Gapminder's Misconception Study 2017 put thirteen fact questions to roughly a thousand people in each of fourteen countries. Twelve were analysed; one, on climate, was set aside because 87 % got it right. Each question offered three options.
| quantity | value |
|---|---|
| mean score, twelve questions | 2.2 (18.3 %) |
| score expected from picking at random | 4.0 (33.3 %) |
| respondents who beat random | 10 % |
| respondents who got all twelve wrong | 15 % |
Rosling made the comparison with chimpanzees picking lettered bananas, and it was taken as a joke about how little people know. The arithmetic says something harder. A respondent answering at random gets all twelve wrong with probability
and 15 % did. That is 19.5 times the rate chance would produce. Scoring 18.3 % on three-option questions is not what not-knowing looks like; not-knowing looks like 33.3 %. To land reliably below chance you must have a model of the world, apply it, and have it be wrong in a consistent direction.
In Global population growth, box by box (TED@Cannes, 2010) Rosling stacked the world's population into boxes and made an argument that runs against the intuition most people bring: you do not slow population growth by letting children die. You slow it by stopping children dying. Families that can rely on their children surviving stop insuring against the loss, and the birth rate falls — after the death rate, not before it.
That is a claim about order, and order is testable.
This volume keeps arriving at South Korea. Ha-Joon Chang, interviewed by Finance & Development in September 2026, described living through it as riding a time machine. Take four World Bank series for Korea, 1960 to 2020, and ask one question of each: in what year had half of the whole sixty-year change already happened?
| series | 1960 | 2020 | half reached |
|---|---|---|---|
| under-5 mortality, per 1,000 | 112.9 | 3.1 | 1970.8 |
| births per woman | 5.99 | 0.84 | 1975.0 |
| life expectancy, years | 53.8 | 83.4 | 1984.7 |
| GDP per capita, constant 2015 US$ | 1,038 | 33,216 | 1999.5 |
Child survival, then fertility, then longevity, then income — in that order, spread over 28.7 years. Korea had banked half its century of child-survival gain before it had banked a sixth of its income gain. The order is Rosling's, and nothing in the computation was chosen to produce it: the four series are the World Bank's, the question asked of each is identical, and ch07-verify.py fails if they ever stop falling in that order.
Chang says life expectancy was 53 when he was born in 1963, and adds that statistically he should have been dead for a decade. The series puts Korean life expectancy at 53.8 in 1960 and 55.0 in 1963.
The figure is real and it is his country's; it is three years older than the sentence carrying it. Nothing turns on it — his point survives 55 exactly as well as 53. It is recorded here because this is the volume about what happens to a number in transit, and a chapter that catches a museum, a bestseller and a public survey while waving through the number in its own epigraph would be the thing it is complaining about.