Few social-science findings arrive with a cleaner moral contrast than this one did. In 2010, four studies reported that people lower on the social ladder were more generous, charitable, trusting and helpful than people with greater status. Here was a memorable counterweight to the assumption that having more makes it easier to give more.
A 2025 meta-analysis in Psychological Bulletin has now reached the opposite overall result. It combined 471 independent studies with 2,340,806 participants and found that higher social class was associated with slightly more prosociality. “Slightly” is doing important work. The link was very small, and it cannot tell us that wealth makes somebody kind.
In my reading, the real correction is not that one class has replaced another as the generous one. It is that social class was never a reliable personality test.
The four studies did not measure one single thing
The original paper, “Having Less, Giving More”, was written by Paul Piff, Michael Kraus, Stéphane Côté, Bonnie Hayden Cheng and Dacher Keltner. Piff, Kraus and Keltner were at the University of California, Berkeley; Côté and Cheng were affiliated with the University of Toronto.
In the first study, 115 participants completed a version of the dictator game, deciding how many points from a small allocation to give to an anonymous partner. Those who placed themselves lower on a social-status ladder gave more. In the second, 88 participants compared themselves with people at the top or bottom of that ladder, then recommended what percentage of income people should give to charity. Those made to feel relatively lower in rank recommended more.
A third study connected lower objective social class with greater trust in an economic game. A fourth found lower-class participants more willing to help a supposedly late laboratory partner, particularly when compassion was involved.
That consistency was striking, but the four adjectives in the headline are not interchangeable measures. Giving points, recommending a charitable percentage, trusting an unseen partner and sacrificing time to help someone are related expressions of prosociality, not four identical replications. Only some involved a real commitment of resources, and the studies used relatively small, specific samples.
Why the explanation was so persuasive
The authors proposed that people with fewer material resources are more dependent on others and more exposed to circumstances they cannot control. That environment might encourage attention to other people, egalitarian values and compassion. People with more resources, by contrast, may enjoy greater independence and have less practical need for mutual reliance.
It is a plausible theory, and it gave the results a coherent shape. It is also easy to stretch beyond the evidence. Social class may be represented by income, education, occupation or a person’s subjective sense of rank. Those indicators overlap, but they are not the same. Likewise, an anonymous allocation game can isolate a choice cleanly without recreating the relationships, tax systems, charitable norms and household constraints of ordinary life.
The paper therefore supported a set of hypotheses under particular conditions. It did not establish that people with less money possess more generous characters across cultures and situations.
The challenge began well before the new meta-analysis
By 2015, the literature was already refusing to cooperate with a simple class story. A large-scale test by Johanna Korndörfer, Boris Egloff and Stefan Schmukle analysed eight studies, including representative German and American datasets with samples as large as 37,136. Across measures of donating, volunteering, helping, trust and cooperation, higher socioeconomic status was generally associated with more prosocial behaviour.
In 2020, researchers reported two preregistered, high-powered direct replications of the first two 2010 studies. Neither recovered evidence that higher status predicted less generosity. Their combined estimate was r = .01, with a 95% confidence interval from -.109 to .131. That interval included no association, but not the original studies’ average estimate of r = -.215.
The 2025 paper did not suddenly overturn an uncontested law. It weighed a body of evidence that had produced conflicting answers for years.
What 2.3 million people add to the picture
Junhui Wu and colleagues gathered 1,106 effect sizes from 471 independent studies conducted between 1968 and 2024. In total, the studies covered 2,340,806 children, adolescents and adults across 60 societies. The researchers included both objective indicators of class, such as income and education, and measured or experimentally manipulated perceptions of social rank.
Across that vast collection, the relationship between social class and prosociality was r = .065, with a 95% confidence interval from .055 to .075. In the way the analysis was coded, the positive sign means higher social class was linked with slightly more prosociality.
The precision of that estimate should not be mistaken for a large effect. As a rough guide to scale, squaring the correlation gives about 0.4 per cent shared variation. Correlations in a complex meta-analysis cannot be reduced perfectly to that calculation, but it makes the central point clear: knowing somebody’s class position tells us very little about how generous or helpful that individual will be.
The result was similar for objective class measures, r = .066, and measured or manipulated subjective class, r = .063. It also did not differ significantly across the broad sociocultural variables the authors tested. That supports a modest overall tilt, not a moral divide.
How generosity is measured changes the answer
The pooled relationship was stronger when prosociality required an actual commitment of resources, r = .079, than when it consisted of an intention, r = .039. It was also larger when behaviour was public, r = .065, than when it was private, where the estimate was close to zero at r = .016.
These differences matter because resources and motives are tangled together. A person with a high income can give a larger absolute sum while sacrificing a smaller share of what they have. Someone with less money may be willing to help but unable to absorb the cost. Public donations may reflect genuine concern, reputational incentives or both. A measure of trust may capture expectations about another person rather than pure generosity.
The meta-analysis calls its interpretation the resource perspective: higher-class people have more material and psychological resources available for prosocial action. Its findings are consistent with that account, especially when behaviour carries a cost. But the pooled correlation does not prove that resources cause generosity, nor does it settle how sacrifice should be compared across unequal budgets.
“Prosociality” is itself an umbrella. Donating money, volunteering time, cooperating, trusting and helping can arise from different circumstances. Combining them reveals the broader direction of the literature while inevitably smoothing over those differences.
What remains after the apparent reversal
The original studies should no longer be treated as evidence for a general rule that lower-status people are more generous. Later direct replications did not recover the pattern, large population studies often pointed the other way, and the most comprehensive synthesis to date found a small positive association between class and prosociality.
That does not make the reverse stereotype sound. The meta-analysis does not show that affluent people are kinder, that every form of giving rises with income, or that the same mechanisms operate in every setting. Its average effect is far too small for confident judgements about an individual. The 2010 explanation may still describe some contexts in which dependence, egalitarian values or compassion encourage cooperation.
The useful next questions are narrower. Researchers can distinguish absolute gifts from the proportion of resources sacrificed, public acts from private ones, capacity from willingness, and correlation from causal effects. On the accumulated evidence, social class does tilt slightly towards greater prosociality at the higher end. It simply does not tilt far enough to turn status into character.