How did 2024 election polls fare? Here’s what the experts say.

Political pollsters still haven’t got a bead on Donald Trump. For the third straight presidential election cycle, public opinion polls underestimated support for the president-elect - although the gap was smaller than in 2016 and 2020.

The continued discrepancy between poll results and the margin of Trump’s win likely reflects broad shifts in how much Americans engage with public-opinion research. Trump supporters may have been less likely to respond to surveys, continuing a trend seen in 2020, according to polling experts who spoke to The Washington Post. And a noticeable swath of Trump voters in 2024 had not consistently voted in the past or were otherwise less politically engaged. His campaign this year zeroed in on turning out lower-propensity voters, including young men.

All of those factors made it more difficult to get a precise picture of the electorate, experts said. While pollsters have gotten better at that since 2016, when Trump was first on the ballot, the full picture of 2024 performance won’t be clear until detailed final election and turnout data is available - giving polling experts a chance to compare it to their pre-election data.

Over the past three election cycles, pollsters have adjusted their methods to address challenges in capturing support for Trump - and to some extent, it has worked. State presidential polls underestimated Trump’s margin against Harris this month by 2.7 percentage points, according to data reporting website 538. That’s down from a 4.2-point underestimate of Trump’s vote margin in 2020 (when he lost to now-President Joe Biden) and 3.2 points in 2016 (when he beat Hillary Clinton). And more than 9 in 10 polls in the seven competitive presidential states came within the margin of error for the actual Harris-Trump vote.

Many still-uncounted votes in California make the exact performance of national polls this cycle unclear, but they appear to have performed similarly to state surveys. The Post Pulse election model forecasts Trump will win the national popular vote by 1.1 points and Edison Research expects his margin to be between 1.5-2 points, while national polling averages found Harris ahead by 1.1 points, suggesting they underestimated Trump’s margin by 2-3 points. Trump overperformed polling averages by 2.3 points across the seven battleground states, winning by about one point in Wisconsin and Michigan, two points in Pennsylvania and Georgia, three points in Nevada and North Carolina and nearly six points in Arizona, with roughly 96 percent of votes counted as of Wednesday evening.

“There’s a sort of unique loyalty to him, not necessarily to Republicans up and down the ballot,” said Chris Jackson, a public opinion researcher at Ipsos.

Across the country in this general election, Trump won in a some states where Republican candidates down the ballot did not perform as well. In Arizona, where Trump won comfortably, Democrat Ruben Gallego defeated Republican Kari Lake, who the president-elect endorsed, for a Senate seat. In Wisconsin, Trump won by about one point while Republican Senate candidate Eric Hovde lost by about the same margin to Democratic incumbent Tammy Baldwin.

This year’s polls are an improvement over those from 2020, which showed the biggest mismatch with election results in decades. Those polls had errors of “unusual magnitude,” according to a report from the American Association for Public Opinion Research (AAPOR).

Among the factors that could explain those errors, AAPOR said, was that Trump had “provided explicit cues to his supporters that polls were ‘fake’ and intended to suppress votes.” The report noted that those comments “could have transformed survey participation into a political act whereby his strongest supporters chose not to respond.”

But whether Trump’s remarks on polling resonated with his supporters would be hard to study and quantify, said Courtney Kennedy, a polling methods expert at Pew Research Center.

Since the 2016 and 2020 election cycles, more pollsters have been weighting for political variables, including partisanship and how poll participants voted in previous elections - something that used to be “taboo” in polling, Kennedy said. But she said those adjustments likely helped improve the representation of Republicans and Trump supporters in 2024 polls.

Lee Miringoff, director of the Marist College Institute for Public Opinion, called the 2.7-percent underestimation this year a “solid number” for “an instrument that’s trying to shoot at a moving target.” Beyond the numbers, Miringoff said this year’s polls did a good job of delineating the top two issues that drove voters to turn out - the economy and immigration.

Jackson, the Ipsos pollster, agreed. Polling as a whole, he said, was useful this year when used as a way to “understand where Americans are” and the specific issues that motivated them to turn out, rather than as a forecasting tool in what appeared to be an exceedingly close race.

Moving forward, experts said the polling industry will have to continue reevaluating its methods, including by considering different weight adjustments such as partisanship and how people voted in previous elections, among other factors, for the next election and expanding options for how people can take surveys beyond online or phone-only methods.

Offline methods, like mailing survey forms, seem better suited to reaching conservatives, Republicans and Trump supporters, Kennedy said.

She added that pollsters also will likely have to continue producing “more elaborate” statistical models to account for low response rates, though these models will look different than traditional polling methods.

“We have to seriously engage with that challenge and really do everything we can to make sure that Republicans, as well as Democrats, are fully represented,” she said.

Mark Blumenthal, a pollster and longtime analyst of industry trends, said polls performed surprisingly well in 2024 “given survey response rates are so low. It’s miraculous they came to within about a point of each candidate’s support.”

“We can still measure the broad strokes of public opinion accurately enough to report those findings with confidence even if our surveys aren’t precise enough to forecast a very close election.”