01
The children get much taller. The gap between countries does not.
Each column is one age. It runs from the shortest country’s average boy up to the tallest country’s average boy.
NCD-RisC, Lancet 2020. Year of measurement, ages 5-19. Boys, 2019, 160 countries. At five the tallest is French Polynesia at 125.66 cm and the shortest Rwanda at 96.06 cm. At nineteen it is Netherlands at 183.78 cm and Timor-Leste at 160.13 cm. Counted in centimetres the gap closes for boys and opens for girls, so the claim here is the proportional one and the axis is the physical one.
02
Where a country ranks at five does not fix where it ranks at nineteen.
Each line is one country, ranked against the others at every age, tallest at the top. Most hold their place. A few cross most of the table.
NCD-RisC, Lancet 2020. Year of measurement, ages 5-19. Boys, 2019, 160 countries. Rank within the 160 countries drawn. Rank at five and rank at nineteen still move together across the whole set, so this says how far one country can travel, not that five tells you nothing.
03
French Polynesia is first at five and 29th at nineteen.
Each row is one age. Each dot is one country’s average boy, shortest on the left. French Polynesia is the red one.
NCD-RisC, Lancet 2020. Year of measurement, ages 5-19. Boys, 2019, 160 countries. Rank within the 160 countries drawn. Against all 200 in the file it is 38th at nineteen rather than 29th, which is the same fact counted against a longer list. It is a territory of a few hundred thousand people, so it is named here beside its own published range of 6.34 cm.
04
Around eleven, girls are taller than boys in almost every country.
Each line is one country: its girls’ average height minus its boys’, at each age. Above the rule the girls are taller. The lead opens almost everywhere and closes everywhere.
NCD-RisC, Lancet 2020. Year of measurement, ages 5-19. 2019, 160 countries. Rwanda: boys 129.21 cm [124.88 to 133.41], girls 139.22 cm [136.62 to 141.86]. Those two published ranges do not overlap, which is why that one is named.
05
The girls’ lead peaks at eleven in almost every country and every year.
Every country in every year of the record, stacked at the age where its own girls’ lead is biggest and shaded by the year it was measured. Twelve of the fifteen ages have never held a peak.
NCD-RisC, Lancet 2020. Year of measurement, ages 5-19. 160 countries across 35 years is 5,600 country-years, of which 395 have no window at all and 5,205 are placed here. The average peak age never leaves 11.061 to 11.143 across the whole span, which is why the 6 Bangladesh years are the only exceptions worth naming.
06
The five-year-olds gained more than the nineteen-year-olds.
Each bar is one age and one sex: centimetres the average country gained between 1985 and 2019. The girls’ bars get shorter at every step down the page.
NCD-RisC, Lancet 2020. Year of measurement, ages 5-19. 160 countries, 1985 and 2019, averaged over countries rather than over people. The girls fall at every one of the fifteen ages. The boys fall from five to nineteen but not at every step, so only the two ends of that row are stated. 12 countries have shorter five-year-old boys than in 1985 and 23 have shorter girls; at nineteen it is 21 and 14.
07
Every country is taller than in 1896, by anything from 0.52 to 20.16 cm.
The page changes source here, and with it the axis. Everything above is children, at each age, by the year they were measured. Everything below is adults at eighteen, by the year they were born. Each stroke is one country and one sex, from its 1896 birth year up to its 1996 one.
NCD-RisC, eLife 2016. Year of birth, height at 18. 200 countries, both sexes. The two sexes are two runs on one centimetre scale, each sorted by its own 1896 height. Smallest gain Marshall Islands men, 0.52 cm. Rank order did not hold: the correlation between 1896 and 1996 is 0.717 for the men and 0.639 for the women, Iran climbing 85 places and South Africa falling 83. The men end further apart than they started and the women do not, so this page says the century did not even the men out and does not say the world diverged.
08
Nearly half the world has already had its tallest generation.
Each dot is one country and one sex whose tallest generation has already been born: the year that generation was born, against how much shorter the latest one is.
NCD-RisC, eLife 2016. Year of birth, height at 18. 200 countries, both sexes. A group counts as past its peak when its tallest birth year came before 1996 and it has lost more than 0.5 cm since. The other 218 groups are still at or near their tallest. Nothing here is compared between countries, only within one.
09
The world’s own average is past its peak.
8 published regions, both sexes, one panel each, on one centimetre scale, earliest peak first. Every series that has fallen is sub-Saharan Africa, the Middle East and North Africa, or the world itself.
NCD-RisC, eLife 2016. Year of birth, height at 18. The 8 aggregate entities published alongside the countries, joined on ISO code: 20,200 of 20,200 country rows match to within 0.000 cm, and the same join on name would drop 1,616 of them without saying so. Still rising at 1996, both sexes: Europe and Central Asia, Latin America and Caribbean, North America, South Asia. Median peak year among the 182 past-peak country groups is 1969, quartiles 1964 and 1974.
10
How precisely a country’s height is known varies 13.2-fold.
Every published height comes with a range around it. Each bar is one country: how wide that range is, averaged over the whole century. A wide bar means nobody measured much.
