Weekly Standard’s Jay Cost analyzes recent polling and methodology
HH: This is the hour of the pollster. Coming up after the break, Steve Shepard of the National Journal is going to spend the rest of the hour with me talking about his article on polling today, in which he relays the viewpoint of Marist and Quinnipiac that no point bothering with, don’t bother us about partisan identification. It doesn’t matter. And I’ll probe that with Steve, but before I do that, I want to talk to Jay Cost, columnist for the Weekly Standard. Morning Jay is his every single day column which you ought to be getting from the Weekly Standard, and no mean pollster and student of such matters. Hello, Jay, welcome back to the Hugh Hewitt Show.
JC: Hi, Hugh, my pleasure to talk to you again.
HH: All right, Jay, at the start, really, these pollsters get so defensive when I talk to them. And I just want to know what the science is. Does it matter what the partisan identification is in a poll?
JC: Of course it matters. Of course it matters, Hugh. I mean, look, you have 90% of Republicans are going to vote for Romney. 89% of Democrats are going to vote for Obama. So if you have, say, let’s say the electorate in Ohio on election day is going to be even between Democrats and Republicans. Then it’s going to come down to the independents. But if you, if you’re doing a poll, and you find that Democrats have 6% more in numbers than the Republicans, then there’s really no way Romney can take a lead in that poll. And that’s exactly what we’re seeing right now.
HH: Well now, the argument that Shepard relays, and we’ll find out if he makes it when it talk to him, but he relays it, is that that’s a soft factor, you know, you come to the end of a polling question, and people might say Republican and maybe they aren’t, and they might say Democrat and maybe they aren’t, and you really can’t weight on the basis of that like you can weight on the basis of ethnicity, age and socio-economic data. What’s your response to that?
JC: My response is that that’s an oversimplification on two counts. First of all, obviously, partisan identification is not nearly as solid as an identification as your race, right? But it is an extremely solid personal quality. You know, there’s gobs and gobs of historical literature going back 60 years in political science which I’ve read in depth that shows that partisan identification tends to be stable over time. In fact, you can see it in…look at elderly people now are overwhelmingly Republican. But 30 years ago, Hugh, they were overwhelmingly Democratic. The reason why is because elderly people 30 years ago lived through the Great Depression, and it turned them into Democrats. I mean, that’s how stable partisan identification is. But even beyond that, you don’t have to do a simple gross weighting. You can do something like what Rasmussen does. You can take a monthly poll of 30,000 people, and say hey, you know, this is how many, this is how many Republicans and Democrats we’re finding in the electorate, and maybe this is a good target. And the other point I want to make, and this is extremely important, right, is that I certainly agree that a simple, naïve weighting toward party identification is a problem. But their methodology creates problems that they tend not to acknowledge, and that is this, is that when you don’t weight or control for party ID, and you have a surge in enthusiasm for one side or another like, oh, I don’t know, two weeks after the Democratic National Convention, Democrats are feeling really good about things, they’re going to break through your likely voter screens, and you’re going to have more Democrats. And these pollsters just shrug and say oh, well, you know, what are you going to do? But what actually is going to happen here is that Democratic enthusiasm, in all likelihood, is going to recede. We’re going to return to our historical 25 year average of about nationwide Democratic advantage of 3 points, and all of these polls are going to see a massive swing back to Romney in a couple of weeks.
HH: Now today, the Washington Post has a poll out that says Ohio, among likely voters, is 37% for Obama, and 30% for Romney. In 2010, Ohio voted 36% for Democrats, 37% for Republicans. That was their turnout. And I’m sorry, the turnout model, I want to restate that. The turnout model in the poll today had 37% Democrats, 30% Republicans. In 2010, 36% were Democrats, 37% were Republicans. In 2008, 39% were Democrats, 31% were Republicans. So on that basis, with those numbers, is the sample a good sample in the Washington Post poll today?
JC: No, I don’t think so. I mean, look, I think it’s a fine sample in terms of what trying to capture voter sentiment right now, when Democrats are, you know, that poll was taken about two weeks after the Democratic convention. Democrats are enthused. You can look at a whole host of metrics, for example, economic confidence in Gallup has spiked only because Democrats are feeling enthusiastic. So yeah, it’s probably a good metric right now. But it’s a bad indicator of where we are going to be on election day. I mean, I think at this point, we’re going to see something roughly in between 2008 and 2004. And if we see that, it’s probably going to be, in Ohio, a 1-2 point party identification edge for the Democrats. And here, the Washington Post has a 7 point party identification edge.
