Feed of findings
Each card is one finding from the data, with numbers and a source. Swipe up for the next one; More turns the card over: plain words, how to read the chart and how far each forecast was. Mark what is interesting: the feed adapts to you, and everyone's marks show what people find interesting and where opinions split.
Without JavaScript the best findings are shown as a list. Swiping, voting and seeing what others find interesting need JavaScript.
ForecastsWorld
537GDP downturns across countries
28of them expected by the IMF a year ahead
2008–2025, April IMF forecasts
Nearly all foreseen downturns were crises already under way (Equatorial Guinea, Venezuela). Even in April of the same year the IMF saw only 57% of them.
More
Every April the IMF publishes a forecast of economic growth for all countries for this year and the years ahead. Here are all the years when a country’s economy actually shrank, and for each, what the IMF had expected a year earlier.
By year: blue, the number of countries whose GDP fell (IMF latest data); orange, how many of them the April IMF forecast made a year earlier already showed falling. The peaks are 2009 and 2020.
A downturn is a year when a country’s real GDP fell (growth below −0.05%, i.e. −0.1% or less once rounded) in the IMF’s latest data (April 2026). The forecast is the April World Economic Outlook of the year before; “expected a fall” means forecast growth below zero. Countries and years 2008–2025 with both a forecast and an outcome.
Misses the other way are rare too: false alarms (a fall expected, growth happened) numbered 35, against 2,651 years of growth. Forecasts almost never show a downturn in advance. These forecasts did not anticipate the pandemic downturn of 2020 (153 downturns): the 3 matches are countries already in decline. Without 2020: 25 of 384.
A question for this picture
Why do forecasts almost never show a downturn when downturns happen every year?
PopulationChina and United States
39.1China
38.0United States
the age half the people are older than, 2023
Half the people here are older than 39.1 years, against 38.0 in the US. In 1950: 22.2 against 29.0. It became older than the US in 2020; in 2019 they were level.
More
The median age splits a population in two: half the people are younger, half older.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates (World Population Prospects 2024). Populations age both because people live longer and because fewer children are born.
A question for this picture
Who will pay pensions in a country ageing faster than the rich ones?
FertilityChina and Japan
1.00China
1.20Japan
children per woman, 2023
Now 1.00 children per woman, against 1.20 in Japan. In 1950: 5.81 against 3.62. It fell below Japan in 2020.
More
The total fertility rate: how many children a woman would have on average at the age-specific birth rates of the year.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates and national statistics for recent years where published. The rate of a year also moves when births are put off to later ages. Where statistics print one decimal, the lead must be at least 0.05.
A question for this picture
What changed faster: incomes, cities, or the idea of how many children is “normal”?
FertilityTurkey and France
1.42Turkey
1.56France
children per woman, 2025
Now 1.42 children per woman, against 1.56 in France. In 1950: 6.47 against 2.97. It fell below France in 2020.
More
The total fertility rate: how many children a woman would have on average at the age-specific birth rates of the year.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates and national statistics for recent years where published. The rate of a year also moves when births are put off to later ages. Where statistics print one decimal, the lead must be at least 0.05.
A question for this picture
What changed faster: incomes, cities, or the idea of how many children is “normal”?
PopulationThailand and United States
39.7Thailand
38.0United States
the age half the people are older than, 2023
Half the people here are older than 39.7 years, against 38.0 in the US. In 1950: 17.5 against 29.0. It became older than the US in 2017.
More
The median age splits a population in two: half the people are younger, half older.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates (World Population Prospects 2024). Populations age both because people live longer and because fewer children are born.
A question for this picture
Who will pay pensions in a country ageing faster than the rich ones?
PopulationPoland and United Kingdom
41.3Poland
39.8United Kingdom
the age half the people are older than, 2023
Half the people here are older than 41.3 years, against 39.8 in the UK. In 1950: 24.8 against 33.9. It became older than the UK in 2017.
More
The median age splits a population in two: half the people are younger, half older.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates (World Population Prospects 2024). Populations age both because people live longer and because fewer children are born.
A question for this picture
Who will pay pensions in a country ageing faster than the rich ones?
FertilityMali and Japan
951kMali
747kJapan
births in 2023
Japan has 5.2 times as many people (124M against 23.8M), yet more babies were born here in 2023: 951k against 747k. It overtook Japan in births in 2019.
