Does Jev know where people trust each other?

For each of about 100 countries, what share of people say most people can be trusted in its latest World Values Survey or European Values Study, and does Jev know?

result109 questions

Asked what share of people in each of 109 countries said "most people can be trusted" in the World Values Survey, Jev is off by 8 points on average and ranks the countries moderately like the surveys (rank correlation 0.58). Its guesses huddle around a default: it says 25% for Trinidad and Tobago, Indonesia and Ghana (real: 3-5%) and 25% for China too (real: 63%).

How to read this: Each dot is a country. Across: the share who said most people can be trusted in its latest survey. Up: Jev's guess. Most of Jev's dots sit in a flat band between 15% and 30%, while the real shares spread much wider.

109 countries; 90% interval on the mean error [1.53, 4.7] points.

In short

Jev guesses the share of trusting people in 109 countries within 8 points on average, and knows the Nordic countries are high.
Its guesses bunch in the middle: 20 countries are under 10% in the surveys, but Jev never guesses below 10%.
The biggest misses run both ways: China at 25% against a surveyed 63%, Ghana and Indonesia at 25% against 5%.

What the data shows

Man Holding Cardboard Sign meme: 25% of people trust each other. in every country.

Close on average: 8 points off, and only slightly too high overall (+3 points).
It knows the high-trust countries: 65% for Norway, Denmark, Finland and Sweden, where the surveys found 63% to 74%.
But its guesses are squeezed. They spread less than the real shares (a spread of 12 points against 16), with many countries at a default-looking 20-30%.
It has a floor. Jev never guesses below 10%, but 20 countries are under 10% in the surveys: Brazil (20% against 7%), Colombia (20% against 5%), the Philippines (25% against 5%), Trinidad and Tobago (25% against 3%), Indonesia and Ghana (25% against 5%).
Too low for a few high-trust countries: China (25% against 63%), Belarus (15% against 40%), Yemen (15% against 38%).

What it means, and what it doesn't

Jev has a rough idea of where trust is high and low, and it knows the Nordic countries well, but it won't believe how low trust gets: in a fifth of the countries, fewer than one person in ten says others can be trusted, and Jev never guesses that low. Its picture is least reliable at the extremes.

It doesn't mean Jev doesn't know these countries; the survey question is a coarse one and small misses are within its sampling error.

Caveats

Different years. Each country's figure is from its latest survey since 2010, so years differ, and the question names the year. Trust can shift with events between surveys.
A known quirk of the question. The survey question is a blunt either-or ("most people can be trusted" or "you need to be very careful"), and countries may read it differently. Some surprises in the real data, like China's high share, may reflect how the question is understood as much as how much people trust each other.
One number per country. The published share has a sampling error of a few points, so small misses (under 5 points) mean little.

Jev on this experiment

Would a person find it interesting to read?: Yes78%
Does it describe you?: No61%
Would you have predicted it?: Yes60%
How fair is the comparison?: The comparison is reasonable
How much should a reader rely on it?: A little
Which caveat matters most?: A known quirk of the question72%

Why ask this

Do most people in a country think others can be trusted? The answer ranges from a few percent to over 70%, and it's one of the most studied numbers in social science.

A model's guess about trust in a place reveals whether it knows the world country by country, or projects one middling picture onto all of them. That picture seeps into anything it writes about a country's politics, business culture or daily life: whether strangers lend each other money, leave doors unlocked or trust official records.

How this was done

The people and the data

The World Values Survey and the European Values Study interview representative samples of adults in dozens of countries every few years. One question asks whether "most people can be trusted" or "you need to be very careful in dealing with people". The measure is the share choosing "can be trusted", as compiled by Our World in Data from the combined surveys, at each country's latest survey since 2010. That gives 109 countries, surveyed in different years between 2010 and 2023, with real shares from 2% (Zimbabwe) to 74% (Denmark).

What Jev was asked

In the 2022 World Values Survey or European Values Study in China, what share of people answered "most people can be trusted" (rather than "you need to be very careful in dealing with people")?
0% · 5% · 10% · ... · 95% · 100%

(The survey's answer: 63%. Jev said 25%.) Each question was asked with the answer steps in three shuffled orders, and the answers averaged.

How it was measured

Jev's median guess against the published share for each country: how far off on average, in which direction, how widely its guesses spread compared with the real shares, and whether it ranks the countries in the same order (rank correlation: 1 means the same order).

Where these questions live

109 questions across 1 topic of the map. Each opens on the map with every question in it.

Countries

Trust and Happiness by Country109

Every question

All 109 questions behind this result, the telling ones first: the examples the analysis points to, then the ones where Jev misses, biggest gap first.

Jev’s own answer

Showing 0 of 109