President Donald Trump’s chances of winning a second term continue to slip away. With under a week to go, he’s still down about 9 percentage points nationally, and there’s little sign of any real movement in either direction. Early last week, I speculated that there was still enough time for significant changes to the race.
That’s much less true now. With the debates over, it’s hard to imagine anything that would spark a shift of more than a percentage point or two. And not only is Trump being badly outspent in the final days by former Vice President Joe Biden, but the current spike in the coronavirus, an issue that plays very badly for the incumbent, also is unlikely to help him as the few remaining undecided voters make up their minds.
As Nate Silver put it: “We’re sort of getting to the point where the only way Trump can win is with a major polling error, bigger than in 2016 (or if the election is stolen somehow).” A Trump win is certainly not impossible. Silver’s model pegs the chances at about 12%, while the Economist model puts it at only about 1 in 20. The reason Trump still has any serious chance is because the Electoral College appears to be biased in favour of Republicans this year, as it was in 2016, and perhaps a little more so (FiveThirtyEight gives Trump a 4% chance of winning the most votes; the Economist thinks it’s almost a lock that Biden will do so). At least, that’s the case if the state polls are correct.
As we saw in 2016, final state polling averages are more likely to be wrong — that is, several percentage points off the actual results — than the national average. That’s not surprising. There may be two or three top-quality polls in the battleground states over the next few days, but there will be half a dozen or more good national ones, and plenty of other useful surveys as well.
In case of a polling error
Right now, Trump’s best bet appears to be holding every state he won last time except for Michigan and Wisconsin. To do so, he’d need to take Pennsylvania, where he currently trails by a bit more than 5 percentage points. That’s not impossible; I’d bet that the results will reveal a polling error of at least 6 points in more than one state. But that’s less likely to happen in heavily polled Pennsylvania.
And if it’s just a one-state problem, then Trump is still sunk, because he’s currently losing (according to the New York Times averages) in Arizona, Florida, Iowa, North Carolina and perhaps Georgia. So what Trump needs is for the national polls to be off (in his favour), and for Pennsylvania to be even more off (in his favour) without those other states being off in Biden’s favour.
We could go through that same exercise, carefully counting where the electoral votes fall, with Wisconsin or Michigan. Or perhaps with Nevada or Minnesota, both of which Hillary Clinton won in 2016. But Biden’s polling lead is even larger in those states, and while it’s certainly possible that errors could still turn them for Trump, it’s increasingly unlikely.
How probable is a larger-than-usual polling error this time around? No one knows for sure. But if pollsters did know that there was a particular likely error, they’d correct for it. That’s why I’m fairly sceptical that looking at the errors of 2016 or 2018 will tell us much about this time around. The main thing making such mistakes more likely this year is the pandemic and the resulting changes in how people vote, with way more voters casting ballots early either absentee or in person.
Unusually high turnout
Also different this time? What appears to be an unusually high turnout. Perhaps all that will make it harder to correctly estimate the pool of likely voters, which could cause larger-than-usual errors. That could be worsened by problems with mail delivery or (additional) efforts by the Trump campaign to knock out absentee ballots. Or perhaps the current virus flare-up will discourage people who planned to vote in person on Election Day.
On the other hand? It’s possible that the unusual stability in preferences over the course of this year makes polling easier than normal. Similarly, both high turnout and early voting could, if pollsters understand what’s happening correctly, make the numbers more accurate than usual. And one more time: Any errors, should they happen, could be in either direction. At this point, if the polls are accurate or if Biden does even a little better than they say, Trump is in very deep trouble.
Of course, this is all assuming that we have a normal and more-or-less honest election. Trump has repeatedly threatened otherwise, by (for example) saying that the final result must be tallied by Election Day, despite laws and centuries of practice mandating that many ballots be counted over the next several days, or longer if necessary.
Most likely, local officials will ignore Trump’s bluster and the count will go on as usual. Accuracy and a full count of all legitimate votes, after all, are far more important than speed. And if we’re lucky, and one of the candidates wins by a solid margin, we might still learn who won the election late Tuesday night after all.
Jonathan Bernstein is a Bloomberg Opinion columnist covering politics and policy. He taught political science at the University of Texas at San Antonio and DePauw University