have in mind.
In particular, richer and poorer countries set very different poverty lines in order to measure poverty in a way that is informative and relevant to the level of incomes of their citizens.
For instance, while in the United States a person is counted as being in poverty if they live on less than roughly $24.55 per day, in Ethiopia the poverty line is set more than 10 times lower – at $2.04 per day. You can read more about how these comparable national poverty lines are calculated in this footnote.{ref}Official definitions of poverty in different countries are often not directly comparable due to the different ways poverty is measured. For example, countries account for the size of households in different ways in their poverty measures.
The poverty lines shown here are an approximation of national definitions, harmonized to allow for comparisons across countries. For all countries apart from the US, we take the harmonized poverty line calculated by Jolliffe et al. (2022). These lines are calculated as the international dollar figure which, in the World Bank’s Poverty and Inequality Platform (PIP) data, yields the same poverty rate as the officially reported rate using national definitions in a particular year (around 2017).
For the US, Jolliffe et al. (2022) use the OECD’s published poverty rate – which is measured against a relative poverty line of 50% of the median income. This yields a poverty line of $34.79 (measured using 2017 survey data). This however is not the official definition of poverty adopted in the US. We calculated an alternative harmonized figure for the US national poverty using the same method as Jolliffe et al. (2022), but based instead on the official 2019 poverty rate – as reported by the U.S. Census Bureau.
You can see in detail how we calculated this poverty line in this Google Colabs notebook.
Jolliffe, Dean Mitchell, Daniel Gerszon Mahler, Christoph Lakner, Aziz Atamanov, and Samuel Kofi Tetteh Baah. 2022. Assessing the Impact of the 2017 PPPs on the International Poverty Line and Global Poverty. The World Bank. Available to read at the World Bank here.{/ref}
To measure poverty globally, however, we need to apply a poverty line that is consistent across countries.
This is the goal of the International Poverty Line of $2.15 per day – shown in red in the chart – which is set by the World Bank and used by the UN to monitor extreme poverty around the world.
We see that, in global terms, this is an extremely low threshold indeed – set to reflect the poverty lines adopted nationally in the world’s poorest countries. It marks an incredibly low standard of living – a level of income much lower than just the cost of a healthy diet.
How does the World Bank set the International Poverty Line?
The exact method used by the World Bank to set the International Poverty Line has changed somewhat over past updates. But each time the objective has been broadly the same – to find a “typical standard by which the poorest countries of the world judge their citizens to be impoverished.”{ref}As described by Dean Jolliffe and others (2022).
Jolliffe, Dean Mitchell, Daniel Gerszon Mahler, Christoph Lakner, Aziz Atamanov, and Samuel Kofi Tetteh Baah. 2022. Assessing the Impact of the 2017 PPPs on the International Poverty Line and Global Poverty. The World Bank. Available online here.{/ref}
The method used in the latest update to arrive at a figure of $2.15, measured in 2017 international-$, is based on a set of harmonized national poverty lines produced by Dean Joliffe and others – shown in the chart here.
As you can see, there is a strong correlation between the poverty lines countries set, shown on the Y axis, and their income level – as measured here by GDP per capita, and plotted along the X axis.
The International Poverty Line is calculated as the median national poverty line adopted among low-income countries – using the World Bank’s income classification system. These are the countries shaded in red in the chart and found in the bottom left corner.
Although the International Poverty Line is by far the most prominent international line, the same method is also used by the World Bank to set two higher poverty lines that reflect the national definitions adopted in lower-middle and upper-middle income groups shown in green and purple respectively. The median poverty line among these two groups of countries are $3.65 and $6.85, and these form the the World Bank's lower-middle income and upper middle-income poverty lines.
You can read more about the methodology used to set these lines in the World Bank's flagship report on poverty, Poverty and Shared Prosperity.
A key question to ask of the World Bank’s updated poverty lines is the extent to which their real value – in terms of the goods and services they can buy – have been maintained, or if instead they represent a shifting of the goalposts.
One way to answer this question, building on the explanation of change from 2011 to 2017 international-$ given above, is to compare the increase in the numbers to the rate of inflation in the US between these two years. An increase in-line with US inflation would indicate that the poverty lines have the same purchasing power. A rise greater than US inflation would indicate that the real value had increased. This is just a direct consequence of the definition of international-$ for a given base year – where one international-$ is defined as the value of goods and services that one US dollar could buy in that base year.
Prices in the US rose by 9% between 2011 and 2017. In the case of both the International Poverty Line and the lower-middle income poverty line, the increase across the update was roughly similar – from $1.90 to $2.15 (13%) and from $3.20 to $3.65 (14%) respectively. That the rise was close to the US inflation rate implies that the purchasing power of the lines was similar before and after the update. And, as demonstrated above, this is reflected in the data: global poverty estimates according to these two poverty lines are very similar across the old and updated methodologies.
The upper-middle poverty line, however, rose by considerably more than US inflation – from $5.50 to $6.85, a 25% increase. Again, this is reflected in the data – with notably higher global poverty estimates resulting from the updated methodology.
Why would the World Bank ‘shift the goal posts’ by using a higher poverty line? This answer lies in the method the World Bank uses to set the line, just explained in the box above. The upper-middle income country poverty line was set at $5.50 in 2011 international-$ because this was the median poverty line observed among this group of countries at the time of the last update to the World Bank methodology.{ref}As documented by Dean Jolliffe and Espen Beer Prydz (2016), in which they introduced the method now applied by the World Bank.
Jolliffe, Dean, and Espen Beer Prydz. 2016. Estimating International Poverty Lines from Comparable National Thresholds. Washington, DC. Available online here.{/ref}
Since that time, the national poverty lines of some countries in this group have been revised upwards as incomes generally increased. This dynamic was not present for the lower income groupings of countries: in these cases those countries with a rising poverty line were also graduating into a higher income group, thereby maintaining the median poverty line within their former income group roughly constant.{ref}See the paper by Dean Jolliffe and others (2022) that introduces the updated World Bank poverty lines and provides a breakdown of the different factors contributing to the change in the poverty lines.
Jolliffe, Dean Mitchell, Daniel Gerszon Mahler, Christoph Lakner, Aziz Atamanov, and Samuel Kofi Tetteh Baah. 2022. Assessing the Impact of the 2017 PPPs on the International Poverty Line and Global Poverty. The World Bank. Available online here.{/ref}
For a fuller understanding of international-$ and PPPs in the context of global poverty measurement you may find some of the following articles helpful.
On the 2017 PPPs:
And some more general discussions, in the context of past PPP rounds:
To track progress towards its goal of eradicating extreme poverty by 2030, the UN relies on World Bank estimates of the share of the world population that fall below the International Poverty Line.
\n\n\n\nIn September 2022, the figure at which this poverty line is set shifted from $1.90 to $2.15. This reflects a change in the units in which the World Bank expresses its poverty and inequality data – from international dollars given in 2011 prices to international dollars given in 2017 prices.
\n\n\n\nIn this article we consider what the World Bank’s revised methodology means for our understanding of global poverty. We then also explain the updated methodology: what international dollars are, what the change to 2017 prices mean, and how this change affects the International Poverty Line and the World Bank’s poverty estimates.
\n\n\n\nThe crucial fact to know about the World Bank’s updated International Poverty Line is that, whilst the number has risen – from $1.90 to $2.15 per day – in terms of the goods and services that it affords, the value is broadly the same.
\n\n\n\nThis is because the units in which the old and new figures are counted have changed: the $1.90 figure was expressed in 2011 international-$, and the new figure of $2.15 is expressed in 2017 international-$.
\n\n\n\nBelow we explain in much more detail what international dollars are and what the change in base year means. But broadly speaking, the difference in the two units reflects inflation over time: Because prices have gone up, one 2017 international-$ buys a smaller quantity of goods and services than one 2011 international-$. This leaves the real value of the revised International Poverty Line broadly equal to what it was before.
\n\n\n\nThis continuity is reflected in the data: Overall the update only very modestly changes our understanding of the extent of extreme poverty globally. We see this in the two charts below that show the share and number of people in extreme poverty using the old and updated methodologies.
\n\n\n\nEstimates of the share of people living in extreme poverty globally for 2019 – the latest available year – are slightly lower using the updated methodology: 8.4% as compared to 8.7%. That translates to 20 million fewer people living in extreme poverty.
\n\n\n\nWe also see a slightly faster decline in poverty over recent decades using the updated methodology. Looking back to the 1990s, estimates of the share in extreme poverty using the updated methodology are around 1.5% points higher – or around 80 million people.
\n\n\n\nThere are only small changes to the World Bank’s extreme poverty estimates at the global level. But the updated methodology has resulted in more substantial revisions to poverty estimates for some countries. These changes reflect a revision to the adjustments made to account for differences in the cost of living across countries, as explained in more detail below.
\n\n\n\nThese revisions go in both directions: some countries see higher poverty estimates and others see lower estimates. These opposing differences are cancelling each other out when we look at the global figures.
\n\n\n\nNigeria is one country that has seen a significant downward revision, with 16 million fewer people estimated to be living in extreme poverty in 2018. You can read more about this particular change in this World Bank blogpost.
\n\n\n\nIn Indonesia, the revision has gone in the other direction – the latest available estimates for 2021 show an additional 4 million people living in extreme poverty when calculated with the new methodology.
\n\n\n\nDespite significant differences in some individual countries, our overall understanding of the global distribution of extreme poverty has changed very little though. This can be seen from a comparison of the two map charts below which show the share of population in extreme poverty according to the old and updated World Bank methodologies.
\n\n\n\nThe International Poverty Line is set by the World Bank to be representative of national definitions of poverty adopted in the world’s poorest countries.
\n\n\n\nIn addition to this very low poverty line the World Bank also sets two higher global poverty lines for measuring poverty: one that reflects the definitions of poverty adopted in lower-middle income countries, and one that reflects the definitions adopted in upper-middle income countries. Within the updated methodology, these lines are set at $3.65 and $6.85 in 2017 international-$, replacing the previous $3.20 and $5.50 lines expressed in 2011 international-$.
\n\n\n\nHow have poverty estimates around the world changed with the revision of these higher lines?
\n\n\n\nAs with the International Poverty Line, the real value of the lower-middle income poverty line is roughly the same across the two methodologies. Accordingly, global poverty estimates are again very similar across the old and updated methodologies – as we see in the first chart below.
\n\n\n\nIn contrast, the update of the upper-middle income poverty line represents a real increase in the value of the line: This revised higher line buys more goods and services than the previously used figure. This real-terms increase is due to a rise in the national poverty lines to which the World Bank anchors this line, as discussed in more detail below.
\n\n\n\nSince the real value of the poverty line is significantly higher, unsurprisingly the prevalence of poverty when measured with the updated methodology is higher – as shown in the second chart.
\n\n\n\nIn this data explorer, you can compare the World Bank’s 2011 and 2017 international-$ estimates for a range of key poverty metrics.
\n\n\n\n\n\n\n\nAt the heart of the update is a change in the units that the World Bank uses to count and compare the incomes of households in different countries at different times.
\n\n\n\nThese units are called international dollars – a hypothetical currency that adjusts for inflation and differences in the cost of living between countries.
\n\n\n\nTo understand the update to the World Bank’s methodology we need a good understanding of what these units mean and how they are calculated. In the box below we give a summary.
\n\n\n\tMuch of the economic data we use to understand the world – for instance on the goods and services bought or produced by households, firms and governments, or the incomes they receive – is initially recorded in terms of the units in which these transactions took place. That means this data starts out being expressed in a variety of local currencies – as so many rupees, US dollars, or yuan, etc. – and without adjusting for inflation over time. This is known as being in ‘current prices’, or in ‘nominal’ terms.
\n\n\n\nBefore these figures can be meaningfully compared, they need to be converted into common units.
\n\n\n\nInternational dollars (int.-$) are a hypothetical currency that is used for this. It is the result of adjusting both for inflation within countries over time and for differences in the cost of living between countries.
\n\n\n\nThe goal of international-$ is to provide a unit whose purchasing power is held fixed over time and across countries, such that one int.-$ can buy the same quantity and quality of goods and services no matter where or when it is spent.
\n\n\n\nThe price level in the US is used as the benchmark – or ‘numeraire’ – so that one 2017 int.-$ is defined as the value of goods and services that one US dollar would buy in the US in 2017.
\n\n\n\nThe year 2017 here indicates two things, related to the two adjustments mentioned. Firstly, it tells us the base year used for the inflation adjustment within countries. This is the year whose prices are chosen to be the benchmark. If prices are higher than this benchmark year, nominal data will be adjusted downwards. If prices are lower, nominal data will be adjusted upwards. In the base year itself, the nominal and inflation-adjusted figures are the same by definition.
\n\n\n\nSecondly, 2017 indicates the year in which the differences in the cost of living between countries was assessed.
\n\n\n\nConverting data in local currencies to international-$ means dividing the figures by a set of ‘exchange’ rates, known as Purchasing Power Parity (PPP) rates. Unlike the exchange rates between currencies you would see at the foreign exchange counter, these account for differences in the cost of living between countries.
\n\n\n\nIf you have ever shopped or eaten in a restaurant abroad, you may have noticed a country as being a particularly expensive or particularly cheap place to live. A given amount of your own currency, when exchanged for another country’s currency, may buy you considerably more or less there than it would have done at home.
\n\n\n\nThe goal of PPP rates is to account for these price differences. They express, for each country, the amount of local currency that is needed to buy the same goods and services there as 1 US dollar buys in the US.
\n\n\n\nYou can read more about this in our article What are PPP adjustments and why do we need them?
\n\n\n\nThe calculation of PPP rates is the task of the International Comparison Program (ICP), which gathers data on the prices of thousands of goods and services in each country in a particular year.
\n\n\n\nThe ICP does not calculate PPP rates every year, but rather conducts its work in ‘rounds’ that are several years apart. The most recent round was conducted in 2017 and the previous round was conducted in 2011.
\n\n\n\nIn converting economic data to international-$, which round of PPPs are used to adjust for cost-of-living differences between countries is, in principle, a separate issue to the base year used to adjust for inflation over time. By convention, however, the same year tends to be chosen for both. When converted to 2017 international-$, nominal local currencies are first adjusted for inflation to local 2017 prices, and are then adjusted to US prices using the PPPs calculated in the ICP’s 2017 round. Likewise, 2011 international-$ adjust for inflation using 2011 local prices, and then use the 2011 PPPs to adjust for cost-of-living differences.
\nThe World Bank’s shift from measuring household incomes in 2011 to 2017 international-$ has affected the data in two ways, relating to the two different adjustments made in the calculation of international-$.
\n\n\n\nFirstly, it reflects the impact of inflation between 2011 and 2017. Secondly, it reflects a revised assessment of how the cost of living compares in different countries, as presented in the 2017 Purchasing Power Parity rates published by the International Comparison Program.
\n\n\n\nAs explained in more detail in the box above:
\n\n\n\nThe amount of goods and services that one US dollar bought in the US changed over these years. There was inflation of around 9%, meaning that one US dollar bought 9% less in the US in 2017 than it did in 2011.{ref}According to World Bank data on the Consumer Price Index, the price level (relative to 2010) in 2011 was 103.2% and the price level in 2017 was 112.4%. The increase in prices was therefore 112.4/103.2 = 1.089, or approximately 9%. These are the same figures that the World Bank uses to adjust its data on household incomes for inflation within each country.{/ref}
\n\n\n\nThis also means that one 2017 int.-$ buys around 9% less than one 2011 int.-$, and this is one key way in which the two units differ. Expressing incomes in 2017 international-$ generally results in higher numbers than when expressed in 2011 international-$.
\n\n\n\nYou can see this general increase in the chart here which compares the average income or consumption per person, as expressed in 2011 international-$ (shown on the x-axis) and in 2017 international dollars (on the y-axis).
\n\n\n\nThe 45 degree line indicates the point where the two numbers would be equal. You see that in almost all countries, average incomes are ‘higher’ when expressed in 2017 international-$ than in 2011 international-$.
\n\n\n\n\n\n\n\nBut we also see in the chart above that the increase is not uniform across countries. In fact, some countries fall below the 45 degree line: their incomes count as fewer 2017 international-$ than they do 2011 international-$.
\n\n\n\nYou can see these differences in more detail in this chart, which plots the average income or consumption expenditure of countries over time – as measured in both 2011 and 2017 international-$.
\n\n\n\nThe incomes are plotted on a log scale to show proportional differences more clearly. Again we see that the switch to 2017 international-$ generally shifts the incomes data up by a certain proportion – but the proportion varies for each country. In the chart we have selected particular countries to demonstrate the range of differences we see in the data, but you can change which countries or regions are shown using the ‘Add country’ button.
\n\n\n\nAt one end of the spectrum are countries such as Angola whose figures shift up by a large amount. The figure for the average level of consumption in Angola is 80% higher when expressed in 2017 international-$ than in 2011 international-$. At the opposite end of the spectrum is Ghana, whose figures fall by around 25%.
\n\n\n\nIn the US – the ‘numeraire’ country for the purposes of the PPP adjustment – the rise is 9%, in-line with US inflation as discussed above.
\n\n\n\nThe different shifts we see across countries means that, with the new 2017 international-$, we have an updated understanding of how incomes compare across countries. Those countries that see a bigger jump in their income figures when measured in the new units now appear relatively richer than countries with a smaller, or even negative jump.
\n\n\n\nBut what explains these such very different jumps, and how are we to interpret them? Is it that our previous understanding, based on the old 2011 PPPs, was wrong?
\n\n\n\nPart of the complication of interpreting the switch to 2017 international-$ is that two factors are operating at the same time.
\n\n\n\nFirstly, the change in PPPs reflects changes in the methods and underlying price data used by the ICP. Producing these estimates, in as comparable a way as possible, is a huge and very complex task and not surprisingly the ICP adapts and refines the methods they use in different rounds. Moreover, the quality and coverage of the country price data that go into these calculations is also generally improving over time. It is not necessarily the case that the quality of the estimates improve in every aspect or for every country, but in general the new PPPs do indeed give us a better, truer picture of how incomes compare around the world.{ref}You can read about these changes and a summary of the ICP’s current methods in its 2017 report. Dean Joliffe and others (2022) provide a helpful comparison of the ICPs methods in the 2011 and 2017 rounds in the context of global poverty measurement.{/ref}
\n\n\n\nSecondly, the change in PPPs reflect actual changes in relative price levels across countries. Inflation happens at different rates in different countries. In part, such changes are accounted for by the inflation adjustment made when converting data to international-$, as discussed above. But the way that PPP rates are calculated means that the change in PPPs need not exactly track these various national inflation rates. When a new cross-country assessment of price levels is made, there can be a gap with what would have been expected based on the previous set of PPPs extended forward with national inflation data.{ref}A helpful – although quite detailed and technical – explanation of why this gap can exist can be found in this methodological note from the team behind the Maddison Project Database – another cross-country dataset of incomes that uses international-$ adjustments.
At the heart of the problem is that, when comparing the overall price level in two places or two years you need to account not just for the prices of each product but also the quantity of that product bought or produced in each country – the weights attached to each product when calculating the average price level. Both the prices and the weights differ between countries. This means that straightforward comparisons of the price level between different pairs of countries needn’t be consistent with each other – the difference in prices between country A and C, need not be the same as the combined differences between country A and country B and country B and country C (in technical terms, they are not transitive). For the purposes of making international comparisons this is clearly an unhelpful property. In calculating PPPs, the IPC use a method for comparing price levels that yields transitive rates. A consequence of this is that there can be inconsistency between the change in price levels found across various rounds of PPPs and the inflation rates recorded in individual countries.{/ref}
Because the change in PPPs in part pick up ‘real’ changes in the relative price levels between countries, some data providers expressing income data in international-$ – such as the Penn World Tables – use PPPs from multiple ICP rounds and adopt techniques to interpolate, or smooth out the jumps between the rounds. An important downside of this approach, however, is that the trends in income calculated with this approach do not then necessarily match up with the trends observed in the original national data, prior to conversion to international-$.
