Here’s a question that sounds like the setup to a psychology experiment, which is exactly what it was.
You’re walking down the street. You find a wallet on the ground. There’s no one else around. The wallet contains identifying information — a business card, some photos — so you could return it if you wanted. It also contains some cash. What do you do with it?
The standard economic prediction, for decades, was clear. Rational actors pursue their self-interest; honesty is costly; the more money in the wallet, the more costly returning it would be; therefore people should return wallets less often when they contain more money. This was, at minimum, a strong implication of standard economic assumptions, and it seemed intuitively plausible. Given the choice between returning a wallet with $5 in it and one with $100, you’d expect more people to keep the $100 one.
In 2019, three researchers — Alain Cohn, Michel André Maréchal, and colleagues at the University of Michigan and University of Zurich — ran the largest real-world experiment on honesty ever conducted to test this prediction. The results surprised almost everyone, including the researchers, and they tell us something worth knowing about the actual shape of human honesty.
The experiment
The design was elegant and expensive. Research assistants in 40 countries, working in 355 cities, “lost” 17,303 wallets by handing them in to staff at banks, theatres, hotels, post offices, police stations, and public institutions. They said they had found the wallet outside, couldn’t stay, and asked the staff member to return it to its owner.
Each wallet contained identifying information — a name, an email address, and a business card so a finder could contact the owner. Some wallets contained no money. Some contained the local equivalent of US$13.45. Some contained the equivalent of US$94.15. The researchers then waited to see which wallets got reported to their supposed owners via the email addresses.
The prediction from standard economic theory, and from much earlier research, was that wallets with more money would be returned at lower rates. The incentive to keep the wallet increases with the amount; honesty should decrease correspondingly.
The actual result was the opposite. In 38 of the 40 countries studied, wallets containing more money were returned at higher rates than wallets containing less money, which in turn were returned at higher rates than wallets containing no money at all. The pattern was large and robust. Globally, the big-money wallets were returned about 72 per cent of the time; the small-money wallets about 51 per cent of the time; the no-money wallets about 40 per cent. More money meant more honesty, not less.
The finding held up across countries with very different income levels, cultures, political systems, and levels of corruption. Honesty rates varied significantly between countries — Scandinavian countries had the highest rates overall, and some countries in Central Asia had the lowest — but within almost every country, the pattern of more money producing more return was the same.
Why
The researchers included follow-up experiments to understand what was producing the result. They tested variations with different wallet contents, different items inside (a key instead of a sum of money, which affected return rates even though the key had no cash value to the finder), and survey questions probing participants’ reasoning.
The emerging explanation, consistent across the experiments, was that honesty is largely not about money. It’s about how people feel about themselves. When a person considers keeping a wallet with a small amount of money, they can easily tell themselves they’re not really doing anything wrong — it’s just a few dollars, the owner probably won’t miss it, the gain to them is small enough not to count as theft. When they consider keeping a wallet with a substantial amount, the psychological framing changes. Keeping it would mean confronting themselves as someone who took meaningful money from a specific stranger. The identity cost of that self-image — I am the kind of person who takes $100 from a stranger — is substantial. The financial gain isn’t worth the cost to their own self-concept.
This is sometimes called, in the behavioural-ethics literature, the identity-protective theory of honesty. People are honest not primarily because of rational calculation about costs and benefits, but because their self-concept depends on seeing themselves as honest. Small dishonesty that doesn’t threaten the self-concept is common. Larger dishonesty that would, is rare. The threshold isn’t the amount that would be financially worthwhile — it’s the amount that would be identity-changing.
The broader research on honesty
The Cohn-Maréchal findings fit within a wider research tradition on how honesty actually operates.
The American behavioural economist Dan Ariely — whose work, we’ll come back to, has itself been troubled by replication problems — had earlier proposed what he called the fudge factor theory: people are willing to cheat a little, because cheating a little doesn’t damage their self-image, but few people cheat a lot, because cheating a lot does. Ariely’s findings suggested that most people, given opportunities to cheat on laboratory tasks, did so modestly but not to the maximum extent possible. They maintained enough honesty to feel good about themselves.
A related research programme by the Israeli economist Shaul Shalvi has shown that people’s honesty depends heavily on how the opportunity to cheat is framed. When the dishonesty is indirect — benefiting from an ambiguous situation rather than actively stating a falsehood — rates go up. When the dishonesty is direct and explicit, rates go down, even when the financial incentives are identical. The framing matters because direct dishonesty produces a larger self-image cost than indirect dishonesty does.
