Where Polling Went Wrong In The 2020 Presidential Election
DAVID GREENE, HOST:
We got to ask, what is the future of political polling after the failures we saw in this election? Joe Biden won the presidential election, but it was a lot closer than polls suggested. And there's evidence that political polling this year was even further off than it was in 2016. So let's talk this through with our own senior political editor and correspondent, Domenico Montanaro. Good morning, Domenico.
DOMENICO MONTANARO, BYLINE: Morning, David.
GREENE: So it seemed pretty clear on election night that the polls were off in the presidential race. I mean, now we've had a little more time to count votes. I mean, what can we say about how far they were off?
MONTANARO: Well, they were off particularly in the Midwest, where there are lots of white voters without college degrees, and in many traditional Republican states. You know, some of the biggest misses in swing states were in Wisconsin, Ohio and Iowa, where the polls were off by about seven points each. Republican states like Montana, Indiana and Missouri, they were also polling in single digits before the election and all wound up being blowouts.
You know, national polls were off, too. They had Biden winning by eight. It's only three. It might get to about four or so when all the counting is done. But Biden's number, you know, was actually pretty close. It was Trump's support that was consistently underestimated. The NPR/Marist poll, by the way, had Biden taking a double-digit lead over Trump in mid-October. That was a pretty long way off, obviously, from what actually happened.
GREENE: So are there explanations for this stuff?
MONTANARO: We don't have all the answers yet. But there are some theories that are starting to take shape. And pollsters are starting to look into them. Some pollsters I've talked to are pointing to the huge and unprecedented share of early votes. They said that when they're looking at their pre-election samples, they're seeing that they had a higher share of people who were voting early than actually did. And we know that those voters leaned heavily toward Biden.
They're also asking if the reporting on the huge early vote leads for Democrats actually wound up turning out more of Trump's supporters in the final days than might have shown up if that didn't happen. You know, there's some evidence for that because in Arizona, for example, where the gap closed considerably in Trump's favor, were all votes that were dropped off the day before Election Day or on Election Day itself. And the other thing that they're looking at is simply whether Trump supporters just don't want to talk to pollsters.
GREENE: Oh, yeah. That theory has been out there, the whole idea of the shy Trump voter. I mean, is there actually evidence of that? Or is it just a theory?
MONTANARO: Yeah. And it's not quite the shy Trump voter thing. You know, there are shy Trump voters. I was talking to a Republican pollster who said shy Trump voters obviously do exist. They tend to be women. They're reluctant to say in social settings that they're supporting Trump. But I asked him, does that mean that they're not telling pollsters that they support him? And he said, no. It's more about them just not wanting to talk to pollsters at all.
The Pew Research Center actually ran an experiment on this after the 2016 election. They had people talk to live questioners and take an online survey. The thinking was that maybe they'd be more honest online since they didn't have to say their preference out loud. But the experiment actually found no significant difference. Basically, the results were the same. Trump is clearly, though, the big factor here. And pollsters, you know, they're scratching their heads. They're wondering if this was a black swan event that only applies when Trump is on the ballot or if there's something more systemic going on.
GREENE: Do you still trust polls?
MONTANARO: (Laughter) You know, look; I've said often I think polls are best used when broadly looking at attitudes and trends. I put a lot less stock in horse race polls this year than in 2016. And I'll look even more skeptically at them in coming cycles. You know, we have to remember that politics is a combination of art and science. Data has a place. Pollsters argue if we don't have an objective baseline, all we'd have is spin from campaigns. But talking to voters is still very important.
GREENE: NPR's Domenico Montanaro. Thanks.
MONTANARO: You got it.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.