\n\n\n\nThis ‘multiple benchmarks’ approach is not the route that the World Bank takes. It uses just one set of PPPs – from the latest available ICP round – applied back over the whole time span of its data. This makes the trends over time consistent with the original data, but it means that the adoption of a new set of PPPs can generate significant revisions to our understanding of income levels in different countries – both now and in the past.
\n\n\n\n\n\n\n\nA key aspect of the World Bank’s updated methodology is that, along with the change in the units by which it measures household incomes, it has also updated the International Poverty Line against which it measures extreme poverty.
\n\n\n\nTo help explain this change, in the box below we describe what the International Poverty Line is and how it is calculated, along with the higher poverty lines set by the World Bank.
\n\n\n\tThere is no single definition of poverty. Our understanding of the extent of poverty and how it is changing depends on which definition we have in mind.
\n\n\n\nIn particular, richer and poorer countries set very different poverty lines in order to measure poverty in a way that is informative and relevant to the level of incomes of their citizens.
\n\n\n\nFor instance, while in the United States a person is counted as being in poverty if they live on less than roughly $24.55 per day, in Ethiopia the poverty line is set more than 10 times lower – at $2.04 per day. You can read more about how these comparable national poverty lines are calculated in this footnote.{ref}Official definitions of poverty in different countries are often not directly comparable due to the different ways poverty is measured. For example, countries account for the size of households in different ways in their poverty measures.
\n\n\n\nThe poverty lines shown here are an approximation of national definitions, harmonized to allow for comparisons across countries. For all countries apart from the US, we take the harmonized poverty line calculated by Jolliffe et al. (2022). These lines are calculated as the international dollar figure which, in the World Bank’s Poverty and Inequality Platform (PIP) data, yields the same poverty rate as the officially reported rate using national definitions in a particular year (around 2017).
\n\n\n\nFor the US, Jolliffe et al. (2022) use the OECD’s published poverty rate – which is measured against a relative poverty line of 50% of the median income. This yields a poverty line of $34.79 (measured using 2017 survey data). This however is not the official definition of poverty adopted in the US. We calculated an alternative harmonized figure for the US national poverty using the same method as Jolliffe et al. (2022), but based instead on the official 2019 poverty rate – as reported by the U.S. Census Bureau.
You can see in detail how we calculated this poverty line in this Google Colabs notebook.
\n\n\n\nJolliffe, Dean Mitchell, Daniel Gerszon Mahler, Christoph Lakner, Aziz Atamanov, and Samuel Kofi Tetteh Baah. 2022. Assessing the Impact of the 2017 PPPs on the International Poverty Line and Global Poverty. The World Bank. Available to read at the World Bank here.{/ref}
\n\n\n\nTo measure poverty globally, however, we need to apply a poverty line that is consistent across countries.
\n\n\n\nThis is the goal of the International Poverty Line of $2.15 per day – shown in red in the chart – which is set by the World Bank and used by the UN to monitor extreme poverty around the world.
\n\n\n\nWe see that, in global terms, this is an extremely low threshold indeed – set to reflect the poverty lines adopted nationally in the world’s poorest countries. It marks an incredibly low standard of living – a level of income much lower than just the cost of a healthy diet.
\n\n\n\nHow does the World Bank set the International Poverty Line?
\n![\""\""](\""https://owid.cloud/app/uploads/2022/10/Five-income-distributions-national-poverty-and-IPL-1-656x550.png\"")
The exact method used by the World Bank to set the International Poverty Line has changed somewhat over past updates. But each time the objective has been broadly the same – to find a “typical standard by which the poorest countries of the world judge their citizens to be impoverished.”{ref}As described by Dean Jolliffe and others (2022).
Jolliffe, Dean Mitchell, Daniel Gerszon Mahler, Christoph Lakner, Aziz Atamanov, and Samuel Kofi Tetteh Baah. 2022. Assessing the Impact of the 2017 PPPs on the International Poverty Line and Global Poverty. The World Bank. Available online here.{/ref}
The method used in the latest update to arrive at a figure of $2.15, measured in 2017 international-$, is based on a set of harmonized national poverty lines produced by Dean Joliffe and others – shown in the chart here.
\n\n\n\nAs you can see, there is a strong correlation between the poverty lines countries set, shown on the Y axis, and their income level – as measured here by GDP per capita, and plotted along the X axis.
\n\n\n\nThe International Poverty Line is calculated as the median national poverty line adopted among low-income countries – using the World Bank’s income classification system. These are the countries shaded in red in the chart and found in the bottom left corner.
\n\n\n\nAlthough the International Poverty Line is by far the most prominent international line, the same method is also used by the World Bank to set two higher poverty lines that reflect the national definitions adopted in lower-middle and upper-middle income groups shown in green and purple respectively. The median poverty line among these two groups of countries are $3.65 and $6.85, and these form the the World Bank’s lower-middle income and upper middle-income poverty lines.
\n\n\n\nYou can read more about the methodology used to set these lines in the World Bank’s flagship report on poverty, Poverty and Shared Prosperity.
\nA key question to ask of the World Bank’s updated poverty lines is the extent to which their real value – in terms of the goods and services they can buy – have been maintained, or if instead they represent a shifting of the goalposts.
\n\n\n\nOne way to answer this question, building on the explanation of change from 2011 to 2017 international-$ given above, is to compare the increase in the numbers to the rate of inflation in the US between these two years. An increase in-line with US inflation would indicate that the poverty lines have the same purchasing power. A rise greater than US inflation would indicate that the real value had increased. This is just a direct consequence of the definition of international-$ for a given base year – where one international-$ is defined as the value of goods and services that one US dollar could buy in that base year.
\n\n\n\nPrices in the US rose by 9% between 2011 and 2017. In the case of both the International Poverty Line and the lower-middle income poverty line, the increase across the update was roughly similar – from $1.90 to $2.15 (13%) and from $3.20 to $3.65 (14%) respectively. That the rise was close to the US inflation rate implies that the purchasing power of the lines was similar before and after the update. And, as demonstrated above, this is reflected in the data: global poverty estimates according to these two poverty lines are very similar across the old and updated methodologies.
\n\n\n\nThe upper-middle poverty line, however, rose by considerably more than US inflation – from $5.50 to $6.85, a 25% increase. Again, this is reflected in the data – with notably higher global poverty estimates resulting from the updated methodology.
\n\n\n\nWhy would the World Bank ‘shift the goal posts’ by using a higher poverty line? This answer lies in the method the World Bank uses to set the line, just explained in the box above. The upper-middle income country poverty line was set at $5.50 in 2011 international-$ because this was the median poverty line observed among this group of countries at the time of the last update to the World Bank methodology.{ref}As documented by Dean Jolliffe and Espen Beer Prydz (2016), in which they introduced the method now applied by the World Bank.
\n\n\n\nJolliffe, Dean, and Espen Beer Prydz. 2016. Estimating International Poverty Lines from Comparable National Thresholds. Washington, DC. Available online here.{/ref}
\n\n\n\nSince that time, the national poverty lines of some countries in this group have been revised upwards as incomes generally increased. This dynamic was not present for the lower income groupings of countries: in these cases those countries with a rising poverty line were also graduating into a higher income group, thereby maintaining the median poverty line within their former income group roughly constant.{ref}See the paper by Dean Jolliffe and others (2022) that introduces the updated World Bank poverty lines and provides a breakdown of the different factors contributing to the change in the poverty lines.
Jolliffe, Dean Mitchell, Daniel Gerszon Mahler, Christoph Lakner, Aziz Atamanov, and Samuel Kofi Tetteh Baah. 2022. Assessing the Impact of the 2017 PPPs on the International Poverty Line and Global Poverty. The World Bank. Available online here.{/ref}
For a fuller understanding of international-$ and PPPs in the context of global poverty measurement you may find some of the following articles helpful.
\n\n\n\nOn the 2017 PPPs:
\n\n\n\nAnd some more general discussions, in the context of past PPP rounds:
\n\n\n\nHow do reusable blocks work?
I don't know
How do reusable blocks work?
\n\n\n\nI don’t know
\nOur World in Data presents the data and research to make progress against the world’s largest problems.
Find all our data and research on poverty on our Poverty topic page.
One of the most important trends that we can learn from the work of social scientists and historians is that extreme poverty, as measured by people's level of consumption, has fallen considerably around the world in the last two centuries. But why should we care? Is it not the case that poor people might have less consumption but enjoy their lives just as much—or even more—than people with much higher consumption levels?
One way to find out is to simply ask. Subjective views are an important way of measuring welfare.
This is what the Gallup Organization did. The Gallup World Poll asked people around the world what they thought about their standard of living—not only about their income. The following chart compares the answers of people in different countries with the average income in those countries. It shows that, broadly speaking, people living in poorer countries tend to be less satisfied with their living standards.
This suggests that economic prosperity is not a vain, unimportant goal but rather a means for a better life. The correlation between rising incomes and higher self-reported life satisfaction is shown in our entry on happiness.
This is more than a technical point about how to measure welfare. It is an assertion that matters for how we understand and interpret development.
First, the smooth relationship between income and subjective well-being highlights the difficulties that arise from using a fixed threshold above which people are abruptly considered to be non-poor. In reality, subjective well-being does not suddenly improve above any given poverty line. This makes using a fixed poverty line to define destitution as a binary 'yes/no' problematic. Therefore, while the International Poverty Line is useful for understanding the changes in living conditions of the very poorest of the world, we must also take into account higher poverty lines reflecting the fact that living conditions at higher thresholds can still be destitute.
And second, the fact that people with very low incomes tend to be dissatisfied with their living standards shows that it would be incorrect to take a romantic view on what 'life in poverty' is like. As the data shows, there is just no empirical evidence that would suggest that living with very low consumption levels is romantic.
A disregard for or disinterest in poverty estimates that are calculated on the basis of low consumption and income levels is partly explained by the fact that it can be very difficult for people to imagine what it is like to live with very little. Even economists who think a lot about income and poverty find it difficult to understand what it means to live on a given income level. It is just hard to picture what life is like when all you know is a ""dollar-per-day"" figure.
To address this, Anna Rosling Rönnlund put together a captivating, visual project at Gapminder.org in which she portrays the living conditions of people living at different income levels. In Dollar Street you can find portraits of families and see how they cook, what they eat, how they sleep, what toilets they have available, what their children's toys look like, and much more.
","{""id"": ""wp-52784"", ""slug"": ""what-do-poor-people-think-about-poverty"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""Our World in Data presents the data and research to make progress against the world’s largest problems."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""text"": ""Find all our data and research on poverty on our "", ""spanType"": ""span-simple-text""}, {""children"": [{""url"": ""https://ourworldindata.org/poverty"", ""children"": [{""text"": ""Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""spanType"": ""span-bold""}, {""text"": "" topic page.\n"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One of the most important trends that we can learn from the work of social scientists and historians is that extreme poverty, as measured by people's level of consumption, has fallen considerably around the world in the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/share-in-poverty-relative-to-different-poverty-thresholds-historical?country=~OWID_WRL"", ""children"": [{""text"": ""last two centuries"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". But why should we care? Is it not the case that poor people might have less consumption but enjoy their lives just as much—or even more—than people with much higher consumption levels?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One way to find out is to simply ask. Subjective views are an important way of measuring welfare."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is what the "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Gallup Organization"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" did. The "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.gallup.com/analytics/232838/world-poll.aspx"", ""children"": [{""text"": ""Gallup World Poll"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" asked people around the world what they thought about their standard of living—not only about their income. The following chart compares the answers of people in different countries with the average income in those countries. It shows that, broadly speaking, people living in poorer countries tend to be less satisfied with their living standards."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Dissatisfaction with standard of living vs GDP per capita{ref} GDP per capita data from the World Bank. Survey data on the satisfaction with living standards is from the "", ""spanType"": ""span-simple-text""}, {""url"": ""http://www.gallup.com/services/170945/world-poll.aspx"", ""children"": [{""text"": ""Gallup World Poll"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". The idea for this chart is taken from Deaton (2010) – "", ""spanType"": ""span-simple-text""}, {""url"": ""https://rpds.princeton.edu/sites/rpds/files/media/deaton_price_indexes_inequality_and_the_measurement_of_world_poverty_aer.pdf"", ""children"": [{""text"": ""Price Indexes, Inequality, and the Measurement of World Poverty"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". In American Economic Review, 100, 1, 5--34. The lightly-shaded circles are for 2006, the darker circles for 2007, and the darkest circles are for 2008.{/ref}"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""alt"": """", ""size"": ""wide"", ""type"": ""image"", ""filename"": ""dissatisfied-vs-income.png"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This suggests that economic prosperity is not a vain, unimportant goal but rather a means for a better life. The correlation between rising incomes and higher self-reported life satisfaction is shown in our "", ""spanType"": ""span-simple-text""}, {""url"": ""/happiness-and-life-satisfaction/"", ""children"": [{""text"": ""entry on happiness"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This is more than a technical point about how to measure welfare. It is an assertion that matters for how we understand and interpret development."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""First, the smooth relationship between income and subjective well-being highlights the difficulties that arise from using a fixed threshold above which people are abruptly considered to be non-poor. In reality, subjective well-being does not suddenly improve above any given poverty line. This makes using a fixed poverty line to define destitution as a binary 'yes/no' problematic. Therefore, while the International Poverty Line is useful for understanding the changes in living conditions of the very poorest of the world, we must also take into account higher poverty lines reflecting the fact that living conditions at higher thresholds can still be destitute."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""And second, the fact that people with very low incomes tend to be dissatisfied with their living standards shows that it would be incorrect to take "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.google.co.uk/search?q=poor+but+happy&safe=off&source=lnms&tbm=isch&sa=X&ved=0ahUKEwi3jtfEo4rQAhVJIcAKHSjuAoQQ_AUICCgB&biw=2560&bih=1272&gws_rd=cr&ei=lfcZWPrQCezQgAa07ZmQCw"", ""children"": [{""text"": ""a romantic view"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" on what 'life in poverty' is like. As the data shows, there is just no empirical evidence that would suggest that living with very low consumption levels is romantic."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A disregard for or disinterest in poverty estimates that are calculated on the basis of low consumption and income levels is partly explained by the fact that it can be very difficult for people to imagine what it is like to live with very little. Even economists who think a lot about income and poverty find it difficult to understand what it means to live on a given income level. It is just hard to picture what life is like when all you know is a \""dollar-per-day\"" figure."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To address this, Anna Rosling Rönnlund put together a captivating, visual project at Gapminder.org in which she portrays the living conditions of people living at different income levels. In "", ""spanType"": ""span-simple-text""}, {""children"": [{""url"": ""https://www.gapminder.org/dollar-street/matrix"", ""children"": [{""text"": ""Dollar Street"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}], ""spanType"": ""span-italic""}, {""text"": "" you can find portraits of families and see how they cook, what they eat, how they sleep, what toilets they have available, what their children's toys look like, and much more."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""What do poor people think about poverty?"", ""authors"": [""Max Roser""], ""excerpt"": ""Some have a romantic view of what life on an extremely low income must be like. But we see from survey data that this is romantic view is not shared by those living in poverty themselves."", ""dateline"": ""April 8, 2017"", ""subtitle"": ""Some have a romantic view of what life on an extremely low income must be like. But we see from survey data that this is romantic view is not shared by those living in poverty themselves."", ""sidebar-toc"": false, ""featured-image"": ""dissatisfied-vs-income.png""}, ""createdAt"": ""2022-09-08T10:15:41.000Z"", ""published"": false, ""updatedAt"": ""2022-09-08T09:55:07.000Z"", ""revisionId"": null, ""publishedAt"": ""2017-04-08T09:15:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""unexpected wp component tag"", ""details"": ""Found unhandled wp:comment tag image""}], ""numBlocks"": 12, ""numErrors"": 1, ""wpTagCounts"": {""html"": 1, ""image"": 1, ""heading"": 1, ""paragraph"": 9}, ""htmlTagCounts"": {""p"": 10, ""h6"": 1, ""div"": 2, ""figure"": 1}}",2017-04-08 09:15:00,2024-02-16 14:22:46,1RdWJsqh16BxYqHLloK8RHnbuVYgVX2eHOlDodo9wIGk,"[""Max Roser""]",Some have a romantic view of what life on an extremely low income must be like. But we see from survey data that this is romantic view is not shared by those living in poverty themselves.,2022-09-08 10:15:41,2022-09-08 09:55:07,https://ourworldindata.org/wp-content/uploads/2017/01/dissatisfied-vs-income.png,{},"Our World in Data presents the data and research to make progress against the world’s largest problems. Find all our data and research on poverty on our **[Poverty](https://ourworldindata.org/poverty)** topic page. One of the most important trends that we can learn from the work of social scientists and historians is that extreme poverty, as measured by people's level of consumption, has fallen considerably around the world in the [last two centuries](https://ourworldindata.org/grapher/share-in-poverty-relative-to-different-poverty-thresholds-historical?country=~OWID_WRL). But why should we care? Is it not the case that poor people might have less consumption but enjoy their lives just as much—or even more—than people with much higher consumption levels? One way to find out is to simply ask. Subjective views are an important way of measuring welfare. This is what the _Gallup Organization_ did. The [Gallup World Poll](https://www.gallup.com/analytics/232838/world-poll.aspx) asked people around the world what they thought about their standard of living—not only about their income. The following chart compares the answers of people in different countries with the average income in those countries. It shows that, broadly speaking, people living in poorer countries tend to be less satisfied with their living standards. ## Dissatisfaction with standard of living vs GDP per capita{ref} GDP per capita data from the World Bank. Survey data on the satisfaction with living standards is from the [Gallup World Poll](http://www.gallup.com/services/170945/world-poll.aspx). The idea for this chart is taken from Deaton (2010) – [Price Indexes, Inequality, and the Measurement of World Poverty](https://rpds.princeton.edu/sites/rpds/files/media/deaton_price_indexes_inequality_and_the_measurement_of_world_poverty_aer.pdf). In American Economic Review, 100, 1, 5--34. The lightly-shaded circles are for 2006, the darker circles for 2007, and the darkest circles are for 2008.{/ref}Our World in Data presents the data and research to make progress against the world’s largest problems.
Find all our data and research on poverty on our Poverty topic page.\n
One of the most important trends that we can learn from the work of social scientists and historians is that extreme poverty, as measured by people’s level of consumption, has fallen considerably around the world in the last two centuries. But why should we care? Is it not the case that poor people might have less consumption but enjoy their lives just as much—or even more—than people with much higher consumption levels?
\n\n\n\nOne way to find out is to simply ask. Subjective views are an important way of measuring welfare.
\n\n\n\nThis is what the Gallup Organization did. The Gallup World Poll asked people around the world what they thought about their standard of living—not only about their income. The following chart compares the answers of people in different countries with the average income in those countries. It shows that, broadly speaking, people living in poorer countries tend to be less satisfied with their living standards.
\n\n\n\n![\""\""](\""https://ourworldindata.org/app/uploads/2017/01/dissatisfied-vs-income.png\"")
This suggests that economic prosperity is not a vain, unimportant goal but rather a means for a better life. The correlation between rising incomes and higher self-reported life satisfaction is shown in our entry on happiness.
\n\n\n\nThis is more than a technical point about how to measure welfare. It is an assertion that matters for how we understand and interpret development.
\n\n\n\nFirst, the smooth relationship between income and subjective well-being highlights the difficulties that arise from using a fixed threshold above which people are abruptly considered to be non-poor. In reality, subjective well-being does not suddenly improve above any given poverty line. This makes using a fixed poverty line to define destitution as a binary ‘yes/no’ problematic. Therefore, while the International Poverty Line is useful for understanding the changes in living conditions of the very poorest of the world, we must also take into account higher poverty lines reflecting the fact that living conditions at higher thresholds can still be destitute.
\n\n\n\nAnd second, the fact that people with very low incomes tend to be dissatisfied with their living standards shows that it would be incorrect to take a romantic view on what ‘life in poverty’ is like. As the data shows, there is just no empirical evidence that would suggest that living with very low consumption levels is romantic.
