Harris is ahead of Trump and it could be a landslide victory, says a leading data scientist

Now, Miller is back on the scene for the first time since his highly unconventional predictions. He is using a similar system to influence the supposedly super-close presidential election, which will be decided in just 48 days. His totally off-the-wall prediction will impress pollsters, pundits and Assetsreaders alike. But Miller's view deserves special attention for two basic reasons: First, it is based on numbers that are arguably far more scientific than the exit polls that are almost always used to determine the course of the campaign, and second, it reached peak accuracy four years ago.

In all three elections in 2020, Miller beat virtually all pollsters and modelers analyzing multiple voter surveys. He missed Biden's Electoral College victory by just 12 votes, marking all states except Georgia in the correct column. For the two Senate runoffs, Miller refined his method of sorting the Peach State data and scored again. A week before Election Day on December 6, 2020, polls showed Republican David Perdue with a large lead over Democrat Jon Ossoff and showed a neck-and-neck race between the Republican Party's Kelly Loeffler and her opponent Raphael Warnock. In contrast, Miller's numbers predicted a big loss for Loeffler and a modest victory for Ossoff. Once again, the recalcitrant academic hit the nail on the head: Miller was only 0.2% below Warnock's 2.0% lead and was spot on with his prediction of Ossoff's 1.0% lead at the ballot box.

Harris vs. Trump in the polls and prediction markets

Miller's approach differs significantly from most political forecasting because it relies not on polls but on prices set by Americans. use their own dollars on the candidates they believe are most likely to win. “Political betting sites are best at predicting the wisdom of the crowd,” he said Assets. He says that while polls tell us something about the past, the odds on betting sites tell us something about the future. “Polls are a snapshot of the recent past,” he adds. “They usually take small samples of 500 to 1,500 people. And pollsters ask respondents who they intend to vote for at that moment, which may have changed a few days later when the results are released. Most polls are about four to five days behind.”

That delay, he says, introduces a lot of noise or variability that obscures the true picture. Another problem with polls: Miller says a better question than “Tell us which candidate you'll vote for?” would be “Which candidate do you think will win?” And while pollsters don't ask that question, it's simply how bettors build a market.

For the 2020 Biden-Trump matchup, Miller used odds posted on the largest U.S. political betting site, PredictIt. He took PredictIt's odds in each of the 56 individual precincts, tracked the hourly changes, and used his own model to turn the data into daily odds that were extremely up-to-date given that people on the site were placing bets on one candidate or the other around the clock. For the Senate race, Miller took a different approach. He gathered a group of about 1,200 Georgians, whom he lured by paying them a few dollars to participate. They also got extra dollars if they named the most likely winner – not necessarily the one they planned to vote for – and predicted the winner's margin. The method he developed, a so-called “prediction poll” that used the best parts of the polls and the betting market, led Miller to a near-perfect estimate of vote shares.

For the 2024 presidential election, Miller is relying on PredictIt again. He praises the site for having “a more stable group of investors” than the demographics studied by pollsters. “Tens of thousands of people are betting on the site at any given time of day,” he says. “The maximum contract is $850, and on average 37,000 'shares' are traded daily.” Put simply, Miller sees PredictIt as a highly liquid market, similar to a stock exchange for, say, small-cap stocks or high-yield bonds, gathering a huge number of buyers and sellers. “Financial markets are forward-looking and process information instantly, and PredictIt offers the same benefits,” Miller says.

Is Harris ahead of Trump? This model says yes

To estimate the renewed battle between Biden and Trump, Miller developed a “generalized linear model” based on the results of the last sixteen presidential elections since 1960. Starting with the Kennedy-Nixon race that year, each presidential race was decided by 538 electoral votes. “The model is based on a lot of historical data,” Miller says. It shows that daily prices on PredicitIt translate closely to the share of the vote going to each candidate. Suppose that on a given day, bettors give Candidate A odds of 52 cents, which corresponds to a 52% chance of winning. By Miller's calculation, that 52% should also mean that at that time, the best prediction is that 52% of likely voters plan to cast their ballot for Candidate A.

Each day, Miller plotted the prices released at midnight on a chart overlaying the key events that influenced the election. The chart, titled “Impact of Current and Campaign-Related Events on Projected Electoral College Votes,” can be found on his website, The Virtual Tout (virtualtout.com). Overall, the chart tracks the number of electoral votes that PredictIt odds indicate. It is shaped like a giant “U,” reminiscent of a steep ski slope rising to a giant mountain at the bottom.

Long before the Biden-Trump debate, the Republican was well ahead; in Miller's chart, he was at the top of the ski slope. In mid-June, Trump seemed to be drawing almost 400 of the 538 electoral votes. After the duel on stage on June 27, Trump became even more dominant, reaching a peak of almost 500. The Democrats' chances rose sharply after Biden's withdrawal on July 21, and by the time Harris secured the nomination on August 6, the Democrats were in a solid lead of about 325 electoral votes. They continued to gain ground over the next two weeks, reaching the 400-vote mark in a stunning turnaround by the end of their convention on August 22.

Then Trump staged a comeback. In the days leading up to the Trump-Harris debate on Sept. 10, Harris was ahead, but Trump had nearly caught her. “At that point, the race was basically neck and neck,” Miller notes. “The Democrats' projection was 288.” It was the battle on the stage in Philadelphia that undid the 78-year-old former president, according to Miller's numbers. Within a day of the candidates leaving the podium, Harris had won exactly over 400 electoral votes. Taylor Swift's endorsement, given to Harris on the day of the debate, likely helped derail Trump's chances, Miller said. Since then, Harris has maintained her total of over 400 votes.

A second chart on The Virtual Tout shows how the share of the popular vote affects the outcome of these 16 elections. Remember that Miller uses PredictIt prices as the equivalent of the popular vote: a candidate who is priced at 51% on Election Day is likely to receive 51% of all votes cast. He points out that Democrats traditionally need a lead of more than two percentage points or more to win, and that Republicans can prevail with as little as around 48%.

On September 16, PredictIt shows odds of 55 cents for Harris and 45 cents for Trump, the opposite of the scenario before Biden's departure. According to Miller's model, these odds mean 55% of the vote for the Democrat again. If the situation remains this way, Trump faces an absolute defeat. “That would be somewhere between the defeats of Barry Goldwater against Lyndon Johnson in 1964 and Bob Dole against Bill Clinton in 1996,” says Miller. “We're talking about an overwhelming victory in which Harris gets over 400 electoral votes and wins Michigan, Pennsylvania, North Carolina and all the other swing states.”

Miller notes that America has never seen anything like such a dramatic reversal of fortune, at least in recent history. “It went from a drastic landslide for Trump to a drastic landslide for Harris,” he marvels. The gap is now so wide that only another massive swing could put Trump back in the race, and Miller predicts that right now it looks like Harris will win big on November 5. As a closing remark, he recalls a slogan the Johnson campaign used to denigrate Goldwater: “In your gut, you know he's crazy.” Miller's market-based analysis assumes that people betting their own money are correct in predicting that by the time the candidates drop out on September 10, millions of voters likely to support Donald Trump will have abandoned the ex-president, setting off the shockwaves that could trigger an avalanche for Harris that few currently see building.