More
Live births in a calendar year, for two countries: one has far fewer people yet more babies.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates (World Population Prospects 2024) and national statistics where available; population on 1 July of the same year.
A question for this picture
How many of these children will look for work in 20 years, and where?
EconomySouth Korea and Japan
$55,700South Korea
$48,300Japan
GDP per person, PPP, 2025
In 1990 GDP per person was $14,400 against $36,100 in Japan, in 2021 dollars. It overtook Japan in 2014. Now $55,700 against $48,300.
More
GDP per person at purchasing-power parity: goods and services produced per resident, in prices brought to the same 2021 dollars.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Bank (ICP). GDP per person is not any household’s income; recent years may be revised. Purchasing-power parities are uncertain by several percent, so the lead must be at least 5%.
A question for this picture
Did people feel this overtaking in their pockets?
EnergyUnited States
19%wind and solar
16%coal
share of electricity generation, 2025
In 2000 coal made 52% of the electricity, wind and sun 0.2%. Wind and sun overtook coal in 2024.
More
Shares of a year’s electricity generation: how much came from wind and solar plants and how much from coal. The rest is gas, nuclear, hydro and others.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
This is a share of electricity, not of all energy: transport, heating and industrial fuels are not included. Imported electricity is not counted, so countries generating less than 5 TWh a year are not shown. Ember yearly data; the latest year may be revised.
A question for this picture
What left with coal, and what replaced it on still nights?
FertilityAfghanistan and Russia
1.47MAfghanistan
1.26MRussia
births in 2023
Russia has 3.5 times as many people (145M against 41.5M), yet more babies were born here in 2023: 1.47M against 1.26M. It overtook Russia in births in 2021.
More
Live births in a calendar year, for two countries: one has far fewer people yet more babies.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
- Afghanistan: Births thousand births per year
384 (1950) → 1,469 (2023)
UN DESA, World Population Prospects 2024
- Russia: Births thousand births per year
2,885 (1950) → 1,264 (2023)
UN DESA, World Population Prospects 2024 → Rosstat: Rozhdaemost, smertnost i estestvenny prirost (demo21_2023.xlsx); TFR: Russian Statistical Yearbook 2025, table 4.14
- Afghanistan: Population thousand persons, 1 July
41,455 (2023)
UN DESA, World Population Prospects 2024
- Russia: Population thousand persons, 1 July
145,440 (2023)
UN DESA, World Population Prospects 2024
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates (World Population Prospects 2024) and national statistics where available; population on 1 July of the same year.
A question for this picture
How many of these children will look for work in 20 years, and where?
FertilityDR Congo and United States
4.37MDR Congo
3.6MUnited States
births in 2023
The US has 3.2 times as many people (343M against 106M), yet more babies were born here in 2023: 4.37M against 3.6M. It overtook the US in births in 2018.
More
Live births in a calendar year, for two countries: one has far fewer people yet more babies.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
- DR Congo: Births thousand births per year
564 (1950) → 4,370 (2023)
UN DESA, World Population Prospects 2024
- United States: Births thousand births per year
3,698 (1950) → 3,596 (2023)
UN DESA, World Population Prospects 2024 → NCHS, Births: Final Data for 2024 (NVSR vol 75 no 2, 9 Jun 2026), Tables 1-2
- DR Congo: Population thousand persons, 1 July
105,790 (2023)
UN DESA, World Population Prospects 2024
- United States: Population thousand persons, 1 July
343,477 (2023)
UN DESA, World Population Prospects 2024
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates (World Population Prospects 2024) and national statistics where available; population on 1 July of the same year.
A question for this picture
How many of these children will look for work in 20 years, and where?
PopulationDR Congo and Germany
106MDR Congo
84.5MGermany
people, 2023
In 1950 it had 12.3M people against 69.8M in Germany. It overtook Germany in 2016.
More
Population on 1 July, UN estimate, for two countries measured the same way.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Population Prospects 2024, estimates to 2023. Population changes with births, deaths and migration; the card does not separate them.
A question for this picture
What made the overtaking: many children in one country or few in the other?
FertilityThailand and Germany
1.21Thailand
1.39Germany
children per woman, 2023
Now 1.21 children per woman, against 1.39 in Germany. In 1950: 6.36 against 2.19. It fell below Germany in 2015.
