The most cynical observers of the 2016 American elections are in it for the money—and many of them don’t even live in the United States.
Ten months before Election Day, members of the U.K.-based gambling site Betfair have spent over $4.4 million betting on the outcome of the U.S.’s primary and general elections.
It’s a market in which Vice President Joe Biden, who swore off a presidential run in October, currently has better odds to win the presidency than Chris Christie or Martin O’Malley. On some sites that accept bids on the 2016 election, Elizabeth Warren, Al Gore, and Michael Bloomberg currently have a far better shot at the presidency than a 509-to-1 bet on Ben Carson, the Republican polling at 11 percent in Iowa.
According to gamblers, betting markets are more accurate than polls for predicting elections. A majority of Betfair users correctly predicted the winners of the 2004, 2008, and 2012 general elections after the races were reduced to two candidates. Whereas polls fluctuate by week, by location, or by polling agency, betting markets tend to exhibit more stable trends.
But this election cycle has thrown even the most statistically-minded gamblers for a loop.
Today, a dollar bet on a Donald Trump primary victory will turn a profit of $1.52 on Betfair. His odds are ranked the best of any GOP candidate on the site.
Trump’s share price to win the entire election reached its all-time high Wednesday on PredictIt, a site that allows users to trade futures on candidates’ chances. Trump’s price sits tied with Bernie Sanders for second place at 27 cents, and 16 cents behind leader Hillary Clinton.
At 48 cents per share at press time, Trump holds a 23-cent lead over Marco Rubio for the Republican nomination. Trump is now surging on the site, up 11 cents week over week and 5 cents on Wednesday alone.
But while Trump’s poll numbers have charted sky-high since his summer campaign launch, the betting markets have been slow to invest. Earlier this summer, a dollar bet on Trump would earn a return of nearly $350 on Betfair, while first Jeb Bush then Marco Rubio led as the betting man’s favorite.
“For 30 or 40 weeks in a row, we saw Jeb Bush ahead, then Rubio pulled ahead of him. Around Dec. 1, trading on the site went up significantly and, recently, Trump pulled ahead,” said PredictIt spokesperson Brandi Travis. “It says that people didn’t really think he would get the nomination until around now.”
According to prediction markets experts, these late-in-the-game lead changes reflect unusual volatility in the betting market, one that suggests a deeper divide in GOP politics.
“I think that what’s different about this year is that you have people with strong but incompatible opinions,” says Rajiv Sethi, a professor at Barnard College who has published a study on patterns in political betting markets.
“There are people who have felt, people who have publicly expressed the view that Trump couldn’t possibly be the nominee. They felt very strongly about that and very comfortable betting against that outcome. And really that view has started to change. The rise of Trump in the prediction markets has been driven by the breakdown of the narrative that he couldn’t possibly win.”
For a market that prides itself on objectivity, political gambling has tripped over the same judgement as television hosts: that something would inevitably lead to Donald Trump’s downfall, be it his attacks on immigrants, his criticism of John McCain’s war record, his anti-Islam rhetoric, or that he once tried and failed to sell “extremely greasy” frozen steaks at the Sharper Image.
David Rothschild, an economist who runs prediction tracking site PredictWise, says the betting markets finally came to terms with Trump at the beginning of the new year.
“You really see this kind of crossing sometime around Jan. 1, where Trump started to really shoot up,” Rothschild says.
According to PredictIt’s internal data, Trump and Sanders both moved into 2nd and 3rd place respectively right around Christmas. Trump has enjoyed a consistent lead over Ted Cruz for the GOP nod since Jan. 6.
Rothschild links Trump’s upshot to his comeback after a brief polling drop in Iowa.
“In the polling, for a little bit, I think people saw it as, ‘It’s finally going to happen,’ the first time he lost after leading something—but then he bounced back in the polls in Iowa and continued to hold firm in New Hampshire,” he said. “I think in many senses, it’s just a matter of time slipping away.”
Trump’s new acceptance in betting markets also indicates a trend toward a broader acceptance of anti-establishment Republican candidates, Rothschild believes. He classifies the GOP race as not one election, but two “semifinal” battles between establishment candidates like Jeb Bush and Marco Rubio, and anti-establishment candidates like Trump and Ted Cruz.
“You saw a sense of stability in the sense that as Jeb Bush went down and Rubio went up. They kind of traded off that weight,” he says of earlier betting markets. “And similarly you’ll see now that as Trump has held on firm in the last few weeks against Cruz’s charge, you see them kind of trade off support in the market. Overall, you see the markets tend to give incrementally more weight to the non-establishment as the candidates have progressed into the year 2016.”
As time dwindles in the primary competition, a bet against Trump is becoming a true gamble. And even those who believe he will fail in the general election (where Betfair values Hillary Clinton at overwhelming odds) would rather bet on Trump than risk on Rubio.
But for those looking to predict the next president, Sethi cautions that reading the betting markets can be a “chicken and egg” situation.
“We don’t know yet whether the prices moved first or the change in public perceptions of the race,” he says. “It’s an interesting question but will take some time and careful empirical work to answer.”
It’s worth noting that these sites skew toward a particular kind of person: an Internet-savvy gambler with some money to spare and an interest in national politics.
“We skew heavily male, with most of our users between the ages of 20 and their late thirties,” said PredictIt’s Travis.
According to data sent to The Daily Beast from PredictIt, 9,939 of the site’s 19,247 trades—or 52 percent—were made by users between the ages of 20 and 30. Over 4,300 of those trades came from users in either New York or California.
Even with its homogenous userbase, Travis says PredictIt is a much better indicator than one might think. It is, after all, the only U.S.-based online prediction market, since InTrade’s shutdown in 2013. And since real money is on the line, Travis says she’s learned users are more likely to bet with their wallets than with their hearts.
“I talk to the traders who show up to [PredictIt] events and they say, ‘There’s no way I would ever vote for, say, Hillary Clinton, but that’s who I would put my money on,’” she said.
And the numbers back that up: A 2008 University of Iowa study concluded that prediction markets are “74 percent closer to the actual outcome” than polling, and the results get even more accurate 100 days out from a vote.
“It’s what you think is going to happen,” said Travis. “Not what you want.”