Politics and Polls

Thank you for having this program. Some questions for your guests.

1. How do polling organizations account for the increasing number of voters who do not have land line telephones and are unable to be included in polls? Up to 1/3 of people age 18-38 do not have home land line telephones.

2. How do polls account for the increasing number of people who do not answer calls where caller ID shows either an unknown number, or a blocked number. Call screening is a reality in this country.

3. Is the reliability of telephone polling compromised as the population that will answer the phone and actually respond to a poll question decreases . Is that still considered a random sample of probable voters.

4. How are samples of probable voters identified.

I worry that given the impact poll numbers have on politics including financial support and donations to influencing voters actual decision making [people want their vote to count or want to support a winner] combined with polls possibly being less valid means that the newsmedia has an increasing responsibility to not just report poll results but to point out their limitations beyond 'margin of error'

Thank you
Richard Friedman

Given that most polls are conducted using home phone numbers, and fewer and fewer people answer their phones at home, why do the pundits and media continue to have such faith in polls? There is a huge variance between the result of the polls and the popularity of the candidates like Ron Paul in social media such as YouTube and Facebook. Has either Andy Smith or David Moore done any studies to better understand any correlations between social media sentiment and actual voting patterns? Isn't it possible that in this election cycle, engagement with a candidate in social media might be a better predictor of election results.

Andy Smith mentioned that the media use polls to help direct their resources during the campaign. I actually think that this is one of the major problems with polls since the media tend to focus on just a few candidates relatively early in the process. In France, the media are required to provide equal coverage to all the candidates during the official campaign period. Don't you think such a requirement serves the democratic system better than relying on polls that tend to favor candidates that have more celebrity ?

I called the Exchange today(12/3/2007), wanting to comment on two types of polls I've been involved in. The first, Zogby polls, was extensively discussed by the panelists, but I did not get a chance to speak about the second.

1. The first is a Zogby poll. I have been on a volunteer panel polled periodically by Zogby for more then two years. I am chosen for certain polls and not for others, and I assume this is because of my location and basic demographics. I have read some of the material that Mr. Zogby has written in defense of using internet polling (of a random statistically valid sample drawn from a large panel based on location) as an alternative to telephone polling. I am convinced that what he is doing is statistically valid and based on good modelling techniques, based on my rudimentary knowledge of statistics. It would be worth your while to contact the Zogby organization to find out what their reply to your panelists dismissal of their methods would be.

2. I received a poll from a marketing organization which seems to include actual disinformation in its presentation. It is available on line at: http://www.iloyalty.net/survey/election/ . I found the process of taking the poll both infuriating and intriguing. Candidates' pictures are associated with sets of three opinions, some of which they actually hold and some of which they don't. We are asked to choose a candidate based on the assumption (which may be true or false) that the candidates hold the opinions presented. They also ask some "reality check" questions. It's apparently not meant to be statistically valid, but to be useful in gathering "marketing" data that would be useful to candidates, things like "Is the war so important as to override all other issues in choosing a candidate?", "Do people vote based on the appearance of the candidate or his/her opinions?"

During the show today, I believe David Moore stated that he felt that the sampling methodology was valid even though many people don't answer calls for polls, because the results of the elections were so close to the poll. Unfortunately, I think this is an example of circular reasoning. While he can state the outcome validates the poll, I can validly say because of the 'bandwagon effect, or the impact of people choosing to back a winner the poll is influencing the outcome. How do we know which has the more influence on the outcome?

Rich Friedman

Towards the end of Monday's show there was a discussion of exit polling in which the Edison-Mitofsky hypothesis that Bush voters refused to talk with pollsters (differential non-response) was used to explain perceived exit poll problems in 2004. That explanation has been thoroughly debunked (start here: http://electionarchive.org/ucvAnalysis/US/Exit_Polls_2004_Mitofsky-Ediso...). The discussion left the incorrect impression that the exit polling rather than vote counting was the problem.