In 1980, John Aristotle Phillips declared his candidacy for the United States Congress, in a district in Connecticut. He would turn twenty-five, the minimum age to hold the seat, a few months before the election. Four years earlier, while a junior at Princeton, Phillips—the son of a Yale engineering professor—had become an unlikely media sensation, thanks to a paper he wrote on how to build an atomic bomb. The F.B.I. confiscated it; after a Pakistani official asked Phillips for a copy, he alerted a senator he admired, William Proxmire of Wisconsin, who later referenced the incident in a speech. “John Phillips is an astonishing young man,” Proxmire said, after Phillips announced his candidacy; President Jimmy Carter and Senator Ted Kennedy, then challenging Carter in the Democratic Presidential primary, reportedly sought Phillips’s endorsement. In November, Phillips lost to his opponent, a four-term Republican incumbent, by twenty-five points. He ran again two years later, and lost by double digits once more.
Phillips had miscalculated his prospects. But that failure helped him to detect an unmet need: predictive political data. In 1983, he and his M.I.T.-graduate brother, Dean, founded Aristotle Industries, which helped introduce computer software to politics. Their début program, Campaign Manager, analyzed polls, campaign contributions, and fund-raising strategies. Within a year, nearly two hundred members of Congress were customers, including the Republican who’d drubbed Phillips twice. Aristotle moved from South Norwalk, Connecticut, to a block away from the U.S. Capitol. “Democracy is a growth business,” the company’s chief executive said.
Thirty years later, Aristotle was hired to help with what Phillips calls “the most interesting project I’ve ever been involved with.” Economists at Victoria University of Wellington, in New Zealand, were studying markets that allowed traders to buy and sell futures contracts tied to political events, and they created an experimental exchange called iPredict. But New Zealand politics attracted a fairly small number of traders, so, in 2014, they requested permission from the Commodity Futures Trading Commission, in the U.S., to operate markets related to American elections and economic indicators. A couple of years before, the C.F.T.C. had rejected a request from a registered U.S. exchange called Nadex to offer political-event contracts, ruling that such contracts met a federal definition of gambling: “the staking or risking by any person of something of value upon the outcome of a contest of others.” In 2013, the C.F.T.C. had effectively shut down Intrade, an Ireland-based exchange that had accepted millions of dollars in bets on U.S. elections.
But there was a small, noncommercial exception to this rule: the University of Iowa, with permission from the C.F.T.C., has overseen the Iowa Electronic Markets—which accept trades from students and faculty on participating campuses—since 1993. Victoria University proposed expanding on the Iowa model by opening its new exchange to the public, conducting limited advertising, and allowing people to bet as much as eight hundred and fifty dollars on any given contract offered on the exchange, an increase from the Iowa project’s limit of five hundred. The C.F.T.C. replied to the proposal with a so-called no-action letter, indicating that “an academic exercise demonstrating the information gathering and predictive capabilities of markets” would be allowed, so long as Victoria didn’t profit from it. Victoria outsourced management of the new online exchange to Aristotle. The company dubbed it PredictIt.
PredictIt now has around eighty thousand active users. Ahead of the 2020 elections, nearly a hundred and fifty million dollars flowed through its markets. It has shared market data with dozens of researchers around the world, and sometimes created markets at the request of scholars. But, behind the scenes, Aristotle has been at odds with regulators over its desire to keep expanding. The C.F.T.C. has held that election markets are akin to gaming and that they are against the public interest, because traders might be tempted to base voting decisions on which candidate will make them more money. (One might argue that voters do this already.) Phillips insists that the exchange has positive civic effects, pointing, for comparison, to the obsessive engagement of sports gamblers, who, according to one study, watch about twice as many games as non-bettors do. “If it takes people having a little bit of skin in the game who don’t normally pay attention to something and now they’re reading the newspaper more, or focussing more on absorbing and critically analyzing what they’re seeing on TV, that’s great,” he told me. Critics, of course, say that this amounts to treating politics as a game. One dedicated PredictIt user, Trevor Boeckmann, a thirty-four-year-old public defender in New York, told me that his pastime has made him “a better-informed citizen.” Boeckmann, who grew up in Iowa, has always been interested in politics, but researching trades has made him fluent in Senate cloture rules and other political minutiae, he said.
