ROBERT SIEGEL, HOST:
Over the past few weeks, we've been checking in regularly with pollster Andrew Kohut, president of the Pew Research Center. After the Democratic Convention, the Pew poll showed President Obama well in front. Then after the first debate, Mitt Romney took the lead. Last week, the Pew poll showed a dead heat. And yesterday Pew's last poll of the campaign came out and, Andy Kohut, President Obama has regained the lead but not by all that much?
ANDREW KOHUT: No. We have 50 percent of our likely voters saying they are going to vote for Obama and 47 percent for Romney. That's a statistically significant lead given the large size of this sample - well, close to 2,800 likely voters. It's the same lead that we had for President Bush in 2004. And remember, that was a close race, but that's about what President Bush won by. Now, things are different. There has been more ups and downs in the polling.
SIEGEL: Well, let me ask you about this. You had, over the past month or perhaps a little bit more than that: 51-43, Obama; 49-45, Romney; now 47-47, tied; and now a lead. Have 10 percent of the electorate really been changing their minds during all this time? Or is it luck of the draw when you happen to call out on a given day?
KOHUT: No. I think it really does represent change. I mean, Obama did slip in support after that first debate. He has come back, and the poll that we conducted over the past four days shows that his handling of the hurricane helped him, even a plurality of Republicans think he did a good job and 63 percent of the swing voters. And while people have been thinking about this race for a long time, we still have 11 percent that says, you know, I just might change my mind, which is consistent with what we see in the exit polls just about in every election. Eight, 10 percent will decide in that last week.
SIEGEL: This year, the polls have been criticized by the Republicans, essentially, when the polls have shown President Obama in the lead. And there have been lots of questions raised about methodology, about how you figure out what the sample is, what do you infer from the last presidential race using cell phone. How much more complicated - how much more difficult or how much more easy is polling a presidential race today than it was for you eight, 12, 16, 20 years ago?
KOHUT: Robert, it's a blunter instrument than it was in campaigns in the past, particularly 10 or 12 years ago. First of all, our response rates are lower. We now have the complications of drawing our sample from two separate frames - people who have landlines, people who have cell phones, and then we have to deal with the people or most of the electorate who use both. So it's a more complicated process.
SIEGEL: And you have caller ID, which shows us that someone we don't know is calling us at that moment.
KOHUT: Yes. And over the course of time, response rates have fallen, but fortunately, up until this election at least, our accuracy in predicting presidential elections and the popular vote in national elections has actually been as good, if not better, than it was years ago. But this has been a particularly difficult election, and I think it has to do with the candidates. I'm going to blame these candidates because people are conflicted by them. There are things about Obama that they just have a hard time accepting, and there are things about Romney that they have a hard time accepting. So they've done back and forth.
SIEGEL: Well, you know, I was thinking back to - 43 years ago, I was a journalism school and I had an assignment crafted by Lou Harris - the pollster, actually - where I had to go to a neighborhood where everybody fit the same demographic. Get inside 10 households and question them and have an interview with these people. And on that basis, report on what that model precinct might do. As arduous as that was, it felt a little bit more palpable to me what I was learning than, say, some company that might make 10 robocalls.
KOHUT: Yeah. The robocalls are really questionable because, first of all, they missed everyone that has a cell phone and they make judgments about what the party affiliation distribution should be. And many of us in the field who've been in this for a long time and take a social science approach see party affiliation not as a fixed characteristic such as sex or age or education but an attitude that varies with how people are feeling about candidates and so on.
SIEGEL: Andy, I look forward to spending many hours with you tomorrow night.
KOHUT: We shall. Hopefully not too many.
SIEGEL: It's Andrew Kohut, who's president of the Pew Research Center. Transcript provided by NPR, Copyright NPR.