ad info

 
CNN.comTranscripts
 
Editions | myCNN | Video | Audio | Headline News Brief | Feedback  

 

  Search
 
 

 

TOP STORIES

Bush signs order opening 'faith-based' charity office for business

Rescues continue 4 days after devastating India earthquake

DaimlerChrysler employees join rapidly swelling ranks of laid-off U.S. workers

Disney's GO.com is a goner

(MORE)

MARKETS
4:30pm ET, 4/16
144.70
8257.60
3.71
1394.72
10.90
879.91
 


WORLD

U.S.

POLITICS

LAW

TECHNOLOGY

ENTERTAINMENT

 
TRAVEL

ARTS & STYLE



(MORE HEADLINES)
 
CNN Websites
Networks image


Saturday Morning News

Are the Presidential Polls Really Accurate?

Aired September 16, 2000 - 9:35 a.m. ET

THIS IS A RUSH TRANSCRIPT. THIS COPY MAY NOT BE IN ITS FINAL FORM AND MAY BE UPDATED.

MILES O'BRIEN, ANCHOR: Presidential polling, it's a controversial subject. Lot of people think the polls are biased one way or another. Depends on where you sit in things.

The man who keeps track of polls here at CNN for us is Keating Holland. He joins us live now from our Washington bureau to take your phone calls and e-mails.

And as it relates to those phone calls, we invite you still to call us at 404-221-1855. We should let you know that is not a toll- free number, but we do think Mr. Holland is well worth it.

Keating, thanks again for being with us.

KEATING HOLLAND, CNN POLLING DIRECTOR: Good to be here.

O'BRIEN: All right. Let's go right to the e-mail, shall we? A lot of questions about how polls are conducted, how you keep the bias out of them. This one comes from Norman Walker, who is in Farmville, Virginia. "Can you explain exactly how you conduct a poll?" This is like Polling 101 now, I guess, Keating. "Who is contacted, and where? Is it truly nationwide, or is it regional?"

HOLLAND: All right, when we conduct a poll, we do a nationwide poll. There are other polling organizations that just poll in individual states, but we do all the continental United States. How we conduct a poll, I guess the basic question is, how we choose the people who we talk to. That's done randomly. A computer picks telephone numbers at random.

So the only thing we know about these people is their telephone number and because of the area code, the state that they live in. Everything else, we have no idea until we actually get them on the phone.

The whole point of random digit dialing, is what it's called, the whole point of random digit dialing is, everyone in the country has an equal chance of being selected to be in our poll. And as long as you make sure that everyone has an equal chance of being selected, it turns out to be representative of the entire country.

O'BRIEN: All right, Keating, let's move along. Mr. Boyd Dodson in Dyer, Arkansas, has this for you. "What is to prevent polling agencies from slanting or outright rigging their questions to get the desired results?"

HOLLAND: Well, one thing is that we're in a competitive environment. CNN does its polling, ABC does its polling, CBS does its polling. And if we ever rigged a poll, and we never have, but if we ever did, ABC, CBS, all of the competition would be able to call us on it almost immediately.

One important thing is that in many instances when we have a new question that we've never tried before, we will pretest it, pilot it, ask it of 50 or 75 people, and we have to pay to do that, but we think it's well worth the money. Ask it of a few people to make sure that there isn't an unintentional bias in the poll question.

Sometimes we'll find one and eliminate it. Normally, however, we think that the basic plain-vanilla question is one that will get a good response. We usually try to stick to those plain-vanilla questions.

O'BRIEN: All right. Here's another e-mail question for you, Mr. Holland, this one from Mr. Helton, Dan Helton in Riverview, Michigan. Question is this. "CNN just displayed the results of a tracking poll showing Gore over Bush 49 to 42. My question is whether CNN is also tracking the state numbers, and if so, what the likely outcome is in terms of electoral votes?"

HOLLAND: We don't do our own polling on a state-by-state basis, but we certainly pay a lot of attention to the statewide polls. Just yesterday, for example, there were polls conducted or released in Michigan, Illinois, Pennsylvania, and Ohio. In three of those states, in Michigan, Illinois, and Pennsylvania, Gore had a lead. In Ohio, Bush had a lead.

Actually the one state that I pay a lot of attention to, because it has voted for the winner in every election since 1948, is Delaware. There was a poll released in Delaware yesterday that showed Gore up by 10 points.

The electoral college map, it depends on who you talk to. Oftentimes you'll see political junkies who keep a few states in the toss-up column, other guys, like Bob Novak on CNN "INSIDE POLITICS" on Thursday gives every state to one candidate or another. Most of those analyses seem to indicate at least a mild Gore advantage in the electoral college.

But CNN itself doesn't do that. We just kind of look at what other people are doing and do a -- in our own minds form an analysis of where we think it's heading based on what other people's polls are showing us.

O'BRIEN: All right. Let me ask you this one, this also an e- mail. Apparently we're having a few problems with our phone lines, folks, so we apologize for that if you're trying to call in right now.

This one comes in from Joanna Haas in Short Hills, New Jersey. "If you have 10 well-designed polls and Gore is ahead in nine of them, but each one is a statistical dead heat, isn't that evidence that Gore is really ahead?"

HOLLAND: Yes. In fact, oftentimes you can say that every poll showing the same thing gives additional robustness to the data, so to speak. Another important thing to bear in mind with a tracking poll is, we've been doing it for about 10 or 11 days now. That means we've got a lot of data. We look at it in three-day chunks, is how we're reporting it, but we can also look at four days, five days, six days.

The larger number of interviews you throw in, the smaller the margin of sampling error is. And we've seen Gore have a solid lead for about four or five days. That makes it increasingly likely that we can say statistically that Gore actually is in the lead.

O'BRIEN: All right. Still having some problems with our phone lines, our apologies to viewers. Some gremlins obviously got the best of us there.

Let's go to one final e-mail, this one from Mike Poliszuk. Don't know where he is, but that really doesn't matter too much, he's out there on the Internet somewhere. "The response to polls depends greatly on the questions asked, and yet the media tries to make broad assertions of public preference based on poll results. I believe that when reporting poll results, the media should report the exact questions asked for each poll result."

Don't we do this, Keating?

HOLLAND: Yes, we do, on our allpolitics.com, cnn.com, you can go there and find the full text of every question. It is an important thing. We -- it is something that we want our viewers to be able to take a look at and decide for themselves whether the question has any bias in it.

On television, normally, with a -- particularly with a tracking poll, we're just showing the results of the horse race question.

O'BRIEN: Yes, and because of brevity, we don't often restate the entire question, just to answer your question a little bit further, Mike. If you are curious about the specific question, we invite you to check out the Web site, as Keating just mentioned.

Keating, we're out of time, sorry about that. If we took a poll, I bet we would say we'd like to have more time with you.

HOLLAND: (inaudible) me back.

O'BRIEN: But maybe we'll hear from you again before the election season is out. Thanks so much.

HOLLAND: Good.

O'BRIEN: All right.

TO ORDER A VIDEO OF THIS TRANSCRIPT, PLEASE CALL 800-CNN-NEWS OR USE OUR SECURE ONLINE ORDER FORM LOCATED AT www.fdch.com

 Search   


Back to the top  © 2001 Cable News Network. All Rights Reserved.
Terms under which this service is provided to you.
Read our privacy guidelines.