More than six in ten of the Americans and Chinese in a preregistered survey said people have grown less cooperative across a sixty-year span, a belief the authors set against two cross-temporal meta-analyses reporting rises in cooperation in laboratory social dilemmas over the decades those series cover

Sixty-five per cent of the American respondents and 62.1 per cent of the Chinese ones said people have become less considerate, helpful or cooperative over a sixty-year span reaching a few years into the future. Twenty-three per cent of Americans and 28.5 per cent of Chinese said the opposite. The rest expected no change. Split by country, the difference between the two distributions did not reach significance (chi-square(2) = 5.36, P = 0.069).

That is one of the more familiar things anyone believes about other people, and it is now measured in two countries the authors describe as differing in numerous ways in history, governance and culture.

The guess they were paid to get right

Yi Liu, Giuliana Spadaro, Sümeyye Ergün, Yingling Li and Paul Van Lange put 628 American and 449 Chinese adults through a preregistered survey published in Communications Psychology in March 2026. The core task was not an opinion question. Participants read the rules of a prisoner’s dilemma (defect against a cooperator and you take 60 points, mutual defection pays 20 each, mutual cooperation pays 40), then estimated what percentage of people alive in a given year would have chosen to cooperate.

The estimates were incentivised. Come within the paper’s stated tolerance of plus or minus 5 per cent of a benchmark rate and you were paid a small bonus. The benchmark was archived experimental data for the past years, and for 2030 an extrapolation from the relationship between time and cooperation. That detail matters twice over. It gives people a reason to answer carefully, and it means the study is asking for a factual guess rather than a mood.

Each participant judged four years, and the years were split across two groups. One group rated 1960, 1980, 2000 and 2020; the other rated 1970, 1990, 2010 and 2030. Both drew a downward line. Older participants drew a steeper one.

Both samples came from professional online panels, which the authors note can over-represent urban, wealthier and better-educated people; re-running the analysis with the sample weighted on age, education and income produced similar patterns for the declining-cooperation belief. The Chinese sample fell short of its 600-participant target after five months of recruitment. Twelve per cent of Chinese responses and four per cent of American ones were dropped as incomplete.

The counter-evidence the authors cite

The reason this is framed as a paradox rather than a finding is a body of data the authors did not collect. The Cooperation Databank holds 2,636 studies of human cooperation run between 1958 and 2017 across 78 societies. In two cross-temporal meta-analyses drawing on that literature, Yuan and colleagues analysed 511 studies with US participants between 1956 and 2017 and 254 with Chinese participants between 1999 and 2019, covering prisoner’s dilemma, public goods and resource dilemma games, and reported that cooperation rose: by 21 per cent in the United States over 61 years, and by 36 per cent in China over 20. The authors of this survey call those increases small.

Those two numbers belong to that earlier work, not to this survey, and this paper reports them as percentage rises without unpacking in its own text whether they are percentage points or relative change. That archive is also the reason the guessing task could be incentivised at all: the same records supplied the answers the bonuses were paid against. The authors are explicit that the historical rates were used only to pay participants and never to test a hypothesis.

So the comparison is real but it is second-hand, and it is a comparison between what people guessed about strangers in a laboratory game and what strangers actually did in laboratory games. It is not evidence about whether anyone holds a door open more or less often than they did in 1970.

Where the two publics part company

The headline agreement hides a split that is the most interesting thing in the paper. The Chinese sample held a steeper belief about cooperation in the game (β = −0.31, 95% CI [−0.34, −0.28]) than the American one (β = −0.22, [−0.25, −0.20]), and both were clearly negative.

Underneath, the two countries were pessimistic about different qualities. Participants rated people of each era on warmth, morality, assertiveness and competence. The Chinese respondents saw the sharper fall in warmth and morality; the American respondents saw the sharper fall in assertiveness and competence. On competence the Chinese estimate ran fractionally the other way, at the edge of significance (β = 0.04, [0.01, 0.07], P = 0.050), and the authors summarise it as Chinese participants reporting similar levels of competence across the years.

Everyone who said cooperation had changed, in either direction, was then asked to pick up to three reasons from a list of twelve and rank them; the people who said it had held steady were not asked at all. Among the decline-believers, both samples put trust first — 1.26 as the paper’s overall figure for both countries, on a nought-to-three weighted rank score — followed by stress (0.80), wealth (0.77) and individualism (0.77). Underneath that shared first place the two publics weighted the list differently: the American respondents were more likely to name social media, face-to-face contact, technology and religiosity; the Chinese respondents gave more weight to trust, stress, wealth and education. The same story about their neighbours, and each side blaming the thing most visible from where it stands.

The twelve reasons were written by the researchers from prior literature, not generated by participants, and the authors spell out what follows from that: self-reported attributions can be post-hoc rationalisation, misremembering what actually drove a judgement, or an echo of whatever narrative about decline is already in the air. The rankings describe folk explanations. They do not identify the mechanism. The authors add that their participants may not have cleanly separated trust from trustworthiness, which are neighbouring ideas the study leans on separately.

Where the decline stops being detectable

The line does not keep going down — or not detectably. On the game estimates, the difference between 2020 and 2030 was not statistically significant (β = −0.10, 95% CI [−0.28, 0.09], P = 0.104), and on warmth the difference between the two years was effectively nil (beta below 0.01, 95% CI [−0.11, 0.12], P = 0.930). For morality, assertiveness and competence the authors report participants rating the future above the present.

The interval on the game estimate is wide enough to contain a real further decline, so this is a difference the study did not detect rather than an absence it demonstrated. And 2020 and 2030 were rated by two different groups of people, which the authors flag themselves: nobody in this study was asked to look forward from a present they had also judged.

Read carefully, the shape is still striking. The same publics that believe the slide has been running for six decades do not, on average, put 2030 below 2020 by any amount this study can detect. The authors’ suggestion is that two habits collide: a functional pessimism about how others behave, and the unrealistic optimism people tend to hold about their own future.

Everything above is a survey of what people believe, read closely and reported at the precision the paper prints. It is not a measurement of how anyone actually behaves, and this piece has no standing to tell anyone what their neighbours are really like. If a belief that other people have stopped showing up is keeping someone from reaching out, a counsellor is a better place to take it. The paper itself notes that doubting others’ good intentions can stop people asking for help they need, and can become the thing that makes the doubt true.

The measurement that would settle it is the one the authors say is missing: cooperation as it happens outside economic games, tracked over years rather than reconstructed from memory. Until somebody builds that record, the most defensible thing to say is that the belief in a sixty-year decline is now documented on two continents, and that the decades-long series of game data the authors cite move the other way.