The Uses and Abuses of Prediction Markets

Prediction markets allow people to bet money on political outcomes. Can they help us guess what will happen in the future? 

Tekst: Olof Fägerstam
Illustration: Thea Montevere Haakonsen

One of humanity’s greatest problems is that we are exceptionally bad at predicting the future. Tens of thousands of analysts work full-time at understanding the world, yet most major political events come as complete surprises. The fall of the Soviet Union, the 2008 financial crisis and the election of Donald Trump are just a few events that almost none of the people who were supposed to be “in the know” managed to foresee. This leads to a huge waste of resources, as we worry about developments that never materialize only to be taken by surprise by disasters we failed to predict. 

Is there a way to improve humanity’s forecasting skills? This is what prediction markets set out to do. A prediction market is basically a stock market for political events. In a stock market you buy shares in companies you think will do well in the future and sell shares in companies you think will do worse. In a prediction market you buy shares in events you believe will happen and sell shares in events you think will not happen. 

Betting on Donald 

This sounds more complicated than it is, so here comes an example. Imagine a prediction market on the statement “Donald Trump will be re-elected in 2020”. A set of a thousand contracts is created with a payout of $1 each. This means that if Trump wins the 2020 election, traders will be paid $1 per contract, while if Trump loses the contracts will be worthless. At what price should you trade such a contract?  

That depends on how likely you think it is that Donald Trump will win. If you’re only 50% certain that Trump will win, you are 50% sure that the contract will pay one dollar, so you’ll be willing to pay $0.50 for it. On the other hand, someone who is 90% certain that Trump will win is going to value the contract at $0.90. She will be willing to buy shares priced up to $0.90, while you’ll be happy to sell your shares at any price above $0.50. The price of the share will move up and down as traders change their expectations on Trump’s re-election, just like the value of a stock moves up and down depending on investors’ expectations on a company. 

“There are three major advantages of prediction markets. They aggregate the wisdom of the crowds, force us to pay for our beliefs and train us to become better forecasters.” 

The Power of Money 

There are three major advantages of prediction markets. They aggregate the wisdom of the crowds, force us to pay for our beliefs and train us to become better forecasters. The ‘wisdom of the crowds’ argument says that a group of people usually makes better predictions than any single individual within the group. If the price of our imagined contract settles at $0.67, the crowd believes Donald Trump has 67% chance of being re-elected. Prediction markets consistently beat both polls and analysts at producing forecasts. For example, researchers Sveinung Jakobsen and Ole Jakob Bergfjord have shown that prediction markets outperformed polls in predicting both the 2008 and 2012 US presidential elections. 

Second, prediction markets force people to put their money where their mouth is. Today’s experts are rewarded for making predictions that sound interesting and convincing, while accuracy takes a distant third place. As the American political scientist Philip E. Tetlock shows in his book Expert Political Judgement, most of the “experts” we see on TV perform just as well as a random coin flip. In fact, Tetlock found that the more famous an expert is, the worse they perform on prediction tests. 

Experts are in the business of giving recommendations, meaning that other people are taking on the risk of trying out the experts’ advice. If the advice turns out to be wrong, it doesn’t matter; the people who listened pay the price while the experts continue to advertise their ideas. Moving risk over to experts will make them more careful about what they say. They’ll have to show that they’re willing to take on part of the risk by buying contracts on their predictions. The observant reader of course recognizes that – ironically – this article suffers from the same problem. It’s argued that prediction markets are useful, but the writer will not pay any price if he turns out to be wrong. 

Third, prediction markets allow researchers to find the best forecasters and learn from them how to make better predictions. Philip E. Tetlock happens to run The Good Judgement Project, a website based on prediction markets where users compete for virtual score instead of money. He has found that some users consistently make more reliable forecasts. Tetlock labels these people superforecasters. According to Tetlock, a team of superforecasters performs about 30% better than intelligence analysts who make use of classified information.  Tetlock’s superforecasters are not exceptionally intelligent or educated; they are normal people who have found reliable ways to overcome their own biases. If we can teach these methods to politicians and other decision makers, they will be able to get more reliable results from their policies. 

Ethical issues 

There is no doubt that prediction markets are instrumentally useful. However, some ethical arguments have been raised against them. A common objection is that betting on political outcomes may lead people to interfere in the political process, just like sports betting encourages bettors to influence the outcome of the game. For example, betting on a politician’s death may encourage people to assassinate her, or betting on when a subway station will be finished could incentivise policymakers to delay its opening. 

The issue of unethical behaviour requires more research, but clearly some events should remain outside the purview of prediction markets. It is also questionable whether knowing when certain people will die provides much value to society. On the other hand, elections provide good material for prediction markets. There are already such strong incentives at play, and so much money being spent on campaigns, that the addition of prediction markets will not make a noticeable change. Another important use could be to forecast whether mass killings will unfold in unstable regions. This will allow governments and human rights groups to respond more quickly and accurately to future atrocities. 

Are insiders a problem? 

Insider trading is another potential risk. Some questions have a strong component of information asymmetry, for example “Will Audi close down any factories within the coming year?” Since the top managers at Audi will have more accurate information on how talks are going, they may buy up stocks ahead of announcing their decision. On the other hand, this sort of trading becomes a form of releasing insider information to the public. If the price of the contract is unexpectedly high, this should alert the public that Audi managers are planning a shutdown. Speculators will also learn from experience not to invest in contracts that are prone to insider trades. 

These objections show that the side effects of political prediction markets are, ironically, quite unpredictable. Prediction markets are currently regarded as a form of gambling and are therefore illegal or highly regulated in most parts of the world. Deregulation should be approached with restrictions on the sort of questions that people can bet on. On the other hand, we should be willing to experiment and take some risks in finding out which areas to open up for prediction markets. The value of better forecasts can hardly be overstated; we will save resources by predicting political developments, save lives by predicting disasters and save our sanity by ignoring overconfident “experts”. 

Could you be the next superforecaster? Good Judgement Open is a website which lets you make predictions, compete against other players and improve your forecasting skills. Check it out on:

Bookmark the permalink.