ROBERT SIEGEL, HOST:
It's something we hear in every election season. Don't obsess over polls. Go tell it to Donald Trump.
(SOUNDBITE OF ARCHIVED RECORDING)
DONALD TRUMP: So CNN came out 33 for Trump; 20 for Cruz. That's good.
UNIDENTIFIED MAN: Yeah.
TRUMP: Quinnipiac just came out 31 percent for Trump; 29 for Cruz. It's a little too close for comfort, folks. The PPP...
SIEGEL: That was Trump at an Iowa event earlier this month. He loves polls. He loves talking about polls, at least the ones that show him winning. And NPR political editor Domenico Montanaro loves polls, too, and he's here to guide us through the poll numbers that we're going to see in this next week before the Iowa caucuses next Monday. Hi, Domenico.
DOMENICO MONTANARO, BYLINE: Hi, and I do love polls, maybe to not the specificity that Donald Trump does...
SIEGEL: (Laughter).
MONTANARO: ...And maybe to a little bit broader range, but we can talk about that.
SIEGEL: OK. A week away from the Iowa caucus - what's different about the polls we're going to see?
MONTANARO: Right now, pollsters are switching over from what's known as a registered voter model to a likely voter model, which means that they're essentially testing enthusiasm and whether or not people will go vote. And those lucky voter screens are making a lot of volatility right now. You can't compare what you saw five month ago to what you're seeing today 'cause it's a completely different universe of people.
SIEGEL: Because deciding who likely voters are, unlike who's a registered voter, it's a judgment call.
MONTANARO: It is. And part of that is because people are telling survey interviewers whether or not they voted, and it's not something that is based in fact because they're not using what's known as a voter file of real evidence of whether or not you voted.
SIEGEL: The rules for caucuses in Iowa are different for Democrats and Republicans. So the polls, as I understand - this could actually mean something different in terms of how well they predict the result.
MONTANARO: In a primary, a lot of times, poll track a little better. In a caucus, it's a lot harder because it's all about organization. And that's what's happening in this final week. Part of the problem is - is that, within certain precincts in Iowa, for example, you could have a concentration of support for Bernie Sanders in three counties where there are big universities. Well, if they don't spread out those folks, Hillary Clinton could wind up with more delegates in the same way Barack Obama wound up with more delegates in rural places than Hillary Clinton did out of big cities.
SIEGEL: It's like a little electoral college system system you describe...
MONTANARO: It kind of is, yeah...
SIEGEL: ...On the Democratic side.
MONTANARO: ...On the Democratic side. On the Republican side, it's very different. It's a little bit more like you would see in a primary system - more informal, though. People just fold over pieces of paper, hand them in. And both sides do make arguments for their candidates, but they don't have the same kind of process as Democrats.
SIEGEL: Four years ago in the Republican caucuses, Rick Santorum won by a whisker. Mitt Romney was second and Ron Paul third. That one didn't seem to be predicted by the most respected poll in Iowa.
MONTANARO: That's right. The most respected poll in Iowa that a lot of people talk about is the Des Moines Register Bloomberg Poll that's conducted by J. Ann Selzer. And her poll showed Ron Paul winning. Now, here's a caveat. She actually put a warning out about her own poll because this poll was conducted over four days, and she found, in the first two days - they had found Ron Paul in the lead. But by days three and days four, she saw a trend upward - an upward tick toward Rick Santorum, the former Pennsylvania senator who got a lot of support from evangelicals in the western part of the state who appeared to be deciding late in the race more so toward Santorum.
SIEGEL: Are there any comparable internal trends that you can see, say, this time around in the Sanders-Clinton race?
MONTANARO: Well, one of the things I think is important to look at are trends, and I think one of the things you're seeing on both sides - I mean, on the Clinton-Sanders race, you're seeing them pretty much neck and neck. They're within the margin of error. On the Republican side, you've seen Donald Trump and Ted Cruz, over a long period of time now, be basically one and two. Who's to say who's first or second? It's very difficult to tell. We shouldn't go with the kind of specificity that Donald Trump might. We should look at the broader trend over time.
SIEGEL: And next week, Domenico, we can take all those polls and, well, grade them, would be a polite way of putting it.
MONTANARO: That's true, and I think people should just take a step back right now. Let everything settle with the volatility because Iowa voters will actually decide and tell us without a margin of error, very much so, who the winner is (laughter).
SIEGEL: That's NPR's political editor Domenico Montanaro.