In 2019, Phillips wrote to the C.F.T.C., arguing that election markets collect information “for the benefit of both participants and non-participants,” and pleading for regulators not only to define the “rules of the road” but also to allow more than “small-scale, academic initiatives.” He never received a response. Instead, this past August, the C.F.T.C. abruptly withdrew its no-action letter, announcing that Victoria had violated the agreement’s terms. The commission didn’t specify which terms had been violated; Bloomberg, citing multiple anonymous sources, reported that C.F.T.C. staff believed that the agency “had allowed PredictIt to stray too far from the original remit of being a small-scale market designed to facilitate academic research.” (The C.T.F.C. declined multiple requests for comment for this story.) It ordered PredictIt to close existing markets by mid-February of next year. Richard Shilts, a former director of the C.F.T.C.’s Division of Market Oversight who now advises Aristotle, told me that he was surprised at the drastic step and by the absence of a specific stated reason. “They could have amended the no-action letter in some way” if they thought it needed revisiting, he said. Phillips told me that he was blindsided by the C.F.T.C.’s decision, calling that day in August “one of the worst days of my life.” He said that his company pleaded with the C.F.T.C. for guidance on how to close markets related to the 2024 Presidential race—roughly fifty million shares have been traded on those markets already.
Three weeks later, the C.F.T.C. made another surprising announcement: it would accept public comment on a request from a startup, Kalshi, to offer contracts on which party will control the House and Senate after the midterm elections. Kalshi’s chief executive, Tarek Mansour, has declared the opportunity “the holy grail of events trading.” Kalshi allows traders to invest up to twenty-five thousand dollars on a given contract, well beyond what PredictIt is allowed to accept. If its request were approved, Kalshi—which met three dozen times with regulators to discuss, among other things, whether “trading” is a euphemism for gaming and whether election markets are democratically redeeming—would become the first regulated U.S. exchange ever to host election markets.
Meanwhile, Aristotle, which supports Kalshi’s request, has sued the C.F.T.C. over the closure of PredictIt, with traders—including Boeckmann—and scholars joining as plaintiffs. “The Commission took this step with no reasoned explanation for its decision, no explication of facts that would support its decision, no transition plan for addressing scores of existing contracts held by more than ten thousand traders, and no consideration of any alternatives to the chaotic, disruptive, and economically damaging wind-down of the Market its decision forces,” the suit alleges. The C.F.T.C. has yet to respond to the specific allegations; Aristotle has asked for a preliminary injunction and expects a ruling on that request in the coming weeks. The lawsuit could help determine the future of political betting in the U.S.
It is perhaps not surprising to learn that there is a long tradition of betting on American politics. What may come as a surprise is that mainstream political reporting once devoted considerable space to that pastime. As early as the eighteen-sixties, American newspapers reported on betting markets pertaining to Presidential contests, covering them almost daily as elections neared. About half the trading took place in New York, outside Wall Street offices or on the steps of the stock exchanges. The economists Paul Rhode and Koleman Strumpf have found that such trading peaked in 1916, when President Woodrow Wilson defeated the former Supreme Court Justice Charles Evans Hughes; adjusted for inflation, about two hundred and seventy million dollars were wagered on the election in the organized New York markets that year, more than double what was spent on the Presidential campaigns themselves. (On some days, investors risked more money on politics than on stocks and bonds.) “The Wall Street betting favorite has always won a Presidential election,” the Times noted on Election Day after Hughes closed in the New York markets a 10–8 favorite, despite a rush of late money for Wilson. The oil tycoon Edward Doheny reportedly made up to half a million dollars by betting on Wilson’s reëlection.
These betting markets “vanished out of existence right around World War Two,” Strumpf told me. The disappearance coincided with the rise of scientific polling, which gained credibility in 1936, after George Gallup predicted Franklin Roosevelt’s reëlection, while the influential Literary Digest reader survey picked his opponent, Alf Landon, to win. It wasn’t that polls did better than the markets, which also favored Roosevelt. But a number of Americans disapproved of all forms of gambling, and polls now appeared to offer an alternative predictive measurement. A crackdown on sports gambling likely helped drive political-betting markets underground, as did passage of the 1936 Commodity Exchange Act, which established regulations for futures trading.
