The Cobra Effect: When Incentives Backfire
The cobra effect describes rewards that produce behavior opposite to their goal; the famous colonial-era story is often repeated, but its historical attribution is contested.
A city pays a contractor for each pothole repaired. If the contract rewards the count but not durability, the contractor may favor many quick fixes over a smaller number of lasting repairs. The incentive changes what work is attractive.
If a school rewards only the number of books students finish, some may choose very short books or rush through them. Reading counts can encourage practice, but comprehension and choice need attention too.
The cobra effect is a name for incentives that produce outcomes opposite to the intended goal. A familiar story says a colonial government paid for dead cobras, prompting people to breed them for the bounty, but the story’s historical sourcing is contested.
The mechanism does not depend on that anecdote being verified. When a reward attaches to an observable proxy, people may find new ways to produce the proxy without improving the underlying condition, especially when oversight is weak.
A related failure occurs when a rule shifts the problem rather than solving it. A ban on a visible nuisance may move it to a less monitored place, while a quota can encourage people to stop just short of a threshold rather than reduce harm.
Not every unintended consequence is a cobra effect. A policy may fail because its assumptions were wrong, resources were insufficient, or outside conditions changed. Calling every disappointing result perverse can obscure which mechanism actually caused it.
A repair contract could pay partly for fixes that remain intact after a defined period, with independent checks and a way to report repeat failures. That design better connects payment to lasting service, though inspections still need to be fair.
Before adding a reward, ask how people can respond strategically, what the measure omits, and who can verify outcomes. Consider whether the target is durable improvement or merely an easily counted action.
Design incentives around outcomes where possible, and pair them with checks that detect displacement or low-quality work. Feedback from the people affected can reveal harms that a dashboard or contract report misses.
When the consequences are uncertain, pilot the policy and make it reversible. Set a review date, monitor side effects, and create a route for reporting problems without requiring people to prove intent before they are heard.
A good incentive should not assume that participants are either saints or villains. People adapt to constraints, so a robust system makes the desired behavior feasible and makes gaming less rewarding than doing the underlying work well.
What unintended response would you test for before rewarding a behavior, and who is most likely to notice it first?