Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations
Our spring member drive has ended, but it's not too late to give. You have the power to help fund the essential journalism that keeps us all informed. Help us close the gap on our spring fundraising goal! GIVE NOW

Daybreak Poll Correctly Predicted Trump Would Beat Clinton

RENEE MONTAGNE, HOST:

Ahead of yesterday's election, popular polls had consistently showed Hillary Clinton as the likely winner of the presidential race. With President-elect Donald Trump's victory, many people this morning are wondering how could pollsters have gotten this so wrong. One pollster did get it right. The USC/LA Times Daybreak Poll initially seemed an outlier this election cycle. It predicted Trump ahead of Clinton for much of the year and especially in the last weeks of the campaign. And joining us is Jill Darling She's director of that poll. Good morning.

JILL DARLING: Good morning.

MONTAGNE: What was different about your survey that led you and it to such different predictions?

DARLING: Well, our poll actually is very different. It's an experimental alternative to traditional polling. So our poll uses a probability method. We ask the questions of a panel of respondents once a week. We asked the same people over time. And instead of asking an up and down question about which candidate they support, we asked them what percentage likelihood they have of voting for the candidates - Trump, Clinton or someone else - and then their percent likelihood of voting.

MONTAGNE: In other words, you're measuring lessening enthusiasm or greater enthusiasm in a way that's not normally measured in a poll. Is that right? And why would that make a difference?

DARLING: Yeah, so the questions that we ask are, what percent likelihood would you have of voting for Hillary Clinton, Donald Trump or someone else? And so, for example, if someone is a 90-percent Trump supporter and then something occurs, they might feel more like a 70-percent Trump supporter. But that would show up in a traditional polling method as a Trump supporter, period. But in ours, if there's enough of an accumulation of those shifts, you will see the percentage for Clinton going down.

MONTAGNE: And now, to just break in a little bit here, 3,200 American households, supposed to encompass the full diversity of America, long-term study, not just a one-time poll, this sounds interesting. But it was harshly criticized by other pollsters, by some news outlets, including The New York Times. People said this just wasn't going to give you the right answers. For instance, you had one young African-American man who's been seen by other pollsters as skewing the poll.

DARLING: So this - this one African-American man was heavily weighted because he represented a group who is very difficult to reach, young, African-American and a certain economic status. And so his weight was very high. And the high weighting of him was enough to change our overall estimate by a little bit less than a point. And it was, you know, an exciting headline for The New York Times to run. But it wasn't entirely accurate.

MONTAGNE: So is there a lesson here in what other pollsters should be doing going forward?

DARLING: Well, that's exactly what it is that we've set out to learn. And depending on what this outcome is, I'm not sure that we nailed this outcome either. We'll have to wait and see how this goes. But what we're looking at is just an alternative method that allows people to sort of express a little bit more uncertainty.

So we have taken an unprecedented step - fairly unprecedented - of sharing our data and our methods with other researchers, which is why we have had so much scrutiny, which we actually very much welcome and hopefully, you know, make polling more accurate as we go.

MONTAGNE: Jill Darling of the University of Southern California's Center for Economic and Social Research. Thank you very much.

DARLING: Thank you. Transcript provided by NPR, Copyright NPR.