With only three days left before Election Day, I am excited to announce the final Political Safari presidential forecast. The page will be locked and the numbers will not update past today.
The forecast uses the expanded Political Safari model. We run the data from this year’s and past elections through a support vector machine. We then add expert ratings and polling data and simulate tens of thousands of times to get the average margin and probability. Its hard to do testing with the expanded model, but k-fold cross validation with the base forecast (without expert ratings and polling) gives us an accuracy of about 94%, so similar to most other forecasting sites.
Today, Vice President Kamala Harris’ win probability sits at 53.8%, just slightly higher than former president Donald Trump, at 46.1%. The remaining 0.1% is the probability that they tie at 269. On average, Harris receives 272 electoral votes in our simulations, while Trump gets an average of 266. Her win probability is down significantly from a peak of 67.3% in late August but she remains a slight favorite.
Out of the 50 states up for grabs (plus DC and a few districts in Nebraska and Maine), there are 7 that you should keep your eye on, all with a less than 75% probability of going to the favored candidate.
We’ve got two Lean Rs on the map this year: Arizona and North Carolina. For both, Trump is in the lead, but Democrats have a solid shot. The x-factor could lie in the other statewide elections, with Arizona having a senate election and North Carolina electing their governor this year. In both, we expect Democrats to perform extremely strongly, with polling suggesting that North Carolinians don’t like Mark Robinson for being racist and Arizonans don’t like Kari Lake for being Kari Lake. The presidential margins will be close, and while we don’t think that it will be enough for the Democrats, this does give them a slight benefit.
Nevada has the smallest electoral vote count of the tossups, at only 6, but it will likely be the closest state this year. While many give Trump the upper hand, our forecast suggests that Harris should be able to pull away with a slim victory. On the other hand, Georgia will probably be a slight Trump victory. While Harris is performing significantly better than Biden in the state, it still looks to go back to the GOP this year, and we give Trump a 55.7% chance of victory.
At the end of the day though, this will probably come down to the Rust Belt. While we give Harris a 64.3% chance of winning Michigan, the uncommitted movement remains a thorn in her side. Wisconsin remains tough to predict, with big polling errors in the last few years. Our base forecast has Trump winning by a point, which is likely to happen if Wisconsin polls struggle once again. Pennsylvania is the biggest swing state, and Harris has a 56.7% chance of taking its 19 electoral votes. The recent comments at a Trump rally regarding Puerto Rico and the ensuing backlash gives Harris a late boost in the key state.
So there you have it. The culmination of a project I started when I got bored a few months ago. Odds are, I won’t get all of them correct (the hope right now is to get 54), but hopefully, in 2028, I have time to do more. Maybe county-by-county regressions rather than state-by-state.
Update: Following this evening's Iowa polls from Selzer and Emerson College, we have made the decision to factor those numbers into our forecast. Iowa's rating has been changed from Very Likely R to Likely R, and the projected margin and win probability for the state have been updated accordingly. We have also finalized our ratings for all gubernatorial races.
If you want more detailed info about the presidential election, here is a spreadsheet with the full results of the forecast, along with the archived forecast from when President Biden was the nominee: https://docs.google.com/spreadsheets/d/1SNzjpb6CimCmx61V7gk_Eii34KCEtYvBP8MBkTBB8Us/edit?usp=sharing
Comments