Washington Wednesday - Parsing the presidential polls

MARY REICHARD, HOST: It’s Wednesday the 9th of September, 2020. Glad to have you along for today’s edition of The World and Everything in It. Good morning, I’m Mary Reichard.

NICK EICHER, HOST: And I’m Nick Eicher. First up: taking the pulse of the presidential campaign

Pollsters started tracking the race to the White House months ago. But with less than eight weeks to go until Election Day, analysts are beginning to take those numbers much more seriously. And with new polls from across the country coming out almost daily, there’s plenty of data to pore over.

REICHARD: As of today, most of the polls show Democrat Joe Biden with a comfortable lead over President Trump. Comfortable, yet definitely not insurmountable: About 7 points nationally, but only around 3 points in an average of battleground states. And as veteran campaign watchers will tell you, political races can turn on a dime.

EICHER: There’s also that little issue of accuracy. In 2016, most polls predicted Hillary Clinton was headed for the White House. So, how seriously should we take the polls this time around?

REICHARD: Well, it’s Washington Wednesday. And joining us to talk about polling and polling data is Mark Caleb Smith. He’s director of the Center for Political Studies at Cedarville University, a Christian college in Cedarville, Ohio. 

Professor, welcome back to the program!

MARK CALEB SMITH, GUEST: Oh, it’s always my pleasure to be with you.

REICHARD: I’d like to start by talking about how polls are taken. Back in the day pollsters used to call people on their landlines, usually during dinner! That’s changed, I presume? Gone online?

SMITH: To some degree they have. I mean, we have some issues with online polling because it isn’t necessarily as representative as we might like it. I have a feeling that in our conversation I’ll use that word “representative” quite a bit. But what we’re looking for is a poll that represents the population that it’s trying to survey. So, a representative sample is one that really looks like the demographics that we see in a population. 

So, if you’re looking at just a landline, for example, these days the kind of person that’s going to sit at home, answer a landline phone call, and take a poll is a certain kind of person. The kind of person that might do an online poll is a different kind of person. And so they have robopolls where you have a computer message that plays on a telephone. They have landline. They have cell phone. They have mixtures of these things. And so pollsters use a variety of methods. And, of course, all of those come with their own pluses and minuses.

REICHARD: Does the method of polling influence on the outcome? If a poll relies on calling landlines, for example, I would think that would skew the respondents to an older demographic.

SMITH: Well, the method has the potential to skew the sample one way or another. So, for example, if I’m relying on a landline phone survey, then the kind of person who’s most likely to be sitting at home will probably be an older voter who still has a landline and who’s still willing to answer the phone and take a survey from a relative stranger. And so that’s a certain kind of voter. That’s a certain kind of demographic that you’re getting. 

It doesn’t mean you can’t construct a survey based on that, but it’s limited in some ways. If you go online, for instance, we’re worried about who has access to the internet and who doesn’t have access. Is this really biased toward those who may have more or less financial means? It might be skewed that way.

REICHARD: We know the poll data have been notoriously unreliable in the last few elections. And not just in the United States. Why is that? Have pollsters made adjustments to improve accuracy?

SMITH: I mean, we have to be really careful when we talk about polls being unreliable. I’m not a pollster. I study surveys to some extent for my research. But a poll can be reliable but still be wrong. So, we have to establish that upfront. So, whenever you do a poll—even if it’s very well-done—it has what we call a margin of error attached to it. 

So, if it’s a 3 percent margin of error, for example, plus or minus 3 percent, then if it says the Republican in a race has 51 percent with a margin of error of 3, well, they’re saying that number could be 54, it could be 48, we feel pretty comfortable it’s somewhere in between that range. And so if the results happen and the Republican is somewhere within that range of 48 to 54, that’s a pretty accurate poll. We would say that captured the race really well. 

But for your sort of average consumer of the news, they might look at it and say, well, the Republican was polled at 51, they got 49 and they lost the contest in a close race. That means the polls were wrong. Well, they’re not really wrong. They’re just not as precise as we might like to think that they are.

So, now, granted, your question is more than that. We have polls that sort of bounce all over the place and that’s normal. What you’re talking about is really a systematic error, where we see an error that’s headed in one direction, like in 2016 in the United States. The polls consistently underreported President Trump’s support and they overestimated Hillary Clinton’s support. And because of that we ended up with some skewed—especially statewide—polls. And there are a good number of reasons for that. Pollsters have adjusted, I think. We’ll see whether their adjustments are accurate.

REICHARD: What are the major polls tracking the U.S. presidential election? Which of those, if any, do you recommend people keep an eye on?

SMITH: Right now, the polls are pretty consistent, as you said in your introduction, showing Vice President Biden with a strong lead nationally. It isn’t an insurmountable lead. It’s 7 points or so. Could change pretty dramatically over the space of a couple of months. But it’s a consistent lead. If you look at the polling throughout the entire campaign, it’s actually been pretty stable. He’s been ahead and it’s been a significant lead for several months now. In the middle of May, we saw the two campaigns come a little bit closer in polling. It was more like a 4, 5 point difference at that point. But now Biden is really consistently ahead.

I think, you know, when you look at polls, I think that it’s unwise, honestly, to look at one poll or to choose one poll or another poll and say, you know, I’m waiting for that Washington Post poll or I’m waiting for that Rasmussen poll. That’s really not the right way to go about it, because those polls are more or less accurate. Again, they’re going to have their own margin of error attached to them. And putting too much weight on a single poll will give you, I think, will mislead you.

The best approach is to take a variety of polls and then look at them over time. And then you sort of aggregate those and say, OK, based on these 10, 15, 20 different polls that come out regularly, what does the race show on average. So, if you look at sites like Real Clear Politics, for example, or if you look at FiveThirtyEight.com, they regularly have these sort of aggregated polling totals. And that, I think, is really the best way to get a clear handle on what’s happening in the race right now.

REICHARD: Let’s talk about where the polls stand now. We’re a little less than eight weeks from Election Day. Joe Biden still has a pretty comfortable lead in most of the national polls. In any normal year, we would be watching for the things that can typically change the direction of a campaign: the debates, for example, and the infamous October surprise.

But 2020 is not a normal year! Is it just too early to put much stock in what people are telling pollsters now?

SMITH: That’s a great question, because normally we would expect to see some variability in the numbers between now and election day. We would expect to see a fairly healthy number of undecided voters—say, between 10 and 15 percent. And we would track those to see how they break in one direction or another over the next couple of months. You said 2020 is not a normal year. That’s true, but if you look at how we’ve been polling and if you look at President Trump’s approval rating, for example, over the past several years, it’s stayed pretty consistent. The dynamics of this campaign have stayed pretty consistent and pretty durable in a way that’s almost unnerving, I think.

And so even though we’ve had a pandemic, even though we’ve had an economic downturn, I’m not sure those are going to be enough to radically change what we’re looking at in the electorate. We have a very polarized political system right now and I’m not sure even those events are even shaking large groups of voters one way or another. Also, if you look at undecideds, that’s a pretty small group right now. And so I’m not even sure that group will be large enough to shift the election by the time we get closer to the election itself.

REICHARD: Mark Caleb Smith is director of the Center for Political Studies at Cedarville University, a Christian college in Cedarville, Ohio. Thanks for joining us today!

SMITH: It’s my pleasure. Thank you.

(AP Photo) In this combination photo, president Donald Trump, left, speaks at a news conference on Aug. 11, 2020, in Washington and Democratic presidential candidate former Vice President Joe Biden speaks in Wilmington, Del. on Aug. 13, 2020. 

WORLD Radio transcripts are created on a rush deadline. 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 WORLD Radio programming is the audio record.