The polls underestimated Trump's support -- again. Here's why

President-elect Donald Trump presents unique challenges for pollsters. Polling has underestimated Trump's support in the last three presidential elections.

Changes in approach to surveys — reaching people on cellphones, online and in different languages — have not corrected for Trump's numbers.

National polls did accurately reflect support for Vice President Harris, though (around 47%).

Trump's result — which looks like it will land at around 50% of voters when all ballots are counted — is about 3 points higher than what the polls showed before the election.

The difference is within the margin of error, roughly +/- 3 or 4 points, but it's significant because there's a consistent 3 point undercount across the seven swing states.

One reason Trump supporters are undercounted could be that Trump disparages pollsters and the media, creating a sense of distrust that could dissuade people from participating in a survey.

It could also simply be that the shape of this election was unique, and pollsters were unable to account for it. White voters went up as a share of the electorate for the first time in decades, for example. But late deciders also broke for Trump by double digits, which could be part of the polling miss, too.

This election result is also a reminder that it's hard to predict who will ultimately turn out to vote.

What the polls didn't show

This year, the electorate defied historical trends.

"There was a reversal of the long-standing trend of the electorate getting less white and more people of color," said Lee Miringoff, director of the Marist Institute for Public Opinion, which conducts polling for NPR. "In 1992, it was 87% white and 13% people of color. This was the first reversal — to 71% white."

The importance of that can't be overstated. Republicans have dominated with white voters in the 21st century. If that voting bloc turns up at a bigger margin than expected, that means good things for the GOP.

White voters without college degrees are especially critical for Trump. Two-thirds of them voted for Trump in the last three presidential elections. But they were 5 points higher as a share of the electorate this year than 2016, and 4 points higher than 2020.

That defies the fact that they are a shrinking share of the eligible voting population — and that they are among the least likely to vote in presidential elections.

"The so-called low-propensity voters, which Trump gambled on showing up, did vote for him," Miringoff added. "Likely voter models, which incorporated enthusiasm, understated these low-propensity voters. And although it looks like there were an unusually high number of ticket splitters, it was really 'Bullet Voters' — folks who voted for Trump and then left. This resulted in Democrats doing better in Senate contests compared to the top of the ticket."

Another factor that could have thrown pollsters off of Trump's scent: Trump won late deciders by double-digits.

Those who said they made up their minds in the last few days broke for Trump by 6 points, while those who said they made up their minds in the past week, cast their ballots for him by 12, according to publicly available exit polls conducted by Edison Research and paid for by the broadcast networks. (NPR did not pay for exit polling.) Their last-minute decisions wouldn't have been reflected in earlier pre-election polling.

Pollsters say that surveys are snapshots of a particular moment in time and not meant to be predictive.

What the polls did show

That's not to say the polls didn't have value. They showed storylines that played out in this election.

For example:

Harris, and Biden before her, lagged with Latinos and younger voters all cycle.

Trump made inroads with younger men and younger men of color.

Harris surged after first getting in, but then her leads evaporated after a month of negative advertising from Trump allies.

The economy and immigration were vital to voters — and they preferred Trump on those issues.

Toward the very end of the race, the polls, like the final NPR/PBS News/Marist poll before Election Day, found that the gender gap might not be as wide as earlier polls suggested.

"In the whole, they provided a useful narrative that the battleground states were always close," Miringoff said.

But the country is in a time of political realignment. Blue-collar voters who had traditionally been Democratic voters have switched to being Republicans, by and large. On the other hand, wealthier and more educated voters have moved toward Democrats.

And when that kind of crack up happens, it produces volatility — and makes polling that much more difficult.