\n\n\n\nA disregard for or disinterest in poverty estimates that are calculated on the basis of low consumption and income levels is partly explained by the fact that it can be very difficult for people to imagine what it is like to live with very little. Even economists who think a lot about income and poverty find it difficult to understand what it means to live on a given income level. It is just hard to picture what life is like when all you know is a “dollar-per-day” figure.
\n\n\n\nTo address this, Anna Rosling Rönnlund put together a captivating, visual project at Gapminder.org in which she portrays the living conditions of people living at different income levels. In Dollar Street you can find portraits of families and see how they cook, what they eat, how they sleep, what toilets they have available, what their children’s toys look like, and much more.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Some have a romantic view of what life on an extremely low income must be like. But we see from survey data that this is romantic view is not shared by those living in poverty themselves."", ""protected"": false}, ""date_gmt"": ""2017-04-08T09:15:00"", ""modified"": ""2022-09-08T10:55:07"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Max Roser""], ""modified_gmt"": ""2022-09-08T09:55:07"", ""comment_status"": ""closed"", ""featured_media"": 10844, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2017/01/dissatisfied-vs-income-150x105.png"", ""medium_large"": ""/app/uploads/2017/01/dissatisfied-vs-income-768x538.png""}}" 52710,Biological and Chemical Weapons,biological-and-chemical-weapons,page,publish,"This page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/1e4MZvBeyg5FxGb8sOKSjFRwHz3LIMbgf7UVa2b24ed0/preview
","{""id"": ""wp-52710"", ""slug"": ""biological-and-chemical-weapons"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""This page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/1e4MZvBeyg5FxGb8sOKSjFRwHz3LIMbgf7UVa2b24ed0/preview"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Biological and Chemical Weapons"", ""authors"": [""Bastian Herre"", ""Max Roser""], ""excerpt"": ""Which countries have pursued, possessed, or used biological and chemical weapons recently and historically? Which countries have joined the international treaties to eliminate them?"", ""dateline"": ""October 5, 2022"", ""subtitle"": ""Which countries have pursued, possessed, or used biological and chemical weapons recently and historically? Which countries have joined the international treaties to eliminate them?"", ""sidebar-toc"": false, ""featured-image"": ""Biological-Chemical-Weapons.png""}, ""createdAt"": ""2022-09-06T16:00:28.000Z"", ""published"": false, ""updatedAt"": ""2023-07-18T20:12:19.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-10-05T09:50:05.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 1, ""numErrors"": 0, ""wpTagCounts"": {""paragraph"": 1}, ""htmlTagCounts"": {""p"": 1}}",2022-10-05 09:50:05,2024-02-16 14:22:41,,"[""Bastian Herre""]","Which countries have pursued, possessed, or used biological and chemical weapons recently and historically? Which countries have joined the international treaties to eliminate them?",2022-09-06 16:00:28,2023-07-18 20:12:19,https://ourworldindata.org/wp-content/uploads/2022/09/Biological-Chemical-Weapons.png,{},This page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/1e4MZvBeyg5FxGb8sOKSjFRwHz3LIMbgf7UVa2b24ed0/preview,"{""id"": 52710, ""date"": ""2022-10-05T10:50:05"", ""guid"": {""rendered"": ""https://owid.cloud/?page_id=52710""}, ""link"": ""https://owid.cloud/biological-and-chemical-weapons"", ""meta"": {""owid_publication_context_meta_field"": [], ""owid_key_performance_indicators_meta_field"": {""raw"": ""Fewer countries use, possess, and pursue biological and chemical weapons."", ""rendered"": ""Fewer countries use, possess, and pursue biological and chemical weapons.
\n""}}, ""slug"": ""biological-and-chemical-weapons"", ""tags"": [225], ""type"": ""page"", ""title"": {""rendered"": ""Biological and Chemical Weapons""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/52710""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/page""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/49"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=52710"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=52710"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=52710"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=52710""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/pages/52710/revisions"", ""count"": 12}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/52742"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 57906, ""href"": ""https://owid.cloud/wp-json/wp/v2/pages/52710/revisions/57906""}]}, ""author"": 49, ""parent"": 0, ""status"": ""publish"", ""content"": {""rendered"": ""\nThis page is now maintained in Google Docs: https://owid.cloud/admin/gdocs/1e4MZvBeyg5FxGb8sOKSjFRwHz3LIMbgf7UVa2b24ed0/preview
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Which countries have pursued, possessed, or used biological and chemical weapons recently and historically? Which countries have joined the international treaties to eliminate them?"", ""protected"": false}, ""date_gmt"": ""2022-10-05T09:50:05"", ""modified"": ""2023-07-18T21:12:19"", ""template"": """", ""categories"": [44, 55], ""menu_order"": 36, ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre""], ""modified_gmt"": ""2023-07-18T20:12:19"", ""comment_status"": ""closed"", ""featured_media"": 52742, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/09/Biological-Chemical-Weapons-150x79.png"", ""medium_large"": ""/app/uploads/2022/09/Biological-Chemical-Weapons-768x403.png""}}" 52633,About PIP data explorer block,about-pip-data-explorer-block,wp_block,publish,"All the data included in this explorer is available to download in GitHub, alongside a range of other poverty and inequality metrics.
This data explorer is collated and adapted from the World Bank’s Poverty and Inequality Platform (PIP).
The World Bank’s PIP data is a large collection of household surveys where steps have been taken by the World Bank to harmonize definitions and methods across countries and over time.
There is no global survey of incomes. To understand how incomes across the world compare, researchers need to rely on available national surveys.
Such surveys are partly designed with cross-country comparability in mind, but because the surveys reflect the circumstances and priorities of individual countries at the time of the survey, there are some important differences.
Income vs expenditure surveys
One important issue is that the survey data included within the PIP database tends to measure people’s income in high-income countries, and people’s consumption expenditure in poorer countries.
The two concepts are closely related: the income of a household equals their consumption plus any saving, or minus any borrowing or spending out of savings.
One important difference is that, while zero consumption is not a feasible value – people with zero consumption would starve – a zero income is a feasible value. This means that, at the bottom end of the distribution, income and consumption can give quite different pictures about a person’s welfare. For instance, a person dissaving in retirement may have a very low, or even zero, income, but have a high level of consumption nevertheless.
The gap between income and consumption is higher at the top of this distribution too, richer households tend to save more, meaning that the gap between income and consumption is higher at the top of this distribution too. Taken together, one implication is that inequality measured in terms of consumption is generally somewhat lower than the inequality measured in terms of income.
In our Data Explorer of this data there is the option to view only income survey data or only consumption survey data, or instead to pool the data available from both types of survey – which yields greater coverage.
Other comparability issues
There are a number of other ways in which comparability across surveys can be limited. The PIP Methodology Handbook provides a good summary of the comparability and data quality issues affecting this data and how it tries to address them.
In collating this survey data the World Bank takes a range of steps to harmonize it where possible, but comparability issues remain. These affect comparisons both across countries and within individual countries over time.
To help communicate the latter, the World Bank produces a variable that groups surveys within each individual country into more comparable ‘spells’. Our Data Explorer provides the option of viewing the data with these breaks in comparability indicated, and these spells are also indicated in our data download.
Along with data for individual countries, the World Bank also provides global and regional poverty estimates which aggregate over the available country data.
Surveys are not conducted annually in every country however – coverage is generally poorer the further back in time you look, and remains particularly patchy within Sub-Saharan Africa. You can see that visualized in our chart of the number of surveys included in the World Bank data by decade.
In order to produce global and regional aggregate estimates for a given year, the World Bank takes the surveys falling closest to that year for each country and ‘lines-up’ the data to the year being estimated by projecting it forwards or backwards.
This lining-up is generally done on the assumption that household incomes or expenditure grow in line with the growth rates observed in national accounts data. You can read more about the interpolation methods used by the World Bank in Chapter 5 of the Poverty and Inequality Platform Methodology Handbook.
To account for inflation and price differences across countries, the World Bank’s data is measured in international dollars. This is a hypothetical currency that results from price adjustments across time and place. It is defined as having the same purchasing power as one US-$ would in the United States in a given base year. One int.-$ buys the same quantity of goods and services no matter where or when it is spent.
There are many challenges to making such adjustments and they are far from perfect. Angus Deaton (Deaton, 2010) provides a good discussion of the difficulties involved in price adjustments and how this relates to global poverty measurement.
But in a world where price differences across countries and over time are large it is important to attempt to account for these differences as well as possible, and this is what these adjustments do.
In September 2022, the World Bank updated its methodology, and now uses international-$ expressed in 2017 prices – updated from 2011 prices. This has had little effect on our overall understanding of poverty and inequality around the world. But poverty estimates for particular countries vary somewhat between the old and updated methodology. You can read more about this update in our article From $1.90 to $2.15 a day: the updated International Poverty Line.
To allow for comparisons with the official data now expressed in 2017 international-$ data, the World Bank continues to release its poverty and inequality data expressed in 2011 international-$ as well. We have built a Data Explorer to allow you to compare these, and we make all figures available in terms of both sets of prices in our data download.
This dataset provides poverty estimates for a range of absolute and relative poverty lines.
An absolute poverty line represents a fixed standard of living; a threshold that is held constant across time. Within the World Bank’s poverty data, absolute poverty lines also aim to represent a standard of living that is fixed across countries (by converting local currencies to international-$). The International Poverty Line of $2.15 per day (in 2017 international-$) is the best known absolute poverty line and is used by the World Bank and the UN to measure extreme poverty around the world.
The value of relative poverty lines instead rises and falls as average incomes change within a given country. In most cases they are set at a certain fraction of the median income. Because of this, relative poverty can be considered a metric of inequality – it measures how spread out the bottom half of the income distribution is.
The idea behind measuring poverty in relative terms is that a person’s well-being depends not on their own absolute standard of living but on how that standard compares with some reference group, or whether it enables them to participate in the norms and customs of their society. For instance, joining a friend’s birthday celebration without shame might require more resources in a rich society if the norm is to go for an expensive meal out, or give costly presents.
Our dataset includes three commonly-used relative poverty lines: 40%, 50%, and 60% of the median.
Such lines are most commonly used in rich countries, and are the main way poverty is measured by the OECD and the European Union.
More recently, relative poverty measures have come to be applied in a global context. The share of people living below 50 per cent of median income is, for instance, one of the UN’s Sustainable Development Goal indicators. And the World Bank now produces estimates of global poverty using a Societal Poverty Line that combines absolute and relative components.
When comparing relative poverty rates around the world, however, it is important to keep in mind that – since average incomes are so far apart – such relative poverty lines relate to very different standards of living in rich and poor countries.
Many poor people today, as in the past, rely on subsistence farming rather than a monetary income gained from selling goods or their labor on the market. To take this into account and make a fair comparison of their living standards, the statisticians that produce these figures estimate the monetary value of their home production and add it to their income/expenditure.
","{""id"": ""wp-52633"", ""slug"": ""about-pip-data-explorer-block"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""All the data included in this explorer is available to download "", ""spanType"": ""span-simple-text""}, {""url"": ""https://github.com/owid/poverty-data"", ""children"": [{""text"": ""in GitHub"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", alongside a range of other poverty and inequality metrics."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""About this data"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""text"": [{""text"": ""Where is this data sourced from?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This data explorer is collated and adapted from the World Bank’s "", ""spanType"": ""span-simple-text""}, {""url"": ""https://pip.worldbank.org/home"", ""children"": [{""text"": ""Poverty and Inequality Platform"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" (PIP)."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The World Bank’s PIP data is a large collection of household surveys where steps have been taken by the World Bank to harmonize definitions and methods across countries and over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""About the comparability of household surveys"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There is no global survey of incomes. To understand how incomes across the world compare, researchers need to rely on available national surveys."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Such surveys are partly designed with cross-country comparability in mind, but because the surveys reflect the circumstances and priorities of individual countries at the time of the survey, there are some important differences."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Income vs expenditure surveys"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One important issue is that the survey data included within the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://pip.worldbank.org/home"", ""children"": [{""text"": ""PIP"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" database tends to measure people’s income in high-income countries, and people’s consumption expenditure in poorer countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The two concepts are closely related: the income of a household equals their consumption plus any saving, or minus any borrowing or spending out of savings. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One important difference is that, while zero consumption is not a feasible value – people with zero consumption would starve – a zero income is a feasible value. This means that, at the bottom end of the distribution, income and consumption can give quite different pictures about a person’s welfare. For instance, a person dissaving in retirement may have a very low, or even zero, income, but have a high level of consumption nevertheless."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The gap between income and consumption is higher at the top of this distribution too, richer households tend to save more, meaning that the gap between income and consumption is higher at the top of this distribution too. Taken together, one implication is that inequality measured in terms of consumption is generally somewhat lower than the inequality measured in terms of income."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In our "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/poverty#explore-data-poverty"", ""children"": [{""text"": ""Data Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of this data there is the option to view only income survey data or only consumption survey data, or instead to pool the data available from both types of survey – which yields greater coverage."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Other comparability issues"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are a number of other ways in which comparability across surveys can be limited. The PIP "", ""spanType"": ""span-simple-text""}, {""url"": ""https://datanalytics.worldbank.org/PIP-Methodology/"", ""children"": [{""text"": ""Methodology Handbook"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" provides a good summary of the comparability and data quality issues affecting this data and how it tries to address them."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In collating this survey data the World Bank takes a range of steps to harmonize it where possible, but comparability issues remain. These affect comparisons both across countries and within individual countries over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To help communicate the latter, the World Bank produces a variable that groups surveys within each individual country into more comparable ‘spells’. Our "", ""spanType"": ""span-simple-text""}, {""url"": ""http://ourworldindata.org/poverty#explore-data-poverty"", ""children"": [{""text"": ""Data Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" provides the option of viewing the data with these breaks in comparability indicated, and these spells are also indicated in our data "", ""spanType"": ""span-simple-text""}, {""url"": ""https://github.com/owid/poverty-data"", ""children"": [{""text"": ""download"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Global and regional poverty estimates"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Along with data for individual countries, the World Bank also provides global and regional poverty estimates which aggregate over the available country data."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Surveys are not conducted annually in every country however – coverage is generally poorer the further back in time you look, and remains particularly patchy within Sub-Saharan Africa. You can see that visualized in our chart of the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/data-deprivation-poverty-surveys-per-decade?time=latest"", ""children"": [{""text"": ""number of surveys included in the World Bank data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" by decade."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In order to produce global and regional aggregate estimates for a given year, the World Bank takes the surveys falling closest to that year for each country and ‘lines-up’ the data to the year being estimated by projecting it forwards or backwards."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This lining-up is generally done on the assumption that household incomes or expenditure grow in line with the growth rates observed in national accounts data. You can read more about the interpolation methods used by the World Bank in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://datanalytics.worldbank.org/PIP-Methodology/lineupestimates.html"", ""children"": [{""text"": ""Chapter 5"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of the Poverty and Inequality Platform Methodology Handbook."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How does the data account for inflation and for differences in the cost of living across countries?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To account for inflation and price differences across countries, the World Bank’s data is measured in international dollars. This is a hypothetical currency that results from price adjustments across time and place. It is defined as having the same purchasing power as one US-$ would in the United States in a given base year. One int.-$ buys the same quantity of goods and services no matter where or when it is spent."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""There are many challenges to making such adjustments and they are far from perfect. Angus Deaton ("", ""spanType"": ""span-simple-text""}, {""url"": ""https://rpds.princeton.edu/sites/g/files/toruqf1956/files/media/deaton_price_indexes_inequality_and_the_measurement_of_world_poverty_aer.pdf"", ""children"": [{""text"": ""Deaton, 2010"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "") provides a good discussion of the difficulties involved in price adjustments and how this relates to global poverty measurement."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But in a world where price differences across countries and over time are large it is important to attempt to account for these differences as well as possible, and this is what these adjustments do."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In September 2022, the World Bank updated its methodology, and now uses international-$ expressed in 2017 prices – updated from 2011 prices. This has had little effect on our overall understanding of poverty and inequality around the world. But poverty estimates for particular countries vary somewhat between the old and updated methodology. You can read more about this update in our article "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/from-1-90-to-2-15-a-day-the-updated-international-poverty-line"", ""children"": [{""children"": [{""text"": ""From $1.90 to $2.15 a day: the updated International Poverty Line"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To allow for comparisons with the official data now expressed in 2017 international-$ data, the World Bank continues to release its poverty and inequality data expressed in 2011 international-$ as well. We have built a "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/poverty-explorer-2011-vs-2017-ppp"", ""children"": [{""text"": ""Data Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to allow you to compare these, and we make all figures available in terms of both sets of prices in our "", ""spanType"": ""span-simple-text""}, {""url"": ""https://github.com/owid/poverty-data"", ""children"": [{""text"": ""data download"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Absolute vs relative poverty lines"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This dataset provides poverty estimates for a range of absolute and relative poverty lines."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""An absolute poverty line represents a fixed standard of living; a threshold that is held constant across time. Within the World Bank’s poverty data, absolute poverty lines also aim to represent a standard of living that is fixed across countries (by converting local currencies to international-$). The International Poverty Line of $2.15 per day (in 2017 international-$) is the best known absolute poverty line and is used by the World Bank and the UN to measure extreme poverty around the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The value of relative poverty lines instead rises and falls as average incomes change within a given country. In most cases they are set at a certain fraction of the median income. Because of this, relative poverty can be considered a metric of inequality – it measures how spread out the bottom half of the income distribution is."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The idea behind measuring poverty in relative terms is that a person’s well-being depends not on their own absolute standard of living but on how that standard compares with some reference group, or whether it enables them to participate in the norms and customs of their society. For instance, joining a friend’s birthday celebration without shame might require more resources in a rich society if the norm is to go for an expensive meal out, or give costly presents."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Our dataset includes three commonly-used relative poverty lines: 40%, 50%, and 60% of the median."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Such lines are most commonly used in rich countries, and are the main way poverty is measured by the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://data.oecd.org/inequality/poverty-rate.htm"", ""children"": [{""text"": ""OECD"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:At-risk-of-poverty_rate"", ""children"": [{""text"": ""European Union"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""More recently, relative poverty measures have come to be applied in a global context. The share of people living below 50 per cent of median income is, for instance, one of the UN’s "", ""spanType"": ""span-simple-text""}, {""url"": ""https://sdg-tracker.org/inequality#10.2"", ""children"": [{""text"": ""Sustainable Development Goal indicators"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". And the World Bank now produces estimates of global poverty using a "", ""spanType"": ""span-simple-text""}, {""url"": ""https://datatopics.worldbank.org/world-development-indicators/stories/societal-poverty-a-global-measure-of-relative-poverty.html"", ""children"": [{""text"": ""Societal Poverty Line"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" that combines absolute and relative components."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""When comparing relative poverty rates around the world, however, it is important to keep in mind that – since "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/poverty-explorer?tab=map&facet=none&hideControls=false&Metric=Median+income+or+expenditure&Household+survey+data+type=Show+data+from+both+income+and+expenditure+surveys&country=IND~MOZ~BRA~MDG~GHA"", ""children"": [{""text"": ""average incomes"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" are so far apart – such relative poverty lines relate to very different standards of living in rich and poor countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Does the data account for non-market income, such as food grown by subsistence farmers?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 3, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many poor people today, as in the past, rely on subsistence farming rather than a monetary income gained from selling goods or their labor on the market. To take this into account and make a fair comparison of their living standards, the statisticians that produce these figures estimate the monetary value of their home production and add it to their income/expenditure."