Extrinsic incentives bias
People predict that others are driven more by pay and job security, and less by interest and learning, than they report being driven themselves.
People predict that others are driven more by pay and job security, and less by interest and learning, than they report being driven themselves.
The claim is about prediction, and it has two halves. Asked what moves other people at work, respondents rank pay, benefits and job security higher than those same people rank them for themselves. Asked about interest, learning and worthwhile work, they rank those lower for others than for themselves. DeVoe and Iyengar quote Heath's summary of this: in their lay theories, members of the general population assess others' motives using Theory X, while they assess their own motives using Theory Y. The practical worry is about principals designing deals for agents. Someone who believes staff work only for money will build a scheme around money, and the belief rather than the evidence is doing the designing.
Repeats disagree, or the effect holds in some conditions and not others. The other 248.
Mixed The self-other gap reappears in North American samples using other methods, while the one large cross-cultural field test finds Asian managers drawing no such distinction and Latin American ones reversing it.
DeVoe and Iyengar surveyed 185 Citigroup consumer branch managers and 1,760 of their employees across six countries, with response rates of 86% and 95%. The employees are the cases, since each manager rated between 2 and 77 of them, and every regional test is fitted on fewer than 1,760. Motivation was measured with one item for intrinsic and one for extrinsic on a scale from 1 to 9, a limitation the authors flag.
North American managers rated their employees higher on extrinsic than intrinsic motivation, 6.97 against 6.30, F(1, 690) = 10.68, p < .01, partial eta squared .02. North American employees rated themselves the other way, 6.59 intrinsic against 5.47 extrinsic, F(1, 894) = 12.56, p < .01. Asian managers rated the two alike, 7.29 and 7.24, F(1, 188) = 2.54, not significant, which the authors read as holism rather than as the bias failing.
Latin American managers judged their employees more intrinsically than extrinsically motivated, 7.01 against 5.78, F(1, 265) = 5.85, p < .05, which matched what those employees said about themselves. Woolley and Fishbach ran five studies with 1,428 people, the third of them using reward items modified from Heath. In that third study, 208 online participants either predicted what recruiters care about or reported what they themselves cared about as recruiters.
Recruiters valued intrinsic rewards at 1.86 on a scale running from -3 to 3, above the 1.31 that candidates predicted, F(1, 206) = 14.91, p < .001, partial eta squared .07. The prediction for extrinsic rewards was accurate, F(1, 206) = 1.87, p = .173. In that test the gap sat on the intrinsic side alone.
The retention plan and the manager's own answerillustration
A manager designing a scheme to keep staff builds it around a bonus, and says of himself that he stays for the work. One office, one question, and two answers depending on whose motives are being described.
Why the volunteers turn upillustration
A committee budgets for vouchers to keep its volunteers coming, while every volunteer asked says she comes for the company. Nothing here shows that vouchers fail. The point is that the plan rests on a belief about other people.
Explaining a colleague's resignationillustration
A colleague leaves for another firm and the office settles on the salary as the reason. Her own account is the chance to learn a part of the trade she has never done.
Cases marked as illustrations describe nobody in particular. They are there to make the pattern recognisable, not to report that a particular thing happened.
Heath's paper is closed at Elsevier and could not be read for this entry. On 21 September 2026 the publisher answered this session with a Cloudflare bot challenge rather than a paywall notice, and an archived 2018 copy of the same page offers unsubscribed readers the abstract alone. That abstract, read at PubMed, says people predict that others are more motivated than themselves by extrinsic incentives such as job security and pay, and less motivated by intrinsic ones such as learning new things.
The figures usually quoted come at one remove, from DeVoe and Iyengar's description of the paper. On their account Heath found 73% of survey respondents agreeing that large differences in pay were needed to get people to work hard, and 67% agreeing that people would not take extra responsibility at work unless paid extra for it. Those same respondents ranked pay third among aspects of their own jobs, and important work ranked first for 50% of them.
In one laboratory study, again on that account, participants ranked an extrinsic item most important only 22% of the time. The prediction figures count predictors rather than targets: 32% of participants predicted that classmates would name an extrinsic item first, against 54% for bank managers and 85% for bank employees.
1 primary research3 replication and reanalysis2 commentary and critique2 background and reference
The paper names the bias and reports three laboratory studies and one field study. Its abstract also claims the bias is separable from a self-serving bias and runs counter to the usual actor-observer effect.
The claim, and the replication this entry reads its figures from.
No preregistered direct replication of Heath's studies was found. The search was re-run on 21 September 2026 across Crossref, PubMed, the 282 papers Semantic Scholar lists as citing Heath, and general web search. Heath's own paper is closed, so every figure attributed to it here arrives through DeVoe and Iyengar. DeVoe and Iyengar compared managers' ratings of their employees with those employees' self-ratings, which is not the within-person self-other comparison Heath ran.
The bias is defined against self-reports as the standard of truth, and Rynes, Gerhart and Minette argue that self-reports understate how much pay drives behaviour. They set survey rankings beside behavioural evidence, including Judiesch's meta-analysis, an unpublished doctoral dissertation, in which individual pay incentives raised productivity by 43.7% on average. If self-ranking of pay is depressed by social desirability, part of the gap belongs to the self side rather than to the prediction side.
The pattern predates Heath and appears in a different format. Jurgensen ran a thirty-year study of job applicants who ranked ten job factors. On Rynes and colleagues' account pay came fifth for men and seventh for women in their own preferences, and first for both when the same list was ranked for someone just like themselves. The two documents disagree about the sample: the ERIC record says 57,000 applicants, while Rynes and colleagues say more than 50,000.
Each of these is named in this entry's own prose as something it gets taken for. Where the two came out differently, that is worth knowing before you quote either.
This is not the overjustification effect, which entry no. 132 covers. That claim is about what a promised reward does to later behaviour; this one is about what people believe of other people's motives, and it predicts nothing about whether an incentive will work. It is also not Miller and Ratner's finding on self-interest, which entry no. 240 also carries.
Their five studies had participants overestimating what a financial reward would do to peers' willingness to donate blood, and overestimating what group membership would do to peers' attitudes and behaviour, with that tendency largely unrelated to the pull self-interest exerted on the participants themselves. Heath's abstract states that the extrinsic incentives bias is separable from a self-serving bias and runs counter to the actor-observer effect, though it adds that the theoretical explanation is similar.
Nothing here shows that pay fails to motivate. Woolley and Fishbach raise the bias in their 2015 paper and argue against it as the explanation of their own result, which is that people value intrinsic incentives more while doing an activity than before or after it, whereas extrinsic incentives are valued about the same either way. A self-directed version of the question gives a related answer without being the same claim.
Kuratomi and colleagues ran seven experiments in which people worked on long repetitive tasks, and participants were engaged more than they had predicted. In their flanker experiment with 80 UK adults, predicted engagement was 3.26 against actual 3.79, F(1, 76) = 13.09, p < .001, and the Prediction by Reward interaction reached F(1, 76) = 7.99, p < .01, with the shortfall no longer significant once performance-based money was on offer.
Every claim on this page was held against these sources on 2026-09-21. Nothing here is written from memory. Written and checked by Krishna Chagti; how entries are written and corrected. Found a mistake? Report it, with the sentence and a source.