What betting markets got right and wrong about Trump’s victory

THE RISE of political polling and subsequent crackdowns on sports gambling drove political betting markets in America underground after the second world war. They remained in the shadows until last month, when a federal court cleared Kalshi, a betting exchange, to offer political bets to Americans. More than $400m has since been wagered on the site. The sums bet globally on other prediction sites reportedly tot up to several billion dollars. How did the speculators fare in America’s election?

At first glance, rather well. Although Kalshi has been operational for only a month, two other prediction markets, Polymarket and PredictIt, were taking bets throughout the election campaign. In general, markets gave shorter odds to Donald Trump, who won comfortably, than would have been implied by the neck-and-neck polls or political models, like the one published by The Economist. On Polymarket Mr Trump’s chances of victory never dipped far below 50%, despite the euphoria that greeted Kamala Harris’s entry into the race, which saw her pull ahead in many swing-state polls. And even as most political models pegged the race as, in effect, tied on the eve of the election, punters on Kalshi still gave Mr Trump an edge.

Markets should be pretty accurate. When people have to wager real money they tend to reveal their true expectations, as they tend not to when they are simply asked. Aggregating bets should result in a “wisdom of the crowd” effect. Or perhaps speculators knew something others did not: that the polls were underestimating Mr Trump’s support by at least two or three points, as happened in 2016 and 2020.

Yet it would be easy to read too much into the betting markets’ apparent success. Speculators may simply have got lucky. Markets could have been biased towards Mr Trump, rather than perceptive. The fact that Polymarket is populated by crypto-heads probably skewed the betting Trumpwards. Kalshi is used as an exchange for hedging risk, which might affect its sample, too.

The most compelling evidence that betting markets do not know much more than pollsters and political modellers comes from the nature of Mr Trump’s win. At no point did betting markets even get close to divining that Mr Trump was on track to comfortably win the popular vote—an outcome easily inferred from expecting a two- or three-point polling error in his favour. Just before polling stations opened, bettors on Polymarket gave such an event just a 27% chance. Speculators are probably an astute bunch, on average. The fact that they are now allowed to bet on elections provides useful information. But in 2024 they were hardly savants. ■

For more expert analysis of the biggest stories in economics, finance and markets, sign up to Money Talks, our weekly subscriber-only newsletter.