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""About PIP data explorer block"", ""authors"": [null], ""dateline"": ""August 31, 2022"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2022-08-31T13:22:44.000Z"", ""published"": false, ""updatedAt"": ""2023-02-03T09:22:16.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-08-31T12:01:19.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 39, ""numErrors"": 0, ""wpTagCounts"": {""heading"": 7, ""paragraph"": 32}, ""htmlTagCounts"": {""p"": 32, ""h3"": 1, ""h4"": 6}}",2022-08-31 12:01:19,2024-02-16 14:23:03,,[null],,2022-08-31 13:22:44,2023-02-03 09:22:16,,{},"All the data included in this explorer is available to download [in GitHub](https://github.com/owid/poverty-data), alongside a range of other poverty and inequality metrics. ## About this data ### Where is this data sourced from? This data explorer is collated and adapted from the World Bank’s [Poverty and Inequality Platform](https://pip.worldbank.org/home) (PIP). The World Bank’s PIP data is a large collection of household surveys where steps have been taken by the World Bank to harmonize definitions and methods across countries and over time. ### About the comparability of household surveys There is no global survey of incomes. To understand how incomes across the world compare, researchers need to rely on available national surveys. Such surveys are partly designed with cross-country comparability in mind, but because the surveys reflect the circumstances and priorities of individual countries at the time of the survey, there are some important differences. **Income vs expenditure surveys** One important issue is that the survey data included within the [PIP](https://pip.worldbank.org/home) database tends to measure people’s income in high-income countries, and people’s consumption expenditure in poorer countries. The two concepts are closely related: the income of a household equals their consumption plus any saving, or minus any borrowing or spending out of savings. One important difference is that, while zero consumption is not a feasible value – people with zero consumption would starve – a zero income is a feasible value. This means that, at the bottom end of the distribution, income and consumption can give quite different pictures about a person’s welfare. For instance, a person dissaving in retirement may have a very low, or even zero, income, but have a high level of consumption nevertheless. The gap between income and consumption is higher at the top of this distribution too, richer households tend to save more, meaning that the gap between income and consumption is higher at the top of this distribution too. Taken together, one implication is that inequality measured in terms of consumption is generally somewhat lower than the inequality measured in terms of income. In our [Data Explorer](http://ourworldindata.org/poverty#explore-data-poverty) of this data there is the option to view only income survey data or only consumption survey data, or instead to pool the data available from both types of survey – which yields greater coverage. **Other comparability issues** There are a number of other ways in which comparability across surveys can be limited. The PIP [Methodology Handbook](https://datanalytics.worldbank.org/PIP-Methodology/) provides a good summary of the comparability and data quality issues affecting this data and how it tries to address them. In collating this survey data the World Bank takes a range of steps to harmonize it where possible, but comparability issues remain. These affect comparisons both across countries and within individual countries over time. To help communicate the latter, the World Bank produces a variable that groups surveys within each individual country into more comparable ‘spells’. Our [Data Explorer](http://ourworldindata.org/poverty#explore-data-poverty) provides the option of viewing the data with these breaks in comparability indicated, and these spells are also indicated in our data [download](https://github.com/owid/poverty-data). ### Global and regional poverty estimates Along with data for individual countries, the World Bank also provides global and regional poverty estimates which aggregate over the available country data. Surveys are not conducted annually in every country however – coverage is generally poorer the further back in time you look, and remains particularly patchy within Sub-Saharan Africa. You can see that visualized in our chart of the [number of surveys included in the World Bank data](https://ourworldindata.org/grapher/data-deprivation-poverty-surveys-per-decade?time=latest) by decade. In order to produce global and regional aggregate estimates for a given year, the World Bank takes the surveys falling closest to that year for each country and ‘lines-up’ the data to the year being estimated by projecting it forwards or backwards. This lining-up is generally done on the assumption that household incomes or expenditure grow in line with the growth rates observed in national accounts data. You can read more about the interpolation methods used by the World Bank in [Chapter 5](https://datanalytics.worldbank.org/PIP-Methodology/lineupestimates.html) of the Poverty and Inequality Platform Methodology Handbook. ### How does the data account for inflation and for differences in the cost of living across countries? To account for inflation and price differences across countries, the World Bank’s data is measured in international dollars. This is a hypothetical currency that results from price adjustments across time and place. It is defined as having the same purchasing power as one US-$ would in the United States in a given base year. One int.-$ buys the same quantity of goods and services no matter where or when it is spent. There are many challenges to making such adjustments and they are far from perfect. Angus Deaton ([Deaton, 2010](https://rpds.princeton.edu/sites/g/files/toruqf1956/files/media/deaton_price_indexes_inequality_and_the_measurement_of_world_poverty_aer.pdf)) provides a good discussion of the difficulties involved in price adjustments and how this relates to global poverty measurement. But in a world where price differences across countries and over time are large it is important to attempt to account for these differences as well as possible, and this is what these adjustments do. In September 2022, the World Bank updated its methodology, and now uses international-$ expressed in 2017 prices – updated from 2011 prices. This has had little effect on our overall understanding of poverty and inequality around the world. But poverty estimates for particular countries vary somewhat between the old and updated methodology. You can read more about this update in our article [_From $1.90 to $2.15 a day: the updated International Poverty Line_](https://ourworldindata.org/from-1-90-to-2-15-a-day-the-updated-international-poverty-line). To allow for comparisons with the official data now expressed in 2017 international-$ data, the World Bank continues to release its poverty and inequality data expressed in 2011 international-$ as well. We have built a [Data Explorer](https://ourworldindata.org/explorers/poverty-explorer-2011-vs-2017-ppp) to allow you to compare these, and we make all figures available in terms of both sets of prices in our [data download](https://github.com/owid/poverty-data). ### Absolute vs relative poverty lines This dataset provides poverty estimates for a range of absolute and relative poverty lines. An absolute poverty line represents a fixed standard of living; a threshold that is held constant across time. Within the World Bank’s poverty data, absolute poverty lines also aim to represent a standard of living that is fixed across countries (by converting local currencies to international-$). The International Poverty Line of $2.15 per day (in 2017 international-$) is the best known absolute poverty line and is used by the World Bank and the UN to measure extreme poverty around the world. The value of relative poverty lines instead rises and falls as average incomes change within a given country. In most cases they are set at a certain fraction of the median income. Because of this, relative poverty can be considered a metric of inequality – it measures how spread out the bottom half of the income distribution is. The idea behind measuring poverty in relative terms is that a person’s well-being depends not on their own absolute standard of living but on how that standard compares with some reference group, or whether it enables them to participate in the norms and customs of their society. For instance, joining a friend’s birthday celebration without shame might require more resources in a rich society if the norm is to go for an expensive meal out, or give costly presents. Our dataset includes three commonly-used relative poverty lines: 40%, 50%, and 60% of the median. Such lines are most commonly used in rich countries, and are the main way poverty is measured by the [OECD](https://data.oecd.org/inequality/poverty-rate.htm) and the [European Union](https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Glossary:At-risk-of-poverty_rate). More recently, relative poverty measures have come to be applied in a global context. The share of people living below 50 per cent of median income is, for instance, one of the UN’s [Sustainable Development Goal indicators](https://sdg-tracker.org/inequality#10.2). And the World Bank now produces estimates of global poverty using a [Societal Poverty Line](https://datatopics.worldbank.org/world-development-indicators/stories/societal-poverty-a-global-measure-of-relative-poverty.html) that combines absolute and relative components. When comparing relative poverty rates around the world, however, it is important to keep in mind that – since [average incomes](https://ourworldindata.org/explorers/poverty-explorer?tab=map&facet=none&hideControls=false&Metric=Median+income+or+expenditure&Household+survey+data+type=Show+data+from+both+income+and+expenditure+surveys&country=IND~MOZ~BRA~MDG~GHA) are so far apart – such relative poverty lines relate to very different standards of living in rich and poor countries. ### Does the data account for non-market income, such as food grown by subsistence farmers? Many poor people today, as in the past, rely on subsistence farming rather than a monetary income gained from selling goods or their labor on the market. To take this into account and make a fair comparison of their living standards, the statisticians that produce these figures estimate the monetary value of their home production and add it to their income/expenditure.","{""data"": {""wpBlock"": {""content"": ""\nAll the data included in this explorer is available to download in GitHub, alongside a range of other poverty and inequality metrics.
\n\n\n\nThis data explorer is collated and adapted from the World Bank’s Poverty and Inequality Platform (PIP).
\n\n\n\nThe World Bank’s PIP data is a large collection of household surveys where steps have been taken by the World Bank to harmonize definitions and methods across countries and over time.
\n\n\n\nThere is no global survey of incomes. To understand how incomes across the world compare, researchers need to rely on available national surveys.
\n\n\n\nSuch surveys are partly designed with cross-country comparability in mind, but because the surveys reflect the circumstances and priorities of individual countries at the time of the survey, there are some important differences.
\n\n\n\nIncome vs expenditure surveys
\n\n\n\nOne important issue is that the survey data included within the PIP database tends to measure people’s income in high-income countries, and people’s consumption expenditure in poorer countries.
\n\n\n\nThe two concepts are closely related: the income of a household equals their consumption plus any saving, or minus any borrowing or spending out of savings.
\n\n\n\nOne important difference is that, while zero consumption is not a feasible value – people with zero consumption would starve – a zero income is a feasible value. This means that, at the bottom end of the distribution, income and consumption can give quite different pictures about a person’s welfare. For instance, a person dissaving in retirement may have a very low, or even zero, income, but have a high level of consumption nevertheless.
\n\n\n\nThe gap between income and consumption is higher at the top of this distribution too, richer households tend to save more, meaning that the gap between income and consumption is higher at the top of this distribution too. Taken together, one implication is that inequality measured in terms of consumption is generally somewhat lower than the inequality measured in terms of income.
\n\n\n\nIn our Data Explorer of this data there is the option to view only income survey data or only consumption survey data, or instead to pool the data available from both types of survey – which yields greater coverage.
\n\n\n\nOther comparability issues
\n\n\n\nThere are a number of other ways in which comparability across surveys can be limited. The PIP Methodology Handbook provides a good summary of the comparability and data quality issues affecting this data and how it tries to address them.
\n\n\n\nIn collating this survey data the World Bank takes a range of steps to harmonize it where possible, but comparability issues remain. These affect comparisons both across countries and within individual countries over time.
\n\n\n\nTo help communicate the latter, the World Bank produces a variable that groups surveys within each individual country into more comparable ‘spells’. Our Data Explorer provides the option of viewing the data with these breaks in comparability indicated, and these spells are also indicated in our data download.
\n\n\n\nAlong with data for individual countries, the World Bank also provides global and regional poverty estimates which aggregate over the available country data.
\n\n\n\nSurveys are not conducted annually in every country however – coverage is generally poorer the further back in time you look, and remains particularly patchy within Sub-Saharan Africa. You can see that visualized in our chart of the number of surveys included in the World Bank data by decade.
\n\n\n\nIn order to produce global and regional aggregate estimates for a given year, the World Bank takes the surveys falling closest to that year for each country and ‘lines-up’ the data to the year being estimated by projecting it forwards or backwards.
\n\n\n\nThis lining-up is generally done on the assumption that household incomes or expenditure grow in line with the growth rates observed in national accounts data. You can read more about the interpolation methods used by the World Bank in Chapter 5 of the Poverty and Inequality Platform Methodology Handbook.
\n\n\n\nTo account for inflation and price differences across countries, the World Bank’s data is measured in international dollars. This is a hypothetical currency that results from price adjustments across time and place. It is defined as having the same purchasing power as one US-$ would in the United States in a given base year. One int.-$ buys the same quantity of goods and services no matter where or when it is spent.
\n\n\n\nThere are many challenges to making such adjustments and they are far from perfect. Angus Deaton (Deaton, 2010) provides a good discussion of the difficulties involved in price adjustments and how this relates to global poverty measurement.
\n\n\n\nBut in a world where price differences across countries and over time are large it is important to attempt to account for these differences as well as possible, and this is what these adjustments do.
\n\n\n\nIn September 2022, the World Bank updated its methodology, and now uses international-$ expressed in 2017 prices – updated from 2011 prices. This has had little effect on our overall understanding of poverty and inequality around the world. But poverty estimates for particular countries vary somewhat between the old and updated methodology. You can read more about this update in our article From $1.90 to $2.15 a day: the updated International Poverty Line.
\n\n\n\nTo allow for comparisons with the official data now expressed in 2017 international-$ data, the World Bank continues to release its poverty and inequality data expressed in 2011 international-$ as well. We have built a Data Explorer to allow you to compare these, and we make all figures available in terms of both sets of prices in our data download.
\n\n\n\nThis dataset provides poverty estimates for a range of absolute and relative poverty lines.
\n\n\n\nAn absolute poverty line represents a fixed standard of living; a threshold that is held constant across time. Within the World Bank’s poverty data, absolute poverty lines also aim to represent a standard of living that is fixed across countries (by converting local currencies to international-$). The International Poverty Line of $2.15 per day (in 2017 international-$) is the best known absolute poverty line and is used by the World Bank and the UN to measure extreme poverty around the world.
\n\n\n\nThe value of relative poverty lines instead rises and falls as average incomes change within a given country. In most cases they are set at a certain fraction of the median income. Because of this, relative poverty can be considered a metric of inequality – it measures how spread out the bottom half of the income distribution is.
\n\n\n\nThe idea behind measuring poverty in relative terms is that a person’s well-being depends not on their own absolute standard of living but on how that standard compares with some reference group, or whether it enables them to participate in the norms and customs of their society. For instance, joining a friend’s birthday celebration without shame might require more resources in a rich society if the norm is to go for an expensive meal out, or give costly presents.
\n\n\n\nOur dataset includes three commonly-used relative poverty lines: 40%, 50%, and 60% of the median.
\n\n\n\nSuch lines are most commonly used in rich countries, and are the main way poverty is measured by the OECD and the European Union.
More recently, relative poverty measures have come to be applied in a global context. The share of people living below 50 per cent of median income is, for instance, one of the UN’s Sustainable Development Goal indicators. And the World Bank now produces estimates of global poverty using a Societal Poverty Line that combines absolute and relative components.
When comparing relative poverty rates around the world, however, it is important to keep in mind that – since average incomes are so far apart – such relative poverty lines relate to very different standards of living in rich and poor countries.
\n\n\n\nMany poor people today, as in the past, rely on subsistence farming rather than a monetary income gained from selling goods or their labor on the market. To take this into account and make a fair comparison of their living standards, the statisticians that produce these figures estimate the monetary value of their home production and add it to their income/expenditure.
\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}" 52598,The world has recently become less democratic,less-democratic,post,publish,"Many more people than in the past have democratic rights. But now there is growing concern that this progress is currently being partially undone.{ref}See for example:
Lührmann, Anna, and Staffan Lindberg. 2019. A third wave of autocratization is here: what is new about it? Democratization 26(7): 1095-1113.
Haggard, Stephan, and Robert Kaufman. 2021. The Anatomy of Democratic Backsliding. Journal of Democracy 32(4): 27-41.{/ref} Is this true? Has the world become less democratic recently?
This article shows that the leading approaches to measuring democracy indicate that this is true: the world has become less democratic in recent years.
Democracy is in decline, whether we look at big changes in the number of democracies and the people living in them; at small changes in the extent of democratic rights; or at medium-sized changes in the number of, and people living in, countries that are autocratizing.{ref}I use ‘autocratizing’ and ‘becoming less democratic’ as synonyms.{/ref}
The extent of this decline is substantial, but it is also uncertain and limited. We can see it clearly across democracy metrics: the world has fallen from all-time democratic highs to a level similar to earlier decades. But the extent of this decline depends on which democracy measure we use. And it is limited in the sense that the world remains much more democratic than it was even half a century ago.
Finally, the recent democratic decline is precedented, and past declines were reversed. The world underwent phases of autocratization in the 1930s and again in the 1960s and 1970s. Back then, people fought to turn the tide, and pushed democratic rights to unprecedented heights. We can do the same again.
The simplest way to explore whether the world has recently become less democratic is by looking at how many countries are democracies.
In the chart I rely on the Regimes of the World (RoW) classification{ref}Lührmann, Anna, Marcus Tannnberg, and Staffan Lindberg. 2018. Regimes of the World (RoW): Opening New Avenues for the Comparative Study of Political Regimes. Politics and Governance 6(1): 60-77{/ref} as a measure for whether a country is a democracy. You can read more about the data in our article on it.
Using the RoW data, the chart shows that the world has become less democratic in recent years. The number of democracies in the world reached an all-time high in 2016, with 96 electoral democracies. In 2022, their number has fallen to 90 countries.
RoW is just one of the leading approaches of identifying which countries are democracies. But the findings are similar when we look at other approaches to measuring democracy: you can see this in versions of the same chart based on the Lexical Index, Boix-Miller-Rosato, Freedom House, and the Economist Intelligence Unit.{ref}I do not include the regime classifications by Polity and Bertelsmann Transformation Index here because Polity currently ends in 2018, and the Bertelsmann Transformation Index does not cover countries they consider to be consolidated democracies.{/ref} You can find all these charts and more in our Democracy Data Explorer.
The same is true of liberal democracies. Their number has fallen from 44 countries in 2009 to 32 in 2022.
The number of democracies does not tell us how many people enjoy democratic rights. But when we look at this data, the findings are the same.
The number of people that have democratic rights has recently plummeted: between 2016 and 2022, this number fell from 3.9 billion to 2.3 billion people.
Similarly, the number of people living in liberal democracies fell from 1.2 billion in 2012 to 1 billion a decade later.
Examples of people losing democratic rights — according to the RoW data — are the 1.4 billion people in India,{ref}The reclassification in RoW is the result of recent changes in the V-Dem data, which identify declines in the autonomy of the election management body, the freedom and fairness of elections, and especially the freedom of expression, the media, and civil society. You can read more in V-Dem’s 2021 annual report Autocratization Turns Viral.{/ref} the 84 million people in Turkey, and the 28 million people in Venezuela.{ref}Different approaches to measuring democracy also identify democratic declines in these countries.{/ref}
Instead of the absolute number of democracies and people living in them, we may be interested in the share of democratic countries and the share of the world population living in them. The world is also becoming less democratic based on these metrics.
Looking at the number of democracies and the people living in them tells us about big changes worldwide: when countries become democracies, or stop being liberal or electoral democracies altogether.
But countries and their citizens may experience smaller changes in their democratic rights that fall short of changing the overall regime type. This is because the extent of democratic rights differs across democracies and non-democracies, but also within them. Such changes within regimes would not show up in the previous charts.
Therefore another way of looking at whether the world has recently become less democratic is to look at how democratic countries are, treating democracy as a spectrum.
We identify how democratic countries are with the electoral democracy index from the Varieties of Democracy (V-Dem) project.{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. V-Dem [Country-Year/Country-Date] Dataset v13. Varieties of Democracy (V-Dem) Project.{/ref} It captures the extent to which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed. You can read more about the data in our article on it.
Using V-Dem’s electoral democracy index, the first chart shows that countries, on average, have become less democratic: in 2022, countries received an electoral-democracy score of 0.50 on the scale from 0 to 1. This score is slightly lower than the highest ever average of electoral democracy (0.53) in 2014.
The chart shows one of several leading approaches. But recent changes look similar if we instead use V-Dem’s other democracy indices, such as its liberal democracy index, or the Economist Intelligence Unit’s data.{ref}I do not include the indices by Polity and the Bertelsmann Transformation Index here because Polity currently ends in 2018, and the Bertelsmann Transformation Index does not cover countries they consider to be consolidated democracies.{/ref}
The second chart instead tells us about how many democratic rights people have by weighing a country’s democracy score by its population.{ref}This means that populous countries matter more when calculating the average than countries with small populations.{/ref} Calculated this way, peoples’ democratic rights have decreased as well: countries received a population-weighted average score of 0.40 for electoral democracy in 2021. This is down from an all-time high of 0.5 in 2004.
Global averages in the extent to which countries are democratic and people enjoy democratic rights tell us about smaller changes in democracy that fall short of regime change.
But they do not tell us which countries are becoming more or less democratic.
One more way of studying global changes in democracy is therefore identifying how the number of countries that are autocratizing or democratizing is changing.
We identify which countries are becoming less or more democratic with data from the Episodes of Regime Transformation (ERT)-project.{ref}Edgell, Amanda, Seraphine Maerz, Laura Maxwell, Richard Morgan, Juraj Medzihorsky, Matthew Wilson, Vanessa Boese, Sebastian Hellmeier, Jean Lachapelle, Patrik Lindenfors, Anna Lührmann, and Staffan Lindberg. 2022. Episodes of Regime Transformation Dataset (v13). Varieties of Democracy (V-Dem) Project.{/ref} We use their data to identify which countries are autocratizing, democratizing, and which countries are not clearly moving in either direction.{ref}Based on ERT, a country is autocratizing from when V-Dem’s electoral democracy index decreases by 0.01, until the score increases or remains unchanged for four years, and the total decrease between start and end amounts to a decrease of at least 0.10. Democratizing countries are classified analogously. We exclude Ukraine from 2002 to 2004 and El Salvador from 2015 and 2017 because for them democratizing and autocratizing episodes happen to overlap.{/ref} This chart shows how each country has been classified for the end of each year since 1900.
