An election expert who used to say that Vice President Kamala Harris would win by a huge margin now says that former President Donald Trump will win.
Thomas Miller, a data scientist who correctly predicted the outcome of the 2020 presidential election, had earlier backed Harris to win in 2024. In September, he told Newsweek that his model showed the Democrat would likely get more than 400 electoral votes.
Based on betting odds instead of polling data, the model now says that Trump is the favorite. Its most recent estimate is that the Republicans could win 345 Electoral College votes, Newsweek reported.
Newsweek has drawn up what a possible win for the former president with 345 Electoral College votes would look like. FiveThirtyEight says that in this case, he wins all the states where Harris has a lead of less than ten points, including the South and the Midwest.
The model first said that Trump would win by a small margin on October 7, when betting odds started to move in his favor.
Miller’s model puts betting odds ahead of survey data and uses what he calls “fundamentals,” or past election trends, to balance the data. This is something that most other models also do.
Miller is using data from the betting site PredictIt to fuel his model for the 2024 election. This is the same data that he used to make his correct guess in 2020.
The facts and the fundamentals, on the other hand, point to different outcomes.
As of October 20, Miller wrote on his website that “technical and fundamental analyses are not in alignment,” which means that past trends point to a Democratic victory while betting odds say Trump will win.
Polls still show that the race is close, but betting odds have continued to favor Trump. For example, Polymarket odds give Trump a 60.3% chance of winning nationally, while Harris only has a 39.7% chance.
In the past, betting odds have always been right about who would win the presidential race. However, they are not based on the same representational statistics as polls, so they are more likely to be wrong.
Miller told Newsweek that he fixes his model for any possible biases, which right now are tilted toward Trump.
“Prediction markets have a Republican bias,” Miller said. “We assessed the degree of Republican bias in 2020, and we currently correct for that degree of bias. What is uncertain, however, is the level of Republican bias in prediction markets in 2024. We are conducting studies across prediction markets that we hope will shed light on the degree of bias in 2024.”
Miller’s most recent estimate is a little better for Harris than the last one, but she’s still losing.
On October 26, the model said the vice president would only win 171 Electoral College votes, which was the worst showing for a candidate since 1996, when Sen. Bob Dole ran as a Republican.
“The 2024 race for the presidency has gone from toss-up to Republican landslide, to toss-up, to a possible Democratic landslide, to toss-up, and now to a possible Republican landslide,” Miller said.
“Could prediction markets and associated election forecasts turn again with less than two weeks before the close of voting on November 5? Yes. We expect to see increased trading and high volatility in the final week of the race,” Miller added.
Since 1980, there have been 11 presidential elections.
The only one where the winning candidate had worse chances than the losing candidate was 2016, when both betting markets and regular polls didn’t see Trump winning.
