Margin of Error

This is a partial transcript of "Special Report With Brit Hume," Nov. 1, 2004, that has been edited for clarity.

Watch "Special Report With Brit Hume" weeknights at 6 p.m. ET!

BRIT HUME, HOST: CBS News/"New York Times" poll says the president is up three points. The FOX News Opinion Dynamics tracking poll says no, no, Senator Kerry is up two points. So what accounts for this much variability this late in the race?

Well, for answers, we turn to Dr. Daron Shaw, professor of political science at the University of Texas (search), Austin, who is also a member of the FOX News decision team, which will be helping to call those races tomorrow night.

Daron, welcome. Nice to have you.

DR. DARON SHAW, PROFESSOR, UNIVERSITY OF TEXAS: Thank you very much.

HUME: Help me out here with these two polls. I’m just citing these really, almost randomly as examples. CBS News, Bush up three. FOX News, Bush down two. It sounds like the polls are all over the place. That’s the cliche, in fact.

SHAW: Right.

HUME: Are they?

SHAW: No. I mean in fact, I think the polls have been remarkably consistent over this last weekend. The difficulty, of course, is people forget about margin of error. And the reality is, is that a poll of about 1,000 likely voters have a margin of error of about three points. So you know, if we’re saying Bush is 50...

HUME: That’s plus or minus three?

SHAW: Plus or minus three.

HUME: So you can talk about a swing as much as six.

SHAW: Absolutely in the margin. Absolutely.

HUME: Right. But basically that says that a pollster surveying 1,000 people will get a result. And that statistically speaking, that may be off by any number in that poll may be off — I mean either number in that poll may be off by as much as three in either direction.

SHAW: Absolutely. And so it’s — from a political science perspective, it’s always infuriating when people come on and talk about, you know, well, this poll is showing a trend. Or breaking based on one percentage point difference, you know, from Thursday night to Friday night.  The fact is those results are statistically indistinguishable.

HUME: Really?

SHAW: Absolutely.

HUME: So even the result between CBS News (search) and FOX News?

SHAW: Absolutely.

HUME: So if out pollster, our chief pollster sat down with CBS’ chief pollster, you’d basically be considering that you’re talking about having made the same finding.

SHAW: Yes. Absolutely. And so when I say the national polls look remarkably consistent. What I’m saying is that there is absolutely no statistical difference, not only between the FOX/Opinion Dynamics poll, CBS/"New York Times," ABC/"Washington Post." They’re basically giving you the same result.

HUME: Now, there’s a new batch of state polls out there today. We have some new ones. One of them, an interesting contrast is between Florida — between the two Florida polls. Our FOX News Florida poll, this was not a tracking poll. This was a full survey, 49-44 for Kerry.

Now, the Bush people think they’re doing great in Florida. They believe that the Quinnipiac poll (search), which came out, was more like it. That shows Bush up above 50 percent, up eight. Two pollsters sitting down together to compare these two polls, are they going to ask each other why are they so different.

SHAW: Yes. This one is a little harder, because these are sort of borderline margin of error. But believe it or not, these actually are pretty much the same poll. You know, it starts getting — when you talk about a statewide poll, you’re usually talking about smaller samples.

HUME: Right.

SHAW: So 600 to 800 compared to 1200.

HUME: Now what does that do to the margin of error?

SHAW: That means margin of error increases. So in both of these polls — I think the Quinnipiac is a little larger. It’s about 1000 or so, so that’s still plus or minus three. But the FOX poll is about 800, I believe.

HUME: Something like that.

SHAW: Which means the error is plus or minus four. Those errors actually make these results, you know, statistically indistinguishable.  Now, you know, the truth also is that we’re less confident in our ability to accurately sample states than we are the nation.

HUME: Just because you haven’t done it that much.

SHAW: We haven’t done it that much. There are all sorts of problems.  And Florida, of course as you know, has all sorts of complications.  Ranging from, you know, not only some of the things that happened this weekend, Halloween, you know, all sorts of football games.

HUME: Talk to me about that whole question. Because we’ve seen what looked like a lot of fluctuations this weekend, and you always hear pollsters fretting about weekend polling. Why?

SHAW: The idea is, is that some people are less likely to be at home and answer the phones and to be represented accurately until their poll.  And you know, you can spin this one of two ways on Halloween, right? Well, you’re less likely to get younger voters because they’ve been partying on Saturday night and are not going to be, you know, willing to talk to pollsters on Sunday.

Most of these calls did take place during the day. No one is adventuresome enough to try to poll on Halloween eve. But you know, I have kids, and we begin preparing the kids for Halloween, you know, noon to 2:00, 3:00 in the afternoon. So other people say oh, you’re going to under represent married people. All those things.

HUME: Because they don’t answer the phone or don’t want to talk to a pollster?

SHAW: Exactly. Well, you know, if you’re getting your kids’ costumes on, and getting them ready and getting the candy ready, you may, you know, be less willing to talk. And so Republicans will scream and say you know, you’re under representing married people, they’re a good constituency for us.

The truth is we don’t know how much of that is going on in here. And you know, the polls come out, the demographics looks great, you know, we put them out there. But these things are complicated and they can result in problems.

HUME: So when you look at all this whole batch of polls now that are coming out, they don’t look surprising to you. And you don’t feel like there’s a ton of fluctuations, correct?

SHAW: I thing that what you really see — you know, what you guys have been doing, a lot of people do is to average across the polls; I think that’s reasonable. Variance is high. We know it’s close.

Content and Programming Copyright 2004 Fox News Network, L.L.C. ALL RIGHTS RESERVED. Transcription Copyright 2004 eMediaMillWorks, Inc. (f/k/a Federal Document Clearing House, Inc.), which takes sole responsibility for the accuracy of the transcription. ALL RIGHTS RESERVED. No license is granted to the user of this material except for the user's personal or internal use and, in such case, only one copy may be printed, nor shall user use any material for commercial purposes or in any fashion that may infringe upon Fox News Network, L.L.C.'s and eMediaMillWorks, Inc.'s copyrights or other proprietary rights or interests in the material. This is not a legal transcript for purposes of litigation.