What does it mean for a country to be autocratizing or democratizing? ERT seeks to strike a balance between large and small changes in how democratic countries are. {ref}Seraphine Maerz, Amanda Edgell, Matthew Wilson, Sebastian Hellmeier, Staffan Lindberg. 2021. A Framework for Understanding Regime Transformation: Introducing the ERT Dataset. Varieties of Democracy Institute: Working Paper No. 113. University of Gothenburg.{/ref} It captures smaller changes in democracy that fall short of regime change. At the same time, it only codes a country as autocratizing when there is a substantial decrease in its democracy score. This is because very small decreases may be fleeting and not indicate broader shifts towards less democracy, or overstate changes altogether because the measurement is uncertain.{ref}This means, however, that some countries are not classified as autocratizing even though their score visibly declines. One example is the United States in the 2010s, whose decline between 2015 and 2020 fell just barely short of the ERT threshold.{/ref} ERT also allows for temporary stagnation because autocratization may not happen abruptly in one year, but slowly over several years.{ref}Bermeo, Nancy. 2016. On democratic backsliding. Journal of Democracy 27(1): 5-19.{/ref}
As shown in the chart, the number of countries that are autocratizing has been increasing: in 2022, 40 countries were autocratizing. This is an all-time high. For a long time, the number of autocratizing countries was offset by a larger number of democratizing countries. But since 2012, the number of countries that are becoming less democratic has been higher.
The number of people living in autocratizing countries is increasing as well: in 2022, 3.2 billion people lived in countries that were becoming less democratic. This is also an all-time high. The number has been trending upward since the late 1980s, with a large jump in 2000 when India is reclassified as becoming less democratic.{ref}This move away from democratic rights looks very similar if instead of looking at absolute numbers, we look at the share of autocratizing countries or the world population living in them.{/ref}
Other examples of people living in autocratizing countries according to the ERT data are the 214 million people in Brazil, the 274 million people in Indonesia, and the 38 million people in Poland.{ref}Different approaches to measuring democracy also identify democratic declines in these countries.{/ref}
The data shown here indicate that the world has recently become less democratic. But how much less democratic has the world become?
Based on the data here, the democratic decline has been substantial. We can readily identify it in the different charts above. We can do so even though the charts put the recent trends into the historical context of the last two centuries. The world has fallen from all-time democratic highs to now — depending on the specific metrics — look more like the 2000s, the 1990s, or even the late 1980s.
But the extent of the democratic decline is also more uncertain; its precise degree depends on the approach of measuring democracy.
The clearest example of this is India. Since India is home to 1.4 billion people, its classification has a big impact on some global trends. RoW and ERT classify India as an electoral autocracy and as substantially autocratizing in recent years. The Boix-Miller-Rosato data, however, disagrees, and continues to classify the country as a democracy. This means that using their data, the number of people in the world living in democracies has not declined, but stagnated.
At the same time, we should not overstate the differences between the approaches either. All other approaches{ref}This includes the Lexical Index, Freedom House, the Bertelsmann Transformation Index, and the Economist Intelligence Unit.{/ref} still identify a democratic decline in India. The approaches also tend to agree that metrics which look at people instead of countries see larger declines — because countries with large populations have tended to lose more democratic rights. Even the Boix-Miller-Rosato data sees an increase in the number of people living in non-democracies. The choice of approach therefore does not matter so much that it would question all characteristics of the recent global democratic decline.{ref}The same goes for uncertainty in the measurement of democracy itself.{/ref}
And while the democratic decline has been substantial, it so far has been limited. The world is still much closer to all-time democratic highs than to historical levels: this clearly is the case for the 19th and early 20th centuries, when democratic rights and democracies were nearly absent. But it also holds for large parts of the later 20th century, when democratic rights were heavily concentrated in some parts of the world.
A decline in democratic rights might seem like an unprecedented step backwards. But it is not. It has happened before: the world underwent prolonged phases of autocratization in the 1930s and again in the 1960s and 1970s.
In the 1930s, a similar number of democracies as recently broke down — though back then that meant more than half of the world’s previously democratic countries. And the decline in the global democratic average was much larger than it has been recently.
In the 1960s and 1970s, many countries were becoming less democratic, with as many as 20 countries autocratizing at any given time. And several hundred million people lost democratic rights when India’s democracy eroded.
But these declines in democracy were temporary. People in India and around the world gained more democratic rights than ever before.
People turned previous autocratic tides by advocating relentlessly for governing themselves democratically. We have done it before, and can do it again.
I thank Hannah Ritchie and Max Roser for their very helpful comments and ideas about how to improve this article.
","{""id"": ""wp-52598"", ""slug"": ""less-democratic"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""url"": ""https://ourworldindata.org/democratic-rights"", ""children"": [{""text"": ""Many more people than in the past have democratic rights"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". But now there is growing concern that this progress is currently being partially undone.{ref}See for example:"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Lührmann, Anna, and Staffan Lindberg. 2019. A third wave of autocratization is here: what is new about it? Democratization 26(7): 1095-1113."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Haggard, Stephan, and Robert Kaufman. 2021. The Anatomy of Democratic Backsliding. Journal of Democracy 32(4): 27-41.{/ref} Is this true? Has the world become less democratic recently?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This article shows that the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracies-measurement"", ""children"": [{""text"": ""leading approaches"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" to measuring democracy indicate that this is true: the world has become less democratic in recent years."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Democracy is in decline, whether we look at big changes in the number of democracies and the people living in them; at small changes in the extent of democratic rights; or at medium-sized changes in the number of, and people living in, countries that are autocratizing.{ref}I use ‘autocratizing’ and ‘becoming less democratic’ as synonyms.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The extent of this decline is substantial, but it is also uncertain and limited. We can see it clearly across democracy metrics: the world has fallen from all-time democratic highs to a level similar to earlier decades. But the extent of this decline depends on which democracy measure we use. And it is limited in the sense that the world remains much more democratic than it was even half a century ago."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Finally, the recent democratic decline is precedented, and past declines were reversed. The world underwent phases of autocratization in the 1930s and again in the 1960s and 1970s. Back then, people fought to turn the tide, and pushed democratic rights to unprecedented heights. We can do the same again."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""Recently, the number of democracies has declined… "", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The simplest way to explore whether the world has recently become less democratic is by looking at how many countries are democracies."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the chart I rely on the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=L%C3%BChrmann%2C+Anna%2C+Marcus+Tannnberg%2C+and+Staffan+Lindberg.+2018.+Regimes+of+the+World+%28RoW%29%3A+Opening+New+Avenues+for+the+Comparative+Study+of+Political+Regimes.+Politics+and+Governance+6%281%29%3A+60-77.%7B%2Fref%7D&btnG="", ""children"": [{""text"": ""Regimes of the World (RoW) classification"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""{ref}Lührmann, Anna, Marcus Tannnberg, and Staffan Lindberg. 2018. Regimes of the World (RoW): Opening New Avenues for the Comparative Study of Political Regimes. Politics and Governance 6(1): 60-77{/ref} as a measure for whether a country is a democracy. You can read more about the data "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/regimes-of-the-world-data"", ""children"": [{""text"": ""in our article on it"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""Using the RoW data, the chart shows that the world has become less democratic in recent years. The number of democracies in the world reached an all-time high in 2016, with 96 electoral democracies. In 2022, their number has fallen to 90 countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""RoW is "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracies-measurement"", ""children"": [{""text"": ""just one of the leading approaches"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" of identifying which countries are democracies. But the findings are similar when we look at other approaches to measuring democracy: you can see this in versions of the same chart based on the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Lexical+Index&Metric=%C2%AD%C2%ADPolitical+regime&Sub-metric=Number+of+democracies"", ""children"": [{""text"": ""Lexical"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Lexical+Index&Metric=%C2%AD%C2%ADPolitical+regime&Sub-metric=People+living+in+democracies"", ""children"": [{""text"": ""Index"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Boix-Miller-Rosato&Metric=Democracy&Sub-metric=Number+of+democracies"", ""children"": [{""text"": ""Boix"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""-"", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Boix-Miller-Rosato&Metric=Democracy&Sub-metric=People+living+in+democracies"", ""children"": [{""text"": ""Miller-Rosato"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Freedom+House&Metric=%C2%AD%C2%AD%C2%AD%C2%ADFreedom&Sub-metric=Number+of+free+countries+and+territories"", ""children"": [{""text"": ""Freedom"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Freedom+House&Metric=%C2%AD%C2%AD%C2%AD%C2%ADElectoral+democracy&Sub-metric=%C2%ADNumber+of+democracies"", ""children"": [{""text"": ""House"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", and the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Economist+Intelligence+Unit&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADPolitical+regime&Sub-metric=%C2%AD%C2%AD%C2%ADNumber+of+democracies"", ""children"": [{""text"": ""Economist"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Economist+Intelligence+Unit&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADPolitical+regime&Sub-metric=%C2%AD%C2%AD%C2%ADPeople+living+in+democracies"", ""children"": [{""text"": ""Intelligence Unit"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}I do not include the regime classifications by "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN&Dataset=Polity&Metric=%C2%AD%C2%AD%C2%ADDemocracy&Sub-metric=Main+index"", ""children"": [{""text"": ""Polity"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN&Dataset=Bertelsmann+Transformation+Index&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADDemocratic+features&Sub-metric=Main+index"", ""children"": [{""text"": ""Bertelsmann Transformation Index"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" here because Polity currently ends in 2018, and the Bertelsmann Transformation Index does not cover countries they consider to be consolidated democracies.{/ref} You can find all these charts and more in our "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy"", ""children"": [{""text"": ""Democracy Data Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The same is true of liberal democracies. Their number has fallen from 44 countries in 2009 to 32 in 2022."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/countries-democracies-autocracies-row?stackMode=absolute&country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""…and fewer people are living in democracies"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The number of democracies does not tell us "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""how many people"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" enjoy democratic rights. But when we look at this data, the findings are the same."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The number of people that have democratic rights has recently plummeted: between 2016 and 2022, this number fell from 3.9 billion to 2.3 billion people."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Similarly, the number of people living in liberal democracies fell from 1.2 billion in 2012 to 1 billion a decade later."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Examples of people losing democratic rights — according to the RoW data — are the 1.4 billion people in India,{ref}The reclassification in RoW is the result of recent changes in the V-Dem data, which identify declines in the autonomy of the election management body, the freedom and fairness of elections, and especially the freedom of expression, the media, and civil society. You can read more in V-Dem’s 2021 annual report"", ""spanType"": ""span-simple-text""}, {""url"": ""https://web.archive.org/web/20220130230849/https://v-dem.net/static/website/files/dr/dr_2021.pdf"", ""children"": [{""text"": "" Autocratization Turns Vir"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://www.v-dem.net/files/25/DR%202021.pdf"", ""children"": [{""text"": ""al"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref} the 84 million people in Turkey, and the 28 million people in Venezuela.{ref}Different approaches to measuring democracy also identify democratic declines in these countries.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Instead of the absolute number of democracies and people living in them, we may be interested in the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=~OWID_WRL&Dataset=Regimes+of+the+World&Metric=%C2%ADPolitical+regime&Sub-metric=Number+and+share+of+democracies"", ""children"": [{""text"": ""share of democratic countries"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?stackMode=relative&facet=none&country=~OWID_WRL&Dataset=Regimes+of+the+World&Metric=%C2%ADPolitical+regime&Sub-metric=People+living+in+democracies"", ""children"": [{""text"": ""share of the world population"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" living in them. The world is also becoming less democratic based on these metrics."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/people-living-in-democracies?country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""In recent years, countries and people have fewer democratic rights"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Looking at the number of democracies and the people living in them tells us about big changes worldwide: when countries become democracies, or stop being liberal or electoral democracies altogether."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But countries and their citizens may experience smaller changes in their democratic rights that fall short of changing the overall regime type. This is because the extent of democratic rights differs "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democratic-world"", ""children"": [{""text"": ""across democracies and non-democracies, but also within them"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Such changes within regimes would not show up in the previous charts."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Therefore another way of looking at whether the world has recently become less democratic is to look at "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""how"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" democratic countries are, treating democracy as a spectrum."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We identify how democratic countries are with the electoral democracy index from the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.v-dem.net/vdemds.html"", ""children"": [{""text"": ""Varieties of Democracy (V-Dem) project"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Carl+Henrik+Knutsen%2C+Staffan+I.+Lindberg%2C+Jan+Teorell%2C+David+Altman%2C+Michael+Bernhard%2C+Agnes+Cornell%2C+M.+Steven+Fish%2C+Lisa+Gastaldi%2C+Haakon+Gjerl%C3%B8w%2C+Adam+Glynn%2C+Ana+Good+God%2C+Sandra+Grahn%2C+Allen+Hicken%2C+Katrin+Kinzelbach%2C+Joshua+Krusell%2C+Kyle+L.+Marquardt%2C+Kelly+McMann%2C+Valeriya+Mechkova%2C+Juraj+Medzihorsky%2C+Natalia+Natsika%2C+Anja+Neundorf%2C+Pamela+Paxton%2C+Daniel+Pemstein%2C+Josefine+Pernes%2C+Oskar+Ryd%C3%A9n%2C+Johannes+von+R%C3%B6mer%2C+Brigitte+Seim%2C+Rachel+Sigman%2C+Svend-Erik+Skaaning%2C+Jeffrey+Staton%2C+Aksel+Sundstr%C3%B6m%2C+Eitan+Tzelgov%2C+Yi-ting+Wang%2C+Tore+Wig%2C+Steven+Wilson+and+Daniel+Ziblatt.+2023.+V-Dem+%5BCountry-Year%2FCountry-Date%5D+Dataset+v13.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG="", ""children"": [{""text"": ""V-Dem [Country-Year/Country-Date] Dataset v13."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" Varieties of Democracy (V-Dem) Project.{/ref} It captures the extent to which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed. You can read more about the data "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/vdem-electoral-democracy-data"", ""children"": [{""text"": ""in our article on it"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}, {""spanType"": ""span-newline""}, {""spanType"": ""span-newline""}, {""text"": ""Using V-Dem’s electoral democracy index, the first chart shows that countries, on average, have become less democratic: in 2022, countries received an electoral-democracy score of 0.50 on the scale from 0 to 1. This score is slightly lower than the highest ever average of electoral democracy (0.53) in 2014."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The chart shows one of "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracies-measurement"", ""children"": [{""text"": ""several leading approaches"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". But recent changes look similar if we instead use V-Dem’s other democracy indices, such as its "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~OWID_WRL&Dataset=Varieties+of+Democracy&Metric=Liberal+democracy&Sub-metric=Main+index"", ""children"": [{""text"": ""liberal democracy"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=~OWID_WRL&Dataset=Varieties+of+Democracy&Metric=Liberal+democracy&Sub-metric=%C2%ADMain+index+weighted+by+population"", ""children"": [{""text"": ""index"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", or the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~OWID_WRL&Dataset=Economist+Intelligence+Unit&Metric=%C2%AD%C2%AD%C2%AD%C2%ADFreedom&Sub-metric=Civil+liberties+%28score%29"", ""children"": [{""text"": ""Economist Intelligence Unit’s"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://ourworldindata.org/explorers/democracy?country=~OWID_WRL&Dataset=Economist+Intelligence+Unit&Metric=%C2%AD%C2%AD%C2%AD%C2%ADFreedom&Sub-metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADMain+index+weighted+by+population"", ""children"": [{""text"": ""data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}I do not include the indices by "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN&Dataset=Polity&Metric=%C2%AD%C2%AD%C2%ADDemocracy&Sub-metric=Main+index"", ""children"": [{""text"": ""Polity"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN&Dataset=Bertelsmann+Transformation+Index&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADDemocratic+features&Sub-metric=Main+index"", ""children"": [{""text"": ""Bertelsmann Transformation Index"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" here because Polity currently ends in 2018, and the Bertelsmann Transformation Index does not cover countries they consider to be consolidated democracies.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/electoral-democracy?tab=chart&country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The second chart instead tells us about how many democratic rights "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""people"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" have by weighing a country’s democracy score by its population.{ref}This means that populous countries matter more when calculating the average than countries with small populations.{/ref} Calculated this way, peoples’ democratic rights have decreased as well: countries received a population-weighted average score of 0.40 for electoral democracy in 2021. This is down from an all-time high of 0.5 in 2004."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/electoral-democracy-popw-vdem?country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""Recently, more countries are autocratizing, and more people live in them"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Global averages in the extent to which countries are democratic and people enjoy democratic rights tell us about smaller changes in democracy that fall short of regime change."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But they do not tell us "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""which"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" countries are becoming more or less democratic."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""One more way of studying global changes in democracy is therefore identifying how the number of countries that are autocratizing or democratizing is changing."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We identify which countries are becoming less or more democratic with data from the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://www.v-dem.net/ertds.html"", ""children"": [{""text"": ""Episodes of Regime Transformation (ERT)-project"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{ref}Edgell, Amanda, Seraphine Maerz, Laura Maxwell, Richard Morgan, Juraj Medzihorsky, Matthew Wilson, Vanessa Boese, Sebastian Hellmeier, Jean Lachapelle, Patrik Lindenfors, Anna Lührmann, and Staffan Lindberg. 2022. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=dgell%2C+Amanda%2C+Seraphine+Maerz%2C+Laura+Maxwell%2C+Richard+Morgan%2C+Juraj+Medzihorsky%2C+Matthew+Wilson%2C+Vanessa+Boese%2C+Sebastian+Hellmeier%2C+Jean+Lachapelle%2C+Patrik+Lindenfors%2C+Anna+L%C3%BChrmann%2C+and+Staffan+Lindberg.+2022.+Episodes+of+Regime+Transformation+Dataset+%28v13%29.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG="", ""children"": [{""text"": ""Episodes of Regime Transformation Dataset (v13)"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Varieties of Democracy (V-Dem) Project.{/ref} We use their data to identify which countries are autocratizing, democratizing, and which countries are not clearly moving in either direction.{ref}Based on ERT, a country is autocratizing from when V-Dem’s electoral democracy index decreases by 0.01, until the score increases or remains unchanged for four years, "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""and"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" the total decrease between start and end amounts to a decrease of at least 0.10. Democratizing countries are classified analogously. We exclude Ukraine from 2002 to 2004 and El Salvador from 2015 and 2017 because for them democratizing and autocratizing episodes happen to overlap.{/ref} "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/political-regime-ert"", ""children"": [{""text"": ""This"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" chart shows how each country has been classified for the end of each year since 1900."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""What does it mean for a country to be autocratizing or democratizing? ERT seeks to strike a balance between large and small changes in how democratic countries are. {ref}Seraphine Maerz, Amanda Edgell, Matthew Wilson, Sebastian Hellmeier, Staffan Lindberg. 2021. A Framework for Understanding Regime Transformation: Introducing the ERT Dataset. Varieties of Democracy Institute: Working Paper No. 113. University of Gothenburg.{/ref} It captures smaller changes in democracy that fall short of regime change. At the same time, it only codes a country as autocratizing when there is a substantial decrease in its democracy score. This is because very small decreases may be fleeting and not indicate broader shifts towards less democracy, or overstate changes altogether because the measurement is uncertain.{ref}This means, however, that some countries are not classified as autocratizing even though their score visibly declines. One example is the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~USA&Dataset=Varieties+of+Democracy&Metric=Electoral+democracy&Sub-metric=Main+index"", ""children"": [{""text"": ""United States"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in the 2010s, whose decline between 2015 and 2020 fell just barely short of the ERT threshold.{/ref} ERT also allows for temporary stagnation because autocratization may not happen abruptly in one year, but slowly over several years.{ref}Bermeo, Nancy. 