More
The total fertility rate: how many children a woman would have on average at the age-specific birth rates of the year.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
UN estimates and national statistics for recent years where published. The rate of a year also moves when births are put off to later ages. Where statistics print one decimal, the lead must be at least 0.05.
A question for this picture
What changed faster: incomes, cities, or the idea of how many children is “normal”?
PopulationMexico and Japan
130MMexico
124MJapan
people, 2023
In 1950 it had 27.6M people against 86.4M in Japan. It overtook Japan in 2020.
More
Population on 1 July, UN estimate, for two countries measured the same way.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Population Prospects 2024, estimates to 2023. Population changes with births, deaths and migration; the card does not separate them.
A question for this picture
What made the overtaking: many children in one country or few in the other?
PopulationTanzania and Italy
66.6MTanzania
59.5MItaly
people, 2023
In 1950 it had 7.63M people against 46.4M in Italy. It overtook Italy in 2020.
More
Population on 1 July, UN estimate, for two countries measured the same way.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Population Prospects 2024, estimates to 2023. Population changes with births, deaths and migration; the card does not separate them.
A question for this picture
What made the overtaking: many children in one country or few in the other?
PopulationJapan
12.8Mpeople aged 80+
4.14Mchildren under 5
on 1 July 2023
In 1950 there were 31 children under 5 for every person aged 80+. The over-80s outnumbered the under-5s in 2004.
More
Two head counts of the same country: people who have reached 80 and children not yet five, both UN estimates for 1 July.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
Five-year age groups from World Population Prospects 2024, added up unchanged. The number of people over 80 grows with longevity and with the size of the generations born 80 years earlier.
A question for this picture
How many nursery teachers and how many carers will this country need in ten years?
EnergyGermany
45%wind and solar
21%coal
share of electricity generation, 2025
In 1995 coal made 54% of the electricity, wind and sun 0.3%. Wind and sun have been ahead since 2020, except in 2021.
More
Shares of a year’s electricity generation: how much came from wind and solar plants and how much from coal. The rest is gas, nuclear, hydro and others.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
This is a share of electricity, not of all energy: transport, heating and industrial fuels are not included. Imported electricity is not counted, so countries generating less than 5 TWh a year are not shown. Ember yearly data; the latest year may be revised.
A question for this picture
What left with coal, and what replaced it on still nights?
HealthMorocco and Russia
75.3Morocco
73.3Russia
life expectancy, 2023
Life expectancy is 75.3 years, against 73.3 in Russia, though GDP per person is $9,500 against $42,400. It overtook Russia first in 1995, and without a break since 1999.
More
Life expectancy at birth: how long a newborn would live at the mortality of the year. For contrast, GDP per person at purchasing-power parity.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Bank, from UN and national estimates; for countries without complete death registration this is the UN model and the country’s own statistics may differ, so the lead must be at least two years. GDP per person in 2021 PPP dollars. Income is not the only thing life expectancy depends on.
A question for this picture
What buys years of life more cheaply than money: diet, habits, doctors or safety?
EconomyPoland and Russia
$46,900Poland
$42,400Russia
GDP per person, PPP, 2025
In 1990 GDP per person was $12,800 against $29,600 in Russia, in 2021 dollars. It overtook Russia in 2019. Now $46,900 against $42,400.
More
GDP per person at purchasing-power parity: goods and services produced per resident, in prices brought to the same 2021 dollars.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Bank (ICP). GDP per person is not any household’s income; recent years may be revised. Purchasing-power parities are uncertain by several percent, so the lead must be at least 5%.
A question for this picture
Did people feel this overtaking in their pockets?
HealthUnited States
41%adults with obesity
16%use tobacco
share of adults, 2024
41% of adults live with obesity; 16% use tobacco. In 2000 it was the reverse: 39% used tobacco and 28% lived with obesity. The lines crossed between 2000 and 2005. Tobacco is estimated every few years.
More
Two shares of adults: those living with obesity (body-mass index of 30 or more, aged 18+, an FAO estimate) and those using any tobacco (aged 15+, a WHO estimate standardised to a standard age structure). The methods differ, so only countries where obesity leads at least one and a half times are shown.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
Obesity is an FAO estimate (FAOSTAT), tobacco a WHO estimate via the World Bank. Both are modelled curves, not a survey every year. Tobacco is not estimated every year, so the crossing falls between two estimates.
A question for this picture
Which of the two costs health care more?
