The Five-Dollar Study Everyone Quotes

A 2008 experiment on 46 people became a fact nobody cites. A preregistered replication found nothing on the original measure. What survived, and why.

Forty-six people got an envelope. Inside was either five dollars or twenty, and an instruction: spend it before five o'clock today. Half were told to spend it on themselves, the other half on someone else. That evening they reported how happy they felt.

The ones who had spent it on someone else were happier. The paper ran in Science in 2008, and the finding went everywhere — TED talks, corporate wellness decks, a thousand articles with the same headline. Giving makes you happier than getting. It became one of those facts that stops needing a citation.

Forty-six people. That is the part that almost never travels with the claim.

What happened when someone ran it again

In 2022 a separate team published a close replication in PLOS ONE. Same manipulation, preregistered, 133 participants instead of 46. On the original outcome measure the effect was not there. Not weakened — gone. The people who spent on others came out very slightly below the ones who spent on themselves, and nowhere near significantly so: F(1,130) = 0.35, p = .557.

They also ran an exploratory analysis using a single direct happiness question, and there the effect appeared, small but present, 4.36 against 4.04. The authors called their own results mixed, which is the honest word for them.

If the story stopped there this would be a tidy little cautionary tale about small samples, and I would rather it did, because tidy is easier to write. It doesn't stop there.

The claim survives in a smaller, conditional form

In 2020 the original authors ran a registered replication themselves. One of their paradigms went like this: participants learned they had a small windfall and could buy a bag of treats at a steep discount — chocolate, that sort of thing — either for themselves or for a sick child at a local children's hospital they would never meet. That version worked. The people buying for the child reported feeling significantly happier.

Two years later, Aknin, Dunn and Whillans went through all fifteen published preregistered experiments on the question. Their count: of seven adequately powered studies — at least 200 people per condition — six found the effect. Of eight underpowered ones, four did. The effect sizes in the large experiments ran from 0.06 to 0.49, mostly landing near 0.3 or 0.4.

So the effect is real, and it is conditional. The conditions turn out to be where the whole thing lives.

What actually makes the difference

The review names three amplifiers, and money is not among them. Having some genuine choice about whether and how to give. Understanding what your generosity actually did. And giving in the moment, rather than being asked to recall a time you gave — the recall studies are consistently the messiest, which is its own quiet finding about how memory launders an experience.

Read that list backwards and you get a description of most charitable giving as it is actually practised. A monthly direct debit you set up in 2019 and have not thought about since. A donate button with no name behind it. The annual report arrives six months later, a PDF nobody opens. Every condition that makes giving feel like something has been engineered out, usually in the name of efficiency.

The goody bag works for the opposite reason. You choose. There is a person, even an anonymous one, and you know what arrives.

Why this matters for a map

We built Good People Map around a bet: that help works better when it is local and attached to a named person. Most of our argument for that bet has been about trust and logistics — you can walk there, you can see who they are, someone in the thread vouches for them.

The prosocial spending literature suggests a second argument, and it is one a marketing page would skip. If the emotional payoff of helping depends on choice, visible impact and immediacy, then the distance between a donation form and a neighbour two streets away is the whole effect.

That cuts against us as much as for us. A karma counter is an abstraction. A generic request from an account with no photo and no follow-up is the direct debit, not the goody bag. The parts of this platform that behave like a form will produce what forms produce.

What I would not claim

A d of 0.3 is a real effect and a modest one. Anyone telling you that helping others is a reliable route to your own wellbeing is quoting a 2008 study about forty-six people and skipping the eighteen years since.

The better reason to help someone is still the oldest one, which is that they needed help and you were nearby. If it also lifts your mood by a fraction of a standard deviation, take it. Just don't build your life around a number that small.

Sources: Dunn, Aknin & Norton, Science 2008, 319(5870):1687–1688. Kim, Adams, Diaw, Celly, Nelson & Jung, "Prosocial spending encourages happiness: A replication of the only experiment reported in Dunn, Aknin, and Norton (2008)", PLOS ONE, 7 September 2022. Aknin, Dunn, Proulx, Lok & Norton, "Does spending money on others promote happiness? A registered replication report", Journal of Personality and Social Psychology, 2020. Aknin, Dunn & Whillans, "The Emotional Rewards of Prosocial Spending Are Robust and Replicable in Large Samples", Current Directions in Psychological Science, 2022. Retrieved 24 August 2026.