How did the polls do in 2024?

President-elect Donald Trump has long presented pollsters with a challenge. We look at what polling underestimated and what it accurately foreshadowed in this election.

Transcript

ARI SHAPIRO, HOST:

President-elect Donald Trump outperformed polling in all seven swing states this election. He has kept pollsters on their toes ever since he first ran in 2016, and they still haven't quite accounted for his impact. Here to talk about what we learned and we didn't from polls this year is NPR senior political editor and correspondent Domenico Montanaro. Hey there.

DOMENICO MONTANARO, BYLINE: Hey, Ari.

SHAPIRO: To start big, what is your top takeaway from the accuracy, or lack thereof, of polls this election?

MONTANARO: Well, they weren't terrible, but they underestimated Trump again. I mean, that's been the story in each of the last three elections when Trump has been on the ballot, and this was no exception. The polls, I have to say, did get Kamala Harris' number right. The final FiveThirtyEight polling had the vice president at 48, and it looks like she's going to get 48 in the popular vote. But Trump was at 47 in the pre-election polling average, and he's probably going to be about three points higher than that when the popular vote is all counted. And by the way, we should remember those leads in the pre-election polls are not, you know, statistically significant, really, because there are margins of error. They're between three and four points, and that means that those leads could have been three points lower or three points higher.

SHAPIRO: Yeah, but the difference wasn't scattered. They were all underestimating Trump by about three points. So what are pollsters saying happened?

MONTANARO: Yeah. We can only make some informed guesses at this point after looking at the numbers and chatting with our pollsters. But a couple things - first, it's possible that the polls were pretty dead on but weren't picking up who late deciders were going to be breaking for. You know, the exit polls found that people who decided days or a week before the election voted for Trump by double digits. It's also possible that the uniqueness of the shape of this electorate was the real issue. Here's Lee Miringoff, director of the Marist Institute for Public Opinion, which conducts polling for us.

LEE MIRINGOFF: The electorate shifted in an opposite direction from what it had been over the last eight elections. In 1992, when Bill Clinton was elected, the electorate was 87% white and 13% people of color. And those numbers have steadily changed, with the white group going down and people of color increasing election by election until this election.

MONTANARO: So in 2020, 67% of the electorate was white. This time, it was 71%, and that meant more Trump voters. More specifically, white voters without college degrees, two-thirds of whom voted for Trump, were four points higher than in 2020. They usually vote at lower rates, but Trump's team put a lot of work and effort into getting them to the polls. They gambled on them, and it paid off. That meant all those likely voter models showing Harris with an advantage before Election Day wound up being flawed. They didn't account for that increase because it would defy trends and was unlikely to happen, but it did.

SHAPIRO: And yet this is the third election in a row where Trump has presented this kind of unique problem for pollsters.

MONTANARO: Absolutely. I mean, you know, pollsters, you know, have gotten better since 2016 in how they're conducting these polls. There's more by cellphone, text and online, in English and in Spanish. You know, they're trying to reach people really where they are. But when Trump's on the ballot, they've clearly had problems figuring out who's going to actually turn out to vote.

Maybe this isn't really surprising, I mean, given that we're in the middle of a political realignment in America. Trump has attracted more blue-collar voters who used to lean Democratic. On the other hand, college-educated and wealthier voters have migrated toward Democrats for the most part. And those blue-collar voters are a group that has certainly leaned Democratic, but those more white-collar voters are the ones that had leaned Republican for decades. Those kinds of dynamics create a lot of volatility in society and in polling.

SHAPIRO: Is the lesson here just don't trust polls?

MONTANARO: No. That's not exactly the lesson because, you know, I have been saying for some time that horse race is not the best use of polling, especially when things are this close all the time. But they did show us that even though Harris had the momentum after she was picked, she wasn't able to sustain that lead.

So there were some lessons that the polls did teach us. You know, they, you know, were valuable because they showed just how much Harris was trying to swim upstream. I mean, the vice president - it was very difficult for her to separate herself from Biden. His approval ratings were very low, the polls had shown. Inflation and immigration showed up repeatedly as important to voters, and they trusted Trump more to deal with those issues. And against Trump, Harris was lagging with Latinos, men of color and younger voters at what wound up happening in the end. So we should always take what we're seeing in the polls with a grain of salt and really embrace that we're not - they're not meant to be as specific or predictive as people want them to be.

SHAPIRO: NPR's Domenico Montanaro. Thank you.

MONTANARO: You're welcome. Transcript provided by NPR, Copyright NPR.