HealthBrazil
30%adults with obesity
12%use tobacco
share of adults, 2024
30% of adults live with obesity; 12% use tobacco. In 2000 it was the reverse: 23% used tobacco and 12% lived with obesity. The lines crossed between 2007 and 2010. Tobacco is estimated every few years.
More
Two shares of adults: those living with obesity (body-mass index of 30 or more, aged 18+, an FAO estimate) and those using any tobacco (aged 15+, a WHO estimate standardised to a standard age structure). The methods differ, so only countries where obesity leads at least one and a half times are shown.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
Obesity is an FAO estimate (FAOSTAT), tobacco a WHO estimate via the World Bank. Both are modelled curves, not a survey every year. Tobacco is not estimated every year, so the crossing falls between two estimates.
A question for this picture
Which of the two costs health care more?
InternetThailand
91%use the internet
27%toilet waste safely treated
share of people, 2024
91% are online. By the strict WHO/UNICEF bar, 27% have a toilet whose waste is safely treated or disposed of. The rest are not necessarily without a toilet: toilets whose waste goes untreated do not count.
More
The internet: people online in the last three months on any device (ITU), the lowest bar. Safely managed sanitation is the strict WHO/UNICEF bar: a toilet not shared with other households whose waste is safely disposed of on site or removed and treated; safely managed water: a clean source at home, available when needed.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
The bars differ: going online once is enough for the internet, while sanitation and water are counted at the top rung. Lacking such a service does not mean having no toilet or no water at all. Both are shares of all people.
A question for this picture
Why was it easier to bring the internet than sewage treatment?
MigrationIndia
18.5Mborn here, living abroad
4.8Mborn abroad, living here
everyone alive, 2024
Residents, not a year’s flow. In 1990, 6.58M natives lived abroad and 7.21M foreign-born people lived in the country. Both numbers include people born in another part of the former common country. The balance flipped between 1990 and 1995.
More
Two head counts: natives of the country living elsewhere, and people born elsewhere living in it. These are everyone alive now, built up over decades, not a year’s flow and not arrivals minus departures.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
Early in the series the foreign-born include many people born before the partition of British India in what are now neighbouring countries. The UN counts by country of birth, or by citizenship where birthplace is not recorded. Refugees are included. Mid-year estimates every five years and for 2024. Countries where the balance flipped only because the other side shrank (people who had moved within a former common country left or died, refugees went home) are not shown.
A question for this picture
What changed first: jobs at home, jobs abroad, or borders?
ForecastsWorld
43countries expected above 5%
149countries where it came in above 5%
inflation in 2022, IMF forecast of April 2021
Prices outran the April 2021 forecast in 178 of 189 countries. It goes the other way too: in 2015 inflation came in below the forecast in 140 of 186.
More
Inflation is how much consumer prices rose over a year on average. Every April the IMF forecasts the next year for all countries; here is how many countries it expected to see high inflation and how many did.
By year: blue, the number of countries where the April IMF forecast a year earlier expected inflation above 5%; orange, the number where it turned out above 5% (latest data).
Inflation: the annual average rise in consumer prices, World Economic Outlook. Forecast: the April edition of the year before; outcome: April 2026. Countries with both; years with at least 100 such countries.
The 5% threshold marks “noticeable” inflation; another threshold changes the numbers but not the picture. An error shared by many countries at once points to an event outside the assumptions (the pandemic, energy and food prices), not to each country separately.
A question for this picture
Why do forecasts err the same way in almost every country at once?
PopulationWorld
The peak was 146M births in 2012; in 2023, 132M, although there are 12% more people on Earth. There are also fewer children under five than at the 2017 peak: 654M against 698M.
More
Births: how many babies were born on Earth in a year, as estimated by the UN. Women have had fewer children for decades, but while the number of young women grew, so did births. Since about 2012 the decline in fertility has outweighed it.
The line is births in the world by year; the dot is the peak.
- Births thousand births per year
146,055 (2012) → 132,110 (2023)
UN DESA, World Population Prospects 2024
- Population thousand persons, 1 July
7,201,202 (2012) → 8,091,735 (2023)
UN DESA, World Population Prospects 2024
- Children under 5 thousand persons, 1 July
697,928 (2017) → 654,028 (2023)
UN DESA, World Population Prospects 2024
UN estimates (World Population Prospects 2024) for 237 countries and territories, the last estimated year being 2023. The peak is the largest value of the series; it counts if at least five years have passed since.
