The final prediction for the US election

The gap between high-quality and Trump-partisan polls has widened, with Harris at 61.5% in the former model and Trump at 73.2% in the latter.

Without further ado, here are the results from the prediction models for the U.S. Presidential election:

The model based on high-quality polls indicates that Harris has a 61.5% chance of winning the presidency. She is favored in the Midwest swing states of Michigan, Pennsylvania, and Wisconsin, as well as in Nevada.

By contrast, the Trump-partisan polls model forecasts a decisive victory for Trump. It assigns him a 73.2% probability of winning, with predictions showing him leading in all swing states and maintaining a slight edge in the national vote.

Model Comparison

While the models yield significantly different predictions, some similarities do exist. The forecast is similar for Arizona and Georgia and now to a lesser extent for North Carolina. For all other states (the Midwest and Nevada) the discrepancies between the models have grown since the last update.

It is important to note that the races remain tight. A typical polling error of about 3% could swing the outcome in favor of either candidate in the high-quality poll model. Even a slight polling error favoring Harris in the pro-Trump partisan model could tilt the election in her direction.

High-Quality Polls with a Pro-Trump Midwest Bias Remain a Toss-Up

Previously, I introduced a third model that serves as a middle ground between the high-quality and Trump-partisan polls. This “high-quality polls with a pro-Trump Midwest bias” model incorporates high-quality poll data while adjusting for a pro-Trump bias in Michigan, Pennsylvania, and Wisconsin, reflecting the average polling errors from the past two elections.

This model still predicts a toss-up, with Trump slightly ahead at 51.4% to Harris’s 47.9%. Notably, the model shows Harris leading in Michigan with a 65.2% probability and in Wisconsin with a 53.9% probability, while Trump holds a slight advantage in Pennsylvania at 54.3%.

More details on the model specifications and previous updates are here and here.

As the election draws near, the final verdict on the accuracy of these models will soon be revealed.