How prediction markets saw something the polls and pundits didn’t

A version of this story appeared in Business’ Nightcap newsletter. To get it in your inbox, sign up for free, here.

New York

  

In the days leading up to the US election, pollsters had the race deadlocked. The vote was essentially a coin flip.

But over on the betting platform Polymarket, the odds were much more solidly in former President Donald Trump’s favor. On Monday, Trump led Vice President Kamala Harris 58% to 42% a lead that, by Wednesday morning, proved to be a much more accurate reflection of reality.

For free market purists, the success of betting sites like Polymarket, Kalshi and PredictIt isn’t surprising at all. The basic theory behind prediction markets is that a lot of people with money on the line can better predict an outcome than any one expert. Even if those people are not well informed, collective wisdom emerges from everyone’s aversion to losing money.

“Financial markets are generally pretty efficient, and the evidence suggests that the same is true of prediction markets,” Eric Zitzewitz, an economics professor at Dartmouth, tells me. “There’s no virtue-signaling in an anonymous market when you’re betting.”

The way it works: You can login to just about any of these sites (technically US bettors aren’t allowed to trade on crypto-powered Polymarket) and wager on the outcome of any event. The presidential election was a popular market, but there are plenty of others, like, will Taylor Swift and Travis Kelce break up in 2024? (A 12% chance, per Polymarket bettors.) Or, will America ban TikTok this year? (Highly unlikely, at 3%.) Who will win the Super Bowl? (The Chiefs are the current favorite, with a 19% chance, followed by the Lions at 17% and Ravens at 13%.)

You’re essentially buying a share tied to a particular outcome. The shares trade between $0 and $1, and once the event is resolved, shares tied to the correct outcome pay out a dollar. If you bought Trump shares on Monday, when they were 58 cents, you can expect to make 42 cents on the dollar.

And in the case of one Polymarket trader known as Fredi9999, who bought millions of Trump shares, the payday is expected to net out at $85 million, according to Bloomberg.

The emergence of that so-called Trump whale sparked intense scrutiny of Polymarket over the past several weeks, because it sure looked possible that an anonymous deep-pocketed investor was trying to manipulate the market and swing favorability in Trump’s direction. How accurate could this platform’s predictive power be if one zealot can go all-in and distort the odds?

Polymarket’s CEO, Shayne Coplan, addressed those concerns in an interview with CNBC on Thursday.

“If someone takes a really big position on Trump … there is someone on the other side, a counterparty,” taking a big position on Harris, Coplan said. “When you see the odds on Polymarket, it is not a function of how much money was put on either side, it is a function of market price at that moment. … Some trade that someone made two weeks ago doesn’t have bearing on what the market price is right now.”

Zitzewitz also noted that big price movements tend to get people’s attention.

“If prices really are getting distorted away from what they should be, then it’s profitable to take the other side,” he said. “Maybe I can’t take all of the other side of some guy with $20 million to bet, but me and 1,000 other people can.”

Historically, betting markets have been pretty good at predicting the outcome of US elections. One study showed that in the 15 elections between 1884 and 1940 the candidate with the best odds as of mid-October won 11 times. (The same study notes that election betting fell out of favor after 1940, as scientific polling grew more sophisticated and reliable.)

But just like polls, prediction markets are far from perfect.

In 2016, bookmakers told the British public that the smart money was overwhelmingly on the United Kingdom rejecting “Brexit” and remaining in the European Union. Later that fall, the popular betting market PredictIt gave Hillary Clinton an 82% chance of beating Donald Trump.

Turns out the smart money was wrong. The UK very much Brexited, and Clinton very much lost.

Those kinds of high-profile misses mean the Nate Silvers and Ann Selzers of the world probably won’t be out of a job anytime soon.

Still, I asked Zitzewitz what a market can tell me that, say, FiveThirtyEight’s polling average can’t.

Essentially, he told me, FiveThirtyEight reflects one person’s approach to turning data into a probability. Markets are looking at the same polls, and turning everyone’s interpretations into a probability.

“What you’re seeing with the market is some average of all of those different opinions, weighted by their willingness to put their money where their mouth is.”