Many countries do not register every birth, and the UN fills the gaps from censuses and surveys, so these are estimates. Children under five: the population on 1 July.
A question for this picture
When will deaths first outnumber births in the world?
InternetIndonesia
73%use the internet
30%clean water at home when needed
share of people, 2024
73% are online; 30% have clean water from their own source at home whenever needed. The rest are not necessarily without water: sources with interruptions or untested for contamination do not count.
More
The internet: people online in the last three months on any device (ITU), the lowest bar. Safely managed sanitation is the strict WHO/UNICEF bar: a toilet not shared with other households whose waste is safely disposed of on site or removed and treated; safely managed water: a clean source at home, available when needed.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
The bars differ: going online once is enough for the internet, while sanitation and water are counted at the top rung. Lacking such a service does not mean having no toilet or no water at all. Both are shares of all people.
A question for this picture
Why was it easier to bring the internet than sewage treatment?
ForecastsChina
The UN's 1998 projection expected 17.6M births a year on average for 2020–2025; the outcome was 9.67M (−45%). The latest checked one: the UN's 2022 projection expected 10.4M births for 2025; the outcome was 7.92M (−24%).
PopulationGermany
1.03Mdeaths
719kbirths
UN estimate, 2023
By UN estimates, 1.03M people died and 719k were born in 2023. Deaths have clearly outnumbered births since 1972, except in 1988–1990, when they were almost level. In 1950 there were 1.7 times as many births as deaths.
More
Births and deaths are the people born and the people who died in a calendar year, by UN estimates. With more deaths, a population shrinks on its own; it can grow only through people moving in.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Both lines are UN estimates for 1950–2023. «Clearly more» means deaths above births by at least 3%; «almost level», a difference under 3% either way. In the latest year deaths must exceed births by at least 5%. The years in the headline count only consecutive years with a clear margin, rounded down to five. The ring marks where the lines crossed before that.
Both lines are UN estimates (World Population Prospects 2024), made the same way for every country. National statistics can put the turning point in another year, especially where births and deaths were almost level. Deaths also rise because large generations grow old; it need not mean higher mortality. Population change includes migration and revisions.
A question for this picture
What tipped the balance more: fewer children or more old people?
PopulationRussia
1.79Mdeaths
1.3Mbirths
UN estimate, 2023
By UN estimates, 1.79M people died and 1.3M were born in 2023. Deaths have clearly outnumbered births since 2017 (in 2016 they were almost level). In 1950 there were 2.1 times as many births as deaths.
More
Births and deaths are the people born and the people who died in a calendar year, by UN estimates. With more deaths, a population shrinks on its own; it can grow only through people moving in.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
- Deaths thousand deaths per year
1,363 (1950) → 1,795 (2023)
UN DESA, World Population Prospects 2024
- Births thousand births per year
2,885 (1950) → 1,297 (2023)
UN DESA, World Population Prospects 2024
- Population thousand persons, 1 July
145,779 (2016) → 145,440 (2023)
UN DESA, World Population Prospects 2024
Both lines are UN estimates for 1950–2023. «Clearly more» means deaths above births by at least 3%; «almost level», a difference under 3% either way. In the latest year deaths must exceed births by at least 5%. The years in the headline count only consecutive years with a clear margin, rounded down to five. The ring marks where the lines crossed before that.
Both lines are UN estimates (World Population Prospects 2024), made the same way for every country. National statistics can put the turning point in another year, especially where births and deaths were almost level. Deaths also rise because large generations grow old; it need not mean higher mortality. Population change includes migration and revisions.
A question for this picture
What tipped the balance more: fewer children or more old people?
HealthLebanon and Russia
77.8Lebanon
73.3Russia
life expectancy, 2023
Life expectancy is 77.8 years, against 73.3 in Russia, though GDP per person is $10,500 against $42,400. It overtook Russia in 1991.
More
Life expectancy at birth: how long a newborn would live at the mortality of the year. For contrast, GDP per person at purchasing-power parity.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Bank, from UN and national estimates; for countries without complete death registration this is the UN model and the country’s own statistics may differ, so the lead must be at least two years. GDP per person in 2021 PPP dollars. Income is not the only thing life expectancy depends on.
A question for this picture
What buys years of life more cheaply than money: diet, habits, doctors or safety?
