Posted by 
If you were hoping the final polls before Election Day would give a clearer picture of the presidential contest, you were hoping in vain.
Pretty much any theory about what will happen Tuesday got some new piece of supporting polling evidence in final polls — and also some new evidence casting doubt on it.
Are late deciders breaking for Trump or for Harris? Are polls missing Trump voters yet again — or have they overcorrected to the point that they’re significantly underestimating Democrats? Is Harris performing better in the Rust Belt battlegrounds or the Sun Belt?
The final batch of polling does not provide a consensus answer to any of these questions, and the uncertainty about the outcome appears greater than ever.
What Democrats hope is true is that late deciders are breaking for Harris (perhaps nudged by Trump’s controversial Madison Square Garden rally last week). Some new state polls suggest that may be true. The New York Times reports its final state polls found that “among the 8 percent of voters who said they had only recently decided on their vote,” Harris “wins the group by 55 percent to 44 percent.”
And yet the final batch of national polls have actually moved toward Trump, and averages now show Harris’s national lead dropping to 1 point or less — her smallest in months. If there were a national trend in Harris’s favor, we’d expect it to show up in the final national polls, but it isn’t there.
The final state polling averages, meanwhile, show a race that is essentially deadlocked, with a margin of 1 point or less separating the candidates in nearly every swing state. But Nate Silver has argued that there are clear signs of widespread pollster “herding” — that, like sheep, pollsters are adjusting their results to fit an expected close outcome.
Statistical principles suggest that, if the race is truly tied, most polls should show near ties. But there should also be a fair amount of variation with some polls showing clear leads for either candidate, and we’re getting very few of those this year. “The odds are 1 in 9.5 trillion against at least this many polls showing such a close margin,” Silver wrote.
But is the herding hurting one candidate more than the other? Both parties have reason to hope the polls are missing support for their side. Republicans’ reason is that pollsters did underestimate support for Trump in 2016 and 2020. That could happen again: New York Times chief polling analyst Nate Cohn wrote Sunday that, in the final Times polls, “white Democrats were 16 percent likelier to respond than white Republicans,” which “raises the possibility that the polls could underestimate” Trump once more.
Others suspect pollsters have overcorrected to the point they’re now overestimating Trump’s support. The highly respected Iowa pollster J. Ann Selzer shocked the political world Saturday with a poll showing Harris up by 3 in her state, despite the widespread assumption that it was a safe Trump state. Some theorize Selzer has caught on to a shift toward Democrats that other pollsters have missed, but others suspect her poll is just an outlier that won’t actually match the results.
The specifics of the swing state map are also highly uncertain. The averages show Harris having a very slight edge in Michigan and Wisconsin, while Pennsylvania is basically tried. They also generally show a slight Trump edge in Georgia, North Carolina, and Nevada — and a more significant Trump lead in Arizona. This suggests Harris’s best path to victory is by holding the Rust Belt.
But the New York Times/Siena College poll released Sunday threw much of that map up in the air, showing Harris ahead in Georgia, North Carolina, and Nevada as well as Wisconsin — with Pennsylvania and Michigan stuck in a tie. (Arizona still went pretty solidly to Trump.) Even if the tied states went to Trump, those leads would be enough for a narrow Harris Electoral College victory.
How much stock should we put in the Selzer poll?
Election wonks’ general advice for interpreting a flood of polling is to “stick to the averages.” Looking at individual polls can be interesting, but any individual poll can be an outlier, and for partisans, the temptation to cherry-pick and see what you want to see is strong.
As of Monday morning, the New York Times averages show:
- Harris leading by 1 point in Wisconsin
- Harris leading by less than 1 point in Michigan
- A tie in Pennsylvania
- Trump leading by less than 1 point in North Carolina and Nevada
- Trump leading by 1 point in Georgia
- Trump leading by 3 points in Arizona
If those results show up on Election Day, then the election gets determined by whoever wins Pennsylvania. And yet we shouldn’t assume those will be the election results, either. Final state polling averages frequently differ from the results by a few points. And given that so many of these final averages show a 1-point difference or less, the only reasonable takeaway here is: It’s really close.
Now, some number crunchers tend to qualify the “stick to the averages” advice by saying that perhaps there are a few pollsters that stand out above the rest and deserve at least a bit of special attention. That elite class includes Selzer’s Iowa polls, and the New York Times/Siena College’s national polls.
Both have won respect in past election cycles for their lack of herding — for seeing results coming that the national polling averages missed.
In 2016, Selzer’s final poll showed Trump up 7 in Iowa when other polls showed a closer contest. Trump won the state by more than 9 points. In 2020, most pollsters again showed a close contest, but Selzer found Trump up 7 and he won by 8.
So Selzer is a pollster who has not underestimated Trump — she accurately gauged her state’s support for Trump in the past two cycles. And her final poll shockingly shows Harris beating him by 3 points.
Theories to explain this have been flying around the political world. Has Selzer caught onto something unique happening in Iowa — perhaps a backlash against state Republicans’ harsh anti-abortion law? More grandly, some theorize that she could be one of the only ones capturing a national shift toward Democrats, one that those other herding pollsters refuse to believe.
Alternatively, nobody’s perfect, and even the best pollsters will be wrong sometimes due to random chance, so maybe she’s just wrong!
The New York Times/Siena College poll also has a reputation for avoiding herding, but its final swing state polls are a bit of a mixed bag for Harris.
Intriguingly, the Times polls do show Harris up 2 in Wisconsin (the tipping point state of 2020), up 3 in Nevada (a state where some early voting analysts thought Republicans looked strong), and up 2 in North Carolina (a state Trump won in both his previous runs). Trump held four rallies in North Carolina in the campaign’s final days, which some have interpreted as a sign his team is worried about their prospects in the state.
But the Times polls also show Michigan and Pennsylvania tied, perhaps a sign that Harris can’t rely on the Rust Belt after all. They also show a 1-point difference in Georgia that we probably shouldn’t put too much stock in.
Altogether, the picture is clear as mud. The polls are not telling us who will win. We could be set up for a gut-wrenchingly close contest. Or either candidate could outperform their polling by a few points and win pretty solidly. The only way to find out is to count the votes.
