In the final days before the 2020 presidential election, polls have generally indicated a clear victory for incumbent President Joe Biden. But when the votes were counted, it turned out that the polls had overestimated him – Biden won, but by the skin of his teeth. This, of course, was similar to what happened in 2016, when former President Donald Trump significantly outperformed polls in Michigan, Pennsylvania and Wisconsin to win a surprise Electoral College victory despite relatively accurate national polls.

This raises two questions for 2024: First, what would happen if the polls are stopped again? And secondly, how likely is it that the polls will be as high as in 2016 or 2020?

Either Trump or Harris could win comfortably

In 2020, polls overestimated Biden’s margin over Trump about 4 percentage points in competitive states. As of October 30 at 11:30 a.m. Eastern, the margin between Vice President Kamala Harris and Trump in 538 polling averages is less than 4 points in seven states: the notorious seven Arizona, Georgia, Michigan, Nevada, North Carolina, Pennsylvania and Wisconsin. That means if the 2020 polling error repeats itself, Trump would win all seven swing states and 312 votes Electoral College.

Of course, if the polls are off, it won’t necessarily work in Trump’s favor. The direction of the polling error is impossible to anticipateand polls have overestimated Republicans many times in the past. In a scenario where polls overestimate Trump’s margin by 4 points in every state, Harris would win all seven swing states and 319 electoral votes.

Note that none of these are particularly close results, at least in the Electoral College.

Some polling error is normal

Both outcomes – and everything in between – are very much on the table next week. But are these scenarios really likely or more like outside possibilities? Well, that’s where the work we do on our election forecasting model can help. Starting October 30th at 11:30 AM Eastern, our forecast gives Trump a 52 out of 100 chance to win the White House and Harris an almost identical 48 out of 100 chance. The model arrives at this probability by calculating how many Electoral College votes each candidate would win, given certain amounts of voting error in their favor, and then counting how many times each candidate wins among these simulations. (More on that in our methodology.)

Based on how much the polls have been off in the past, our electoral model estimates that the average polling error in competitive states this year will be 3.8 points per margin.* This error is not uniform across states—for example, states with different demographics. tend to have different levels of polling error — but in general, when polls overestimate a candidate, they tend to overestimate him across the board. In other words, the model expects a polling error roughly the size of 2020 — though not necessarily in the same direction as in 2020. (In 50% of the model’s simulations, Trump beats his polls, and in 50% of cases, Harris it does.)

This point is worth insisting. Since our average expectation is that there is a decently large polling error at least half the time, there is actually a very small probability that the polls will be perfect and the election will turn out exactly as the polls suggest. Let’s look at this using the largest lead each candidate has in the seven swing states: Trump’s current 2-point lead in Arizona. Nationally, our model expects the polling error to be greater than 2 points in either direction 62 percent of the time. In other words, there is only a 1 in 3 chance that the polls will miss by less than 2 points (which we would consider historically small polling error).

With all seven key swing states so close, even small polling errors in the same direction can have a big impact on who wins the election. According to the simulations in our model, there is a 60-in-100 chance that each candidate will win more than 300 Electoral College votes — which Harris could do by winning five of the seven swing states and Trump six of the seven. By modern standards, I think it’s fair to call this a blowout victory – given how tightly divided the country is, it’s relatively unlikely that either candidate will win much more than that (even to get to 320 of electoral votes, Trump should win a state like Minnesota and Harris should win a state like Florida).

Of course, the likelihood of an explosion in any case depends largely on the outcome of the popular vote. If Harris wins the national popular vote by 3 points, he is much more likely to win the states that will decide the Electoral College than if he loses the popular vote by 3. This is vividly shown in the chart below, which takes all the simulations in our model and group by popular vote result.

As you can see, Trump is favored to win the election even if he loses the popular vote by 1-2 points, which is national polling average currently suggest. And if national polls turn out to underestimate him, with Trump winning the popular vote by 1-2 points, he would be favored to win in a blowout.

Meanwhile, our model finds that Harris needs to win the popular vote by 2.1 points to be favored to win the election because swing states lean more Republican than the nation as a whole. And if she wins the popular vote by 4.5 points (Biden’s 2020 popular vote margin), she is favored to win in a blowout of her own.

Polls are inherently uncertain. This is why we model.

So far, I’ve said little about the polls themselves. And, in fact, there is reason to believe polls may be more accurate this year than in the past. While the share of polls conducted or sponsored by Republican-aligned organizations is increasing, something we wrote about — the total share of partisan polls is lower than in previous years, and the average rating of 538 polls in the 2024 polls is higher than in past years. All else being equal, that should lead to better polls than in 2016 and 2020. We also saw fewer polls from the firms that overestimated Democrats the most in those years.

However, the news is not good at all. In particular, surveys still report difficulties reaching voters not at all, and Trump supporters may be even less likely to respond to polls – even the high quality ones. This means pollsters are as addictive as ever (or maybe more!) on weighting and modeling to obtain good estimates of public opinion. But her decisions makes it matter a lotand in particular, there appear to be large differences between surveys that attempt to use these techniques for balance the samples by party or previous vote and those who don’t.

And that’s the big fundamental problem with pre-election polling: We don’t know what the demographic and political makeup of the actual electorate will be, so pollsters are just making the best guesses they can. These assumptions always have and always will come with an error attached.

And this is where electoral models like 538 come in really handy. The purpose of creating election forecasting models is not to provide a hyper-accurate, laser-like predictive picture of the election that eliminates all polling error. Rather, it is to give people a good understanding of how polls could be wrong and what would happen if they were. By analyzing possible polling errors and uncertainties, these models help us approach elections with a clearer idea of ​​how likely each side is to win (and by how much).

As we enter the final week of this election, it’s a good time to remember that uncertainty is an inherent part of polls and elections. That’s especially true this year, with races locked in swing states. Given that the polls are imperfect, our expectation is that they will be off by some amount in either direction. And if the polls end up being off, given how close the election is, there’s a pretty wide spread in the range of Electoral College results.

In other words, we can sum up the current state of the race this way: Although Trump and Harris have roughly equal chances of winning the election, the final margin won’t necessarily be close. In fact, there’s a pretty good chance it won’t.

Footnotes

*We simulate potential polling errors for future elections using a fat-tailed distribution – specifically, a Student’s distribution with five degrees of freedom (a parameter that increases or decreases the probability of surprise “tail” events in our simulations). This error of 3.8 points is the spread or sigmaof that distribution — analogous to that standard deviation of a normal distribution. 538’s distributions are slightly wider than those used by other forecast models. This is because our model takes into account that survey failures have increased over the past decade. Therefore, our model predicts more survey error than it would if we assumed a constant level of error over time, as most other forecasts do.