In many other countries, political betting continued; Andrew Leigh, an economist and a member of Australia’s Parliament since 2010, told me, “I always look at prediction markets instead of polls.” But such markets didn’t return to the U.S. in any legal form until 1993, when the University of Iowa began its small-scale experiment. Nowadays, despite the popularity of PredictIt and its strong track record—multiple studies have indicated that PredictIt forecasts election results better than polls do—journalists still seem skeptical of the markets. One Capitol Hill reporter I spoke to said that he doesn’t know a single reporter who follows them, and other journalists I talked with offered similar impressions. “The markets are more vibe-based,” David Weigel, a former political reporter at the Washington Post who now writes for the news startup Semafor, said. “They can be moved by conventional wisdom changing.”
Justin Wolfers, an economist at the University of Michigan who supported himself as an undergraduate in Australia by working for horse-racing bookies, believes that news outlets favor polls in part because their fluctuations keep audiences engaged. Wolfers and other market enthusiasts I spoke with pointed grumpily to the influence of FiveThirtyEight’s Nate Silver, who was hailed as “America’s secular god of predictions” after his poll-based forecasting system correctly called forty-nine states in the 2008 Presidential election. In Silver’s book “The Signal and the Noise,” he describes meeting with Wolfers and a Wolfers protégé, David Rothschild, who had compared prices on Intrade to Silver’s 2008 projections. After adjusting the market prices for known inefficiencies, such as the longshot bias—in the same way polls correct for sampling imbalances—Intrade came out ahead. If markets are superior, Wolfers asked him, why bother publishing models? They’re “intellectually interesting,” Silver replies in his book, “and they help to produce traffic for my blog.”
Silver, who did not respond to multiple requests for comment, has written of being “increasingly skeptical about the wisdom of crowds,” mainly because efficient markets supposedly need independent traders, and today’s traders are in constant communication. But the same dynamics are true of sports gambling, and yet oddsmakers are famously hard to beat. (Research suggests that betting markets predict N.F.L. results more accurately than about ninety-nine per cent of individual handicappers do.) “The same overconfidence that leads people to bet on football also leads people to reject the advice that prediction markets give,” Wolfers told me. For people whose livelihood depends on perceived expertise, he went on, “it’s very hard” to concede that you can’t predict the future as consistently as “a number on a Web page.”
Earlier this year, Trevor Boeckmann, the PredictIt hobbyist, bought roughly twenty-six hundred “yes” shares, at six cents each, anticipating that Republicans would hold forty-nine Senate seats after the November midterms. Share prices on PredictIt range from a penny to ninety-nine cents and can be read as odds, meaning that Boeckmann bought in at a time when Republicans were given just a six-per-cent chance of coming away down a seat. But Boeckmann thought that the upcoming midterms might have a similar dynamic to those in 2018, when, as he put it, “there was a huge wave against Trump, but Republicans managed to hold on to the Senate because they were running in good spots against not-great candidates.” He figured Democrats would need to run the table in Arizona, Georgia, Nevada, and Pennsylvania, with Wisconsin possibly in play.
In June, the Democrats’ prospects improved after the Supreme Court overturned Roe v. Wade, and the “yes” price Boeckmann had bet on climbed to thirteen cents. A few weeks after the abortion earthquake, he checked FiveThirtyEight’s forecast and saw the odds of Democrats picking up a seat pegged at eleven per cent, suggesting that PredictIt traders were now, perhaps, slightly overvaluing Democrats’ chances. Boeckmann sold high, bringing his year-to-date earnings on the site to about three grand. (Last year, he made a little under fifteen thousand dollars on the site.) He timed his trading on the Senate markets deftly; PredictIt traders are once again bullish on Republicans regaining the majority.
With PredictIt’s cap on investments, plus its five-per-cent withdrawal fee and ten-per-cent fee on profits, the puny payoff from betting on shoo-ins isn’t worth the trouble for many traders. As a result, “yes” bets on sure-thing candidates are often undervalued; J. B. Pritzker, for instance, is given a ninety-four-per-cent chance of being reëlected governor of Illinois, but most observers believe that it’s even less of a tossup than that. (FiveThirtyEight projects that he has a greater than ninety-nine-per-cent chance of winning.) Boeckmann has a time-intensive strategy for taking advantage of this inefficiency: using as many credit cards as he can open, he has, in the course of his PredictIt career, deposited about seventy-nine thousand dollars on the exchange; his bets on sure things make a small profit, which he uses to pay his credit-card debts in time to hit signup-bonus thresholds, all while accumulating eye-popping credit-card points. After we spoke this summer, Boeckmann spent a week in Berlin, he said, with hotel and airfare paid for entirely by his PredictIt credit-card use.