Horizon, Host: Ted Simons

October 14, 2010


Host: Ted Simons

Public Opinion Polls


  • Pollsters are the soothsayers of the election season. Arizona State University Pollster Dr. Bruce Merrill tells us all we need to know about polling in this election season.
Guests:
  • Dr. Bruce Merrill - ASU
Category: Elections

View Transcript
Ted Simons: Polling ramps up during the election season. But how do you know which polls to trust? Recently I spoke to Arizona State University Pollster Bruce Merrill, who says that one thing's for sure -- we're seeing fewer polls.

Dr. Merrill: We are seeing fewer polls this year. Number one, polls have become very expensive if they're done right. And there's been some tremendous problems with polling. They have just developed in the last few years, frankly. For instance, about almost one out of every five people now are using their cell phones for their only means of communication. We do not have access to those numbers. And so we're less and less likely each year to get a representative sampling of, for instance, all registered voters, because if you're using your cell phone only, we can't get you.


Ted Simons: So you've got a pretty good sample of folks who still have land lines.


Dr. Merrill: That's it. And we know there are some biases. People that are young, young married people, both working, are people that are much more likely to use only cell phones.


Ted Simons: You mentioned fewer polls because of the expense. Why is it so expensive?


Dr. Merrill: Well, again, what happened in the past is you would draw a random sample, say 400 people in Arizona, and if you can't get them, you have to call back and call back and call back. And so what is happening now is there's, for instance, the two major polls that you've even seen in the last couple months here, are the Rasmussen poll, which is a ROBO poll. It isn't even a phone call where people are talking to them. We know very little about the methodology. Or the Morrison institute, for instance, at ASU has published a couple of polls, but those are polls where you have to have access to the internet to even be interviewed. And my concern with both of those is that there's a tendency for instance, with calls based on the internet, for lower socioeconomic and minorities to not have access to that technology.


Ted Simons: So the best way to do a poll, as you've described it, is to have a bunch of folks with a bunch of phones and just repeatedly call. In the perfect world, you'd also have people calling cell phones as well. Is that feasible anymore?


Dr. Merrill: No. It really isn't. And in fact, polls have been remarkably accurate in the last 50 years. Particularly the national polls. The average error in a national poll predicting the presidency since -- in the 1940s has been less than 1%. So based on good scientific academic sampling methods, polls have been very, very accurate. But the demographics of America are changing, technology is affecting polling, I have some real concerns about the accuracy of polls because it may be that polls create public opinion as well as measure.


Ted Simons: And I want to get to that in a second, but back to the academic sampling aspect, and not get doing deeply involved into that, people will say, no one ever calls me. I saw the sample size, there's only a few hundred, how can they get an accurate representation? No one talks to me about this. How do you respond when people say that?


Dr. Merrill: Well, it isn't intuitively obvious, but what happens is once you have a population over 2500 people, 2500 adults, 2500 registered voters, 2500 most likely to vote, you only need a random sample, according to probability theory, of 400 to generalize that population with plus or minus 5% accuracy. So it doesn't matter. With 400, I can do a statewide poll or a national poll with the same degree of accuracy.


Ted Simons: And it still works, doesn't it?


Dr. Merrill: It could if you're able to get the people that you've randomly drawn, and that's where we're get nothing trouble.

Ted Simons: OK. I just want to mention, that because we get that a lot, how can we get that, I don't know anyone who has answered a poll! You mentioned the fact that polling -- the impact of polling on campaigns and how they could provide narratives or other aspects of pushing the electorate in a certain direction. Talk more about that.


Dr. Merrill: Well, basically the biggest use much polls is probably raising money, for instance. So the indirect effect of polling can be enormous. Let's say you have a candidate and do you a poll and their -- and they're slightly behind. That person can take the poll and go to their supporters and say, I'm behind, I need your money. So really, the decision whether or not to run is generally based on polling. In raising money is a big factor. The other major roll the polls have is actually kind of determining what the message is that resonate with the public. For instance, 1070, you do a poll, we know that about 60% of the people support 1070. Well, if you happen to support it, you're going to run on that issue, which reinforces the 1070 in the public.


