1. World and country figures (WHO)
Country, regional and global rates come from the World Health Organization’s Global Health Observatory, which publishes estimates produced with the NCD Risk Factor Collaboration (NCD-RisC) for every year from 1980 to 2024. We use them exactly as published, rounded to one decimal.
- Age-standardized rates are used whenever we compare or rank countries, because they remove the effect of countries having older or younger populations.
- Crude rates are used when we count people, because they describe the real population.
- Each WHO estimate has a 95% uncertainty interval, shown as shaded bands and “likely range” notes.
- Children and teens are ages 5–19, using WHO’s BMI-for-age reference (overweight above +1 SD, obesity above +2 SD).
2. Numbers of people
To say “about X million adults”, we multiply WHO’s crude rate by the UN World Population Prospects 2024 population for that country and year. The UN publishes population in 5-year age groups, so adults aged 18+ are the 20+ groups plus two-fifths of the 15–19 group (assuming ages within that group are evenly spread). Children 5–19 are the 5–9, 10–14 and 15–19 groups.
Check: for 2022 this method gives 875 million adults and 154 million children and teens with obesity. WHO’s own published figures for 2022 are 890 million and over 160 million (WHO fact sheet). The small difference comes from WHO’s newer data release and the age-group approximation. Our automated tests stop the site from building if the gap grows beyond 3% (adults) or 5% (children).
3. United States: our own survey calculations
For the US we went back to the raw public survey files from the CDC’s National Center for Health Statistics — every survey from NHES I (1959–62) to NHANES (Aug 2021–Aug 2023) — and calculated obesity ourselves using the survey weights and design, with adults age-adjusted to the 2000 US population. The 2017–2018 cycle is left out of trend lines because it overlaps the 2017–March 2020 file. Children are classified with the CDC 2000 growth charts (95th percentile).
To make sure we did it right, we check our results against the figures the CDC itself has published:
| Measure | CDC published | Our calculation |
|---|---|---|
| US adults 20+, obesity, 2017–Mar 2020 (age-adjusted) (NCHS National Health Statistics Report 158) | 41.9% | 41.9% |
| US adults 20+, obesity, Aug 2021–Aug 2023 (age-adjusted) (NCHS Data Brief 508) | 40.3% | 40.3% |
| US children 2–19, obesity, 2017–Mar 2020 (NCHS National Health Statistics Report 158) | 19.7% | 19.7% |
| US adults 20–74, obesity, NHES I 1959–62 (NCHS Health E-Stat, adults 1960–2018) | 13.4% | 13.4% |
| US adults 20–74, obesity, NHANES III 1988–94 (NCHS Health E-Stat, adults 1960–2018) | 23.2% | 23.2% |
| US high-school students, obesity, 2013 (YRBS) (MMWR Surveillance Summaries 63(SS-4)) | 13.7% | 13.7% |
Two technical notes: NHES I weights included about 2 lb of exam clothing (per its codebook), which we subtract; and for 1999–2010 we round BMI to one decimal as NCHS does. Our confidence intervals use a standard method (Taylor series, t-distribution) and may differ slightly from NCHS’s published intervals.
4. US states and teens (self-reported)
State figures come from the CDC’s Behavioral Risk Factor Surveillance System, published on data.cdc.gov. Teen figures are our calculation from the national Youth Risk Behavior Survey files (1999–2013). Both rely on people reporting their own height and weight.
5. School surveys worldwide
We calculated overweight and obesity from the Global School-based Student Health Survey files, using WHO’s growth reference and the survey weights. We only show samples where at least 60% of students reported both height and weight (24 samples); 17 samples fell below that line and 8 had no height and weight answers. National and city or region samples are never combined.
6. Other data
- Calorie supply: FAO Food Balances via Our World in Data, 1961 onward only.
- GDP per person, urban population and under-5 overweight: World Bank World Development Indicators.
- Sugary-drink taxes: World Bank Global SSB Tax Database, August 2023 (no later updates exist).
- The UCI “eating habits” dataset is mostly synthetic; we use it only to show how to spot unreliable data, never as a real rate.
7. Writing the words
Sentences such as a country’s rank are generated automatically from the data, so they always match the charts. Quotations from health organisations are copied word for word, with a link and date.
Limitations
- WHO figures are modelled estimates; for countries with few surveys, the uncertainty ranges are wider.
- BMI is a screening measure and doesn’t measure body fat directly — see the limits of BMI.
- Self-reported data (US states, teens, school surveys) gives lower obesity figures than measured data in the US comparison we show.
- Patterns shown on the site (for example between wealth and obesity) do not by themselves show cause and effect.
Found a problem? Please tell us. Every source is listed on the data sources page.
Last updated