Pattern/Systems Thinking/No. 0153
Cobra Effect
The cobra effect is a perverse incentive that makes producing a problem profitable, worsening what a reward was meant to solve. Popularized by economist Horst Siebert in 2001, the term draws on an unverified story about a cobra bounty in British India.
- Evidence
- Well established
- Read
- 6 min
- Links
- 7 connections
01You've seen this when…
- in life
You pay your child whenever they remind you to turn off an unused light. Soon they start leaving lights on.
- at work
Your team pays a bonus for every software bug fixed. Producing and repairing a stream of small defects earns developers more than preventing bugs.
- out in the world
A town pays residents for invasive plants delivered to a collection depot. An inspector finds a field where someone grows them for the bounty.
02The idea
The reward seems sensible: pay people to remove something you don’t want. But once removing it earns money, producing it can earn money too. The problem becomes a source of income, and the people responding to your policy may have a reason to keep it alive.
That is the cobra effect: an incentive changes behavior in a way that worsens the problem it was supposed to solve. The mistake is evaluating the reward against today’s behavior and overlooking the behavior that becomes worthwhile once the reward exists.
The name comes from a story about a bounty for dead cobras in British India. People supposedly began breeding cobras to collect it. The story’s historical accuracy is uncertain. It is a memorable illustration. Treating it as reliable evidence requires independent support.
The pattern is narrower than unintended consequences, which can arise without any reward. It also differs from Goodhart’s law. A target can corrupt a measure without worsening the underlying problem. Here, the incentive helps manufacture more of the thing you wanted less of.
03Why it happens
- The unwanted thing acquires a price. A pest, defect, or pile of waste was previously a nuisance. A bounty gives it cash value. Someone who can produce it cheaply may earn more than someone who merely finds it.
- Proof of activity replaces proof of improvement. Counting removals is easier than measuring the problem that remains. Administrators see a growing stack of receipts, while nobody checks whether the original problem is shrinking.
- People find routes the designer overlooked. The policymaker imagines an honest participant working within the intended situation. Participants can change that situation: import waste, split incidents, delay prevention, or manufacture eligible cases. Designing around these responses is the work of mechanism design.
- The reward can feed its own demand. More problems generate more paid work. Revenue can finance the production of still more problems. The intended corrective loop turns into reinforcing feedback.
Even honest participants can make the problem worse. A team can simply stop doing unpaid prevention and concentrate on paid repairs. Each local choice makes financial sense, while the overall system gets worse.
04A worked example
In 1902, French colonial authorities in Hanoi faced a rat problem. The city’s new sewer system provided shelter and routes for rats, and officials feared plague. Alongside organized rat killing, they offered a bounty to residents. Payment depended on handing over rat tails.
Historian Michael G. Vann reconstructed the campaign from colonial records. His account describes officials encountering rats without tails and discovering rat-breeding operations supplying the bounty system.
What it looks like Every delivered tail records another rat eliminated. More tails suggest a more successful campaign, and paying residents appears to extend the authorities’ reach.
What’s actually going on A delivered tail is evidence that a tail was removed. The rat may still be alive and able to reproduce. More fundamentally, the bounty creates a market for rats: breeding them can be easier than hunting them. The program pays for a flow of submissions in pursuit of its goal: reducing the living population.
What would have helped Officials could have checked rat populations independently of bounty claims and investigated where submissions came from while emphasizing sanitation and habitat removal. Requiring whole carcasses could address the tailless-rat loophole while leaving breeding possible. The deeper repair is to stop making the production of rats profitable.
This case has an archival basis. The cobra story is less securely documented.
05How to spot it
06What to do about it
- Price the unwanted response before launch. Calculate whether someone could profit by creating a claim. Include importing, relabeling, splitting, and manufacturing cases—not just obvious fraud.
- Make prevention pay. Where feasible, reward a verified reduction in the remaining problem, rather than the number of interventions. This aims for incentive compatibility: serving the goal should also serve the participant’s interests.
- Measure the problem separately from the payout. Use independent sampling, inspections, or outcome checks. A mountain of bounty claims cannot establish a declining pest population.
- Test on a limited scale. Start with a bounded pilot and predetermined warning signs. Investigate unusual increases in submissions before expanding the scheme.
- Ask participants how they would exploit it. Run a pre-mortem with people who know the work. Give them permission to identify profitable behavior that would defeat the policy.
- Plan the withdrawal. If the reward has created stocks, equipment, or dependent businesses, cancellation may create another shock. Arrange safe disposal and a transition without rewarding further production.
07When it isn’t a cobra effect
Determining whether a program created more problems requires evidence beyond a surge in claims. A new reward may reveal an existing backlog or bring previously hidden cases into view. Better detection can make the numbers look worse while conditions improve.
A cobra effect is a specific kind of incentive failure. Paying for repairs can be sensible when providers cannot manufacture demand, repairs are independently verified, and prevention still benefits them. The test is whether the incentive changes behavior so that more of the targeted problem exists than would otherwise exist.
Related concepts capture different failures. Moral hazard concerns taking greater risks when someone else bears the consequences. A cobra effect concerns a supposed remedy making the problem profitable. They can overlap, but neither implies the other.
08Roots
German economist Horst Siebert put a cobra in the title of his 2001 book about economic policy mistakes, Der Kobra-Effekt. The image made a policy-design problem easy to remember: officials offer money for dead snakes, and someone discovers that raising snakes is a business.
In the anecdote, British authorities in Delhi cancel the bounty after discovering the breeding. Breeders then release their now-worthless cobras, leaving the city with more snakes. That final twist gives the story its force. But a secure historical source for the episode has not been established, so the tale should remain labeled as an anecdote.
The pattern predates its name. Hanoi’s rat campaign supplies a firmer early example, tied to the ambitions of colonial modernization: new sewers beneath the city also offered rats a favorable habitat. Vann’s 2003 historical study brought the administrative records of that campaign to a wider audience. Siebert popularized a label for an incentive failure that was already known.
The idea has since traveled through economics, management, and environmental policy. Lambert Schneider’s 2011 analysis of credits for destroying HFC-23, an industrial waste gas, examined rules that encouraged producing more gas eligible for payment. Different setting, same design question: does paying to remove a problem also make producing it worthwhile?
09How solid is this?
Historical records and economic analyses document incentives that encourage producing the problem they target. The famous Delhi cobra story is unverified, and the label alone does not establish that a particular policy caused harm.
10Connections
- Countered byMechanism Design, Pre-Mortem, Incentive Compatibility
- Can follow from Goodhart’s Law
- Part of Moral Hazard, Unintended Consequences, Reinforcing Feedback
11Origin and sources
Popularized by economist Horst Siebert in Der Kobra-Effekt (2001). The name draws on an uncertain historical anecdote about a cobra bounty in British India; documented incentive failures predate the label.
- [1]Siebert, H. (2001). Der Kobra-Effekt: Wie man Irrwege der Wirtschaftspolitik vermeidet. Deutsche Verlags-Anstalt.
- [2]Vann, M. G. (2003). Of Rats, Rice, and Race: The Great Hanoi Rat Massacre, an Episode in French Colonial History. French Colonial History, 4, 191–203.
- [3]Schneider, L. (2011). Perverse incentives under the CDM: an evaluation of HFC-23 destruction projects. Climate Policy, 11(2), 851–864.
Suggest an edit· Updated 2026-10-02