Ted Simons: I know that many critics out there of the Rasmussen poll, which seems it's taken on a life of its own the past couple of campaign seasons, some are suggesting it's actually pro-Republican. Has the GOP tries to set the narrative and takes it from there. Is that a valid criticism?


Dr. Merrill: Well, I don't know, because the Rasmussen poll rarely gives you any methodological way that they've done the poll. For instance, the only way you can tell if a poll is valid is we know that about 32% of the registered voters in Arizona are Republican. So you should be able to look at the poll and it should have 32% Republican. If it has 50%, it's biased. So the problem is with many polls, they should tell you who paid for the poll because people release polls if they're supportive of their position. So the public should know who paid for it, when it was conducted, what the actual wording of the question is, what the population was, and what the sampling error was. And frankly it will media rarely reports any of that.


Ted Simons: Indeed, and the wording of the question can be huge, can't it?


Dr. Merrill: It can -- the outcome of a poll is more than anything else determined by the way the questions are worded. You can get a poll to come out any way you want based upon the questions that you ask.


Ted Simons: So how do you know, you see a poll, it says something, and you're kind of going, I don't know from up or down on this thing. How do you know whether that poll is something you should question or something you should look at and go "interesting"?


Dr. Merrill: Well, the most important thing is to look at the credibility and the credentials of the people doing the polling. If you see a Gallup poll, the people that do Gallup polling are very well trained. Here in Arizona, you have some very good pollsters. Have you Michael Neil, who is a Ph.D. in survey research, they're very credible, they've been doing polls for a long time. We used to do a poll at the University that was a University neutral-based poll that is a shame we don't have that frankly, because it's very important not to have ties to anything. So I think for the average person, look at the people conducting the polls, what are their qualifications, and their credibility.


Ted Simons: And we've talked about polling here, and you've mentioned how it can shape a race and these things. But in the long run, how much influence do polls have and does that influence change as the campaign goes on?


Dr. Merrill: Well, clearly it does. And basically if you have the media publishing polls that show a particular candidate ahead, whether or not they're ahead or not, but let's say have you a poll with 10 days to go, showing somebody ahead, there is a tendency for people in our society to want to be with the winners. We put a lot of emphasis on being the winner. And so my concern is, if the poll is accurate, then that's legitimate. But what if the person really isn't ahead, but the poll shows they are? Could that influence in a very close election the outcome of an election? It potentially could.

Ted Simons: But real quickly, aren't we talking about an enterprise that is -- we're talking free market to the bone here. You get it wrong, no one wants you anymore because most folks want to you get it right.


Dr. Merrill: Well, that's exactly right. And we can't go night here, but with probability theory, a pollster can never be wrong. Because one of the interesting things, whatever a pollster tells you, they only have a 95% probability that they're even close to the true population parameter. So in other words, if I'm wrong in a poll, I can just say, gee, that's one of the 5% of the times I'd get a sample that was more than 5% from the real estimate.


Ted Simons: Interesting. Kind of like a weather caster. You got all bases covered.


Dr. Merrill: Well, absolutely. And we do a lot of it in politics. We just join with the politicians. Absolutely.


Ted Simons: So keep an eye on the polls, but also keep an eye on what they're asking, how they're asking and it always keep another eye out for other things.


Dr. Merrill: Absolutely.


Ted Simons: Very good. Good to see you again.


Dr. Merrill: Good to see you, Ted.

What's on?
  About KAET Contact Support Legal Follow Us  
  About Eight
Mission/Impact
History
Site Map
Pressroom
Contact Us
Sign up for e-news
Pledge to Eight
Donate Monthly
Volunteer
Other ways to support
FCC Public Files
Privacy Policy
Facebook
Twitter
YouTube
Google+
Pinterest
 

Need help accessing? Contact disabilityaccess@asu.edu

Eight is a member-supported service of Arizona State University    Copyright Arizona Board of Regents