The New York Times’ chief political analyst warned Sunday that pollsters may not have solved the underlying issues they encountered in 2016 and 2020 and could once again be underestimating Donald Trump.
Times scribe Nate Cohn, while cautioning on X that he has “no idea whether our polls (or any polls) polls [are] ‘right,’ ” explained that inescapable survey bias may be spoiling the results because Democrats dramatically outpace white Republicans, a Trump voting bastion, when it comes to answering pollsters.
“Four years ago, the polls were thought to underestimate Mr. Trump because of nonresponse bias — in which his supporters were less likely to take surveys than demographically similar Biden supporters,” Cohn wrote in an analysis for the Times.
Cohn 36, one of the foremost election-data gurus, stressed it is difficult to track “nonresponse bias” but noted that he likes to review the level at which Democrats and Republicans respond to pollsters to try to gain insights.
“Across these final polls, white Democrats were 16 percent likelier to respond than white Republicans,” he said. “It raises the possibility that the polls could underestimate Mr. Trump yet again. We do a lot to account for this, but in the end there are no guarantees.”
Cohn also warned that Democrats appear to be suffering from a “diminished” typical early-voting edge, which could help Republicans on Election Day.
He mused that given the recent dichotomy between the two parties when it comes to early voting, it is possible Trump could get a boost on Election Day when GOP voters flock to the polls.
“The pattern is fairly similar across the battlegrounds: Democrats lead in early voting; Republicans lead with what remains, and in each case it’s not by the sweeping margins of four years ago, when the pandemic upended the usual early voting patterns,” he wrote.
“There is a little bit of a leap of faith here for Democrats: They’re counting on a lot of people to vote on Election Day who didn’t in 2020 or 2022. Their track records of voting give plenty of reason to think they will do so, but if not, the result will quickly look very different.”
Trump, 78, has historically railed against early voting, but this election cycle, his campaign has pushed for Republicans to cast their ballots early in a bid to maximize turnout among the base.
The New York Times and Siena College released a fresh batch of battleground-state polling Sunday that pegged six of the seven key close contests as within the margin of error, or too close to call. Arizona, which hewed toward Trump, was the sole exception.
“Usually, the final polls point toward a relatively clear favorite, even if that candidate doesn’t go on to win. This will not be one of those elections,” Cohn said of those results.
One finding in the New York Times/Siena College poll that caught the interest of many polling junkies was evidence that late-deciders could be favoring Vice President Kamala Harris.
Given the jump-ball nature of the race, that could be crucial to the race’s outcome. About 8% of voters claimed to have made their decision recently. Of that group, 55% went for Harris compared to 44% for Trump.
Overall, Cohn underscored that he simply doesn’t know how the election will play out and that “no one does.”
Some survey junkies have been fretting that the pollsters have been “herding” survey results to ensure that their polls aren’t an outlier.
Experts such as Nate Silver have contended that there has been far less variety within polling of the 2024 presidential contest than there should be.
Cohn has similarly raised concerns about poll “herding” and outlined ways in which surveys could be overestimating Republicans as well.
“Many pollsters (not us) have adopted heavy-handed practices that yield more Republican-leaning samples, out of potentially but not necessarily justified fear of systematically failing to reach Trump voters again,” Cohn wrote on X. “The polls are way more sensitive to turnout this cycle.”
Trump campaign pollster Tony Fabrizio penned a memo in response to the latest survey, arguing its samples are historically skewed toward the left.
“The New York Times is polling an electorate that looks far to the left of 2020, despite hard facts of voter registration and early voting indicate a current electorate that has shifted only to the right in every state,” Fabrizio wrote in a memo shared by the campaign.
“The New York Times even helpfully admitted in their story that they had a harder time reaching Republican voters than in their 2020 polls, which were notoriously inaccurate,” he added in reference to Cohn’s analysis.
At the moment, Trump and Harris are tied in the most recent RealClearPolitics aggregate of multi-candidate national polls.
Trump is favored in most of the RCP’s battleground state polling averages.
