Kindness and Happiness

Giving money away: a famous finding, cut down to size

In 2008, Elizabeth Dunn, Lara Aknin and Michael Norton reported in Science that people who spent more of their money on others were happier, both in a national survey and in workers spending a bonus. In a small experiment, people told to spend money on someone else felt happier than people told to spend it on themselves.

A 2013 follow-up found the same link in survey data from 136 countries, rich and poor, and in experiments in Canada, Uganda, India and South Africa.

The honest update came in 2020. The same team ran large, preregistered replications. With 712 students, buying something for a stranger in need did make people happier than buying for themselves. But with 1,950 adults, simply remembering a time you spent on others made no difference. A further study of 5,199 people found a difference, but a small one. A 2025 review by the same research group found that only 19 of 71 experiments on giving and happiness were preregistered, and only about one in five samples (15 of 79) came from outside Western, rich countries. Our reading: giving does tend to lift mood, but the effect is small and depends on how it is done.

Acts of kindness experiments

A 2018 review pooled 27 experiments with 4,045 people. Doing acts of kindness raised the doer's wellbeing by a small to medium amount. The authors found no sign of publication bias, but most studies were small, mostly used Western students, and measured effects only for a short time. The review was funded by kindness.org, an organisation that promotes kindness.

A large 2024 trial in the USA, UK and Australia (4,284 adults) asked people to do at least one kind act a week for four weeks. Loneliness fell a little in the USA and UK, but not in Australia, and there was no effect on depression or positive mood.

In a 2023 trial with 122 adults who had raised anxiety or depression, kind acts beat a thinking-skills exercise on mood and life satisfaction, and beat both comparison groups on feeling connected to others.

Two recent studies add useful detail. In three 2025 trials with people who were feeling depressed, giving away as little as one cent a day online for two months eased their low mood a little more than waiting did, mainly because it lifted their positive feelings. But kindness programmes do not always work: a 2026 trial in 42 Dutch primary schools found that a programme to make classrooms kinder made no overall difference to pupils, and only helped a little where teachers ran it fully as designed.

Does kindness to yourself count?

Two studies disagree. In one (473 adults, six weeks), doing kind things for others raised flourishing, while treating yourself did not. In another (683 adults, one week), kindness to self and to others helped equally.

Self-compassion is different from treating yourself. It means responding to your own failures with warmth rather than harsh criticism. People higher in self-compassion have much lower distress, in adults and in 19 studies of 7,049 adolescents. These are correlations. But a review of 27 randomised trials found that self-compassion programmes reduced depression, anxiety and stress by a moderate amount.

Volunteering

In cohort studies, volunteers report better mood and life satisfaction (though not better physical health), and their risk of death during follow-up was about 22% lower. But the randomised trials did not confirm the mental health benefits. Healthier, happier, better-off people are more likely to volunteer in the first place, so some of the link runs backwards.

If low mood is severe, lasts for weeks, or comes with thoughts of self-harm, talk to a doctor or a trusted adult.

What this means for you

Kindness is worth practising, and it probably helps most through relationships. Expect a gentle lift, not a transformation. Being kind to yourself, in the sense of self-compassion, has the largest effects on this page, though the trials are mostly small and may overstate them.

Prosocial spending

Dunn et al. (2008) combined a nationally representative survey, a windfall field study and a small experiment. Aknin et al. (2013) extended the correlational finding to 136 countries and ran recall and purchase experiments in Canada, Uganda, India and South Africa. Aknin et al. (2020) is the key update: a registered replication report with three studies. Experiment 1 (N = 712, real purchases) replicated. Experiment 2 (N = 1,950, recall paradigm) was null. Experiment 3 (N = 5,199, revised recall) found a small effect. Proulx et al. (2025) reviewed 71 experiments: 19 preregistered, about half of those well powered, 79% testing prosocial spending, and only 19% with non-WEIRD samples, which gave mixed results. Note that both replication papers come from the originating group.

Kindness interventions

Curry et al. (2018): 27 studies, N = 4,045, δ = 0.28 (95% CI 0.16 to 0.41), no moderators, no evidence of publication bias by funnel plot or Begg's test. The mean sample was about 79 per condition, giving power of roughly 0.42 for the pooled effect, and follow-ups were short. Funding: kindness.org. For context, White et al. (2019) re-analysed the two main positive-psychology-intervention meta-analyses and, after correcting for small-sample bias, found wellbeing effects of about r = .10 and no reliable effect on depression. Kindness effects should be read against that background.

Large trials give mixed results. Lim et al. (2024; three RCTs, N = 4,284, waitlist control) found small loneliness reductions (d = -0.13 USA, -0.21 UK, ns in Australia) and no effects on depression, quality of life or positive affect. Dropout exceeded 40%, participants were recruited through the Nextdoor platform and were unblinded, and secondary outcomes were not corrected for multiple testing. Cregg and Cheavens (2023; N = 122, elevated symptoms) found kindness superior to cognitive reappraisal for symptoms and life satisfaction, but not to planned social activities; reduced public self-absorption mediated the effect. A loving-kindness meditation trial (N = 809) found no wellbeing advantage over light exercise, with 82% attrition (Galante et al., 2016).

