Economists have long loved prediction markets. Even niche platforms such as the Iowa Electronic Markets and betting pools inside companies like Ford Motor Co. have been praised for their fortune-telling prowess.
And now, the online emporiums have entered the big leagues with the 2024 presidential election.
Electoral betting is legal for the first time in modern-day US on the platform Kalshi, and wagers worth hundreds of millions of dollars have piled up on it and Polymarket, a crypto-based offshore exchange.
In theory, this means these markets are more liquid than ever. And, therefore, the conclusion they’ve been broadcasting for the past month — that Donald Trump is, to one degree or another, the clear favorite over Kamala Harris — should be seen as more telling than a few notable misfires in recent years, such as the Brexit referendum and 2016’s presidential election. Again, in theory.
Yet even some supporters of these platforms fret that too much faith — among the Trump camp and up and down Wall Street — is being put in their prophetic virtues.
“I’m a bit worried,” said Rajiv Sethi, an economist at Barnard College, Columbia University, who supported Kalshi in its successful effort to overturn a ban on the use of derivatives to wager on the election.
As much as markets likes Polymarket have grown, he said, they’re still places where a handful of deep-pocketed investors can plunk down large bets and distort the odds — in Trump’s favor, in this case. “There’s no reason why somehow the price that emerges from their activity with each other is going to generate an accurate forecast,” said Sethi.
Over the last few days, Harris’s chances have risen in the betting markets as polls released over the weekend indicated she was gaining ground. Most notably she overtook Trump on PredictIt, an inherently limited market hosted by a university in New Zealand.
The prevailing wisdom on Wall Street is that prediction markets have an edge over polls because participants are economically motivated to incorporate every drip of new information faster.
Between a single forecasting model and the wisdom of a crowd that has digested all that information, “I’m going to go with that bigger aggregation of people,” said Eric Zitzewitz, an economics professor at Dartmouth College. “People have fairly strong incentives to do careful analysis.”
There’s been some evidence of that this year. Harris’s odds jumped after Biden’s disastrous debate when pollsters weren’t even paying attention to her potential candidacy. In recent weeks, poll-based forecast models like Nate Silver’s or FiveThirtyEight’s were mostly playing catch-up with Trump’s higher betting odds.
Harris’s gains over the weekend, however, are stirring a debate over whether the biggest betting sites were too skewed in Trump’s favor, either because of a herding mentality or even alleged manipulation.
Polymarket, in particular, has come under scrutiny. That’s in part because trading on the platform is done with cyptocurrencies, an industry for which Trump has voiced support. Also, people in the US are forbidden from trading on Polymarket, though some Americans have found ways to use the platform.
Two weeks ago, Polymarket revealed that a French national spent more than $45 million on bets that Trump will win, pushing the odds of a victory by the former president as high as 66%. Crypto researchers at Chaos Labs have also found other alleged flaws, including that the platform’s election market shows signs of so-called wash trading, which is when users buy and sell shares rapidly to create a false impression of heavy volume and deep liquidity in the market.
“As the researchers themselves note, a single trader taking positions on both sides of a market is hardly unique to Polymarket and not in and of itself problematic,” a representative for Polymarket said by email. “What makes Polymarket unique is that unlike traditional financial markets, Polymarket makes all transactions on its platform transparent and publicly available, including to researchers. Polymarket’s Terms of Use expressly prohibit market manipulation. We strive to provide users with the fairest analysis possible and our transparency allows the market to decide.”
The sway of whale trades is nothing new on election markets. In 2012, a pro-Romney speculator famously buoyed the Republican’s odds on one site even as his momentum evidently faded.
Academic research has generally shown that even small markets wagering on everything from geopolitical events to the box office can be fairly accurate. For elections, however, there’s much dissent over whether prediction markets are better than models that use historical data to convert poll results into probabilities of the outcome.
A 2012 paper concluded that betting markets stopped having much added value after the introduction of polling. In the 2020 and 2022 elections, Sethi found that using model forecasts to bet in prediction markets was profitable, though he stresses those results were very close.
The scrutiny facing electoral betting now even raises the question of whether its odds have somehow become less insightful in a market that’s more active than ever.
Before Kalshi, electoral betting mostly took place on overseas platforms like the UK-based Betfair or academic venues like PredictIt, which limits the number of users on each contract to 5,000 and position sizes to $850.
Contrary to conventional thinking, PredictIt might be more reliable because it can’t be easily swayed by big bets, says Sethi. Harris’s odds on that platform on Friday exceeded Trump’s for the first time in about a month.
In the stock market, a big one-way wager can still move prices, but with enough trading that effect should fade if it’s not supported by fundamentals. Yet as anyone on Wall Street can attest to, that process can take time.
“It’s hard for me to know what to do with the information from the betting market,” said Andrew Gelman, a statistics and politics professor at Columbia who worked on The Economist’s election forecast model. “It would be naive for people to think that just because someone is putting money down that means they’re going to have more information. People make bad investments all the time.”
This article was provided by Bloomberg News.