2016. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Bermeo%2C+Nancy.+2016.+On+democratic+backsliding.+Journal+of+Democracy+27%281%29%3A+5-19.&btnG="", ""children"": [{""text"": ""On democratic backsliding"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "". Journal of Democracy 27(1): 5-19.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As shown in the chart, the number of countries that are autocratizing has been increasing: in 2022, 40 countries were autocratizing. This is an all-time high. For a long time, the number of autocratizing countries was offset by a larger number of democratizing countries. But since 2012, the number of countries that are becoming less democratic has been higher."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/countries-democratizing-autocratizing-countries-ert?stackMode=absolute&country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The number of people living in autocratizing countries is increasing as well: in 2022, 3.2 billion people lived in countries that were becoming less democratic. This is also an all-time high. The number has been trending upward since the late 1980s, with a large jump in 2000 when India is reclassified as becoming less democratic.{ref}This move away from democratic rights looks very similar if instead of looking at absolute numbers, we look at the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/share-democratizing-autocratizing-countries-ert"", ""children"": [{""text"": ""share of autocratizing countries"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" or the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/people-living-in-democratizing-autocratizing-countries-ert?stackMode=relative&country=~OWID_WRL"", ""children"": [{""text"": ""world population living in them"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Other examples of people living in autocratizing countries according to the ERT data are the 214 million people in Brazil, the 274 million people in Indonesia, and the 38 million people in Poland.{ref}Different approaches to measuring democracy also identify democratic declines in these countries.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/people-living-in-democratizing-autocratizing-countries-ert"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""The democratic decline has been substantial, but more uncertain and limited"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The data shown here indicate that the world has recently become less democratic. But "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""how much"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" less democratic has the world become?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Based on the data here, the democratic decline has been substantial. We can readily identify it in the different charts above. We can do so even though the charts put the recent trends into the historical context of the last two centuries. The world has fallen from all-time democratic highs to now — depending on the specific metrics — look more like the 2000s, the 1990s, or even the late 1980s."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But the extent of the democratic decline is also more uncertain; its precise degree depends on the approach of measuring democracy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The clearest example of this is India. Since India is home to 1.4 billion people, its classification has a big impact on some global trends. RoW and ERT classify India as an electoral autocracy and as substantially autocratizing in recent years. The "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Boix-Miller-Rosato&Metric=Democracy&Sub-metric=Main+classification"", ""children"": [{""text"": ""Boix-Miller-Rosato"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" data, however, disagrees, and continues to classify the country as a democracy. This means that "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Boix-Miller-Rosato&Metric=Democracy&Sub-metric=People+living+in+democracies"", ""children"": [{""text"": ""using their data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", the number of people in the world living in democracies has not declined, but stagnated."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""At the same time, we should not overstate the differences between the approaches either. All other approaches{ref}This includes the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~IND&Dataset=Lexical+Index&Metric=%C2%AD%C2%ADPolitical+regime&Sub-metric=Main+classification"", ""children"": [{""text"": ""Lexical Index"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~IND&Dataset=Freedom+House&Metric=%C2%AD%C2%AD%C2%AD%C2%ADFreedom&Sub-metric=Main+classification"", ""children"": [{""text"": ""Freedom House"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", the Bertelsmann "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~IND&Dataset=Bertelsmann+Transformation+Index&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADPolitical+regime&Sub-metric=Main+classification"", ""children"": [{""text"": ""Transformation"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~IND&Dataset=Bertelsmann+Transformation+Index&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADDemocratic+features&Sub-metric=Main+index"", ""children"": [{""text"": ""Index"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", and the "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy?tab=chart&country=~IND&Dataset=Economist+Intelligence+Unit&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADDemocratic+features&Sub-metric=Main+index"", ""children"": [{""text"": ""Economist Intelligence Unit"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref} still identify a democratic decline in India. The approaches also tend to agree that metrics which look at people instead of countries see larger declines — because countries with large populations have tended to lose more democratic rights. Even the Boix-Miller-Rosato data sees an increase in the number of people living in "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""non-"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""democracies. The choice of approach therefore does not matter so much that it would question all characteristics of the recent global democratic decline.{ref}The same goes for uncertainty in the measurement of democracy itself.{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""And while the democratic decline has been substantial, it so far has been limited. The world is still much closer to all-time democratic highs than to historical levels: this clearly is the case for the 19th and early 20th centuries, when democratic rights and democracies were nearly absent. But it also holds for large parts of the later 20th century, when democratic rights were "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democratic-world"", ""children"": [{""text"": ""heavily concentrated"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in some parts of the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""The world has become less democratic before — and turned the tide"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A decline in democratic rights might seem like an unprecedented step backwards. But it is not. It has happened before: the world underwent prolonged phases of autocratization in the 1930s and again in the 1960s and 1970s."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the 1930s, a similar number of democracies as recently broke down — though back then that meant more than half of the world’s previously democratic countries. And the decline in the global democratic average was much larger than it has been recently."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the 1960s and 1970s, many countries were becoming less democratic, with as many as 20 countries autocratizing at any given time. And several hundred million people lost democratic rights when India’s democracy eroded."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But these declines in democracy were temporary. People in India and around the world gained more democratic rights than ever before."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""People turned previous autocratic tides by advocating relentlessly for governing themselves democratically. We have done it before, and can do it again."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/countries-democratizing-autocratizing-countries-ert?stackMode=absolute&country=~OWID_WRL"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""Keep reading on "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""Our World in Data"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/explorers/democracy"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/democratic-rights"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/democratic-world"", ""type"": ""prominent-link"", ""title"": """", ""description"": """", ""parseErrors"": []}, {""text"": [{""children"": [{""text"": ""Acknowledgements"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""I thank Hannah Ritchie and Max Roser for their very helpful comments and ideas about how to improve this article."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}], ""type"": ""article"", ""title"": ""The world has recently become less democratic"", ""authors"": [""Bastian Herre""], ""excerpt"": ""Many more people have democratic rights than in the past. Some of this progress has recently been reversed."", ""dateline"": ""September 6, 2022"", ""subtitle"": ""Many more people have democratic rights than in the past. Some of this progress has recently been reversed."", ""sidebar-toc"": false, ""featured-image"": ""less-democratic2.png""}, ""createdAt"": ""2022-08-30T13:32:38.000Z"", ""published"": false, ""updatedAt"": ""2023-03-14T11:59:49.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-09-06T10:40:47.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [{""name"": ""prominent link missing title"", ""details"": ""Prominent link is missing a title attribute""}], ""numBlocks"": 60, ""numErrors"": 1, ""wpTagCounts"": {""html"": 7, ""heading"": 8, ""paragraph"": 43, ""owid/prominent-link"": 4}, ""htmlTagCounts"": {""p"": 43, ""h4"": 8, ""iframe"": 7}}",2022-09-06 10:40:47,2024-02-16 14:22:54,14ThFloGyQv4uSGdEeZROWq_2yvHC9ggnAFij_9Oyz6g,"[""Bastian Herre""]",Many more people have democratic rights than in the past. Some of this progress has recently been reversed.,2022-08-30 13:32:38,2023-03-14 11:59:49,https://ourworldindata.org/wp-content/uploads/2022/09/less-democratic2.png,{},"[Many more people than in the past have democratic rights](https://ourworldindata.org/democratic-rights). But now there is growing concern that this progress is currently being partially undone.{ref}See for example: Lührmann, Anna, and Staffan Lindberg. 2019. A third wave of autocratization is here: what is new about it? Democratization 26(7): 1095-1113. Haggard, Stephan, and Robert Kaufman. 2021. The Anatomy of Democratic Backsliding. Journal of Democracy 32(4): 27-41.{/ref} Is this true? Has the world become less democratic recently? This article shows that the [leading approaches](https://ourworldindata.org/democracies-measurement) to measuring democracy indicate that this is true: the world has become less democratic in recent years. Democracy is in decline, whether we look at big changes in the number of democracies and the people living in them; at small changes in the extent of democratic rights; or at medium-sized changes in the number of, and people living in, countries that are autocratizing.{ref}I use ‘autocratizing’ and ‘becoming less democratic’ as synonyms.{/ref} The extent of this decline is substantial, but it is also uncertain and limited. We can see it clearly across democracy metrics: the world has fallen from all-time democratic highs to a level similar to earlier decades. But the extent of this decline depends on which democracy measure we use. And it is limited in the sense that the world remains much more democratic than it was even half a century ago. Finally, the recent democratic decline is precedented, and past declines were reversed. The world underwent phases of autocratization in the 1930s and again in the 1960s and 1970s. Back then, people fought to turn the tide, and pushed democratic rights to unprecedented heights. We can do the same again. ## Recently, the number of democracies has declined… The simplest way to explore whether the world has recently become less democratic is by looking at how many countries are democracies. In the chart I rely on the [Regimes of the World (RoW) classification](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=L%C3%BChrmann%2C+Anna%2C+Marcus+Tannnberg%2C+and+Staffan+Lindberg.+2018.+Regimes+of+the+World+%28RoW%29%3A+Opening+New+Avenues+for+the+Comparative+Study+of+Political+Regimes.+Politics+and+Governance+6%281%29%3A+60-77.%7B%2Fref%7D&btnG=){ref}Lührmann, Anna, Marcus Tannnberg, and Staffan Lindberg. 2018. Regimes of the World (RoW): Opening New Avenues for the Comparative Study of Political Regimes. Politics and Governance 6(1): 60-77{/ref} as a measure for whether a country is a democracy. You can read more about the data [in our article on it](https://ourworldindata.org/regimes-of-the-world-data). Using the RoW data, the chart shows that the world has become less democratic in recent years. The number of democracies in the world reached an all-time high in 2016, with 96 electoral democracies. In 2022, their number has fallen to 90 countries. RoW is [just one of the leading approaches](https://ourworldindata.org/democracies-measurement) of identifying which countries are democracies. But the findings are similar when we look at other approaches to measuring democracy: you can see this in versions of the same chart based on the [Lexical](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Lexical+Index&Metric=%C2%AD%C2%ADPolitical+regime&Sub-metric=Number+of+democracies)[Index](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Lexical+Index&Metric=%C2%AD%C2%ADPolitical+regime&Sub-metric=People+living+in+democracies), [Boix](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Boix-Miller-Rosato&Metric=Democracy&Sub-metric=Number+of+democracies)-[Miller-Rosato](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Boix-Miller-Rosato&Metric=Democracy&Sub-metric=People+living+in+democracies), [Freedom](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Freedom+House&Metric=%C2%AD%C2%AD%C2%AD%C2%ADFreedom&Sub-metric=Number+of+free+countries+and+territories)[House](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Freedom+House&Metric=%C2%AD%C2%AD%C2%AD%C2%ADElectoral+democracy&Sub-metric=%C2%ADNumber+of+democracies), and the [Economist](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Economist+Intelligence+Unit&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADPolitical+regime&Sub-metric=%C2%AD%C2%AD%C2%ADNumber+of+democracies)[Intelligence Unit](https://ourworldindata.org/explorers/democracy?facet=none&country=ARG~AUS~BWA~CHN~OWID_WRL&Dataset=Economist+Intelligence+Unit&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADPolitical+regime&Sub-metric=%C2%AD%C2%AD%C2%ADPeople+living+in+democracies).{ref}I do not include the regime classifications by [Polity](https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN&Dataset=Polity&Metric=%C2%AD%C2%AD%C2%ADDemocracy&Sub-metric=Main+index) and [Bertelsmann Transformation Index](https://ourworldindata.org/explorers/democracy?country=ARG~AUS~BWA~CHN&Dataset=Bertelsmann+Transformation+Index&Metric=%C2%AD%C2%AD%C2%AD%C2%AD%C2%ADDemocratic+features&Sub-metric=Main+index) here because Polity currently ends in 2018, and the Bertelsmann Transformation Index does not cover countries they consider to be consolidated democracies.{/ref} You can find all these charts and more in our [Democracy Data Explorer](https://ourworldindata.org/explorers/democracy). The same is true of liberal democracies. Their number has fallen from 44 countries in 2009 to 32 in 2022.Many more people than in the past have democratic rights. But now there is growing concern that this progress is currently being partially undone.{ref}See for example:
\n\n\n\nLührmann, Anna, and Staffan Lindberg. 2019. A third wave of autocratization is here: what is new about it? Democratization 26(7): 1095-1113.
\n\n\n\nHaggard, Stephan, and Robert Kaufman. 2021. The Anatomy of Democratic Backsliding. Journal of Democracy 32(4): 27-41.{/ref} Is this true? Has the world become less democratic recently?
\n\n\n\nThis article shows that the leading approaches to measuring democracy indicate that this is true: the world has become less democratic in recent years.
\n\n\n\nDemocracy is in decline, whether we look at big changes in the number of democracies and the people living in them; at small changes in the extent of democratic rights; or at medium-sized changes in the number of, and people living in, countries that are autocratizing.{ref}I use ‘autocratizing’ and ‘becoming less democratic’ as synonyms.{/ref}
\n\n\n\nThe extent of this decline is substantial, but it is also uncertain and limited. We can see it clearly across democracy metrics: the world has fallen from all-time democratic highs to a level similar to earlier decades. But the extent of this decline depends on which democracy measure we use. And it is limited in the sense that the world remains much more democratic than it was even half a century ago.
\n\n\n\nFinally, the recent democratic decline is precedented, and past declines were reversed. The world underwent phases of autocratization in the 1930s and again in the 1960s and 1970s. Back then, people fought to turn the tide, and pushed democratic rights to unprecedented heights. We can do the same again.
\n\n\n\nThe simplest way to explore whether the world has recently become less democratic is by looking at how many countries are democracies.
\n\n\n\nIn the chart I rely on the Regimes of the World (RoW) classification{ref}Lührmann, Anna, Marcus Tannnberg, and Staffan Lindberg. 2018. Regimes of the World (RoW): Opening New Avenues for the Comparative Study of Political Regimes. Politics and Governance 6(1): 60-77{/ref} as a measure for whether a country is a democracy. You can read more about the data in our article on it.
Using the RoW data, the chart shows that the world has become less democratic in recent years. The number of democracies in the world reached an all-time high in 2016, with 96 electoral democracies. In 2022, their number has fallen to 90 countries.
RoW is just one of the leading approaches of identifying which countries are democracies. But the findings are similar when we look at other approaches to measuring democracy: you can see this in versions of the same chart based on the Lexical Index, Boix–Miller-Rosato, Freedom House, and the Economist Intelligence Unit.{ref}I do not include the regime classifications by Polity and Bertelsmann Transformation Index here because Polity currently ends in 2018, and the Bertelsmann Transformation Index does not cover countries they consider to be consolidated democracies.{/ref} You can find all these charts and more in our Democracy Data Explorer.
\n\n\n\nThe same is true of liberal democracies. Their number has fallen from 44 countries in 2009 to 32 in 2022.
\n\n\n\n\n\n\n\nThe number of democracies does not tell us how many people enjoy democratic rights. But when we look at this data, the findings are the same.
\n\n\n\nThe number of people that have democratic rights has recently plummeted: between 2016 and 2022, this number fell from 3.9 billion to 2.3 billion people.
\n\n\n\nSimilarly, the number of people living in liberal democracies fell from 1.2 billion in 2012 to 1 billion a decade later.
\n\n\n\nExamples of people losing democratic rights — according to the RoW data — are the 1.4 billion people in India,{ref}The reclassification in RoW is the result of recent changes in the V-Dem data, which identify declines in the autonomy of the election management body, the freedom and fairness of elections, and especially the freedom of expression, the media, and civil society. You can read more in V-Dem’s 2021 annual report Autocratization Turns Viral.{/ref} the 84 million people in Turkey, and the 28 million people in Venezuela.{ref}Different approaches to measuring democracy also identify democratic declines in these countries.{/ref}
\n\n\n\nInstead of the absolute number of democracies and people living in them, we may be interested in the share of democratic countries and the share of the world population living in them. The world is also becoming less democratic based on these metrics.
\n\n\n\n\n\n\n\nLooking at the number of democracies and the people living in them tells us about big changes worldwide: when countries become democracies, or stop being liberal or electoral democracies altogether.
\n\n\n\nBut countries and their citizens may experience smaller changes in their democratic rights that fall short of changing the overall regime type. This is because the extent of democratic rights differs across democracies and non-democracies, but also within them. Such changes within regimes would not show up in the previous charts.
\n\n\n\nTherefore another way of looking at whether the world has recently become less democratic is to look at how democratic countries are, treating democracy as a spectrum.
\n\n\n\nWe identify how democratic countries are with the electoral democracy index from the Varieties of Democracy (V-Dem) project.{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. V-Dem [Country-Year/Country-Date] Dataset v13. Varieties of Democracy (V-Dem) Project.{/ref} It captures the extent to which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed. You can read more about the data in our article on it.
Using V-Dem’s electoral democracy index, the first chart shows that countries, on average, have become less democratic: in 2022, countries received an electoral-democracy score of 0.50 on the scale from 0 to 1. This score is slightly lower than the highest ever average of electoral democracy (0.53) in 2014.
The chart shows one of several leading approaches. But recent changes look similar if we instead use V-Dem’s other democracy indices, such as its liberal democracy index, or the Economist Intelligence Unit’s data.{ref}I do not include the indices by Polity and the Bertelsmann Transformation Index here because Polity currently ends in 2018, and the Bertelsmann Transformation Index does not cover countries they consider to be consolidated democracies.{/ref}
\n\n\n\n\n\n\n\nThe second chart instead tells us about how many democratic rights people have by weighing a country’s democracy score by its population.{ref}This means that populous countries matter more when calculating the average than countries with small populations.{/ref} Calculated this way, peoples’ democratic rights have decreased as well: countries received a population-weighted average score of 0.40 for electoral democracy in 2021. This is down from an all-time high of 0.5 in 2004.
\n\n\n\n\n\n\n\nGlobal averages in the extent to which countries are democratic and people enjoy democratic rights tell us about smaller changes in democracy that fall short of regime change.
\n\n\n\nBut they do not tell us which countries are becoming more or less democratic.
\n\n\n\nOne more way of studying global changes in democracy is therefore identifying how the number of countries that are autocratizing or democratizing is changing.
\n\n\n\nWe identify which countries are becoming less or more democratic with data from the Episodes of Regime Transformation (ERT)-project.{ref}Edgell, Amanda, Seraphine Maerz, Laura Maxwell, Richard Morgan, Juraj Medzihorsky, Matthew Wilson, Vanessa Boese, Sebastian Hellmeier, Jean Lachapelle, Patrik Lindenfors, Anna Lührmann, and Staffan Lindberg. 2022. Episodes of Regime Transformation Dataset (v13). Varieties of Democracy (V-Dem) Project.{/ref} We use their data to identify which countries are autocratizing, democratizing, and which countries are not clearly moving in either direction.{ref}Based on ERT, a country is autocratizing from when V-Dem’s electoral democracy index decreases by 0.01, until the score increases or remains unchanged for four years, and the total decrease between start and end amounts to a decrease of at least 0.10. Democratizing countries are classified analogously. We exclude Ukraine from 2002 to 2004 and El Salvador from 2015 and 2017 because for them democratizing and autocratizing episodes happen to overlap.{/ref} This chart shows how each country has been classified for the end of each year since 1900.
\n\n\n\nWhat does it mean for a country to be autocratizing or democratizing? ERT seeks to strike a balance between large and small changes in how democratic countries are. {ref}Seraphine Maerz, Amanda Edgell, Matthew Wilson, Sebastian Hellmeier, Staffan Lindberg. 2021. A Framework for Understanding Regime Transformation: Introducing the ERT Dataset. Varieties of Democracy Institute: Working Paper No. 113. University of Gothenburg.{/ref} It captures smaller changes in democracy that fall short of regime change. At the same time, it only codes a country as autocratizing when there is a substantial decrease in its democracy score. This is because very small decreases may be fleeting and not indicate broader shifts towards less democracy, or overstate changes altogether because the measurement is uncertain.{ref}This means, however, that some countries are not classified as autocratizing even though their score visibly declines. One example is the United States in the 2010s, whose decline between 2015 and 2020 fell just barely short of the ERT threshold.{/ref} ERT also allows for temporary stagnation because autocratization may not happen abruptly in one year, but slowly over several years.{ref}Bermeo, Nancy. 2016. On democratic backsliding. Journal of Democracy 27(1): 5-19.{/ref}
\n\n\n\nAs shown in the chart, the number of countries that are autocratizing has been increasing: in 2022, 40 countries were autocratizing. This is an all-time high. For a long time, the number of autocratizing countries was offset by a larger number of democratizing countries. But since 2012, the number of countries that are becoming less democratic has been higher.
