Photo-Illustration: Intelligencer; Photos: Getty

For political junkies, the midterms have been an unusually stressful roller coaster. Pollsters forecasted a Republican triumph before the summer’s Dobbs decision gave Democrats the opportunity to defy history and pick up seats in an election in which high inflation would normally cause swing voters to rush away from the incumbent’s party. Since Labor Day, it seems the momentum is back on the Republicans’ side. Now imagine following all of this with a decent chunk of money on the line.

That’s the draw of a site called PredictIt, where users like Jon Kimball can place live bets on the results of elections. After betting big on the past two presidential elections — he lost his shirt on Hillary Clinton but said he made more than $10,000 betting on Joe Biden — he’s playing it relatively safe this year. “I don’t have a great feeling about this,” said Kimball, who hosts a gambling podcast called Election Profit Makers and buys domain names for a living. But he has gone in on one long shot: Democrat Tim Ryan beating Republican J.D. Vance. Despite polling averages suggesting a close race in Ohio, PredictIt has Vance as the nearly five-to-one favorite.

Here’s how it works for those who haven’t dabbled in America’s new favorite pastime of gambling. Unlike in sports betting, there is no house determining the odds; user demand sets the market price. In Ohio, for example, as more people wager that Vance will win, the price of his stock goes up, making it cheaper to buy a position on a Ryan victory. With generic polling favoring the GOP in the final days of the race, Vance’s stock has soared to 85 percent, meaning that if one were to buy 100 shares at 85 cents a share, they would make $14 in profit if he wins, minus fees. “I don’t think J.D. Vance has an 80 percent chance of winning, so Ryan dropping down to 20 cents — that’s a good place to buy there, a really good return if he’s able to pull it off,” said Kimball. He stands to win over $450 on a $83 bet if Ryan can beat the odds.

PredictIt has offered this service to American users since 2014 as an “academic research” experiment from the Victoria University of Wellington in New Zealand with the political-data firm Aristotle serving as its service provider. The site was relatively quiet until the 2016 election, when up to 17,000 users were joining each month. The company said it now has 177,000 users (including the account I opened to get a sense of how it all works). “It’s really fun,” said Aristotle CEO John Aristotle Phillips. “We underestimated how completely engaging this is.”

But soon it could all be over. In August, the Commodity Futures Trading Commission ordered PredictIt to close in February 2023, claiming that the site did not adhere to rules laid down eight years ago. The CFTC did not provide specifics in its public announcement, and Phillips said the regulator gave the company no additional information. “We know nothing more than what was put out there,” he said. Since the order, PredictIt has stopped adding new contracts and has not yet determined how markets for 2024 will be resolved if the CFTC order is held up.

To keep the site alive, PredictIt has filed a motion for a preliminary injunction in a U.S. district court in Texas to stop the order to liquidate all its markets next year. But many users are already seeing this year’s midterms as the last hurrah for the site that turned legal gambling on elections into a small phenomenon throughout the political world. “If it goes away, I’ll miss it, but I don’t know if this lifestyle is sustainable, trading all of the time,” said Kimball.

Aside from the money available to win (or lose), PredictIt has been controversial as a potential indicator in an age in which polling sucks, helping to show which candidates people believe in enough to put real money on. Although a majority of users were off on Brexit and had favored Clinton in 2016 by three-to-one odds, the markets successfully forecasted Trump wins in Florida, Texas, Ohio, and North Carolina in 2020 — states that polling suggested had a reasonable chance of going for Biden. Academic researchers have found that there is value in comparing polling with what people are willing to wager in the markets, a prospect FiveThirtyEight founder Nate Silver has called “dumb.” Economist Rajiv Sethi has even suggested that a market that gave Trump a one-in-eight chance of winning the 2020 election nearly a month after it was called was an “advance warning” of the January 6 insurrection.

“I don’t think it’s an overstatement to say it’s an antidote to fake news,” Phillips said. “If you base your positions — your buying and selling — on fake news, you’re going to lose your money.”

PredictIt users have made some impressive calls — like New York City’s Democratic mayoral primary last year, when the Board of Elections accidentally included 135,000 test ballots in the results, skewing the ranked-choice results in favor of Kathryn Garcia. Users were “immediately dismissed” when they brought the issue to the attention of the BOE, said PredictIt’s assistant director, Ethan Rosen. In the few hours before it was corrected, users who caught the error cashed in on the city’s mistake by shorting Garcia’s stock.

As for this year, PredictIt users are much more confident in the odds of a GOP-controlled Congress than polling might suggest. Two weeks out from the election, FiveThirtyEight projects an 81 percent likelihood of a Republican House — similar to the 88 percent shot given in the PredictIt market. But in the Senate, FiveThirtyEight suggests Democrats have a 55 percent chance of winning compared with the 65 percent chance PredictIt users have given Republicans. In Pennsylvania’s senate race, polling averages give a 2.6 point advantage to Democrat John Fetterman, whereas PredictIt had Republican Mehmet Oz up by three points before the October 25 debate. But after the contest showed that Fetterman was still recovering from a stroke, Oz soared to a nearly three-to-one advantage. “Traders are forecasting that there are going to be a lot of ticket splitters in this election, which is somewhat unusual given recent trends in the United States,” said Rosen.

It will take a few weeks to see how close the markets are to the final results, but in the interim, the smart money on the site is ready to make a run. “PredictIt is a platform where you’re betting against people who have strong political opinions, and they’re driving the odds, and you can bet against that,” said a user named JMark. “Many times, their feelings don’t match reality.” Since opening an account close to seven years ago, he claims to have made $50,000 annually on the site, where individual bets max out at $850. “It’s been nice to have.”

Paul, a Ph.D. student with a keen knowledge of statistics who declined to give his last name, said he has made guaranteed returns in markets where you can bet on more than one result, like a primary election. “Say there are ten candidates for Democratic primary,” he explained. “If you bet on all ten of them, all but one of them has to at least win. If you bet no on everyone, at least nine are gonna hit. So if the noes are cheap, you are guaranteed to make money.”

Successful users are lamenting the loss of this easy money at the expense of the masses. “If it does disappear, yeah, I’ll be sad,” said Kimball. “It’s the end of an era. It’s not the end of the era for prediction markets because it’s obvious that the CFTC is willing to work with others that follow certain guidelines that PredictIt must not have.” He’s referring to Kalshi, the CFTC-approved exchange that is the new home for many PredictIt users, where you can bet on everything from the midterms (58 percent favor Fetterman) to the number of TSA check-ins this week (the over-under is at 2.25 million).

And while legal and offshore options will be there, some traders are still holding out the possibility that PredictIt will stay in the game. “There’s another prediction market saying PredictIt has a 25 percent chance to be alive in March,” said Paul.