GEOFF BENNETT: While votes are still being counted in some places from last week's election, Donald Trump is likely to be the first Republican presidential candidate to win the popular vote in the last 20 years.

And yet preelection polls have consistently underestimated his support since he first ran eight years ago.

And that's left some folks spending the last week trying to figure out what happened.

NPR's Domenico Montanaro is here to help us -- walk us through what the preelection polls got right and what they missed.

So let's start, Domenico.

First, it's good to see you.

DOMENICO MONTANARO, Political Editor, NPR: Yes, you too.

GEOFF BENNETT: So help us remember what the polls said, what they said about support for Donald Trump and Kamala Harris nationally and in the battlegrounds.

DOMENICO MONTANARO: Well, I thought it was pretty interesting.

If you look at the national polls, they got Kamala Harris' number exactly right.

If you look at the FiveThirtyEight final polling average before the election, it was 48 for Harris, 47 for Trump.

What did Harris wind up getting?

She's at about 48.2 in the polls nationally or in the final election result, Trump, though, at 50 percent.

So the polls underestimated Trump's support again, which is what we saw in the last two elections and did so by about three points on average.

And when you look at the state polls, it was also the same sort of story.

Across all seven battleground states, Trump's support was underestimated by anywhere from two to four points.

And that -- they got pretty close to what Harris' number was, but there are some things that the pollsters are going to have to look at for the next election.

I feel like we say that every election about what the polls are going to have to do.

GEOFF BENNETT: I remember back in 2016 talking to voters and you would say, who are you going to vote for?

And they'd say, oh, I don't like either of them.

And it would take five or six questions to get people to ultimately say, I'm going to vote for Trump.

Like, the response bias is real.

Is that what accounts for the undercount this time around?

DOMENICO MONTANARO: No.

I mean, the response bias is real, but that's not what accounts for what was the - - what we saw as the undercount, because the pollsters have really adjusted.

They have tried to really get more people to be able to respond.

Since the 1980s, it's gone way, way down.

They have done a lot of work on mixed modalities.

So it used to be live callers were the key gold standard.

Now they're doing them online.

They're doing them by text.

They're doing them by cell phone in English and in Spanish.

So they're trying multiple different things to try to get people.

But what we really saw here is a couple of different factors.

Number one, when you look at the exit polls, Trump won late deciders by double digits.

When you looked -- when they asked, who did you -- if you were deciding in the last few days, Trump won them by six points.

In the last week, Trump won them by 12.

And that was 7 percent of the electorate overall.

So you do the math, that certainly could mean what the makeup for those three points that we saw for the undercount.

The other thing here, though, that I think is really important is that the white vote went up in this election.

And, OK, you might say, well, why would that even be a thing?

Of course, that could be a possibility.

Not really.

I mean, in the last three decades, we have seen the white eligible voting population go down significantly.

I mean, this is the first election since 1992 that the share of the white vote went up.

GEOFF BENNETT: Really?

DOMENICO MONTANARO: It went from 67 to 71 percent.

And that's because of the growth we have seen with Latinos, the growth we have seen with Asian American voters.

So the pollsters were not factoring in an increase in the white vote.

They would have thought maybe it'd be 66, 65 percent.

And that also could have thrown the polls off a little bit.

GEOFF BENNETT: So what trends did the polls accurately capture ahead of Election Day?

DOMENICO MONTANARO: You know, there were a lot of things that the polls got right in this election, frankly.

I mean, if you were to ask me four-and-a-half weeks ago who's going to win this election, if you were thinking there might be a two- or three-point bias in the polls because of what we saw the last time, you would say Trump's probably going to win, because Harris had a pretty sizable lead, four points or so in the national averages.

But then that just collapsed after about a month of just a relentless attack campaign from the Trump folks trying to define Harris.

They were able to do so and brought her numbers -- or at least brought his numbers up for those undecided voters.

So I think that that was a major part of what we saw in this election, but also the storylines that we saw, the effect that Harris was struggling with Latino voters, struggling with young men in particular, young Black men as well.

And we saw that obviously play out on election night.

GEOFF BENNETT: Despite Donald Trump sweeping the battlegrounds, there are downballot Democrats who won Senate seats in four of those states.

So, in Arizona, for instance, Trump won by six points, but the Democrat in the Senate race, Ruben Gallego, won by 2.5 points.

And then, in Nevada, Trump won by three, but the Senator Jacky Rosen won by almost two.

So what explains this split, where people are voting for Donald Trump, but they're also voting for the Democrat.

DOMENICO MONTANARO: Right.

Well, people would say, oh, ticket-splitting is back, right?

Must be that people voted for Trump, but then also voted for the Democrat.

Not the case.

What we found and in talking to our pollsters, they said that look at the number for Donald Trump and the number for the Senate Republican who ran and vice versa, and what we found is they're calling it bullet voters, people who went in and voted for Donald Trump, but left and didn't vote for the Senate candidate at the bottom of the ticket.

Because, overall -- I mean, I crunched the numbers on the way over here.

In Pennsylvania, for example, the Senate candidate was 144,000 votes fewer than for Donald Trump, 172,000 in Arizona, 96,000 in Michigan.

That's really significant.

And the Democrats who ran didn't do -- almost hit Kamala Harris' number and in some places were higher than what Harris got.

GEOFF BENNETT: Bullet voters, I'm writing that in my notes, because I have never heard that before.

Domenico Montanaro... DOMENICO MONTANARO: Me neither, until I talked to them, yes.

I asked about it.

GEOFF BENNETT: Always good to talk with you, buddy.

Thanks for coming in.

DOMENICO MONTANARO: Good talking to you too.