General information, not financial advice.
The single most quoted finding in the economics of wellbeing is that money stops buying happiness at about $75,000 a year. It comes from a 2010 paper by Daniel Kahneman and Angus Deaton, both of whom have Nobel prizes, published in PNAS, and it has been repeated in roughly every discussion of income and life satisfaction since.
In 2021, Matthew Killingsworth published a paper in the same journal reporting the opposite: using a large smartphone-based dataset of momentary experience, he found that emotional wellbeing kept rising with income well past $75,000, with no plateau visible anywhere in the data.
Two careful studies, the same journal, incompatible conclusions. What happened next is the part worth reading about.
Adversarial collaboration
Rather than trading rebuttals in print, Killingsworth and Kahneman agreed to work the disagreement out directly, with Barbara Mellers of the University of Pennsylvania acting as an arbiter — a format Kahneman had long advocated and called adversarial collaboration. The three of them jointly re-analysed Killingsworth’s data, with the explicit aim of finding where the two results came from rather than which one won.
The joint paper appeared in PNAS in March 2023. The resolution is genuinely interesting and neither side got to claim victory.
What they found
The plateau is real — for some people. Among the least happy portion of the sample, roughly the bottom 20% by wellbeing, happiness rose with income up to around $100,000 and then flattened. More money did not help that group beyond a point.
For everyone else, happiness kept rising. Across the middle of the distribution, wellbeing continued to increase with log income past the plateau point, with no ceiling in the data.
For the happiest group, the relationship accelerated. In the upper reaches of the happiness distribution, the slope above roughly $100,000 was steeper, not flatter.
Average the three curves together and you get something that looks like a flattening. Look at any one group and you see something else. The 2010 plateau was, in large part, a composition effect — a feature of the average rather than of most people’s experience.
Why the original study found what it found
The methods also differed in ways that mattered, and the joint paper identified them.
Kahneman and Deaton’s 2010 measure of emotional wellbeing came from Gallup survey questions with a yes/no format. A person could report the maximum on that scale and still be happier in ways the instrument could not record. That ceiling effect compresses differences at the top of the distribution — exactly where the plateau appeared.
Killingsworth’s data came from an app that pinged people at random moments and asked how they felt right then, on a continuous scale, producing many observations per person. It measured a finer-grained thing, and it could register differences the coarser instrument could not.
The two studies were not measuring quite the same quantity with quite the same precision. Once that was accounted for, both sets of results could be reproduced from the same data.
What is still true
The finding that survives all of this — and that was never really in dispute — is that the relationship runs on a logarithmic scale. Happiness rises with proportional increases in income, not absolute ones.
Going from $30,000 to $60,000 buys roughly the same increment of wellbeing as going from $120,000 to $240,000. Each step requires doubling. That is why “money keeps buying happiness” and “money matters less as you have more” are both correct statements about the same curve, and why they sound like a contradiction when the scale is left out.
The 2023 paper is also explicit that there is a separate group for whom money does not appear to be the operative variable — the miserable minority whose wellbeing flattens. The authors suggest the sources of that unhappiness, whatever they are, are not the kind that income addresses.
Honest limits
This is all correlational. Nobody randomly assigned income. The relationship could partly run the other way — happier people may earn more — or be driven by a third factor such as health or stable employment. The 2023 paper does not resolve causation and does not claim to.
The sample is drawn from app users who opted into an experience-sampling study, which skews younger, more educated and more employed than the population. And the income ranges examined are those of ordinary working people; very little in this literature speaks to the extreme upper tail, where the number of observations collapses.
The actual lesson
The substantive finding about money is useful. The methodological one is more so.
Two of the most careful researchers in the field published contradictory results in the same journal, and the field’s normal machinery — citation, rebuttal, the slow accumulation of further studies — would have taken years to sort it out, if it ever did. Instead the authors sat down together with a neutral arbiter and re-analysed the data jointly, and the disagreement dissolved into a more precise and more interesting finding than either of them had started with.
That process is rare. The 2010 plateau, meanwhile, will probably outlive the correction by decades, because a round number is easier to carry than a set of three curves.
Sources
- Killingsworth MA, Kahneman D, Mellers B. Income and emotional well-being: A conflict resolved. PNAS 2023;120(10):e2208661120.
- Kahneman D, Deaton A. High income improves evaluation of life but not emotional well-being. PNAS 2010;107(38):16489-16493.
- Killingsworth MA. Experienced well-being rises with income, even above $75,000 per year. PNAS 2021;118(4):e2016976118.