What emerges from this body of work is a picture of honesty that’s neither purely Kantian (people being honest because they believe it’s morally required) nor purely economic (people being honest because the costs of lying exceed the benefits). It’s something more complicated. Honesty is substantially protected by the individual’s need to maintain a positive self-concept. The identity defence isn’t perfect — people cheat at the margins where the cheating doesn’t threaten their self-image — but it’s real, and it scales with the size of the dishonesty being considered.
The Ariely problem
A painful complication in this research tradition is that Dan Ariely’s own work, which popularised many of these findings, has been found to include serious methodological problems.
In 2012, Ariely and colleagues published a paper claiming that asking people to sign honesty declarations at the top of forms, before filling them out, produced more honest responses than asking them to sign at the bottom. The finding was widely celebrated and applied in insurance forms, tax documents, and other settings.
In 2021, data sleuths examining the insurance-company data underlying the study found what they called clear evidence of fabrication. The numbers had patterns that couldn’t have come from real respondents. The claim appears to have been built on falsified data. Ariely has denied direct responsibility for the fabrication; the investigation continues. But the finding itself — the specific claim that signatures at the top of forms produce more honesty than signatures at the bottom — appears not to be real.
This matters for how we read the broader research on honesty. Some of Ariely’s other findings have replicated; some haven’t. His popularised framework of small-but-not-large cheating has supporting evidence from other labs, but the specific magnitudes he reported may be unreliable. The Cohn-Maréchal wallet study, by contrast, is a very large-scale pre-registered natural experiment that’s much harder to falsify — the wallets were really lost, the returns were really tracked, the pattern is reliably measurable. This is the more trustworthy version of the honesty-research story.
The Ariely case is also a useful reminder that even well-popularised research findings can turn out to be wrong, sometimes in dramatic ways. The wallet research is currently our best single piece of evidence on ordinary human honesty. It’s probably not going to be overturned by later work, because of how it was designed. But the Ariely case is a reminder to hold specific findings with appropriate care, and to give more weight to findings that have been replicated independently than to findings that rest on single studies.
The national variation
One of the most interesting aspects of the Cohn-Maréchal data is the cross-country variation. Return rates varied from about 80 per cent in Switzerland, Norway, and Denmark to about 15 per cent in Kazakhstan, Morocco, and Peru. The variation was strongly correlated with national measures of civic honesty, trust in institutions, and economic development.
This matters because it suggests that honesty isn’t purely an individual trait. It’s substantially shaped by the social context you’re in. In societies with high civic honesty, honesty is easier — the social norms support it, the expected behaviour of others reinforces it, the assumption that other people are honest makes it easier to be honest yourself. In societies with lower civic honesty, individual honesty becomes harder, because the social supports are weaker and the assumption that others will be honest becomes less warranted.
This fits with the situationist research we looked at earlier. What looks like individual character is often substantially a product of situation — and the social-cultural environment is one of the most powerful situational variables. A given person is probably more honest in one culture than another, not because they’ve changed but because the cultural context has changed.
The implication is that maintaining honest institutions and honest cultural norms has effects beyond the individual level. The person who’s honest in a generally honest society is partly honest because of that society. Preserving the conditions that support collective honesty is, in a specific sense, a contribution to everyone’s capacity to behave well.
The question that remains
The deepest thing the wallet research teaches, beyond any specific numbers, is something clarifying about human moral psychology. Honesty isn’t primarily a product of rational calculation. It isn’t primarily a product of fear of punishment. It’s substantially a product of identity — of most people’s need to see themselves as the kind of person who doesn’t steal from strangers, even when they could.
This is encouraging because it means that the ordinary machinery of human self-regard is, on balance, working in favour of honest behaviour most of the time. It’s also humbling because it means that honesty is more fragile than moral philosophy often suggests. The circumstances that protect people’s sense of themselves as honest can be disrupted — by cultural breakdown, by economic desperation, by gradual normalisation of dishonesty in a community. When the identity protection weakens, the behaviour weakens with it.
The question worth carrying, the next time you notice yourself considering a small dishonesty:
If the amount involved were large enough that keeping it would mean changing how you see yourself — would you still do it?
Key research referenced: Alain Cohn, Michel Maréchal and colleagues’ 2019 Science paper on civic honesty (with David Tannenbaum and Christian Lukas Zünd); Dan Ariely’s earlier honesty research, and the 2021 data-fabrication investigation that has since cast doubt on parts of it; Shaul Shalvi’s research on the framing of dishonesty; the broader identity-protective theory of honesty.