Three preregistered waitlist-controlled RCTs (total N = 883) of a daily online micro-donation intervention in adults with elevated depressive symptoms found reductions in depression of d = -0.19 to -0.46, mediated by gains in positive emotion; the lack of an active control means expectancy effects cannot be ruled out (Zhang et al., 2025).

Self versus other

Nelson et al. (2016; N = 473) found other-focused kindness raised flourishing via positive and negative emotions, while self-focused acts did not differ from control. Rowland and Curry (2019; n = 683) found no difference between kindness to strong ties, weak ties and self. Tashjian et al. (2021; N = 97 adolescents) found no overall condition differences, with benefits only for more altruistic teens.

Correlational synthesis

Hui et al. (2020): 201 independent samples, N = 198,213, r = .13. Links were stronger for eudaimonic than hedonic wellbeing, for informal than formal helping, and for younger givers (except on physical health, where older and retired givers did better).

A larger three-level meta-analysis (k = 366 effect sizes, N = 222,866) estimated r = 0.26 between prosocial behaviour and mental health, with significant moderation by culture, positive versus negative outcome indicators, measure and cross-sectional versus longitudinal design (Li et al., 2026).

Self-compassion

MacBeth and Gumley (2012): r = -0.54 with psychopathology (20 samples). Marsh et al. (2018): r = -0.55 with adolescent distress (19 studies, N = 7,049, ages 10 to 19), robust to publication-bias correction but highly heterogeneous. In a three-wave study of 1,214 Chinese adolescents, within-person rises in positive self-compassion and in prosocial behaviour both predicted fewer later depressive symptoms, while the scale's negative components behaved differently from its positive ones (Qin et al., 2026). Ferrari et al. (2019): 27 RCTs; g = 0.66 depression, 0.57 anxiety, 0.67 stress, 0.56 self-criticism, 1.37 rumination; the authors flag possible publication bias. In university students, 11 RCTs (N = 803) gave g = -0.85 for anxiety and -0.69 for depression (Sobhaninia et al., 2026). Small trials, waitlist controls and developer-led programmes probably inflate these figures.

Volunteering and mortality

Jenkinson et al. (2013): 40 papers. Cohort studies showed benefits for depression, life satisfaction and wellbeing, but five RCTs did not confirm them; pooled cohort mortality RR 0.78 (0.66 to 0.90). Okun et al. (2013): mortality risk 47% lower unadjusted, 24% lower adjusted, which shows how much confounding matters. Kim et al. (2020; HRS, N = 12,998), adjusting for prior volunteering and prior outcomes, found ≥100 hours a year linked to lower mortality, higher positive affect and purpose, and fewer depressive symptoms, but not to life satisfaction or clinical depression. Nakamura et al. (2025) suggest the mortality link works largely through social contact and physical health. A 2026 review of older-adult volunteering found 11 RCT papers from five trials, all at high risk of bias, and warned that harms may be under-reported (Williams et al., 2026). Kim, Nakamura and VanderWeele are affiliated with Harvard's Human Flourishing Program; the Program is substantially funded by the John Templeton Foundation, although we could not see study-specific funding statements in the texts we read.

Young people and schools

Layous et al. (2012; 19 classrooms, N = 415, ages 9 to 11) found no wellbeing difference versus control, but larger peer-acceptance gains. Schreier et al. (2013; N = 106) found two months of volunteering lowered adolescents' IL-6, cholesterol and BMI; the authors did not report a wellbeing outcome. Universal social-emotional learning programmes raise prosocial behaviour, with a small overall effect (ES = 0.15; 59 studies, N = 83,233; Shi & Cheung, 2024).

In a cluster RCT of 2,337 Dutch pupils across 42 schools, the Meaningful Roles programme showed null intention-to-treat effects, with small benefits for need satisfaction, prosocial climate and school wellbeing confined to a high-fidelity subsample (about 25%). That analysis is no longer randomised, and both authors coordinate the programme's implementation (van Loon & Kaufman, 2026).

Mechanisms

An fMRI study found reward-related brain activity when money went to charity, more when the transfer was voluntary (Harbaugh et al., 2007). That shows reward processing, not a lasting "high". Tested mediators so far are positive emotion, reduced self-focus and social contact.

Shu

The Stanford Encyclopedia of Philosophy describes shu as projecting oneself into another's situation to grasp what they would and would not want, framed negatively as not imposing on others what one does not desire (Analects 12.2, 15.24), and as one facet of ren (Wong, 2023). Csikszentmihalyi (2024) notes the Golden Rule comparison, reports that some scholars treat these passages as later interpolations, and cautions against reading them as an axiom.

Claims we deliberately do not make

That kindness is "proven" to cause lasting happiness; that its effects last for months (most trials measured days to weeks); that science has "proven" Confucius right, since no study tests shu; that giving releases a "helper's high" of endorphins or oxytocin; that volunteering adds years of life (the evidence is observational); that treating yourself works as well as helping others; that you must be kind to yourself before you can be kind to others, which no study we found has tested.