Thanks to the turbulence of the 2016 U.S. national election season, Americans are looking to pundits and pre-election polling in massive numbers for a glimpse of what the political future holds. There may be a less-known but more reliable way for you to find accurate forecasts of uncertain political outcomes, though: prediction markets, and especially prediction market aggregators like PredictWise. These exchanges can give you a clear, percentage-based idea of what's likely to happen in a given election or primary that benefits from some accuracy-enhancing structural dynamics. The above breakdown of Republican presidential candidates' chances of winning the nomination (copied March 22) is just one example.
Prediction markets, like certain financial markets, are essentially places where individuals can make bets and trade those bets with each other. They're unique, however, in that prediction market participants bet on events that happen in the real world — including political developments such as elections (as well as the outcomes of more obscure concerns, such as the gender of the next UN secretary general), as well as the outcomes of sports competitions, entertainment awards shows, and so forth. Some popular prediction markets including PredictIt, the Iowa Electronic Markets, and HyperMind.
PredictWise combines the predictions from these markets and a number of others, plus information from pollsters and professional oddsmakers (or "bookies"), to generate outcome likelihood figures for both political results and other subjects. This rich assortment of input data makes it a robust one-stop shop for prediction market-derived forecasts.
Prediction markets have a number of structural advantages that potentially make them more reliable than the traditional prognostication provided by expert commentators and public polling:
1. Most available information and expert opinions go into the mix.
It can be extremely difficult for lay people to know which pundit to trust in extremely complicated fields like politics, and it's truly impossible for any one person to keep track of all the information relevant to making predictions about such complex subjects — experts included.
Prediction markets handle this problem in a unique fashion: because they're comprised of many people who make their bets with a huge variety of expert opinions and hard data in mind, they essentially incorporate most of the available information on a given subject, weighted by how many people find that information credible and how willing those people are to bet on their beliefs.
This aggregative approach tends to produce more accurate results on average than any one expert or polling service can claim.
2. The cream of the predicting crop rises to the top.
Bettors in prediction markets face a strong incentive to get their predictions right. They earn money if they hit the mark, and they lose money if they mess up. (Contrast this dynamic with professional punditry, where pundits often keep their jobs and reputations even if they're wrong on a regular basis.)
The reverse is also true, in that prediction market bettors who mess up frequently lose money.
This incentive structure produces three effects that make for good forecasts overall:
It encourages accurate predictors to get more involved by rewarding them.
It encourages bad bettors to drop out of the game by costing them money.
It discourages bettors from placing bets based on non-evidence-based judgements (e.g. political preferences, social affiliation, or self-serving beliefs), because if they do, other bettors can profit off of them by betting against their inaccurate predictions.".
3. Cognitive biases come out in the wash.
Like all humans, prediction market actors suffer from cognitive biases — built-in reasoning quirks that can cause people to think and act irrationally. (We offer free mini-courses on some of the biases that can cause people to make bad predictions, such as the Sunk Cost Fallacy and the tendency to misinterpret evidence strength.)
Prediction markets can compensate for their participants' biases in two ways:
First, some unshared biases cancel each other out — for instance, the overconfident bettors in the market will be balanced by the overly cautious ones.
Second, the less biased actors in the market can make money by betting against the more biased folks. This means that less biased participants tend to get more involved, while more biased participants tend to bet less as they keep losing money, and eventually drop out.
While these advantages make prediction markets a powerful forecasting tool, they can still be wrong and are subject to some limitations.
For example, while the variety of people involved in prediction markets can help cancel out cognitive biases that aren't shared by everyone, they're still susceptible to more universal biases. This imperfection is one of the chief drivers of the unstable boom/bust cycle of the stock market, which is itself essentially a prediction market.
Prediction markets are also only capable of aggregating the information that its component humans have available; they can miss the mark completely if an unknown x-factor comes into play. And when prediction markets don't have enough participants making and trading predictions, they can become arbitrary and swing wildly based on who's betting at a given moment.
It's important to keep these shortcomings in mind when looking at prediction market forecasts — but it's also worth noting that other forecasting methods face issues that may be more damaging. And in spite of these problems, prediction markets remain one of the strongest methods available to us for getting a sense of where politics are likely to go this year.