The latest New York Times/Siena College poll has folks on the left feeling more than a little jittery. With Kamala Harris and President Trump running neck-and-neck across crucial battleground states, even seasoned analysts like the Times’ Nate Cohn are urging Democrats to brace for a possible Election Day “surprise.” After all, Trump’s supporters have been known to upend polling predictions in the past, showing up stronger than anticipated in both 2016 and 2020. Could it happen again? Cohn’s assessment suggests that yes, it very well might.
Trump is leading in the majority of the battlegrounds, holding a four-point advantage in Arizona and a narrow lead in Michigan. He’s also slightly ahead in Pennsylvania, Georgia, and even Nevada. According to Cohn, one of the telltale signs that Trump could outperform polls again is a phenomenon called “nonresponse bias.” Basically, Trump’s supporters have historically been less likely to pick up the phone and take surveys, meaning these folks often go under the radar until, well, Election Day itself.
This trend isn’t exactly breaking news, but what’s noteworthy is that Cohn’s team found that white Democrats were 16% more likely to respond to polling than white Republicans. In short, even if the pollsters try to account for this, there’s always a margin of error—especially if Trump’s base remains as elusive to pollsters as they were in previous cycles.
On top of this, Democrats are banking on a turnout from their base that we simply haven’t seen in recent cycles. Early voting patterns suggest that Democrats are leading there, but that’s far from a comfort blanket. With fewer voters than in 2020 casting early ballots, Democrats are putting their faith in the idea that large swaths of occasional voters—who skipped the midterms and didn’t vote for Biden in 2020—will suddenly come out to support Harris. And if they don’t? The map could shift red by the end of the night.
No matter how you slice it, Democrats should probably be sweating—because this race looks a lot tighter than they expected.