"How Pollsters Got New Hampshire Wrong"

SCOTT SIMON, host:

This is WEEKEND EDITION from NPR News. I'm Scott Simon.

Coming up, Jared Diamond on what makes ethnic differences sometimes turn into ethnic violence.

But first, pollsters and pundits are still trying to figure out how they got the pre-election polls in New Hampshire so wrong on the Democratic side. In recent history, most pre-election polls have been accurate in predicting outcome. This case, though, seems to have been an unprecedented failure. Even the candidates' own polls were apparently wrong. The director of polling for ABC News, Gary Langer, has been trying to figure it out and agrees to let us put him on the grill. He joins us from New York.

Gary, thanks very much for being with us.

Mr. GARY LANGER (Director of Polling, ABC News): You bet, Scott.

SIMON: None of your polls are wrong, were they?

Mr. LANGER: That's what I like best about the problem and that it wasn't part of it.

SIMON: Let me go through some of the theories that have been booted around, okay, and just get your response to them one by one.

Mr. LANGER: Sure.

SIMON: One theory holds that a lot of people made up their mind at the last minute, especially because of Senator Clinton's performance in the debate and her manifest emotion.

Mr. LANGER: Yeah. It's one you hear a lot in - I don't think that works in this case for a couple of reasons. One, there were nine pre-election polls - all wrong. All had Obama ahead - he lost. But three of them were done right through Monday night. And if there had been any shift to Clinton after that event, we would have seen it in those polls and we didn't. And the exit poll itself asked people when they made their decision. And people who said that made their decision on election day were no better for Clinton than the group overall. So again, there's no support for that theory in the data we see so far.

SIMON: Mm-hmm. Another theory: ballot order. In New Hampshire, they list candidates alphabetically. Clinton was higher on the ballot than Obama.

Mr. LANGER: Really interesting. This work by Professor John Crosnik(ph) at Stanford University, who's found that when the ballots are listed in, in this case, alphabetical order, the best-known candidate toward the top gets perhaps about three points more support than he or she would otherwise. The pre-election polls almost all randomized the order of the candidate. That might have contributed to their own misstatement. It can't have accounted for all of it, but it could be some.

SIMON: Mm-hmm. Another theory, of course, holds that there's a history of black candidates polling higher than the number of people who actually support them.

Mr. LANGER: The problem with this theory, I think, is that it stems from six elections, many years ago, 15 to 25 years ago, in which polls, pre-election polls, understated support for white candidates in a white versus black candidate race. And this led to suggestions that maybe some whites are reluctant to express support for a white candidate perhaps for fear of being received as racist in a biracial race. You know, there's one study in 1989 that suggested maybe there might be something there. I'd like to see more than that.

Another problem is that there've been plenty of races since with a black versus white candidate that were perfectly accurate in terms of pre-election polling, six such races in Senate and governor contests in 2006 alone. That seems to me right now as theory in search of data.

SIMON: Was there an Iowa effect - I mean, people, if I may, in your business were saying, well, you know, the New Hampshire primary follows so closely. It used to be that the candidate winning in Iowa would come out with a bump, that it would have a chance to go down, but not this time.

Mr. LANGER: Yeah. You know, this one is probably the most appealing to me personally. Iowa happened right on top in New Hampshire and it could be that there is a wave of enthusiasm after that for Obama, among his supporters in New Hampshire and maybe some demotivation among Clinton supporters. They may have been deflated, perhaps less likely to express their support for her and it might have kept them out of these likely voter models.

SIMON: Is it possible that all of these theories could be insufficient, but each of them explains 1 percent of the vote then you have the reasons.

Mr. LANGER: You know, that itself is the perfect storm theory and it's also entirely possible, that it can be a little piece of each one of these that we're talking about. I think we will have solid answers. Hopefully, the pollsters that were involved in these estimates will give out their data sets and let them be subjected to some really close scrutiny. Every of the scenarios we've discussed can be tested in an empirical way and I think we will come to an answer. And I think, again, it's essential that we do.

SIMON: Gary Langer, director of polling for ABC News.

Gary, thanks so much.

Mr. LANGER: You bet, Scott.