HealthJordan and Russia
77.8Jordan
73.3Russia
life expectancy, 2023
Life expectancy is 77.8 years, against 73.3 in Russia, though GDP per person is $10,800 against $42,400. People have lived longer than in Russia since 1983, except in 1986–1987.
More
Life expectancy at birth: how long a newborn would live at the mortality of the year. For contrast, GDP per person at purchasing-power parity.
Blue is the country of the headline, orange the one it is compared with; one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both share.
Well-known countries against every country of sufficient size. An overtaking counts if the country was well behind at the start, ahead in no more than a fifth of the years before it, and it lasts to the latest shared year (at least two or three years in a row). A short dip of up to two years does not end a lead and is named. At most three overtakings per well-known country, and each country in at most one card of a kind.
World Bank, from UN and national estimates; for countries without complete death registration this is the UN model and the country’s own statistics may differ, so the lead must be at least two years. GDP per person in 2021 PPP dollars. Income is not the only thing life expectancy depends on.
A question for this picture
What buys years of life more cheaply than money: diet, habits, doctors or safety?
PopulationWorld
35%Africa
4.7%Europe
share of the world’s births, 2023
In 2023, 35% of all births on Earth: 46M out of 131M. In 1950, 12%, less than in Europe (13%). Now Europe accounts for 4.7% of the world’s babies.
More
Of all babies born on Earth in a year, how many were born in African countries and how many in European ones (Russia included). In Africa fertility falls more slowly and the number of young women keeps growing; in Europe the opposite.
Blue is Africa’s share of the world’s births, orange Europe’s; one scale. The ring marks the year Africa overtook Europe.
The sum of UN estimates for the region’s countries divided by the sum over all countries and territories, 1950–2023. Regions follow the UN classification: 58 countries and territories in Africa, 50 in Europe.
Many African countries do not register every birth; the UN estimates them from censuses and surveys.
A question for this picture
How will schools, labour markets and migration change when half of the world’s babies are born in Africa?
FertilitySouth Korea
After the low of 2023 (230k): 238k in 2024, 254k in 2025, 10.6% more. Children per woman: 0.72 → 0.80. Marriages are up 24%. It is still far below the 1960 peak (1.1M).
More
Births per year after they reached their lowest level since 1950. A rise can come from women having more children and from there being more women around the age of 30.
The line is births per year; the dot is the low.
- Births thousand births per year
230 (2023) → 254 (2025)
National Data Office (국가데이터처, formerly Statistics Korea), '2025 Birth Statistics' (final), released 2026-08-26, Table 1 and appendix table 1
- Total fertility rate children per woman
0.72 (2023) → 0.80 (2025)
National Data Office (국가데이터처, formerly Statistics Korea), '2025 Birth Statistics' (final), released 2026-08-26, Table 1 and appendix table 1
- Marriages thousand marriages a year
194 (2023) → 240 (2025)
national: Statistics Korea / Ministry of Data and Statistics, 2025 marriage and divorce statistics (table 1970-2025); by date of report (registration); rate per 1,000 mid-year population
National statistics where available, otherwise UN estimates. Selection: countries of at least 3 million with data for 2025 where births rose every year after their lowest point since 1950, for at least two years and by 4% or more in total.
Two or three years of growth are not yet a new trend: such rises happened before, and fertility then fell again.
A question for this picture
Is this postponed births coming back or a real turn?
PopulationItaly
355k births. In 1862, where the series begins: 833k. Even in the worst year of World War I, 1918, 640k were born, 1.8 times more.
More
The number of babies born in a year, from the longest series for this country, since 1862. In 2025 there were fewer than in any earlier year, wars and epidemics included, although the population is now larger.
The line is births per year; the dot, the record low; the band, the years of the war with the fewest births.
Births: Human Mortality Database; the latest years after its series ends, national statistics. Population then: births divided by births per 1,000 people of the same series. Selection: series of at least 100 years where a year since 2020 is below every earlier year.
The country’s borders may have changed over 150 years, and 19th-century birth registration was less complete, so births then were, if anything, more than recorded.
A question for this picture
What would have to change for births to rise again: incomes, housing or the wish to have children?
InternetMexico
83%use the internet
43%clean water at home when needed
share of people, 2024
83% are online; 43% have clean water from their own source at home whenever needed. The rest are not necessarily without water: sources with interruptions or untested for contamination do not count.
