The Obvious Answer Is Wrong
Ask someone what they'd do with a surprise $100 and most people give some version of the same answer: buy something nice for themselves. The logic seems self-evident. Happiness comes from getting what you want, and money lets you get more of it.
Elizabeth Dunn, a psychologist at the University of British Columbia, suspected the logic was flawed. In 2008, she and colleagues Lara Aknin and Michael Norton published three studies in Science that tested whether how people spend money matters more than how much they have, and the answer was yes, but not in the direction anyone predicted.
The National Survey
The first study was straightforward. Dunn's team surveyed 632 Americans about their income, monthly expenses, and happiness, splitting spending into two categories: personal (bills, expenses, gifts for yourself) and prosocial (gifts for others, charitable donations). Income had no relationship to happiness, and personal spending showed no connection either, but the amount people spent on others was a significant predictor of how happy they reported being.
People who devoted a larger share of their money to others reported being happier, regardless of how much they earned. Not by a trivial amount. The magnitude was striking in a specific way: spending an extra dollar on someone else moved the happiness needle about as much as earning an extra dollar in salary.
The Windfall Experiment
Correlations leave room for doubt. Maybe happy people are just more generous, not the other way around. So Dunn ran an experiment.
Forty-six University of British Columbia students received envelopes containing either $5 or $20, with instructions to spend the money on themselves or on someone else by five o'clock that day. That was the entire intervention. Researchers called each participant that evening to measure their happiness.
Participants assigned to prosocial spending reported significantly greater happiness than those assigned to personal spending (p = .042), and the amount made no difference: five dollars spent on a stranger produced the same happiness boost as twenty dollars spent on yourself.
The Prediction That Proves the Point
Dunn asked a separate group of 109 people to predict which condition would make someone happiest, and sixty-three percent chose personal spending while eighty-six percent chose the $20 condition over the $5 one. Both predictions were wrong, revealing that people are not just mistaken about the source of spending-related happiness but are confidently, doubly mistaken about it.
This misprediction matters because it reveals the mechanism keeping people stuck: if you believe buying yourself a nicer dinner will feel better than buying a colleague coffee, you will keep allocating money the less effective way, and the error perpetuates itself because nobody tests the alternative.
From Vancouver to 136 Countries
A finding among Canadian undergraduates could be a cultural artifact. Nobody would bet that people in rural Uganda or sub-Saharan Africa, many of whom struggle to feed themselves, would feel happier giving money away. Aknin and colleagues tested this in 2013 by examining Gallup World Poll data covering 234,917 individuals across 136 countries. After controlling for income, social support, perceived freedom, and national corruption, the correlation between prosocial spending and happiness was positive in 120 of 136 countries. The relationship was significant in every world region, including sub-Saharan Africa, South Asia, and Latin America. It held everywhere.
Separate experiments in Canada, Uganda, India, and South Africa confirmed the causal direction. When people were randomly assigned to spend on others versus themselves, prosocial spenders reported greater happiness across all four nations, despite vast differences in wealth and culture. Even in countries where many people could not meet their own basic needs, giving money away was associated with greater happiness than keeping it.
What the Meta-Analysis Shows
By 2018, enough studies had accumulated for a formal synthesis. Oliver Scott Curry and colleagues at the University of Oxford identified 27 experimental studies meeting strict inclusion criteria, with a combined sample of 4,045 participants. Multi-level modeling yielded a pooled effect size of ฮด = 0.28, classified as small-to-medium. The effect was not moderated by sex, age, participant type, specific intervention, control condition, or outcome measure. No exceptions. Funnel plot analysis found no indication of publication bias. Clean data.
For context, ฮด = 0.28 is comparable to the effect of cognitive behavioral therapy on anxiety, or the benefit of regular exercise on depressive symptoms, and while it will not transform anyone's life, the effect is real, consistent, and robust to the analytic choices researchers made across two decades of independent studies.
The Strongest Counterargument
A 2022 registered replication by Kim, Adams, and colleagues at NYU attempted to reproduce the original windfall experiment with 133 participants. Using the same analysis as Dunn, they found no statistically significant difference between prosocial and personal spending on their composite happiness measure. When they used a more direct happiness item, the effect reappeared, but it was smaller than the original. The honest reading: the original study's sample of 46 was too small to reliably detect an effect of this size. Curry's meta-analysis requires roughly 200 participants per condition for 80% power at ฮด = 0.28. By that standard, the original study's success was more surprising than any single replication failure.
A separate PNAS study by Ellingsen and colleagues examined real-money, high-stakes prosocial choices and found that initial positive feelings faded over months and even reversed in some conditions, suggesting that the happiness boost from prosocial spending may be genuine but transient: a reliable emotional pulse that surfaces when you give but dissipates before it becomes a permanent feature of your well-being.
What We Didn't Prove
The original Dunn experiment used 46 Canadian undergraduates, a sample too small to reliably detect the effect that the meta-analysis later confirmed at ฮด = 0.28. The 632-person survey was cross-sectional. Happy people may simply remember their generosity more vividly than unhappy ones. The Gallup World Poll used a single yes/no item ("Did you donate to charity in the past month?"), which cannot distinguish between someone who gave $5 and someone who gave $5,000. Most experiments measured happiness within hours of spending; whether prosocial spending produces lasting changes in well-being over weeks or months remains largely untested. Nearly all studies relied on self-reported happiness, which is sensitive to demand characteristics: participants told to be generous may report feeling good about it because that is the socially desirable answer, not because they actually feel better.
The Bottom Line
Across national surveys, randomized experiments in four countries, and a meta-analysis of 27 studies encompassing 4,045 participants, the evidence converges on a finding that most people would never predict: spending money on others reliably produces a small-to-medium boost in the spender's own happiness, while personal spending has no detectable relationship to well-being at all. The effect is modest and probably temporary, but it does not require large sums, and five dollars appears to work as well as twenty.
What You Can Do
Redirect a small, fixed amount weekly toward someone else, because based on this evidence, ten dollars a month split between a friend's coffee and a local cause would produce more emotional return than an equivalent upgrade to your own routine. When you receive a bonus or windfall, spend a portion on someone specific before spending on yourself, since the Dunn data suggest the sequence matters for how the money feels. If you want to test this personally, try a structured two-week experiment: spend one week buying small things for others and the next spending the same amount on yourself, tracking your mood each evening with a simple 1-to-10 rating. You will generate your own data on a question where human intuition is demonstrably unreliable, and the answer may change how you think about every dollar that comes after.