This weekend saw the release of a final batch of high-quality polling that generally confirms what polling has said all along: Democratic presidential nominee Joe Biden is in the lead.
Evidence for that proposition comes from a tetralogy of New York Times polls conducted in conjunction with Siena College, which were released Sunday morning. They show Biden ahead in Arizona, Pennsylvania, Florida, and Wisconsin — easily enough states to give him the win. But even a Des Moines Register poll conducted by the legendary Ann Selzer that was released on Saturday evening, which showed Trump with a 7 percentage point lead in Iowa, is actually not very good news for President Donald Trump.
The only genuinely solid result for Trump was an ABC News/Washington Post poll showing him up 2 points in Florida. But the same pollsters simultaneously found him down by 7 points in Pennsylvania.
Overall, the message of the polls is crystal clear — Trump is losing the election, including in the key swing states, and the margin is not small. That doesn’t mean he won’t win, as large polling errors do sometimes occur, but you’d be well-advised to bet fairly heavily against it.
Trump is losing in the latest polls
To sum it all up, here are the latest poll results:
- Arizona: Biden 49, Trump 43 (NYT/Siena)
- Florida: Biden 47, Trump 44 (NYT/Siena)
- Pennsylvania: Biden 49, Trump 43 (NYT/Siena)
- Wisconsin: Biden 52, Trump 41 (NYT/Siena)
- Iowa: Biden 41, Trump 48 (DMR/Selzer)
- Pennsylvania: Biden 51, Trump 44 (Washington Post/ABC)
- Florida: Biden 48, Trump 50 (Washington Post/ABC)
Given that there is essentially no time left in the election, these polls are very bad news for Donald Trump. He has been lagging badly in national head-to-head polling the whole race, but the contest for Pennsylvania has generally been closer — FiveThirtyEight’s average of recent polls in Pennsylvania has Biden ahead by 5 points — and there are good odds that Pennsylvania will be the decisive state in the Electoral College.
These two Pennsylvania polls do show the state slightly closer than national polling averages, which Biden leads by more than 8 points, but it’s still just not that close. Polling errors of 6 to 7 percentage points in magnitude do happen, but it would be unusual. Or to put it another way, even if the polls in Pennsylvania are off this year as much as they were in 2016, then Trump would still lose.
A 6 percentage point lead in Arizona, according to the Times/Siena poll, gives Biden a backup path to victory if he does somehow fall short in Pennsylvania. An average of recent polls in Arizona shows Biden with a smaller, 3-point lead there.
And while the Biden campaign would be disappointed to lose Iowa by 7 percentage points after having looked close there in many other polls, this result is simply not as good as it superficially seems for Trump. He won the state by 9 percentage points two years ago, and if he loses 2 percentage points of support across the Northern battlegrounds he will lose the election. Iowa, with its six Electoral College votes, isn’t a big prize for either candidate. But where the Iowa poll does matter is in the race for the Senate, where Theresa Greenfield’s efforts to unseat Joni Ernst are looking worse than they did a month ago in a race that could potentially decide whether Democrats or Republicans control the chamber.
Finally there is the ABC/Post Florida poll, which has genuinely good news for Trump. By their numbers, he is not only winning the state, he’s winning it by a slightly larger margin than he did in 2016. That’s the kind of result Trump needs to win the election nationally, but of course, that poll also has him losing Pennsylvania rather badly — in which case Florida alone won’t get it done for Trump. And Florida polling averages from FiveThirtyEight show Biden with a narrow 2-point lead.
Only a historic error will deliver Trump the win
Hillary Clinton was leading in the polls in 2016 and then lost the election, dealing a blow to the public’s confidence in pollsters and forecasting.
But much of the flawed overconfidence in Clinton coming from sites projecting her odds of winning as 90 percent or more was based on bad modeling, not on bad polling. She led in all the swing states, but led by small margins. Small polling errors happen all the time (indeed, they happened in 2012; it’s just that nobody remembers that Obama won by a bit more than expected), and it’s only moderately unlikely that a candidate down by 2 or 3 percentage points will lose.
Trump, however, is down by 6 or 7 percentage points in Pennsylvania, and separately is down by smaller amounts in North Carolina and Arizona — which have distinct regional and demographic mixes meaning that poll errors in those states would be only partially correlated with ones in Pennsylvania. None of this is to say that he can’t win the election, but if he does it will involve a much larger polling error than we saw last time around, and raise some fundamental questions about whether the major public pollsters can do reliable surveys at all.