IRA FLATOW: What would an election season be without the polls?

SPEAKER 1: The latest Quinnipiac poll is showing Ben Carson surging past front runner Donald Trump in Iowa.

SPEAKER 2: There’se also new polling from the Des Moines register. Donald Trump has five points ahead of Ted Cruz in that polling.

SPEAKER 3: Last month about a group organizing Iowa’s first online straw poll, and now the results are in. Donald Trump is leading GOP candidates with 34 percent of the vote. Rand Paul comes in close second in this straw poll with 18 percent.

IRA FLATOW: Yeah, those polls all missed their mark. And for every voter type, state, caucus, and candidate, it seems like there’s a poll for that. And with the proliferation of polls, online surveys, and big data, you’d think that someone would be able to get it right. Why is it so hard?

Well, my next guest says that election polling is in near crisis. Cliff Zukin is professor of public policy and political science at Rutgers University of New Brunswick, New Jersey. Also, founding director of the Center for Public Interest Polling. Welcome to Science Friday.

CLIFF ZUKIN: Thanks. Very nice to be with you.

IRA FLATOW: Why do you say that polling is in a crisis? How bad is it out there?

CLIFF ZUKIN: Well, now I want to make a distinction first. I’m not saying the polling is in crisis, I’m saying that election polling is in a crisis. I think, in fact, most public opinion polling is still quite good. But, election polling is becoming more difficult to do, as is all polling.

And it’s a combination of two or three things. First is this tremendous growth of cell phones that we didn’t have to worry about in 2008 and 2012. Maybe 10 years ago the country was 6% cellphone only, now it’s 43%. And the country’s 60% cellphone only or mostly. And cellphones or harder for us to do, they’re much more expensive for us to do given federal regulation.

IRA FLATOW: And you mentioned in your work that the response rate has just dropped like a rock since like 1970.

CLIFF ZUKIN: Yeah, when I first started doing polling we would get an 80% or 85% response rate, we would worry about the 15% we didn’t get. When you looked at all the surveys done by the Pew Research Center over the last year or so, the average was 9% response rate.

IRA FLATOW: From 80% to 9%.

CLIFF ZUKIN: Yeah, and so–

IRA FLATOW: So is there any reason why you’re not getting an accurate result, with 9% just 9% response rate.

CLIFF ZUKIN: You know, it’s really something that no one understands. When we compare results of general public surveys on smoking and health behaviors that we do in our normal surveys with the gold standard surveys of the American Community Survey in the health and the census, where participation is 100% because it’s mandated, surprisingly the surveys are still pretty accurate. But I have no idea why that works. If you had told me that we’d be just as accurate with a 10% response rate as an 80% response rate, I’d say you’re crazy.

But what it has done is to drive up the cost of polling, to making many, many more calls. And cell phones are more expensive calls. And that’s really what’s changed the economics of the profession in the last 5 to 10 years.

IRA FLATOW: I’m Ira Flatow, this is Science Friday from PRI, Public Radio International. Talking about Cliff Zukin about political polling. So, how do you make it better? I mean, or you just can’t make it better? Or more reliable?

CLIFF ZUKIN: Well, I don’t think we’re going to be ready in time for 2016. I think the internet is the future. And telephone polling is going to be so expensive to do compared to the cheapness of internet polling that I think that’s where we’re going to go.

But we have problems with internet sampling right now. First, not everybody has internet. It’s older people who don’t have internet coverage and those people are the most likely to vote. And then second, we don’t have a good way to sample. Since this is Science Friday, it’s nice to be with a group that you can say there’s a scientific basis for public opinion polling and sampling, based in inferential statistics. And that’s what we use, we use probability samples. There’s no good way right now to do an internet based probability sample where people opt in. We just don’t have that developed yet for 2016.

IRA FLATOW: And we’re not anywhere close to it, yet.

CLIFF ZUKIN: No, I think we will by 2020. I know there’s an awful lot of experimentation going on. This is not evil intention, or anybody trying to pull any of the wool over– there a lot of people in the industry working very hard to figure out how to do a non-probability sample that’s accurate. It’s just, we’re not ready for this election, I don’t think.

IRA FLATOW: Could it be that the questions on the poll just aren’t the questions that people are thinking about, so they don’t answer them?

CLIFF ZUKIN: No, I don’t think so. These are questions that have worked pretty well. You know, so, we were very, very accurate in 2008 and pretty accurate in 2012, and I don’t think the electorate has changed that much. I think in Iowa a number of things happened. There was some movement before, but polls are very–

The third thing that’s very different about election polling is that it’s incredibly difficult to figure out who’s actually going to vote from those people who tell you they’re going to vote. And in Iowa we had about a 20% turnout. And when you look at exit polls it seems that the evangelical turnout was very strong and that benefited Ted Cruz and Rubio. And Trump supporters were less likely to go to the caucus, and more likely first time caucus goers that didn’t know the rules.

And then polls are just horrible at measuring the ground game of get out the vote, and organizational effort, because we don’t really talk to that when we do random samples of the electorate. Also, Iowa’s an entrance poll, it’s not an exit poll, so it gets people when they go in. But finding likely voters is the biggest problem of election polling.

IRA FLATOW: A word that gets thrown out a lot by pollsters and analysis is outperforming, or underperforming, if the results don’t match the poll. Isn’t that sort of blaming the people for the inaccuracy of your poll?

CLIFF ZUKIN: Yeah, and it’s an unhealthy dynamic in the presidential nomination race in that what you have is the end results, where one person got 24 and the other 21, or whatever, being judged not by x percent of delegates selected to the nomination convention, but by the expectations of the polls. And so now you see, because Cruz did better than expectations that were set by polls, he’s able to command more attention in New Hampshire and set his own agenda.

And so, in some ways, the expectation set by election polls is an unhealthy dynamic in the way Americans choose a president. No question about that.

IRA FLATOW: Well, there’s something more to think about. We’re going to check back in with you every now and then, OK Cliff?

CLIFF ZUKIN: Yep. Happy to talk with you.

IRA FLATOW: Cliff Zukin, professor public policy and political science at Rutgers University in New Brunswick, New Jersey.

Copyright © 2016 Science Friday Initiative. All rights reserved. Science Friday transcripts are produced on a tight deadline by 3Play Media. Fidelity to the original aired/published audio or video file might vary, and text might be updated or amended in the future. For the authoritative record of ScienceFriday’s programming, please visit the original aired/published recording. For terms of use and more information, visit our policies pages at http://www.sciencefriday.com/about/policies.