More
The internet: people online in the last three months on any device (ITU), the lowest bar. Safely managed sanitation is the strict WHO/UNICEF bar: a toilet not shared with other households whose waste is safely disposed of on site or removed and treated; safely managed water: a clean source at home, available when needed.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
The bars differ: going online once is enough for the internet, while sanitation and water are counted at the top rung. Lacking such a service does not mean having no toilet or no water at all. Both are shares of all people.
A question for this picture
Why was it easier to bring the internet than sewage treatment?
Markets and valuationsBrazil
December 2022 → September 2026, share price index in local currency, without inflation adjustment or dividends.
More
How the market changed since late 2022.
December 2022 = 100. Monthly values, normally means of daily closes.
Change = (latest value / December 2022 - 1) × 100. Methodology breaks are not spliced.
The annual axis selects the last published month of each year, labelled with its actual date. The current year is incomplete; this is neither a full-year mean nor a live quote. OECD national indices, normally monthly means of daily closes, 2015 = 100. Index composition differs across countries; dividends are not added. Index changes are not total-market capitalisation growth. Monthly averages smooth intramonth falls.
A question for this picture
What changes after inflation and currency movements?
Markets and valuationsChina
December 2022 → September 2026, share price index in local currency, without inflation adjustment or dividends.
More
How the market changed since late 2022.
December 2022 = 100. Monthly values, normally means of daily closes.
Change = (latest value / December 2022 - 1) × 100. Methodology breaks are not spliced.
The annual axis selects the last published month of each year, labelled with its actual date. The current year is incomplete; this is neither a full-year mean nor a live quote. OECD national indices, normally monthly means of daily closes, 2015 = 100. Index composition differs across countries; dividends are not added. Index changes are not total-market capitalisation growth. Monthly averages smooth intramonth falls.
A question for this picture
What changes after inflation and currency movements?
Markets and valuationsGermany
December 2022 → September 2026, share price index in local currency, without inflation adjustment or dividends.
More
How the market changed since late 2022.
December 2022 = 100. Monthly values, normally means of daily closes.
Change = (latest value / December 2022 - 1) × 100. Methodology breaks are not spliced.
The annual axis selects the last published month of each year, labelled with its actual date. The current year is incomplete; this is neither a full-year mean nor a live quote. OECD national indices, normally monthly means of daily closes, 2015 = 100. Index composition differs across countries; dividends are not added. Index changes are not total-market capitalisation growth. Monthly averages smooth intramonth falls.
A question for this picture
What changes after inflation and currency movements?
HealthPhilippines
9.2%adults with obesity
3%people undernourished
shares of adults and of all people, 2023
9.2% of adults live with obesity; 3% of people are undernourished. In 2001 it was the reverse: 19% undernourished, 3.5% with obesity. The lines crossed in 2018.
More
Undernourishment is FAO’s modelled share of all people short of dietary energy. Obesity is the share of adults 18+ with a body-mass index of 30 or more. The denominators differ; both are FAO estimates.
Blue is the card’s first number, orange the second; they share one scale. The ring marks where the lines crossed; the gap after it is shaded. The large numbers are the latest year both lines share.
Only years where both series are published; gaps are not filled. An overtaking counts if the new leader was behind at the start and ahead in no more than a fifth of the years before it, and it leads in the latest shared year and in at least two shared years in a row. A short dip of up to two years does not end a lead and is named; if the leader was ahead once before, that year is named too. Size thresholds pick a story and do not show that one causes the other.
Undernourishment published as «<2.5%» is a bound, not a value; such years are not counted as an overtaking. FAO year labels can stand for three-year windows.
A question for this picture
Can hunger and obesity live in the same family?
PopulationJapan
709k births. In 1873, where the series begins: 842k. Even in the worst year of Russo-Japanese War, 1904, 1.5M were born, 2.1 times more.
More
The number of babies born in a year, from the longest series for this country, since 1873. In 2024 there were fewer than in any earlier year, wars and epidemics included, although the population is now larger.
The line is births per year; the dot, the record low; the band, the years of the war with the fewest births.
Births: Human Mortality Database; the latest years after its series ends, national statistics. Population then: births divided by births per 1,000 people of the same series. Selection: series of at least 100 years where a year since 2020 is below every earlier year.
The country’s borders may have changed over 150 years, and 19th-century birth registration was less complete, so births then were, if anything, more than recorded.
A question for this picture
What would have to change for births to rise again: incomes, housing or the wish to have children?