HH: Now when Steven caught up with Lee Miringoff, the director of the Marist College Institute, with whom I spoke, Miringoff said why would pollsters want to look in accurate, he asked rhetorically in an interview. How would you answer that, because you know, that’s saying are you questioning my integrity, which I don’t want to question anyone’s integrity. But I’m just surprised by a willingness to go with so obviously a distorted screen.
JC: Well look, I think what they’re trying to do here, I think pollsters have a couple different incentives. I think one of their incentives is to facilitate news, especially if you’re doing a poll for NBC News. But you know, when we’re talking about accuracy, the only way to judge a poll’s accuracy is to judge its final results, right, its pre-election poll usually taken a week before election day or less. And my contention is that all of these polls are going to swing back to a historical average by that point, at which point all of these pollsters will be able to say correctly, ha ha, we caught it on the nose. The difficulty now is this is the issue. The issue right now is that it is very difficult to get a sense of who the true voters are. We are five weeks away from the election, and we’re just two weeks off the Democratic convention. It’s very difficult to get a sense of who the true voters are. And these pollsters are making a guess. And their defense never acknowledges that in fact what they are doing is making a guess. And moreover, Hugh, it is a contestable guess, because the Rasmussen poll, the Gallup poll, the Purple Strategies poll, and a handful of others do not make the same guesses, and are getting different results. And that, I think, is the core point here, is that this is much more guess work than pollsters are prepared to admit.
HH: And it also, my problem is it has an impact on behavior. Now I want polls to show that Romney’s up six points, because that would crush the Democrats. They’d just give up. They’d go home now. They wouldn’t give money, they’d stay from polling banks. But I also fear that the same kind of bad number on their side has that kind of an impact on the Republican side. Am I right to worry about the present effects of skewed polls on political behavior that has consequences?
JC: Well, I’m not sure. I mean, I think there is something to be said for that. But the question, of course, comes down to the people who are paying close attention to the polls are going to vote anyway. You know, what we’re really worried about here are the marginal conservatives who may or may not vote, who maybe voted in 2004, but stayed home in 2008. Are they paying enough attention to be swayed by this? My guess at this point is probably not.
HH: And so, even though they’re wrong, they don’t matter?
JC: Yeah, that’s what, you know, my basic thesis here is that look, it’s September, right? And we’re two weeks off the Democratic National Convention. This president’s clearly on a bounce. I think these polls are catching this bounce. I don’t think there’s anything wrong with that, per se. I don’t think anybody’s doing anything below board. But I think these pollsters should admit yeah, look, this is probably a bounce. These numbers look really good for Democrats. They don’t square with history. I mean, that’s the thing, Hugh.
HH: All right, then my last question, Jay…
JC: We’ve been taking exit polls for 40 years. We know where the electorate winds up most of the time.
HH: Now we’ve seen a run of very bad stories for the president, beginning with his Univision pratfall, and then his View unfortunate eye candy, and the speech to the UN is not so good, and most of all, the Benghazi attack is being revealed to be a disaster for the United States. How long, if something’s going to impact a poll, when does it show up?
JC: Well, you know, that’s a good question, and again, the issue right now is that if you look across these polls, the advantage that this president has is through Democrats. I mean, for instance, the Gallup economic confidence numbers have jumped about ten points. But Gallup had a graph out today that showed that it was entirely among Democrats. My contention is now, and has been for years, that it’s independent voters who swing elections. And I see no movement among independents. In fact, Hugh, the most recent average that I was able to cull from the RCP polls among nationwide independents shows a tie, with this president stuck at 43% with the independent vote, which is about ten points off of his 2008 margin.
JC: So I mean, and that’s, see that’s the thing, Hugh. How do you get a 4 point Democratic advantage for this president on the top line number with a tie among independents? The only way you get that is through an overwhelming advantage in terms of partisan identification for Democrats.
HH: Jay Cost, that’s why we call you. Thank you from the Weekly Standard, author of a wonderful book, Spoiled Rotten, available at Amazon.com.
End of interview.