What Markets Know: Prediction Markets

How markets where people bet on future events can combine scattered information into forecasts, and where they fall short.

If prices pull together dispersed knowledge, could a market be designed specifically to forecast the future? That is the idea behind prediction markets, where people buy and sell contracts that pay out depending on whether an event happens.

How they work

A typical contract pays 1 dollar if an event occurs, such as a particular candidate winning an election, and nothing if it does not. If the contract trades at 60 cents, the market is roughly saying the event has a 60 percent chance of happening. People who think the true chance is higher buy; people who think it is lower sell. The price moves as new information arrives.

Why they can work

Prediction markets draw on the wisdom of crowds: the idea that combining many independent judgements can produce accurate estimates, because individual errors cancel out. They also reward people who have good information and punish those who do not, through profits and losses.

The Iowa Electronic Markets, run by the University of Iowa since 1988 for research and teaching, allowed small-stakes trading on U.S. election outcomes. Studies found its forecasts were often at least as accurate as opinion polls. Some companies have run internal prediction markets to forecast things like product launch dates.

A company's launch date

A company asks employees to trade contracts on whether a product will launch on time. Managers are officially confident, but engineers know about hidden delays. As engineers sell the "on time" contract, its price falls, giving managers an early warning that official reports did not.

Limits

Prediction markets are not magic. They can be inaccurate when few people trade, when traders share the same biases, or when someone tries to move the price to influence opinion. They work best for clearly defined events where information is widely spread. In many countries, prediction markets on real-money outcomes face legal restrictions because of gambling laws, and regulators have debated how to treat them.

Thinking a 70 percent forecast that misses was wrong

If a market says an event has a 70 percent chance and it does not happen, the market was not necessarily wrong. Events with a 70 percent chance should fail about three times in ten. Forecasts are judged over many predictions, not one.

Key takeaways

Prediction markets let people trade contracts that pay out if an event happens.
Prices can be read as probabilities, updated as information arrives.
They draw on the wisdom of crowds and reward good information.
They can fail with few traders, shared biases or manipulation, and face legal limits in many places.