Every policy is a push on a system, and systems push back in ways their designers did not plan. A rule meant to save fuel can raise fuel use; a reward meant to reduce a pest can multiply it. Studying these surprises is not a counsel of despair. It is a way of designing actions that work with the grain of human behavior rather than against it.

An old warning

In 1850 the French economist Frédéric Bastiat wrote an essay called "What Is Seen and What Is Not Seen." His point, in paraphrase, was that every action has an immediate, visible effect and a train of later, hidden ones, and that a good judge of policy weighs the unseen effects as well as the visible ones. A public works project shows its new bridge; it does not show the things the taxpayers would have bought with the money.

In 1936 the sociologist Robert Merton gave the problem its modern form in an essay on the unanticipated consequences of purposive social action. Besides ignorance and error, he named a source he called the imperious immediacy of interest, by which he meant that a focus on immediate gains can blind people to later effects they would otherwise have seen.

Perverse incentives

The sharpest surprises come from rewards. A perverse incentive is a reward that ends up encouraging the very behavior it was meant to reduce. The best-known story is the so-called cobra effect, a name popularized by the German economist Horst Siebert in 2001. In the tale, colonial officials in Delhi paid a bounty for dead cobras, people began breeding cobras for the money, and when the bounty was scrapped the breeders released their snakes, leaving more cobras than before. The story makes the point well, but the Delhi cobra tale is widely told and poorly documented, and historians have not found records confirming it. A better documented case comes from Hanoi, described in the worked example below.

Risk compensation

People also adjust to protections. Risk compensation is the tendency of people to take more risks when they feel safer. In 1975 the economist Sam Peltzman argued that American car safety rules had saved fewer lives than expected because drivers, feeling protected, drove a little faster and less carefully, putting pedestrians and cyclists at greater risk.

The evidence deserves fair treatment. Many researchers think Peltzman overstated the effect, and later studies found that seat belts and airbags save large numbers of lives overall. Risk compensation usually offsets part of a safety gain, not all of it. The lesson is not that safety measures are pointless but that their designers should expect people to respond to them.

When people route around rules

In 1989 Mexico City, choking on smog, introduced Hoy No Circula, "today it doesn't drive," which banned each car from the roads one weekday according to the last digit of its license plate. It seemed sure to cut traffic by a fifth. In 2008 the economist Lucas Davis studied air quality data from before and after the rule. Davis found no measurable improvement in air quality, partly because many households bought an extra car, often an older and dirtier one, so they could drive every day.

And then what?

The simplest defense against such surprises is a habit of mind. Before backing a policy, ask "and then what?" to trace effects beyond the first one. If we pay per rat, what will people do? If we judge teachers on one test, what will they teach? This is sometimes called second-order thinking.

A caution about caution

The idea of unintended consequences can itself be misused. In his 1991 book The Rhetoric of Reaction, the economist Albert Hirschman showed that opponents of reform, from the French Revolution to the welfare state, have used the same few arguments again and again. One he called the perversity thesis: the claim that any attempt at reform will backfire and make things worse. Sometimes that is true. Often it is a reflex used to block change that would work. A systems thinker asks for evidence in both directions: unintended consequences are possible, not guaranteed, and some of them turn out to be good.

Worked example: the rats of Hanoi

Around 1900 the French colonial government in Hanoi built a modern sewer system for the European quarter. The sewers turned out to be ideal homes for rats, and in 1902, with plague spreading in Asia, the authorities hired rat catchers and then offered the public a bounty for each rat killed. To spare officials from handling carcasses, proof was a severed tail. The historian Michael Vann, who studied the colonial records, found that officials soon noticed tailless rats running through the city. People were catching rats, cutting off tails, and releasing the animals to breed more bounty. There were reports of rat farming on the city's edges. The program was abandoned.

Apply the ideas. The visible effect was a growing pile of tails; the unseen effect was a stable or growing rat population. The reward was perverse: it paid for tails, not for fewer rats. An "and then what?" question at the start, what will people do when a tail is worth money, would have pointed to the flaw. The deeper lesson is that the rat catchers were not villains. They responded sensibly to the system the rule had created.