The Cobra Effect
A reader's summary of the idea that an incentive can make the very problem it targets worse: the cobra story and its doubtful pedigree, the well-documented rats of colonial Hanoi, and what both say about rewards that people can manufacture.
The term at a glance
The cobra effect is a perverse incentive: a reward offered to reduce something that instead encourages people to produce more of it. The name comes from a story about venomous snakes in British India. It is the standard shorthand, in economics and policy writing, for a fix whose side effects outweigh the problem, and it turns up in online arguments every time someone proposes to pay for, or ban, a behaviour and a reply points out that people will adapt.
Origin
The tale goes that the colonial government in Delhi, worried about the number of cobras, offered a bounty for every dead one. Residents began breeding cobras to collect it. When officials noticed and cancelled the scheme, the breeders released their snakes, and the wild population rose. The German economist Horst Siebert gave the episode its name in his 2001 book Der Kobra-Effekt, and it spread through popular economics from there.
The pedigree is weak. Researchers who looked for records of the Delhi scheme, including the team behind a 2012 episode of the Freakonomics radio programme, found no document confirming it, and the cobra story is best read as a parable that may never have happened as told.
History and context
A better-documented case comes from Hanoi in 1902, under French colonial rule. Faced with rats in the new sewer system, the authorities paid a bounty on each rat, with a severed tail as proof of death. Officials soon saw tailless rats running around, and reports of rat farms. The historian Michael Vann reconstructed the episode in a 2003 article in French Colonial History, and the scheme was abandoned.
The wider idea is older than either story. In 1936 the sociologist Robert K. Merton published “The Unanticipated Consequences of Purposive Social Action,” arguing that deliberate action regularly produces results nobody intended, from ignorance, error or the way people respond to the action itself. A modern corporate example is the 2016 settlement in which Wells Fargo paid $185 million over accounts opened without customers' consent, in a bank where staff sales targets rewarded the number of accounts opened.
Main ideas
A reward can pay for the wrong thing
An incentive attaches money or status to a proxy, such as a dead cobra or a rat tail, rather than to the goal itself, which is fewer cobras or fewer rats. Once the proxy has a price, producing the proxy becomes a business of its own.
People respond to the rule, not the intention
Whoever designs the scheme pictures a hunter. People who read the same rule see a market: breeding cobras is cheaper than finding them. Nothing needs to be dishonest for this to happen; it is simply the cheapest route to the reward.
Ending the scheme can make things worse
In the version of the story usually told, the government cancelled the bounty and the breeders released their now worthless stock, leaving more cobras than before. A policy that is withdrawn leaves behind the behaviour it created.
A cousin of Goodhart's law
Where Goodhart's law says a measure stops being good once it becomes a target, the cobra effect adds the visible damage: the target is not only gamed, the problem it was meant to shrink grows.
The feed is a bounty system
Platforms that pay for views, shares or replies put a price on attention. Creators answer by producing what attention is cheapest to buy with, such as clickbait, rage-bait and karma farming.
Critique
- The signature story is unverified. The tidy cobra anecdote is repeated far more often than it is sourced, which makes it an example of the thing it describes: a good story that spreads because it is satisfying.
- Incentives often work. Bounties, subsidies and fines do change behaviour in the intended direction in many cases. The cobra effect describes a failure of design, not a law that rewards always backfire.
- It can be used to dismiss any intervention. “It will just create a cobra effect” is cheap to say and hard to refute. A fair use of the idea asks what, specifically, people would do to collect the reward and what it would cost them.
Impact
The cobra effect gives designers a question to ask before launching a reward: how would I collect this if I did not care about the goal? For online platforms the answer is rarely hypothetical, because the reward is paid in public and the strategies are copied within days. The practical lesson is to reward the outcome you want, and to treat any cheaper way of triggering the reward as a design flaw to expect, not as bad behaviour to be surprised by.