THE PRESIDENTIAL election has entered the final stretch. With 11 days until votes are tallied, The Economist’s presidential election forecast puts Donald Trump in an eight-percentage-point hole nationally and losing in all of the important swing states. Our model gives the president just under a 1-in-10 chance of winning re-election. His hopes are kept alive by three main factors.

The first and largest contributor is the chance of large polling errors. Our analysis of polling between 1948 and 2016 reveals that polls tend to miss the eventual Democratic margin in the election by about 3.5 percentage points on average. Statistically, that implies a margin of error of about seven percentage points. That alone puts Joe Biden’s eight-point national lead outside the threshold of the traditional 95% uncertainty interval; in fact, our model today finds that polls virtually guarantee he will win the popular vote majority.

But the president is not elected by the national popular vote. The electoral college, with its votes allocated by the states, does the real deciding. And Mr Biden’s margin in competitive states is much smaller than his support countrywide. Today our model puts him up by between six and seven points in Pennsylvania, Michigan and Wisconsin—the states he needs to secure a victory. But unlike his national margin, these margins are well inside the uncertainty interval for state polls, which are usually less accurate than national polls and carry with them a margin of error closer to nine points. The chance that polls are wrong enough for Mr Trump to win any states if the election were held today is much higher than his chance of winning the popular vote; roughly 1-in-20.

Second, there is the chance that events transpiring over the next 11 days change the minds of voters who have yet to cast their ballots. That gives Mr Trump an extra chance that a truly fundamentally unpredictable event could move the polls his way. Perhaps his administration will announce a vaccine for covid-19 before the election, or maybe Mr Biden could be hospitalised himself. Whatever it is, that additional uncertainty stemming from the time remaining until election day pushes up Mr Trump’s odds from 4% to 8%.

Finally, there is the chance that something outside the scope of our quantitative models could upset the balance of the election. Perhaps the pandemic, poll-watchers or inclement weather will deter voters from going to the polls. Or maybe a historically large share of postal ballots will be rejected. There is also the chance that Mr Trump puts pressure on his appointees on state and federal courts to halt counting in the event of a contested election (though there are still questions over whether such a halt is possible).

Mr Trump’s best hope is to combine substantial gains in the polls with a stroke of luck—that polls have been undercounting his supporters—and the type of electoral chicanery that is impossible to predict beforehand. Even then, it is hard to rate the president as anything but a severe underdog.

Editor’s note: This article first appeared in our weekly “Checks and Balance” newsletter on American politics. You can sign up to receive it here.