Over the past few weeks, I’ve been following the forecasts for the presidential race posted by Thomas Miller, a data scientist at Northwestern University. Miller bases his predictions not on the polls, but on prices displayed on what he deems America’s most trustworthy political betting site, PredictIt. Since June 8, Miller has been updating his forecast once a day. The new numbers are posted at midnight EST on his homepage, the Virtual Tout.
Alert to election watchers: On Oct. 29, Miller unveiled a new site that continuously displays the electoral vote count that the PredictIt odds suggest each candidate is winning. PredictIt updates its prices once a minute. Miller instantaneously translates those odds into a forecast for the number of electoral votes (EVs) that each candidate is winning. “We’re pretty much showing where the election is moving in real time,” he says. “It’s like having your own personal ticker tape for the presidential election.”
Right now, you can click on virtualtout.io to watch what this writer considers among the best data for handicapping the wildly careening 2024 race for the White House. Believe me, it’s addictive.
As I’ve described in previous stories, Miller deployed PredictIt in 2020 to make superaccurate calls both on the presidential race and the two Georgia Senate runoffs that followed. Miller likes PredictIt because, he says, it’s extremely responsive to campaign events. By carefully studying the workings of PredictIt, Miller developed a model that translates prices posted on the site into the popular vote shares that each candidate commands from moment to moment. To get there, Miller makes a number of adjustments. For example, he found that since most PredictIt bettors are men who also frequently wager on sports, the prices show a slight Republican bias. So his framework corrects for the rightward tilt. Likewise, the data scientist discovered that when one contestant pulls far ahead on PredictIt, the odds swing way too abruptly in their favor. Reason: When the race looks like a sure thing, the bettors flood in to back the “winner,” creating a bandwagon effect that pushes prices to “landslide” extremes far larger than the probable margin of victory.
Miller also examined all elections since 1960, and found a close correlation between a candidate’s share of the overall ballots cast nationwide, and their percentage of the 538 EVs that decide the winner. He crafted a proprietary framework that converts the popular vote counts derived from PredictIt prices into electoral votes.
Miller’s real-time number now shows a shift toward Harris
As of 3 p.m. on Oct. 29, Miller’s “ticker tape” showed the following: Trump/Vance 345, Harris/Walz 193. At all times, you’ll find next to the totals for each candidate a red or green arrow tagging where they stand versus their count the previous midnight, and the amount of the change. In our snapshot, Harris/Walz are at +1, and Trump/Vance –1. “Trump’s still way ahead,” says Miller. “But the prices are moving in the vice president’s favor.” He notes that on Oct. 26, Trump hit his all-time high of 367, against 171 EVs for Harris, capping a giant reversal since the close of September, when the former president trailed by a wide margin.
Suddenly, Trump’s momentum ceased, and the race tightened. So far, the narrowing been slight, but several factors indicate that Harris may be mounting a major comeback. “She’s already gained 22 EVs, and at some points on Oct. 29 her gains have been several EVs higher,” declares Miller. The turning point, he says, was the rowdy, six-hour Trump rally at Madison Square Garden on the evening of Oct. 27, where an insult comic branded Puerto Rico “a floating island of garbage in the middle of the ocean.” That crack could be undermining what had been Trump’s rising support among Hispanic voters, or may just have pushed Americans on the fence into the Harris column.
“We’re seeing a considerable increase in Americans’ interest in the campaign on PredictIt,” says Miller. “As many people are entering as are betting on the site right now. The current bettors are changing their views, and the new entrants may have different views than the population that’s been dominating the betting until now.”
Miller notes that the trading volumes on PredictIt have increased from the past average of 37,000 a day to 41,000 over the past week. Heavier trading, he says, means more volatility. “And more volatility makes it harder to predict what’s going to happen on Nov. 5.”
Miller cites still another reason why, despite Trump’s wide lead, the race remains in flux. He notes that betting sites appear to be advertising for customers on social media. “It’s targeted advertising,” he says. “Like all advertisers, they’re targeting the sites where people are ‘most likely to buy their product,’ in other words, where the viewers are most likely to bet on the election. And it’s the Republicans who wager the most.” As a result, he posits, the sites are attracting an excessive number of GOP-leaning bettors, which could be far more than he’s already adjusted for. That phenomenon could at least partially explain Trump’s gigantic surge in October, whereas we didn’t see the kind of milestones that generally trigger such a surge. The bettors on PredictIt could be seeing big odds for Trump on the sites that do lots of advertising, and adjust their own wagers in Trump’s direction to follow the trend.
Or, PredictIt could be right-on. Indeed, the site has long been Miller’s North Star. “I’m not yet changing my call that Trump will probably win,” says Miller. “But the numbers are changing. The volatility, new people entering the betting markets, the effect of the ads, all make the outcome increasingly unpredictable.” To get what may well be the best predictions out there, now posted minute by minute, tune in to Miller’s new ticker tape.
