For pollsters hoping to accurately forecast an election with Donald Trump at the top of the ticket, the third time may have been the charm — but it was still somewhat unsatisfactory.
Opinion researchers swung and missed in 2016 and 2020, systematically underestimating Trump’s level of support. But in 2024, the polls corrected — and both popular vote and swing state results landed within the margin of error for aggregate predictions.
Pollsters argue that they finally cracked the challenge of pinning down Trump voters, an elusive segment of the polling population that has caused polls to veer off the mark in previous election cycles when Trump was on the ticket.
“In the past, we have had a lot of Trump supporters who have simply refused to answer our questions,” said GOP pollster Whit Ayres. “We call, ‘I’m from the New York Times or The LA Times or The Washington Post, and I’m doing a survey,’ and they go, ‘Well, to hell with you,’ click.”
This is not to say there weren’t flubs. Most notably, gold standard pollster Ann Selzer, who is legendary in the industry for being right, ending her long relationship with the Des Moines Register this year after her firm released a poll a few days before Election Day showing Vice President Kamala Harris beating Trump by 3 points in Iowa.
The poll’s results were met with both shock and skepticism: even Harris’ campaign warned against reading too much into the results. Trump ultimately trounced Harris in Iowa by over 13 points.
Since 2016, when the polls predicted a landslide victory for Hillary Clinton, pollsters have faced issues with undercounting Trump’s base, a challenge unique to elections with Trump running, since polling was relatively accurate in the 2018 and 2022 midterms.
Turnout is typically higher in presidential election years — especially among those in Trump’s base who tend to distrust institutions and can be difficult to survey.
“The very same Trump voters who don’t trust experts, don’t trust the media, don’t trust science — also don’t trust pollsters,” said Democratic pollster Paul Maslin. “And we found in several states that simply, they were opting out.”
In order to correct course, researchers this year “jumped through a lot of hoops,” Ayres said, in order to reach greater numbers of Trump voters and more accurately predict the outcome. They adjusted modeling of the likely electorate, weighted more heavily certain demographic groups, and adjusted their outreach strategy to non-college-educated voters, who tend to lean Republican.
That manifested in final polling aggregates showing five of the swing states in a dead heat, with Trump having comfortable leads in Arizona and Nevada. The polls also accurately forecasted Trump’s strength with white voters and Harris’ softness with Black and Latino men, which contributed to Trump’s decisive victory.
The president-elect ended up sweeping all of the battlegrounds, an outcome pollsters attributed to undecided voters splitting unevenly — and perhaps surprisingly — toward Trump.
“If you look at the past, undecideds tended to not favor Republicans,” said GOP pollster Brent Buchanan. “This year, Trump benefited from undecideds.”
The results on election night, while not outside of what pollsters predicted, were still surprising in their sweep. In the lead-up to Nov. 5, many journalists and pundits braced for a long, drawn out process that would take days to resolve due to the closeness of the polls in swing states.
Although the final vote counts were within poll aggregates’ margin of error across the battleground states, they still tended to undercount his support by about 3 points. That could be because Trump benefited from voters who made their decision at the last minute, after the final surveys were conducted.
In the final week of the campaign, Trump’s campaign notched a coveted endorsement from Joe Rogan, to which Maslin attributed Trump’s last-minute surge. Buchanan’s models showed Trump up by 3 points among voters who made their choice in the last week of the election.
Pollsters have largely blamed Trump as the variable driving inaccuracies in 2016 and 2020. “It’s a Trump issue, not a Republican issue,” said Ayres, who has four decades of GOP polling experience. Looking toward 2028 — the first presidential election without Trump in a dozen years — Eyres predicted “we can reasonably expect whatever problems we had getting Trump voters to go away.”
Others weren’t so sure.
“It’s not like, ‘Oh, great, fine, polling was pretty good this time we can put this to bed,’” Maslin said. “No, of course not. It’s an ongoing question. It will continue to be, and it should be.”
