Monday, January 9, 2012

Polls

Our chapter on public opinion discusses the quality of polling, a topic that is much in the news as the primary season gets under way.

The Erie Times-News interview Joseph Morris, associate professor of political science at Mercyhurst College and director of the Mercyhurst Center for Applied Politics:
1 We've seen a lot of presidential polling. And we'll see much more this year. How can people tell whether these polls are legitimate and unbiased?

I always pay close attention to how the poll was conducted. If the poll numbers aren't accompanied by a thorough report that describes the method used to conduct the poll, indicates a margin of error, explains how respondents were selected, and provides the exact wording of the questions that were asked, I tend to look elsewhere.

2 How accurate are presidential polls in predicting winners?

Let me answer that this way: Polls are nothing more than a snapshot of a single moment in time. History and professional experience tell me that they almost always accurately provide accurate pictures of what's going on. Using Iowa as an example here, if you think back to the beginning of the week, on the eve of the Iowa caucuses, more than 40 percent of likely Republican caucus-goers said that they preferred one candidate over all the rest, but that they could change their minds. In these situations, poll results don't have much of a life span. Having said this, I have great confidence in the accuracy of the polls that are conducted well.
 The New York Daily News reports:
You should be careful of polls that survey small groups of people. Fewer than 400 respondents should raise a red flag, Charles Franklin, political science professor at Marquette University and founder of PollsAndVotes.Com, told the Daily News. He cautioned this can be a problem with state primary polls.
Polls also become more reliable the closer you get to a state primary.
"Polls conducted a day or two before the primary have historically been much more accurate than polls conducted a week or so beforehand, a reflection of the fact that voters often settle on their choice only at the last minute," Nate Silver wrote on his New York Times blog FiveThirtyEight.
Polls can produce biased results if leading questions were included. Good pollsters will release the questions they used so you can decide for yourself.