From the archives: On the Folly of Rewarding A, While Hoping for B
Steven Kerr, Academy of Management Journal, December 1975
First published 50 years ago. Still uncomfortable reading?
I’d like to thank Professor Rob Briner for bringing this paper back to our more immediate attention.
In turn it’s worth a re-share and a short summary to tempt you to read the whole thing.
Enjoy, or not!
Karen
Fifty years on, this short paper remains one of the most quoted, and most ignored, pieces ever written about reward. Kerr's argument is simple: organisations routinely set up systems that pay off for behaviour A while loudly hoping for behaviour B, and then the orchestrators act surprised when they get A.
Re-reading it now could be uncomfortable. Almost every example he gives could be dropped into a 2026 reward conversation with only the proper nouns changed.
"Numerous examples exist of reward systems that are fouled up in that behaviours which are rewarded are those which the rewarder is trying to discourage, while the behaviour he desires is not being rewarded at all."
If any of that sounds familiar, the full paper is worth your time: 'Kerr 1975 Folly of Rewarding A'. It is fifteen pages long and an accessible read, if dated by some of his references in places.
The examples Kerr uses and why they hold up
Kerr works through six settings where the stated hope and the actual reward point in opposite directions.
Politics. Voters say they want politicians who are specific about policy. But voters reliably punish candidates who provide that specificity and reward those who stay vague.
War. In World War II, soldiers got home when the war was won. Obeying orders and personal survival pointed the same way. In Vietnam, soldiers went home at the end of a tour regardless of outcome, and mutiny was more likely to be met with rest and rehabilitation than punishment. The system rewarded disobedience while designed for obedience.
Medicine. Diagnosing a healthy patient as ill generates income, repeat business and a reputation for caution. Missing a real illness risks lawsuits and professional embarrassment. The system therefore rewards over-diagnosis while the profession publicly admonishes it.
Orphanages. An orphanage's budget, staffing and director's prestige all scale with the number of children in the orphanage. The stated mission is placing children in good homes. Kerr's observation was it becomes rational to make adoption difficult.
Universities. Society hopes academics teach well. Universities reward research and publications. Teaching credentials are hard to document and don't travel well. Research CVs do. The rational choice for a professor is obvious; it isn't teaching.
Business. Pollution penalties so low that breaking the law is the cheaper option. Training programmes evaluated by the people who sold them. MBO systems that measure what's easy to count and ignore everything that isn't. Short-termism driven by quarterly reporting cycles. The cost-plus contract gets special mention i.e. if next year's budget is a function of this year's spend, you are rewarding spending, not economy.
The detail behind each of these is worth reading in the original paper.
Two company cases
Kerr backs up the theory with two organisational examples. Neither presents a positive picture.
A midwest manufacturer used an Expect Approval survey to ask employees what behaviours they thought management rewarded. Lower-level staff in one division reported that yes-manning, going along with the majority and staying on everyone's good side all brought approval; exactly the 'conservatism and apple-polishing' that senior managers had been complaining about in interviews. Risk-taking on best available information was as likely to be punished as rewarded below grade 8 (the company’s grading system). Management wasn't getting what it hoped for. It was getting what it was perceived to reward.
A large insurance firm's claims division tracked complaints from underpayments but had no equivalent signal for overpayments. Writing to clarify ambiguous claims would breach the target of paying 90% within two days. The working norm became: 'When in doubt, pay it out.' Layered on top, the merit system differentiated so weakly between outstanding, average and negligent performance (5% / 4% / 3%) that almost no one engaged with it. The only rule that cut through was the absence and lateness rule, which forfeited the entire increase after three incidents in six months. The firm hoped for performance and rewarded attendance. It got attendance
.
Why this keeps happening: Kerr's four causes
- Fascination with objectivity. Measuring the wrong thing is preferred to exercising judgement about the right thing, because it looks defensible through the data,
- Overemphasis on the visible. The measurable crowds out the meaningful.
- Hypocrisy. In some cases the rewarder is actually getting what they want, despite what they say. The system isn't broken but the public articulation of what they want is.
- Equity over efficiency. Sometimes fairness, as an example factor, overrides pure incentive design.
Kerr notes that only the first two are genuine design failures. The other two are the system working as intended, if not as communicated.
What Kerr recommends
Kerr considers three remedies.
- Selection: hiring only people whose goals align with the organisation's, he dismisses as empirically weak.
- Training: trying to alter employees once hired, he's sceptical about.
- Altering the reward system itself: fix what is reinforced and you no longer need to rely on good intentions, moral fibre or lucky alignment of personal values.
He closes by quoting B.F. Skinner: where sufficient and right incentives for reinforcement exists, no one needs goodness.
"For an organisation to act upon its members, the formal reward system should positively reinforce desired behaviours, not constitute an obstacle to be overcome."
Why Kerr’s paper still matters in 2026
The mechanisms Kerr names: goal displacement, visibility bias, measurement fetishism, the gap between stated and operating intent, sit beneath a lot of the live discussions in reward right now:
• Short-term incentive design
• ESG metrics
• Executive pay structures rewarding what can be given a number rather than what is consequential i.e. less easy to defend?
• The persistent problem of rewarding individual outputs in work that is actually collective
None of these are new problems.
The most uncomfortable line in the paper isn't really about reward systems at all. It's the observation that if a system makes it irrational to be moral, immorality may not automatically follow, but you are, in his phrase, 'asking for trouble.' Some would argue that fifty years of reward practice has done a reasonable amount to confirm his view.
Read the paper
Kerr's paper is available to download below but is also freely available online. Search 'Kerr 1975 Folly of Rewarding A' or find it via the Academy of Management Journal archive. It's been a much cited paper in management literature.