\n\n\n\n\n\n\n\nThe number of people living in autocratizing countries is increasing as well: in 2022, 3.2 billion people lived in countries that were becoming less democratic. This is also an all-time high. The number has been trending upward since the late 1980s, with a large jump in 2000 when India is reclassified as becoming less democratic.{ref}This move away from democratic rights looks very similar if instead of looking at absolute numbers, we look at the share of autocratizing countries or the world population living in them.{/ref}
\n\n\n\nOther examples of people living in autocratizing countries according to the ERT data are the 214 million people in Brazil, the 274 million people in Indonesia, and the 38 million people in Poland.{ref}Different approaches to measuring democracy also identify democratic declines in these countries.{/ref}
\n\n\n\n\n\n\n\nThe data shown here indicate that the world has recently become less democratic. But how much less democratic has the world become?
\n\n\n\nBased on the data here, the democratic decline has been substantial. We can readily identify it in the different charts above. We can do so even though the charts put the recent trends into the historical context of the last two centuries. The world has fallen from all-time democratic highs to now — depending on the specific metrics — look more like the 2000s, the 1990s, or even the late 1980s.
\n\n\n\nBut the extent of the democratic decline is also more uncertain; its precise degree depends on the approach of measuring democracy.
\n\n\n\nThe clearest example of this is India. Since India is home to 1.4 billion people, its classification has a big impact on some global trends. RoW and ERT classify India as an electoral autocracy and as substantially autocratizing in recent years. The Boix-Miller-Rosato data, however, disagrees, and continues to classify the country as a democracy. This means that using their data, the number of people in the world living in democracies has not declined, but stagnated.
\n\n\n\nAt the same time, we should not overstate the differences between the approaches either. All other approaches{ref}This includes the Lexical Index, Freedom House, the Bertelsmann Transformation Index, and the Economist Intelligence Unit.{/ref} still identify a democratic decline in India. The approaches also tend to agree that metrics which look at people instead of countries see larger declines — because countries with large populations have tended to lose more democratic rights. Even the Boix-Miller-Rosato data sees an increase in the number of people living in non-democracies. The choice of approach therefore does not matter so much that it would question all characteristics of the recent global democratic decline.{ref}The same goes for uncertainty in the measurement of democracy itself.{/ref}
\n\n\n\nAnd while the democratic decline has been substantial, it so far has been limited. The world is still much closer to all-time democratic highs than to historical levels: this clearly is the case for the 19th and early 20th centuries, when democratic rights and democracies were nearly absent. But it also holds for large parts of the later 20th century, when democratic rights were heavily concentrated in some parts of the world.
\n\n\n\nA decline in democratic rights might seem like an unprecedented step backwards. But it is not. It has happened before: the world underwent prolonged phases of autocratization in the 1930s and again in the 1960s and 1970s.
\n\n\n\nIn the 1930s, a similar number of democracies as recently broke down — though back then that meant more than half of the world’s previously democratic countries. And the decline in the global democratic average was much larger than it has been recently.
\n\n\n\nIn the 1960s and 1970s, many countries were becoming less democratic, with as many as 20 countries autocratizing at any given time. And several hundred million people lost democratic rights when India’s democracy eroded.
\n\n\n\nBut these declines in democracy were temporary. People in India and around the world gained more democratic rights than ever before.
\n\n\n\nPeople turned previous autocratic tides by advocating relentlessly for governing themselves democratically. We have done it before, and can do it again.
\n\n\n\n\n\n\n\nI thank Hannah Ritchie and Max Roser for their very helpful comments and ideas about how to improve this article.
\n"", ""protected"": false}, ""excerpt"": {""rendered"": ""Many more people have democratic rights than in the past. Some of this progress has recently been reversed."", ""protected"": false}, ""date_gmt"": ""2022-09-06T10:40:47"", ""modified"": ""2023-03-14T11:59:49"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Bastian Herre""], ""modified_gmt"": ""2023-03-14T11:59:49"", ""comment_status"": ""closed"", ""featured_media"": 52862, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/09/less-democratic2-e1662990170102-150x76.png"", ""medium_large"": ""/app/uploads/2022/09/less-democratic2-e1662990170102-768x389.png""}}" 52489,We just published a new data explorer on the Environmental Impacts of Food,new-env-food-data-explorer,post,publish,,"{""id"": ""wp-52489"", ""slug"": ""new-env-food-data-explorer"", ""content"": {""toc"": [], ""body"": [], ""type"": ""article"", ""title"": ""We just published a new data explorer on the Environmental Impacts of Food"", ""authors"": [""Hannah Ritchie""], ""excerpt"": ""Explore the environmental impacts of hundreds of specific food products."", ""dateline"": ""August 17, 2022"", ""subtitle"": ""Explore the environmental impacts of hundreds of specific food products."", ""sidebar-toc"": false, ""featured-image"": ""Env-Impacts-of-Food-Data-Explorer.png""}, ""createdAt"": ""2022-08-15T16:00:06.000Z"", ""published"": false, ""updatedAt"": ""2022-08-17T10:38:21.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-08-17T10:00:00.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 0, ""numErrors"": 0, ""wpTagCounts"": {}, ""htmlTagCounts"": {}}",2022-08-17 10:00:00,2024-02-16 14:22:54,,"[""Hannah Ritchie""]",Explore the environmental impacts of hundreds of specific food products.,2022-08-15 16:00:06,2022-08-17 10:38:21,https://ourworldindata.org/wp-content/uploads/2022/08/Env-Impacts-of-Food-Data-Explorer.png,{},,"{""id"": 52489, ""date"": ""2022-08-17T11:00:00"", ""guid"": {""rendered"": ""https://owid.cloud/?p=52489""}, ""link"": ""https://owid.cloud/new-env-food-data-explorer"", ""meta"": {""owid_publication_context_meta_field"": {""latest"": true, ""homepage"": true, ""immediate_newsletter"": true}}, ""slug"": ""new-env-food-data-explorer"", ""tags"": [], ""type"": ""post"", ""title"": {""rendered"": ""We just published a new data explorer on the Environmental Impacts of Food""}, ""_links"": {""self"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/52489""}], ""about"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/types/post""}], ""author"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/users/17"", ""embeddable"": true}], ""curies"": [{""href"": ""https://api.w.org/{rel}"", ""name"": ""wp"", ""templated"": true}], ""replies"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/comments?post=52489"", ""embeddable"": true}], ""wp:term"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/categories?post=52489"", ""taxonomy"": ""category"", ""embeddable"": true}, {""href"": ""https://owid.cloud/wp-json/wp/v2/tags?post=52489"", ""taxonomy"": ""post_tag"", ""embeddable"": true}], ""collection"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts""}], ""wp:attachment"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media?parent=52489""}], ""version-history"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/posts/52489/revisions"", ""count"": 2}], ""wp:featuredmedia"": [{""href"": ""https://owid.cloud/wp-json/wp/v2/media/52488"", ""embeddable"": true}], ""predecessor-version"": [{""id"": 52493, ""href"": ""https://owid.cloud/wp-json/wp/v2/posts/52489/revisions/52493""}]}, ""author"": 17, ""format"": ""standard"", ""status"": ""publish"", ""sticky"": false, ""content"": {""rendered"": """", ""protected"": false}, ""excerpt"": {""rendered"": ""Explore the environmental impacts of hundreds of specific food products."", ""protected"": false}, ""date_gmt"": ""2022-08-17T10:00:00"", ""modified"": ""2022-08-17T11:38:21"", ""template"": """", ""categories"": [1], ""ping_status"": ""closed"", ""authors_name"": [""Hannah Ritchie""], ""modified_gmt"": ""2022-08-17T10:38:21"", ""comment_status"": ""closed"", ""featured_media"": 52488, ""featured_media_paths"": {""thumbnail"": ""/app/uploads/2022/08/Env-Impacts-of-Food-Data-Explorer-150x79.png"", ""medium_large"": ""/app/uploads/2022/08/Env-Impacts-of-Food-Data-Explorer-768x403.png""}}" 52472,Text for Plastic Pollution Explorer,text-for-plastic-pollution-explorer,wp_block,publish,"200 years ago, everyone lacked democratic rights. Now, billions of people have them.
But there are still large differences in the degree to which citizens enjoy political rights: most clearly between democracies and non-democracies, but also within these broad political regimes.
To understand the extent of people’s political rights, we shouldn’t only look at whether a country is classified as a democracy or not. We should also look at smaller differences in how democratic countries are.
How democratic have countries been across the world? And how big are the differences between them?
To answer these questions, we need information on countries’ political systems over recent centuries.
Measuring how democratic countries are comes with many challenges. People do not always agree on what characteristics define a democracy. Its characteristics — such as whether an election was free and fair — are difficult to assess. The assessments of experts are to some degree subjective and they may disagree; either about a specific characteristic, or how several characteristics can be reduced into a single measure of democracy.
Because of these difficulties, classifying political systems is unavoidably controversial. I have written more about how researchers deal with the challenges of measuring democracy in another article.
The source we show here is the electoral democracy index from the Varieties of Democracy (V-Dem) project{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. V-Dem [Country-Year/Country-Date] Dataset v13. Varieties of Democracy (V-Dem) Project.{/ref}. In our Democracy Data Explorer, however, we show the data for several leading approaches, so that you can compare how different sources score democracy across the world.
The electoral democracy index from V-Dem tries to capture the extent to which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed.{ref}While we use V-Dem’s data, we expand the years and countries covered. You can find more information in this article.{/ref}
The interactive map shows how democratic each country is at the end of each year, going back in time as far as 1789.{ref}You can download the complete dataset, including supplementary indicators, from GitHub.{/ref} To explore changes over time, you can drag the time-slider below the map.
We see that countries differ in how democratic they are, with some countries close to the index’s maximum of 1, and others close to its minimum score of 0. Most countries are somewhere in the middle.
The world did not always look like it does today. It has become much more democratic over time.
A very clear way of showing this is to look at the distribution of democracy scores at different stages in history.
Here we do this in the form of a bar chart, where electoral democracy is again measured on a scale from 0 to 1. The shortest bars here are the least democratic countries, the highest bars indicate the most democratic.
This means that the area covered by all of the bars gives us a proxy for the extent of democracy globally.
In the data’s earliest available year, 1789, the world was very undemocratic: most of the world’s political leaders were unelected, few people had voting rights, elections were neither free nor fair, and citizens were not able to assemble and speak freely.
Only a few countries, such as France, the United Kingdom, and the United States, had a few democratic characteristics. The United States was the most democratic country according to V-Dem’s assessment, but still only received a score of 0.35.
In this sense, the world was more equal than it is today: democratic rights were very limited everywhere.
Related charts:
By 1900, political institutions had become more diverse, but remained highly undemocratic in most countries.
A fair number of countries in Europe and the Americas now had some democratic features: citizens especially had become freer to associate and express their opinions. A couple of countries, such as Australia, France, and Switzerland, had even developed fairly democratic features, with men now having the right to vote and almost all political leaders being chosen in elections.
The most democratic country, with a score of 0.8, was New Zealand.
But the many other countries, most under colonial rule, had political systems that granted few democratic rights to their citizens: the colonial powers installed unelected leaders, gave no or only few citizens the right to vote, and restricted citizens’ ability to assemble and express their opinions.
In the first half of the 20th century, some countries continued to become more democratic, while progress in most others stalled.
More democratic political institutions in Europe after the First World War were almost completely undone in the two decades that followed, but were then reestablished after the second World War. Some non-European countries such as Canada and the United States also extended the democratic rights of their citizens.
In the rest of the world, however, countries broadened some political rights while remaining overwhelmingly undemocratic. The colonial powers at times expanded voting rights and loosened restrictions on freedoms of expression and association, but local political leaders remained unelected, or the elections choosing them were marred by violence, intimidation, or fraud.
This meant that democratic rights were distributed highly unequally across the world’s inhabitants, shown by the chart’s steep slope.
Many countries then became much more democratic in the second half of the 20th century.
In the 1990s especially, democratic institutions expanded across the world. Countries in South America shed their highly autocratic political systems that had spread in the 1970s. According to V-Dem, a considerable number of countries in Africa became fairly democratic (such as Ghana and South Africa) in the decades after gaining independence. Civil society organizations and political parties could operate more freely, and elections became freer and fairer.
And while many countries in Asia and the Middle East remained decidedly undemocratic, some countries in these regions expanded democratic rights, such as India, Indonesia, Turkey, and South Korea.
Democratic rights therefore became much more evenly distributed across the world.
The spread of democracy has slowed in the 21st century compared to previous decades. While some countries became more democratic according to V-Dem, such as Tunisia and Peru, many stagnated or became less democratic — some, such as Brazil, India, Indonesia, Mexico, and Poland considerably so.
Despite these declines in democracy, almost all countries remain much more democratic than they were at the beginning of the 20th century.
And they remain vastly more democratic than most countries during the 18th and 19th centuries: a score of 0.11 made Denmark one of the most democratic countries in 1789, while the same score made Mexico an average country in 1900, and Yemen one of the least democratic countries in the world in 2021.
While the world has become much more democratic over the last 200 years, there are still large differences between countries.
As the previous chart shows, some countries — mostly located in Europe and the Americas — are highly democratic: they have elected political leaders, elections are free and fair, and most citizens have the right to vote and can associate and express their opinions freely. The most democratic countries were Denmark and Sweden, with scores of 0.92 and 0.90.
Other countries, concentrated in Asia, are highly undemocratic according to V-Dem. This includes countries such as China, North Korea, the United Arab Emirates, and the least democratic country in the world, Saudi Arabia, with a score of just 0.02. In these countries, citizens do not have the right to choose their political leaders in popular elections.
Most countries, often situated in Africa and East and Southeast Asia, fall somewhere in the middle. In these countries, political leaders usually are elected and most citizens have the right to vote, but their rights to associate and express their opinions are limited, and elections are not entirely free and fair.
As mentioned, V-Dem is only one of the leading approaches to measure democracy. And its electoral democracy index is only one main measure it provides alongside other, more comprehensive indices of democracy.
Yet, using another approach or V-Dem index to measure democracy shows a similar development from a highly undemocratic world in the 18th and 19th century, to high democratic inequality in the earlier 20th century, and a much more democratic, and more equally democratic, world in recent decades.
You can see so for yourself by exploring the four charts below, which use the Polity project’s democracy index and V-Dem’s liberal democracy index.
Taken together, the democratic political systems of many countries show that a world where people have much more say in how they are governed is possible.
But the fact that so many countries are still highly undemocratic means that the fight for democratic political rights goes on.
