Consider the kinds of choices that shape a life: taking a higher-paying job that adds an hour to your commute, working extra shifts to fund a wedding, moving across the country for a promotion that pulls you away from family. Each involves weighing income against something else you value. The catch is that you can only live one version of the outcome, which makes it hard to know whether the extra money is actually worth it.
A new NBER working paper asks whether people systematically overestimate how much additional income will improve their happiness, and whether that misperception nudges them toward money-focused decisions they might otherwise reconsider. The short answer, according to the authors: yes on both counts.
The question behind the experiment
Economists and psychologists have spent decades estimating the actual relationship between income and life satisfaction. What has received far less attention is what ordinary people believe that relationship to be. Yet those beliefs, not the underlying statistical truth, are what people rely on when deciding whether to take the new job or work the extra hours.
Ricardo Perez-Truglia and Rafael Macedo Rubião, both at UCLA’s Anderson School of Management, set out to measure those beliefs directly, test whether they can be corrected with scientific evidence, and see whether corrected beliefs change what people say they would choose. Their central concept is what they call the “marginal satisfaction from income,” or the expected bump in life satisfaction (measured on a 0-to-100 scale) from a 20% raise.
Why beliefs about income might be inflated
The researchers built a model to guide their experiment. The intuition is straightforward. When people think about how income affects happiness, they tend to draw on vivid memories: the thrill of a bonus, the excitement of a new car, the relief of paying off a debt. But research on hedonic adaptation and rising aspirations suggests those short-term boosts fade. What felt like a luxury becomes the new normal, and the satisfaction gain shrinks.
If people extrapolate from those vivid short-run experiences without accounting for adaptation, the model predicts they will overestimate the long-run satisfaction effect of higher income, especially for themselves, since their own memories weigh more heavily than what they know about other people.
How the study was set up
The authors recruited 3,002 U.S. respondents through the online research platform Prolific in April 2026. After quality checks and dropping respondents with extreme prior beliefs, they analyzed data from 2,775 people. About 79% of them returned for a follow-up survey roughly a month later.
In the baseline survey, respondents first answered questions about their current or most recent job, including earnings, hours, and commute. They were then asked to estimate how much a 20% raise would raise life satisfaction, both for an average person earning $50,000 and for themselves. To help calibrate the 0-to-100 scale, the survey first showed benchmarks from published research: the death of a spouse is associated with a 7.9-point drop in life satisfaction, becoming unemployed with a 5.2-point drop, and getting married with a 4.6-point rise.
Respondents were then randomly assigned to a treatment group or a control group. The treatment group saw a three-screen module summarizing scientific evidence that a 20% income increase raises life satisfaction by roughly one point on the 100-point scale. The module presented correlational data from the General Social Survey, causal evidence from a study of Swedish lottery winners, and a discussion of adaptation and aspirations. The control group saw no such information.
Afterward, both groups were asked the same belief questions again, along with a series of hypothetical job-choice scenarios tailored to each respondent. In one, respondents chose between their current-style job and one that paid 20% more but required 20% more hours. Others traded higher pay against a longer commute or less sleep.
An AI interviewer for real-world decisions
To get beyond hypothetical scenarios, the researchers added a novel step. Respondents were asked to describe, in their own words, a real decision they were currently weighing that involved a trade-off between income and something else. An AI-led chatbot interviewed each respondent about that decision, asked clarifying questions, and then generated a personalized binary choice for the respondent to consider.
The decisions ranged widely. One respondent was deciding whether to work extra hours to pay for an expensive wedding at the cost of family time and risking burnout. Another was choosing between a higher-paying career in law and a passion for the circus. Others weighed relocation, delaying medical treatment, or leaving a marriage. To identify which option offered higher income, the authors used a large language model to classify each decision, with human coders reviewing the trickiest cases. This produced 1,711 usable real-world decisions.
What the researchers found
On average, respondents believed a 20% raise would raise their own life satisfaction by 6.80 points and the average person’s by 5.20 points. Compared with the scientific benchmark of about one point that the treatment group would later receive, those beliefs were roughly five to seven times larger. Consistent with the model, the overestimate was larger for beliefs about oneself than for beliefs about others.
When treated respondents saw the scientific evidence, their beliefs shifted substantially. Average posterior beliefs about the effect on oneself fell from 6.95 in the control group to 4.23 in the treatment group. Beliefs about the average person fell even more, from 5.81 to 2.50. A month later, roughly a third of that belief change persisted, suggesting respondents genuinely absorbed the information rather than briefly parroting it back.
The more consequential question was whether these belief shifts changed choices. In the hypothetical job scenarios, treated respondents became less willing to pick the higher-income option. For the trade-off between 20% more pay and 20% more hours, willingness to choose the higher-paying job dropped by 0.089 standard deviations relative to the control group. Similar patterns appeared for the commute and sleep scenarios, and for the real-world decisions, though the real-world estimates were noisier because of the smaller sample.
Combining the belief and choice results using an instrumental-variable approach (which isolates the portion of belief change caused by the random assignment to treatment), the authors found that a one-point reduction in a respondent’s belief about their own marginal satisfaction from income translated into roughly a 0.02 to 0.04 standard-deviation shift away from higher-income options across the different scenarios.
Caveats worth noting
A few limitations deserve attention. The choices measured in the study are stated preferences, not observed behavior. Respondents said what they would choose; the researchers did not track whether treated respondents actually changed jobs or turned down promotions. In the follow-up, about 74% of respondents’ baseline preferences matched their eventual real-world decision, which is suggestive but not definitive.
The scientific benchmark of one point per 20% income increase is itself an estimate drawn from specific studies, and the true effect may vary across people and circumstances. The authors are explicit that they compare respondents’ beliefs against this benchmark rather than an unimpeachable ground truth. Their sample also skews younger, more educated, and more left-leaning than the general U.S. population, as is common for online research panels.
The authors interpret their results as evidence that beliefs about the income-satisfaction relationship are measurable, malleable, and behaviorally meaningful. In their framing, people may be pursuing money too aggressively at the expense of family time, sleep, and other non-financial goods, in part because they overestimate how much the money will pay off in felt life satisfaction. Whether that interpretation generalizes to bigger, harder-to-reverse life decisions, like career changes or where to live, remains an open question.
