By Jason Hreha

Glossary

Perverse Incentives

Reviewed by Jason Hreha

Published 5 min read

In 1902, during the third global plague pandemic, the French authorities in colonial Hanoi wanted the city’s rats gone. They began paying residents for every rat tail handed in. The tails arrived by the thousands. The rats did not go away, because some residents had worked out that a rat was worth more alive.

What are perverse incentives?

Perverse incentives are rewards or penalties that encourage the opposite of what their designers intend. In the clearest cases, the incentive pays people to make the underlying problem worse, as the Hanoi bounty, in effect, paid people to raise more rats.

The term is narrower than a failed incentive. An incentive can fail by doing nothing. A perverse incentive does something, and what it does works against the goal. The organizational researcher Steven Kerr described the pattern in his 1975 paper “On the Folly of Rewarding A, While Hoping for B”. In such reward systems, the rewarded behavior is the one the rewarder is trying to discourage, and the desired behavior goes unrewarded.

Perverse incentive effects are sometimes called the cobra effect, after a story about colonial India. That story is better known than it is documented, so this entry starts with the case that is documented.

The Hanoi rat bounty: a documented case

The Hanoi case rests on archival research. The historian Michael G. Vann found the records in the French colonial archives in Aix-en-Provence, published a journal article on the episode in 2003, and expanded it into a 2018 book, The Great Hanoi Rat Hunt.

Vann summarized the sequence in a 2020 interview about the book. A dossier of daily counts for two districts, from April to July 1902, records thousands of rats killed on many days and more than 20,000 on the worst. The city’s Vietnamese sewer workers refused the job of hunting rats underground and went on strike. Officials then offered residents a small bounty, a few cents per tail.

About three months in, officials discovered rat farming: people were cutting off tails, collecting the bounty, and setting the rats free to breed. A smuggling network was also bringing rats into the city from across Tonkin, the surrounding region of northern Vietnam. The authorities eventually turned to quarantines and other invasive public health measures instead.

The bounty paid for evidence of a dead rat, and a tail was that evidence. From a bounty hunter’s point of view, a living rat without a tail was worth more than a dead one, because it could keep breeding.

Why perverse incentives arise

The Hanoi bounty shows four conditions that let it backfire:

  • The reward pays for a proxy, not the goal. Officials wanted fewer rats. They paid for tails, which they could count.
  • The proxy can be produced without serving the goal. Breeding rats produced tails while adding rats.
  • The payer cannot easily see the underlying outcome. Officials could count tails at the police station. They could not count the city’s rats.
  • Producing the proxy is cheaper than the reward. Once that holds, manufacturing the proxy becomes a business.

These conditions connect perverse incentives to Goodhart’s law, the observation that a measure can stop being a good measure once people are rewarded for moving it. The most damaging Goodhart cases are perverse incentives: moving the measure makes the real problem worse.

The cobra effect: a story, not a record

In the usual telling, the British government in Delhi offered a bounty for dead cobras. People began breeding cobras to collect it. When officials discovered the breeding and cancelled the bounty, the breeders released their now worthless snakes, and Delhi ended up with more cobras than before.

The German economist Horst Siebert is usually credited with the name, from his 2001 book Der Kobra-Effekt, on avoiding mistakes in economic policy. The Delhi version, with its breeding farms, cancelled bounty, and mass release, is an anecdote. It is usually told without dates, officials, or sources.

Some of the surrounding history is documented. Governments in British India did pay rewards for killing venomous snakes, and rumors that people bred snakes for the rewards circulated at the time. In 1887 the Government of Bombay asked the Bombay Natural History Society whether snakes were being bred for the reward in two of its districts, Satara and Ratnagiri.

The Society’s secretary, H. M. Phipson, replied that this was “highly improbable”. He noted that the cobra had, to his knowledge, never been known to breed in confinement. He suggested the rumors probably came from people keeping snake eggs they had found until the eggs hatched, so they could claim the reward.

A 2025 review of colonial newspapers and archives by the Friends of Snakes Society, an Indian snake-rescue organization, found breeding allegations in the record but no documented breeding operation or prosecution.

Phipson’s reply also shows when a bounty on a proxy does not backfire. If the cobra could not be bred in captivity, the second and fourth conditions failed: there was no cheap way to manufacture dead cobras. Hatching eggs found in the wild did not add snakes to the countryside, and Phipson judged it a practice that “should be encouraged.” A proxy bounty turns perverse when the proxy can be produced more cheaply than by serving the goal.

Another documented example: carbon credits for a waste gas

Under the Kyoto Protocol’s Clean Development Mechanism, factories making the chemical HCFC-22 could earn emission credits for destroying HFC-23, a greenhouse gas created as a byproduct of that production. Lambert Schneider analyzed data from registered projects in 2011. Plants produced significantly less HFC-23 during periods when no credits could be claimed than during periods when they could. Schneider concluded that some claimed emission reductions may not have been real and that the credits created a perverse incentive to generate more of the gas.

The incentive’s logic follows the fourth condition. Emission credits rise with the amount of gas destroyed, up to a cap set from each plant’s past production, so a plant could earn more credits by generating more HFC-23 to destroy. Whenever the credits were worth more than the cost of producing the extra gas, making more waste paid.

  • Goodhart’s law describes how measures degrade under pressure. Not every Goodhart effect is perverse; a gamed measure may simply become less informative.
  • Motivational crowding out describes a different route to a backfire. In a well-known daycare study, a fine for late pickup was followed by more late pickups. One explanation is that the fine changed what lateness meant. No one was being paid to produce lateness, and the study could not settle the mechanism.

Checking a scheme for perverse incentives

Before launching a reward, bounty, target, or penalty, three questions help:

  1. Could someone earn the reward by creating more of the problem? If the reward is paid on evidence that people can manufacture, assume some will.
  2. What exactly is being paid for, and can the payer verify the real outcome? If only the proxy is visible, add an independent check on the goal itself.
  3. Is producing the proxy cheaper than earning it honestly? The Hanoi tails were cheap to produce from living rats. Wild cobras were not cheap to breed.

The guide to employee incentives applies these questions to bonus and recognition schemes at work, with a worksheet.

References