Tom Miller has just pinpointed the precise moment the presidential race shifted from a solidly favorable rating for Donald Trump to a huge lead for Vice President Kamala Harris, which she has maintained to this day.
“It was obvious to me, but I didn’t notice it at first,” the Northwestern University data scientist told our reporter by phone on Sunday. “I saw this huge jump in support for Harris on July 31, but I didn’t connect it to Trump’s appearance at the National Association of Black Journalists convention that day. That event, not the debate that made things worse for Trump, was the turning point in the campaign.”
Miller’s election predictions are not based on polls, but on candidate prices posted on the PredictIt betting site. He considers PredictIt’s odds more reliable than opinion polls, which reflect voters’ preferences four to five days out. Since polls typically survey 500 to 1,500 likely voters, they reflect a great deal of statistical “noise”—hence the wide variation in the numbers returned by different models.
PredictIt is the most liquid betting market, with an average of about 37,000 bets per day, according to Miller. And because each player is subject to an $850 cap, no individual bettor or group of high-rollers can artificially inflate the odds on one candidate or another.
Trump Leads Before NABJ Disaster
Miller’s model first assumes that PredictIt probabilities closely mirror popular vote shares. Put simply, a candidate who has a 55% chance of winning, or who is priced at 55 cents on PredictIt, is likely to receive a similar share of all votes cast. Second, Miller shows that historically, popular vote shares closely track the share of the 538 electoral votes each candidate receives. He finds that this relationship has been very stable in every race since 1960.
Miller’s Home Page, Virtual promoterThe chart below shows the share of the electoral vote that is shifting toward the Democratic side, as well as the events that moved the odds significantly, and then the fluctuations in the expected electoral vote around the 270 needed to win.
Between July 21—the day President Biden dropped out of the race and endorsed Harris—and July 27, her electoral vote tally surged. After that, her numbers held steady for four days in a row.
“She was still far behind the previous president, and it looked like her numbers had plateaued,” Miller says.
But then, Miller argues, there was a tremor that could turn into a Harris landslide by November. On July 31, Trump falsely suggested at a National Association of Journalists event that Harris had changed the way she described her racial heritage, questioned her mixed-race background, and accused the vice president of “going black” and that Harris “now wants to be known as black.”
Although the inflammatory comments drew ire from the press and among pundits, no one viewed Trump’s NABJ interview as a turning point in the election. Miller notes that the PredictIt market went crazy that day as bettors shifted en masse from Trump to Harris.
“Over 100,000 shares were traded that last day in July, three times the usual number,” he says. “Overnight, the election went from Republican to Democratic, with Harris’s stock rising to over 270. Trump’s NABJ comments proved to be a complete disaster for his campaign. It had nothing to do with anything Harris did. The huge shift was Trump’s doing.”
In the wake of the NABJ debacle, Trump partially bridged the gap—and then came the debate.
Miller’s chart shows Harris’s electoral vote count continuing to climb over the next two weeks, peaking shortly before the Democratic National Convention. But the Windy City celebration itself failed to provide an additional boost. By early September, her numbers had slipped slightly. On Sept. 6, news that Trump’s hush-money trial would be postponed until after the election boosted his numbers. The day before the debate, he was trailing only by a narrow margin.
“At that point, even though Harris was still in the lead, the race was very close,” Miller says. “Surprisingly, most of Trump’s jumps came on good news about his legal issues.”
Then the Philadelphia showdown sent Harris’s projected electoral vote tally skyrocketing. “This surge was a result of the combined debate and Taylor Swift’s endorsement of Harris,” Miller says.
As of September 22, PredictIt prices Harris at 56.3% to 43.7% for Trump, which Miller claims would translate into a landslide victory for the vice president with 43 days to go before the election.
“Major events can change things, wars are raging and can change the race, and candidates can make big mistakes,” he warns.
But right now, he says, Harris is far ahead, and the polls haven’t caught up with the huge win that may be building — which began building the day Trump made those disastrous comments to black journalists, and squandered a lead that he never regained.