Here are a few excerpts -
Read the article here (subscription required)Tucson pollster Margaret Kenski said people can manipulate the numbers by intentionally oversampling demographics that are supportive of a particular candidate, target respondents who are already known to support a particular candidate or ask loaded questions.Good polls are supposed to rotate the order in which candidates are named because some respondents will simply choose the first name they hear, Kenski said. So a pollster can simply put the same person’s name at the top of every list to boost his or her numbers.“We all know in the business of too many instances where people massaged the data to make it look better,” Kenski said.
Republican political consultant Bert Coleman told the paper about the candidate-associated polls, “The polls are released for one motivation, and that is to get people to think they’re winning...."
Some polls of the governor’s race have attracted criticism because of how they determine who is a “likely Republican voter” in the Aug. 26 primary.Pollster Bruce Merrill, a professor emeritus at Arizona State University’s Walter Cronkite School of Journalism and Mass Communication, said people will lie when pollsters ask if they plan to vote in an upcoming election, much like they’ll lie about whether they go to church, beat their children or use pornography.“If it’s a socially acceptable kind of a thing, then people are going to answer it in a socially acceptable way,” Merrill said.That can skew results because it gives a false impression of who is voting, Merrill said, especially in low-turnout elections like primaries. Merrill and other experts say the most reliable way to choose respondents in a primary poll is only to call people with a documented history of voting in recent primary elections.“The lower the turnout, and the people that have voted consistently … in the last three primaries are going to tend be much more conservative,” Merrill said.