In 2016, Democrats headed into Election Day confident they’d win the White House. Many election forecasts were confident of that. But of course, that's not what happened. Donald Trump landed on the lucky side of a normal polling error.
In 2020, the polls were off by even more, but it's easy not to dwell on that since Joe Biden won.
Four years ago, Biden led Trump by over 10 percentage points in mid-October, according to 538’s national polling average. As the election drew nearer, Biden even saw his margin in some swing-state polls hit double digits, such as when The New York Times/Siena College—arguably the country’s best pollster—showed Biden leading by 11 points in Wisconsin.
But Biden went on to win the Badger State by just 0.6 points. And his national popular-vote lead? Barely over 4 points, and roughly half of the 8.4-point polling lead he had on Election Day.
So what does that mean for the 2024 election, in which polling has consistently shown coin-flip odds?
Third time’s a (c)harm
If this year’s polls were to underestimate Trump’s performance as they did in 2016, Harris would lose every swing state but Nevada, and Trump would net around 306 electoral votes. But if polls underestimate him as much as in 2020, Trump would take all the swing states and Nevada too. The margins wouldn’t even be that close.
Of course, the polls could be underestimating Harris, not Trump. Pollsters may have overcorrected for past errors that underestimated Trump’s performance. (Polling is incredibly hard to do well!) And polling errors don’t necessarily have to benefit Republicans, of course.
538’s Nathaniel Rakich found that polls in the final 21 days of a general election cycle overestimated Republicans in two out of the six presidential elections since 2000, each time by about 2.5 points. It’s possible that we’re due for another overestimation of Republicans.
It happened a little in 2022, after all. Following the Supreme Court’s elimination of the constitutional right to abortion in June 2022, Democrats in many statewide races outperformed their polls. For instance, 538’s final polling average showed now-Sen. John Fetterman half a point underwater against Republican Mehmet Oz. But Fetterman ended up winning by a relatively comfortable margin of 4.9 points.
A Pew Research Center survey of national pollsters found that most used different methods in 2022 than they used in 2016 and 2020—and early data suggests it was worth it. Across the whole electoral map, polls in the 2021-2022 cycle were historically accurate, overestimating Democrats by just an average of 0.8 points across Senate, House, and gubernatorial races, according to Rakich’s analysis. That makes for the smallest Democratic bias in an election cycle since the 2009-2010 cycle.
But given Harris’ incredibly narrow leads in battleground states, even a minor error in favor of Trump could tip him into an Electoral College victory. If 2022’s D+0.8 bias were to happen this year at the presidential level, Harris would lose every swing state on the chart above.
Using midterm results to project presidential election results is also fraught. For one thing, voters are generally more open to splitting their ticket between a presidential candidate and a downballot candidate (though that’s becoming less the case.) It’s why Vermont’s Republican governor won reelection in 2022 by 47 points in a state that Biden won two years earlier by over 35 points.
Another reason is that turnout during midterms is much lower than during presidential elections. Midterms also tend to see higher turnout among more highly educated voters—a group that leans ever more Democratic. These factors may help to explain why Democrats have performed close to or better than their polling in the past two midterms but quite a bit worse in the past two presidential elections.
Put simply, it seems unlikely—but not impossible—Harris will score the margins seen on the 2022-related chart above.
Worse for Harris (slash our democracy) is that polls are generally more biased toward Democrats. While the degree of statistical bias is hard to predict from cycle to cycle, polls have been biased an average of 1.2 points toward Democratic presidential candidates since 2000, according to Rakich’s analysis. And if you goose Trump’s national and swing-state polling by that amount, he would win the Electoral College while losing the popular vote. Again.
Left-leaning pundits online often accuse Republican pollsters of “flooding the zone,” or releasing scores of biased polls with the hopes of tipping the polling averages against Democratic candidates. To be sure, you may run across some low-quality pollsters with eye-grabbingly odd results.
But as The New York Times’ Nate Cohn reported on Monday, those partisan pollsters just aren’t having a large effect on the Times’ averages. In fact, not a single state would see a change in the polling leader, and at best, Harris’ lead would grow from less than 1 point to a solid 1 point. (And Cohn isn’t the only one finding these results.)
So what’s the good news?
Harris still has a polling lead, however narrow. That means Democrats need only for the polls to be very, very accurate for her to win.
One good sign for her is her blockbuster fundraising, with her campaign and its affiliated party committees having raised over $1 billion since her entrance in the race. That suggests she has the enthusiasm edge. And the Harris campaign’s ground game is far more developed than Trump’s, which is “relying on well-funded but inexperienced outside groups” for larger demographic outreach, The New York Times reported on Sunday.
The Harris campaign also seems to be targeting people more likely to vote. Meanwhile, Trump's going after the “bro vote,” as Politico put it recently. Going after low-propensity voters is a gamble—and Harris isn’t giving up this demographic without a fight. She may soon hop on the podcast of the patron saint of bro-dom, Joe Rogan.
Better yet, Black women, a key base of the Democratic Party, appear especially enthusiastic to vote. Tom Bonier, a senior adviser to data firm TargetSmart, recently pointed out that the early-vote turnout among Black women in Pennsylvania had starkly increased compared with 2020. In 2020, Biden won 92% of Black voters, according to a study by Pew Research Center, though there are signs that Trump is gaining with voters of color.
But however much support she may be losing with voters of color could be made up by her appeal to other demographics.
“Harris is doing the best for a Democratic presidential candidate among white women in the 21st century,” CNN’s Harry Enten said on Tuesday, adding that her better support among white women, who make up the plurality of the electorate, is likely tied to abortion rights. And as noted above, that was likely a critical factor in 2022, which was the first midterm or presidential election since the Supreme Court overturned Roe v. Wade.
And yet …
The biggest worry for Harris is that despite everything—despite her massive fundraising advantage, her excellent ground game, her obliteration of Trump in their Sept. 10 debate, plus all the Democratic enthusiasm around her—she leads by only 2.0 points nationally. And because of the Republican bias in the Electoral College, that might not be enough.
As we saw in 2000 and again 2016, a Democrat winning the popular vote—whether it’s Al Gore or Hillary Clinton—isn’t enough to guarantee they win the White House.
We can’t currently know whether Trump will outperform his polls once again. But polls are not intended to be the final say on anything. And with high levels of Democratic turnout, they can be beaten.
We need your help if we’re going to defeat Trump, Vance, Project 2025, and Republicans up and down the ballot. Click here to volunteer to write letters so we can increase voter turnout.
