The swing states that will help predict the election outcome

DONALD TRUMP is likely to be a one-term president. At least, that is what The Economist's election-forecasting model predicts will be the outcome of America’s election on November 3rd. Of course, that prediction rests on several assumptions, including that the polls this year are not wildly off, that everyone who plans to vote is able to cast a ballot and that those ballots are counted. Although it is hard for a quantitative election forecast to take these factors into account before the election, early returns can provide some guidance on the final outcome, even before polls close.

To track and update our understanding of the race in real time, The Economist has built a statistical model that updates the results of our pre-election forecast with final election results from the states as they come in. The model operates on the assumption that we can predict how the night will unfold based on how closely the actual election results match our original projections. If the early returns are consistent with the polls, our forecasts will remain largely unchanged. If not, the model will adjust our state-by-state predictions, as well as our prediction of the overall winner.

Such real-time updates will draw heavily on results in states that count both early and election-day ballots rather quickly. These include Florida and Virginia, where polls close at 7pm (midnight GMT), and North Carolina, where voting ends half an hour later. If early results show that the polls are wrong in these states, they are likely to be wrong in other battleground states with similar demographic characteristics. So if, for example, Mr Trump wins in Florida, his chances will improve in similarly competitive states with lots of Hispanics and older white voters (conversely, a win in Florida for Mr Biden all but assures him the White House). A victory in Pennsylvania, where polls close at 8pm Eastern time, would bode well for his chances in other northern battlegrounds, and thus the majority of electoral-college votes. Our chart above gives the probability of either candidate winning the election if he wins one of the highlighted states (NB: this is based on the result of one state; it is not a cumulative probability).

Our real-time model may also help to evaluate claims of election-rigging and other foul play. Votes cast on election day will lean to the right, thanks to Mr Trump’s exhortations to his supporters to vote in person (mail-in ballots are, in his view, “a whole big scam”). If they are counted more quickly than Democratic-leaning mail-in votes, then early returns may create a “red mirage”—the appearance of a Trump victory on Tuesday before a full count is finalised. That could lend plausibility to false claims of a “stolen” election if postal ballots tallied over the next few hours or days ultimately give victory to Joe Biden. Our number-crunching on election day could serve as a check against such allegations. And, along the way, it is a better way to follow the results as they unfold.

Editor’s note: The Economist has open-sourced the code for its live election-night model for anyone to run. Download it here.

Dig deeper:
Read the best of our 2020 campaign coverage and explore our election forecasts, then sign up for Checks and Balance, our weekly newsletter and podcast on American politics.