NCD-RisC, eLife 2016. Year of birth, height at 18. 200 countries, both sexes. Mean width of the published 95% range over all birth years and both sexes. A wide range is not a statement about the people, it is a statement about how little was measured. It is why the 40 countries with the widest ranges are drawn on this page and never named.
11
The record got better until the 1960s and has been getting worse since.
The same ranges again, now by the year a person was born rather than by their country. Thin means well known. It narrows for most of the century and then flares back out. The least known people are the youngest.
NCD-RisC, eLife 2016. Year of birth, height at 18. 200 countries, both sexes. Averaged by year of birth. The last year in the file is 1996, measured at eighteen in 2014. The flare at the end is the youngest people sitting at the edge of the evidence. It is not a change in anybody’s height.
12
Two studies of the same people disagree about which way heights moved.
The same scientists published two sets of national average heights four years apart. Both cover everyone born between 1967 and 1996, at eighteen. Each curve is one study’s account of what happened to the average country. Both start at the same point and one ends below it.
Both releases. They share an axis nowhere else on this page, and they can here because the 2020 rows were converted first: a child measured at eighteen in year Z was born in Z minus 18, so both sides are indexed by year of birth and describe the same people at the same age. That gives 30 shared birth years and 12,000 shared pairs. The join a careless reading would make, 2020 year Z against 2016 birth year Z, is absent from this page and is asserted absent in the pipeline. Row by row the two differ by 1.148 cm on average, 0.917 at the median and 6.47 at the worst. Averaged over countries, never over people.
13
The contradictions are all in the best measured countries.
Each dot is one country in one birth year. Up the page: how far apart the two studies put it. Across: how far the two published ranges reach toward each other. Above the line the ranges miss entirely, so both studies cannot be right.
Both releases, the same 12,000 pairs as above. A pair counts as a contradiction when the two published 95% ranges have no value in common. South Korea is the most contradicted country at 54 of its 60 birth years, then China at 39, Japan at 38 and Italy at 35. The across axis stops at 9 cm and the 879 pairs that reach further are drawn against that edge. None of them contradicts and none could: the widest reach that ever misses is 5.197 cm, against a file that reaches to 17.016.
14
Pick a country and a birth year. The two studies disagree.
The same scientists published two sets of national average heights four years apart, and they do not agree. Both cover everyone born between 1967 and 1996, at eighteen. Pick a country, a sex and a birth year: zero is the 2016 study’s answer and the line above or below it is the 2020 study’s. The shape behind is every other country and birth year they both cover.
sex
Both releases, the same 12,000 pairs as the two sections above and the same converted axis. 139 of the 200 countries are offered by name: a claim that reads one study against the other may only name a country whose published range is narrow in both, and the rest are in the field here and in the two marks above, drawn and unnamed. Heights are the studies’ own, to the hundredth of a centimetre, and country names are as each one publishes them. Among the 139, the two differ by 1.03 cm on the average pair, 0.846 at the median and 5.951 at the worst, which is Saudi Arabia men born 1967: 168.22 cm in one and 162.27 in the other. Over all 12,000 pairs, including the countries this page never names, it is 1.148 cm on average and 6.47 at the worst.
Where this comes from
The two studies
Both come from the NCD Risk Factor Collaboration, and both are modelled national averages rather than measurements of individuals. They are two separate publications, not one series updated, so nothing on this page averages or ranks across them. The last three sections read one against the other, and only after the children's study is converted to the adults' axis: a child measured at eighteen in year Z was born in Z minus 18. The two are never compared on the labels they were published with.
- Children, ages 5 to 19. The 2020 study. Height and body-mass index trajectories of school-aged children and adolescents from 1985 to 2019, The Lancet 2020. CC BY 4.0. Indexed by year of measurement, with age as its own column. The country comparisons use the 2019 numbers; the two sections asking what changed use the whole 1985 to 2019 span.
- Adults, at eighteen. The 2016 study. A century of trends in adult human height, eLife 2016. CC BY 4.0. Indexed by year of birth, reporting height at eighteen for everyone born 1896 to 1996. The last of them was measured at eighteen in 2014.
The countries drawn, and the countries named
Every estimate carries a published 95% range, and those ranges are wildly uneven: from 1.09 cm for Japan to 16.12 cm for Angola in the same year. A country known only to 16.12 cm cannot carry a ranking. So the countries are sorted by the width of their own range and the widest fifth is set aside: 160 of 200 countries are drawn and named, and the 40 widest are drawn in the roll call but never named anywhere on this page. The cutoff falls at 9.75 cm.
Each study has its own ranges and its own widest fifth, so a claim that reads one against the other has to clear the bar twice. 139 of the 200 countries do, and those are the only ones the last section names.
Two things this page will not say
- Who inside a country got taller. Neither study carries income, education, region or city. The unit is a country, a sex and an age, and there is nothing underneath it.
- That the centimetre gap between countries is closing. It is not, for girls: their spread widens from 16.9 cm at five to 19.4 cm at nineteen while the same spread measured against their own height falls from 15.2% to 12.1%. The first finding is stated proportionally for that reason.
Reproducing it
bash pipeline/fetch.sh pulls both published studies. npm run data derives
every number on this page into one file and asserts each one against the frozen findings record
before writing it. There are 200 countries in the adult study and the typical
published range in it is 3.32 cm wide. If a guard and the record ever disagree,
the build stops.