","{""id"": ""wp-52306"", ""slug"": ""countries-around-the-world-have-become-much-more-democratic-but-there-are-large-differences-between-them"", ""content"": {""toc"": [], ""body"": [{""type"": ""text"", ""value"": [{""text"": ""200 years ago, everyone lacked democratic rights. Now, "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democratic-rights"", ""children"": [{""text"": ""billions of people have them"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But there are still large differences in the degree to which citizens enjoy political rights: most clearly between democracies and non-democracies, but also "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""within"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" these broad political regimes."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To understand the extent of people’s political rights, we shouldn’t only look at whether a country is classified as a democracy or not. We should also look at smaller differences in "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""how democratic"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" countries are."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""How democratic have countries been across the world? And how big are the differences between them?"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""To answer these questions, we need information on countries’ political systems over recent centuries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""text"": [{""text"": ""How can researchers measure how democratic a country is?"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Measuring how democratic countries are comes with many challenges. People do not always agree on what characteristics define a democracy. Its characteristics — such as whether an election was free and fair — are difficult to assess. The assessments of experts are to some degree subjective and they may disagree; either about a specific characteristic, or how several characteristics can be reduced into a single measure of democracy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Because of these difficulties, classifying political systems is unavoidably controversial. I have written more about how researchers deal with the challenges of measuring democracy "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/democracies-measurement"", ""children"": [{""text"": ""in another article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The source we show here is the electoral democracy index from the Varieties of Democracy (V-Dem) project{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. "", ""spanType"": ""span-simple-text""}, {""url"": ""https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Carl+Henrik+Knutsen%2C+Staffan+I.+Lindberg%2C+Jan+Teorell%2C+David+Altman%2C+Michael+Bernhard%2C+Agnes+Cornell%2C+M.+Steven+Fish%2C+Lisa+Gastaldi%2C+Haakon+Gjerl%C3%B8w%2C+Adam+Glynn%2C+Ana+Good+God%2C+Sandra+Grahn%2C+Allen+Hicken%2C+Katrin+Kinzelbach%2C+Joshua+Krusell%2C+Kyle+L.+Marquardt%2C+Kelly+McMann%2C+Valeriya+Mechkova%2C+Juraj+Medzihorsky%2C+Natalia+Natsika%2C+Anja+Neundorf%2C+Pamela+Paxton%2C+Daniel+Pemstein%2C+Josefine+Pernes%2C+Oskar+Ryd%C3%A9n%2C+Johannes+von+R%C3%B6mer%2C+Brigitte+Seim%2C+Rachel+Sigman%2C+Svend-Erik+Skaaning%2C+Jeffrey+Staton%2C+Aksel+Sundstr%C3%B6m%2C+Eitan+Tzelgov%2C+Yi-ting+Wang%2C+Tore+Wig%2C+Steven+Wilson+and+Daniel+Ziblatt.+2023.+V-Dem+%5BCountry-Year%2FCountry-Date%5D+Dataset+v13.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG="", ""children"": [{""text"": ""V-Dem [Country-Year/Country-Date] Dataset v13."", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" Varieties of Democracy (V-Dem) Project.{/ref}. In our "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/explorers/democracy"", ""children"": [{""text"": ""Democracy Data Explorer"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", however, we show the data for several leading approaches, so that you can compare how different sources score democracy across the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The electoral democracy index from V-Dem tries to capture the extent to which "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/elected-political-leaders"", ""children"": [{""text"": ""political leaders are elected"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" under "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/suffrage"", ""children"": [{""text"": ""comprehensive voting rights"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/free-and-fair-elections"", ""children"": [{""text"": ""free and fair elections"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "", and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/freedom-of-association"", ""children"": [{""text"": ""freedoms of association"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" and "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/grapher/freedom-of-expression"", ""children"": [{""text"": ""expression"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "" are guaranteed.{ref}While we use V-Dem’s data, we expand the years and countries covered. You can find more information in "", ""spanType"": ""span-simple-text""}, {""url"": ""https://ourworldindata.org/vdem-electoral-democracy-data"", ""children"": [{""text"": ""this article"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref}"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The interactive map shows how democratic each country is at the end of each year, going back in time as far as 1789.{ref}You can download the complete dataset, including supplementary indicators, from "", ""spanType"": ""span-simple-text""}, {""url"": ""https://github.com/owid/notebooks/tree/main/BastianHerre/democracy"", ""children"": [{""text"": ""GitHub"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-link""}, {""text"": "".{/ref} To explore changes over time, you can drag the time-slider below the map. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""We see that countries differ in how democratic they are, with some countries close to the index’s maximum of 1, and others close to its minimum score of 0. Most countries are somewhere in the middle."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/electoral-democracy"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""The world was highly undemocratic in the 18th and 19th centuries"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The world did not always look like it does today. It has become much more democratic over time."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A very clear way of showing this is to look at the distribution of democracy scores at different stages in history. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Here we do this in the form of a bar chart, where electoral democracy is again measured on a scale from 0 to 1. The shortest bars here are the least democratic countries, the highest bars indicate the most democratic. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This means that the area covered by "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""all"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": "" of the bars gives us a proxy for the extent of democracy globally."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the data’s earliest available year, 1789, the world was very undemocratic: most of the world’s political leaders were unelected, few people had voting rights, elections were neither free nor fair, and citizens were not able to assemble and speak freely."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Only a few countries, such as France, the United Kingdom, and the United States, had a few democratic characteristics. The United States was the most democratic country according to V-Dem’s assessment, but still only received a score of 0.35."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In this sense, the world was more equal than it is today: democratic rights were very limited "", ""spanType"": ""span-simple-text""}, {""children"": [{""text"": ""everywhere"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-italic""}, {""text"": ""."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-electoral-democracy-vdem?time=1789"", ""type"": ""chart"", ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""children"": [{""text"": ""Related charts"", ""spanType"": ""span-simple-text""}], ""spanType"": ""span-bold""}, {""text"": "":"", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/country-ranking-electoral-democracy-popw-vdem"", ""type"": ""prominent-link"", ""title"": ""Distribution of democracy weighted by population using V-Dem data"", ""description"": """", ""parseErrors"": []}, {""text"": [{""text"": ""The world was mostly undemocratic in the early 20th century"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""By 1900, political institutions had become more diverse, but remained highly undemocratic in most countries. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""A fair number of countries in Europe and the Americas now had some democratic features: citizens especially had become freer to associate and express their opinions. A couple of countries, such as Australia, France, and Switzerland, had even developed fairly democratic features, with men now having the right to vote and almost all political leaders being chosen in elections."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The most democratic country, with a score of 0.8, was New Zealand."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But the many other countries, most under colonial rule, had political systems that granted few democratic rights to their citizens: the colonial powers installed unelected leaders, gave no or only few citizens the right to vote, and restricted citizens’ ability to assemble and express their opinions."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-electoral-democracy-vdem?time=1900"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""Democratic rights became highly unequal across the world in the first half of the 20th century"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the first half of the 20th century, some countries continued to become more democratic, while progress in most others stalled. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""More democratic political institutions in Europe after the First World War were almost completely undone in the two decades that followed, but were then reestablished after the second World War. Some non-European countries such as Canada and the United States also extended the democratic rights of their citizens. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the rest of the world, however, countries broadened some political rights while remaining overwhelmingly undemocratic. The colonial powers at times expanded voting rights and loosened restrictions on freedoms of expression and association, but local political leaders remained unelected, or the elections choosing them were marred by violence, intimidation, or fraud."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""This meant that democratic rights were distributed highly unequally across the world’s inhabitants, shown by the chart’s steep slope."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-electoral-democracy-vdem?time=1950"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""Democracy spread across the world in the second half of the 20th century"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Many countries then became much more democratic in the second half of the 20th century."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""In the 1990s especially, democratic institutions expanded across the world. Countries in South America shed their highly autocratic political systems that had spread in the 1970s. According to V-Dem, a considerable number of countries in Africa became fairly democratic (such as Ghana and South Africa) in the decades after gaining independence. Civil society organizations and political parties could operate more freely, and elections became freer and fairer."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""And while many countries in Asia and the Middle East remained decidedly undemocratic, some countries in these regions expanded democratic rights, such as India, Indonesia, Turkey, and South Korea. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Democratic rights therefore became much more evenly distributed across the world."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-electoral-democracy-vdem?time=2000"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""Democracy’s spread has slowed in the 21st century"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""The spread of democracy has slowed in the 21st century compared to previous decades. While some countries became more democratic according to V-Dem, such as Tunisia and Peru, many stagnated or became less democratic — some, such as Brazil, India, Indonesia, Mexico, and Poland considerably so."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Despite these declines in democracy, almost all countries remain much more democratic than they were at the beginning of the 20th century. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""And they remain vastly more democratic than most countries during the 18th and 19th centuries: a score of 0.11 made Denmark one of the most democratic countries in 1789, while the same score made Mexico an average country in 1900, and Yemen one of the least democratic countries in the world in 2021."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-electoral-democracy-vdem"", ""type"": ""chart"", ""parseErrors"": []}, {""text"": [{""text"": ""Some countries are much more democratic than others"", ""spanType"": ""span-simple-text""}], ""type"": ""heading"", ""level"": 2, ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""While the world has become much more democratic over the last 200 years, there are still large differences between countries."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As the previous chart shows, some countries — mostly located in Europe and the Americas — are highly democratic: they have elected political leaders, elections are free and fair, and most citizens have the right to vote and can associate and express their opinions freely. The most democratic countries were Denmark and Sweden, with scores of 0.92 and 0.90."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Other countries, concentrated in Asia, are highly undemocratic according to V-Dem. This includes countries such as China, North Korea, the United Arab Emirates, and the least democratic country in the world, Saudi Arabia, with a score of just 0.02. In these countries, citizens do not have the right to choose their political leaders in popular elections."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Most countries, often situated in Africa and East and Southeast Asia, fall somewhere in the middle. In these countries, political leaders usually are elected and most citizens have the right to vote, but their rights to associate and express their opinions are limited, and elections are not entirely free and fair."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""As mentioned, V-Dem is only one of the leading approaches to measure democracy. And its electoral democracy index is only one main measure it provides alongside other, more comprehensive indices of democracy."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Yet, using another approach or V-Dem index to measure democracy shows a similar development from a highly undemocratic world in the 18th and 19th century, to high democratic inequality in the earlier 20th century, and a much more democratic, and more equally democratic, world in recent decades."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""You can see so for yourself by exploring the four charts below, which use the Polity project’s democracy index and V-Dem’s liberal democracy index."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""Taken together, the democratic political systems of many countries show that a world where people have much more say in how they are governed is possible. "", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""type"": ""text"", ""value"": [{""text"": ""But the fact that so many countries are still highly undemocratic means that the fight for democratic political rights goes on."", ""spanType"": ""span-simple-text""}], ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/democracy-polity"", ""type"": ""chart"", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-democracy-polity"", ""type"": ""chart"", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/liberal-democracy"", ""type"": ""chart"", ""parseErrors"": []}, {""url"": ""https://ourworldindata.org/grapher/distribution-liberal-democracy-vdem"", ""type"": ""chart"", ""parseErrors"": []}], ""type"": ""article"", ""title"": ""Countries around the world have become much more democratic, but there are large differences between them"", ""authors"": [null], ""dateline"": ""July 19, 2022"", ""sidebar-toc"": false, ""featured-image"": """"}, ""createdAt"": ""2022-10-05T19:25:44.000Z"", ""published"": false, ""updatedAt"": ""2023-07-20T13:09:40.000Z"", ""revisionId"": null, ""publishedAt"": ""2022-07-19T16:46:30.000Z"", ""relatedCharts"": [], ""publicationContext"": ""listed""}","{""errors"": [], ""numBlocks"": 61, ""numErrors"": 0, ""wpTagCounts"": {""html"": 10, ""heading"": 7, ""paragraph"": 44, ""owid/prominent-link"": 1}, ""htmlTagCounts"": {""p"": 44, ""h4"": 7, ""iframe"": 10}}",2022-07-19 16:46:30,2024-02-16 14:23:03,,[null],,2022-10-05 19:25:44,2023-07-20 13:09:40,,{},"200 years ago, everyone lacked democratic rights. Now, [billions of people have them](https://ourworldindata.org/democratic-rights). But there are still large differences in the degree to which citizens enjoy political rights: most clearly between democracies and non-democracies, but also _within_ these broad political regimes. To understand the extent of people’s political rights, we shouldn’t only look at whether a country is classified as a democracy or not. We should also look at smaller differences in _how democratic_ countries are. How democratic have countries been across the world? And how big are the differences between them? To answer these questions, we need information on countries’ political systems over recent centuries. ## How can researchers measure how democratic a country is? Measuring how democratic countries are comes with many challenges. People do not always agree on what characteristics define a democracy. Its characteristics — such as whether an election was free and fair — are difficult to assess. The assessments of experts are to some degree subjective and they may disagree; either about a specific characteristic, or how several characteristics can be reduced into a single measure of democracy. Because of these difficulties, classifying political systems is unavoidably controversial. I have written more about how researchers deal with the challenges of measuring democracy [in another article](https://ourworldindata.org/democracies-measurement). The source we show here is the electoral democracy index from the Varieties of Democracy (V-Dem) project{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. [V-Dem [Country-Year/Country-Date] Dataset v13.](https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=Coppedge%2C+Michael%2C+John+Gerring%2C+Carl+Henrik+Knutsen%2C+Staffan+I.+Lindberg%2C+Jan+Teorell%2C+David+Altman%2C+Michael+Bernhard%2C+Agnes+Cornell%2C+M.+Steven+Fish%2C+Lisa+Gastaldi%2C+Haakon+Gjerl%C3%B8w%2C+Adam+Glynn%2C+Ana+Good+God%2C+Sandra+Grahn%2C+Allen+Hicken%2C+Katrin+Kinzelbach%2C+Joshua+Krusell%2C+Kyle+L.+Marquardt%2C+Kelly+McMann%2C+Valeriya+Mechkova%2C+Juraj+Medzihorsky%2C+Natalia+Natsika%2C+Anja+Neundorf%2C+Pamela+Paxton%2C+Daniel+Pemstein%2C+Josefine+Pernes%2C+Oskar+Ryd%C3%A9n%2C+Johannes+von+R%C3%B6mer%2C+Brigitte+Seim%2C+Rachel+Sigman%2C+Svend-Erik+Skaaning%2C+Jeffrey+Staton%2C+Aksel+Sundstr%C3%B6m%2C+Eitan+Tzelgov%2C+Yi-ting+Wang%2C+Tore+Wig%2C+Steven+Wilson+and+Daniel+Ziblatt.+2023.+V-Dem+%5BCountry-Year%2FCountry-Date%5D+Dataset+v13.+Varieties+of+Democracy+%28V-Dem%29+Project.&btnG=) Varieties of Democracy (V-Dem) Project.{/ref}. In our [Democracy Data Explorer](https://ourworldindata.org/explorers/democracy), however, we show the data for several leading approaches, so that you can compare how different sources score democracy across the world. The electoral democracy index from V-Dem tries to capture the extent to which [political leaders are elected](https://ourworldindata.org/grapher/elected-political-leaders) under [comprehensive voting rights](https://ourworldindata.org/grapher/suffrage) in [free and fair elections](https://ourworldindata.org/grapher/free-and-fair-elections), and [freedoms of association](https://ourworldindata.org/grapher/freedom-of-association) and [expression](https://ourworldindata.org/grapher/freedom-of-expression) are guaranteed.{ref}While we use V-Dem’s data, we expand the years and countries covered. You can find more information in [this article](https://ourworldindata.org/vdem-electoral-democracy-data).{/ref} The interactive map shows how democratic each country is at the end of each year, going back in time as far as 1789.{ref}You can download the complete dataset, including supplementary indicators, from [GitHub](https://github.com/owid/notebooks/tree/main/BastianHerre/democracy).{/ref} To explore changes over time, you can drag the time-slider below the map. We see that countries differ in how democratic they are, with some countries close to the index’s maximum of 1, and others close to its minimum score of 0. Most countries are somewhere in the middle.200 years ago, everyone lacked democratic rights. Now, billions of people have them.
\n\n\n\nBut there are still large differences in the degree to which citizens enjoy political rights: most clearly between democracies and non-democracies, but also within these broad political regimes.
\n\n\n\nTo understand the extent of people’s political rights, we shouldn’t only look at whether a country is classified as a democracy or not. We should also look at smaller differences in how democratic countries are.
\n\n\n\nHow democratic have countries been across the world? And how big are the differences between them?
\n\n\n\nTo answer these questions, we need information on countries’ political systems over recent centuries.
\n\n\n\nMeasuring how democratic countries are comes with many challenges. People do not always agree on what characteristics define a democracy. Its characteristics — such as whether an election was free and fair — are difficult to assess. The assessments of experts are to some degree subjective and they may disagree; either about a specific characteristic, or how several characteristics can be reduced into a single measure of democracy.
\n\n\n\nBecause of these difficulties, classifying political systems is unavoidably controversial. I have written more about how researchers deal with the challenges of measuring democracy in another article.
\n\n\n\nThe source we show here is the electoral democracy index from the Varieties of Democracy (V-Dem) project{ref}Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2023. V-Dem [Country-Year/Country-Date] Dataset v13. Varieties of Democracy (V-Dem) Project.{/ref}. In our Democracy Data Explorer, however, we show the data for several leading approaches, so that you can compare how different sources score democracy across the world.
\n\n\n\nThe electoral democracy index from V-Dem tries to capture the extent to which political leaders are elected under comprehensive voting rights in free and fair elections, and freedoms of association and expression are guaranteed.{ref}While we use V-Dem’s data, we expand the years and countries covered. You can find more information in this article.{/ref}
\n\n\n\nThe interactive map shows how democratic each country is at the end of each year, going back in time as far as 1789.{ref}You can download the complete dataset, including supplementary indicators, from GitHub.{/ref} To explore changes over time, you can drag the time-slider below the map.
\n\n\n\nWe see that countries differ in how democratic they are, with some countries close to the index’s maximum of 1, and others close to its minimum score of 0. Most countries are somewhere in the middle.
\n\n\n\n\n\n\n\n\n\n\n\nThe world did not always look like it does today. It has become much more democratic over time.
\n\n\n\nA very clear way of showing this is to look at the distribution of democracy scores at different stages in history.
\n\n\n\nHere we do this in the form of a bar chart, where electoral democracy is again measured on a scale from 0 to 1. The shortest bars here are the least democratic countries, the highest bars indicate the most democratic.
\n\n\n\nThis means that the area covered by all of the bars gives us a proxy for the extent of democracy globally.
\n\n\n\nIn the data’s earliest available year, 1789, the world was very undemocratic: most of the world’s political leaders were unelected, few people had voting rights, elections were neither free nor fair, and citizens were not able to assemble and speak freely.
\n\n\n\nOnly a few countries, such as France, the United Kingdom, and the United States, had a few democratic characteristics. The United States was the most democratic country according to V-Dem’s assessment, but still only received a score of 0.35.
\n\n\n\nIn this sense, the world was more equal than it is today: democratic rights were very limited everywhere.
\n\n\n\n\n\n\n\nRelated charts:
\n\n\nBy 1900, political institutions had become more diverse, but remained highly undemocratic in most countries.
\n\n\n\nA fair number of countries in Europe and the Americas now had some democratic features: citizens especially had become freer to associate and express their opinions. A couple of countries, such as Australia, France, and Switzerland, had even developed fairly democratic features, with men now having the right to vote and almost all political leaders being chosen in elections.
\n\n\n\nThe most democratic country, with a score of 0.8, was New Zealand.
\n\n\n\nBut the many other countries, most under colonial rule, had political systems that granted few democratic rights to their citizens: the colonial powers installed unelected leaders, gave no or only few citizens the right to vote, and restricted citizens’ ability to assemble and express their opinions.
\n\n\n\n\n\n\n\nIn the first half of the 20th century, some countries continued to become more democratic, while progress in most others stalled.
\n\n\n\nMore democratic political institutions in Europe after the First World War were almost completely undone in the two decades that followed, but were then reestablished after the second World War. Some non-European countries such as Canada and the United States also extended the democratic rights of their citizens.
\n\n\n\nIn the rest of the world, however, countries broadened some political rights while remaining overwhelmingly undemocratic. The colonial powers at times expanded voting rights and loosened restrictions on freedoms of expression and association, but local political leaders remained unelected, or the elections choosing them were marred by violence, intimidation, or fraud.
\n\n\n\nThis meant that democratic rights were distributed highly unequally across the world’s inhabitants, shown by the chart’s steep slope.
\n\n\n\n\n\n\n\nMany countries then became much more democratic in the second half of the 20th century.
\n\n\n\nIn the 1990s especially, democratic institutions expanded across the world. Countries in South America shed their highly autocratic political systems that had spread in the 1970s. According to V-Dem, a considerable number of countries in Africa became fairly democratic (such as Ghana and South Africa) in the decades after gaining independence. Civil society organizations and political parties could operate more freely, and elections became freer and fairer.
\n\n\n\nAnd while many countries in Asia and the Middle East remained decidedly undemocratic, some countries in these regions expanded democratic rights, such as India, Indonesia, Turkey, and South Korea.
\n\n\n\nDemocratic rights therefore became much more evenly distributed across the world.
\n\n\n\n\n\n\n\nThe spread of democracy has slowed in the 21st century compared to previous decades. While some countries became more democratic according to V-Dem, such as Tunisia and Peru, many stagnated or became less democratic — some, such as Brazil, India, Indonesia, Mexico, and Poland considerably so.
\n\n\n\nDespite these declines in democracy, almost all countries remain much more democratic than they were at the beginning of the 20th century.
\n\n\n\nAnd they remain vastly more democratic than most countries during the 18th and 19th centuries: a score of 0.11 made Denmark one of the most democratic countries in 1789, while the same score made Mexico an average country in 1900, and Yemen one of the least democratic countries in the world in 2021.
\n\n\n\n\n\n\n\nWhile the world has become much more democratic over the last 200 years, there are still large differences between countries.
\n\n\n\nAs the previous chart shows, some countries — mostly located in Europe and the Americas — are highly democratic: they have elected political leaders, elections are free and fair, and most citizens have the right to vote and can associate and express their opinions freely. The most democratic countries were Denmark and Sweden, with scores of 0.92 and 0.90.
\n\n\n\nOther countries, concentrated in Asia, are highly undemocratic according to V-Dem. This includes countries such as China, North Korea, the United Arab Emirates, and the least democratic country in the world, Saudi Arabia, with a score of just 0.02. In these countries, citizens do not have the right to choose their political leaders in popular elections.
\n\n\n\nMost countries, often situated in Africa and East and Southeast Asia, fall somewhere in the middle. In these countries, political leaders usually are elected and most citizens have the right to vote, but their rights to associate and express their opinions are limited, and elections are not entirely free and fair.
\n\n\n\nAs mentioned, V-Dem is only one of the leading approaches to measure democracy. And its electoral democracy index is only one main measure it provides alongside other, more comprehensive indices of democracy.
\n\n\n\nYet, using another approach or V-Dem index to measure democracy shows a similar development from a highly undemocratic world in the 18th and 19th century, to high democratic inequality in the earlier 20th century, and a much more democratic, and more equally democratic, world in recent decades.
\n\n\n\nYou can see so for yourself by exploring the four charts below, which use the Polity project’s democracy index and V-Dem’s liberal democracy index.
\n\n\n\nTaken together, the democratic political systems of many countries show that a world where people have much more say in how they are governed is possible.
\n\n\n\nBut the fact that so many countries are still highly undemocratic means that the fight for democratic political rights goes on.
\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n""}}, ""extensions"": {""debug"": [{""type"": ""DEBUG_LOGS_INACTIVE"", ""message"": ""GraphQL Debug logging is not active. To see debug logs, GRAPHQL_DEBUG must be enabled.""}]}}"![\""\""](\""https://owid.cloud/app/uploads/2022/10/Flu-deaths-thumbnail-768x303.png\"")
![\""Life](\""https://owid.cloud/app/uploads/2020/03/Screenshot-2022-03-14-at-11.51.27-768x443.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2021/12/Population-cartogram_World-768x294.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2019/06/FEATURED-IMAGE-World-Population-Growth-768x547.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2021/11/Screen-Shot-2021-11-21-at-21.06.10-768x277.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2023/04/Peak-pop-featured-01-768x402.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2023/04/Future-population-region-featured-01-768x402.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2019/04/Population-Pyramid-1950-to-2100-768x533.jpg\"")
![\""\""](\""https://owid.cloud/app/uploads/2023/06/pop-momentum-thumbnail-768x402.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2023/06/demographic-transition-thumbnail-01-768x352.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2023/04/Population-sources-featured-01-768x402.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2023/06/population-projections-thumbnail-01-768x402.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2018/02/Famine-victims-and-world-population-since-1860-768x524.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2017/05/Demographic-Impact-of-China’s-Great-Leap-Forward-Famine-768x538.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2022/12/featured-image-Timeline-of-Transformative-Events-1-768x440.png\"")
![\""\""](\""https://owid.cloud/app/uploads/2022/12/featured-image-ai-training-computation-768x440.png\"")