What Political Polls Can Learn From Market Research Surveys

Email Article |  Facebook |  Twitter |  LinkedIn 

Last Thursday at the MRA Conference, Kathleen A. Frankovic, director of surveys for CBS News - and recent winner of the American Association for Public Opinion Research Lifetime Achievement Award - gave a wonderful talk entitled "Market Research & the Presidential Election: Lessons from 2008". She joked that she had accepted the invitation to speak only after saying to herself, "June? Well, the primaries will have been resolved for three months by then."

Political polling is very different than marketing research, but Kathleen said that in the wake of the 2004 problems with exit polls, she had decided to experiment more with MR techniques.

For competitive advantage, CBS News began asking questions in the form, "Would you say you like [your preferred candidate] a great deal better than any other, somewhat better or only a little better?" This question in fact revealed weak support for Clinton in Iowa and strong support for Obama; in retrospect, this was a particularly useful question since in the Iowa caucus a voter can change their vote between rounds, if their earlier preference had garnered less than 15% of the total vote. In New Hampshire, this indicated that Clinton had stronger support than Obama, and in retrospect CBS might have qualified its prediction of a NH primary victory for Obama.

CBS News also began asking the question, "Have you ever recommended someone vote for this candidate?" Again, the results of this question help indicate the relative strength of the support for each candidate in Iowa and New Hampshire.

Kathleen also gave several examples of leading questions used in the past to attempt to measure the American electorate's willingness to vote for a female President. A 1937 poll question was phrased: "Would you vote for a woman for President if she were qualified in every other aspect?" (Thirty-three percent of respondents said that they would.) A Virginia Slim poll of women, in 1970, phrased the question this way: "There won't be a woman President of the United States for a long time and that's probably just as well." (Sixty-seven percent of respondents agreed.) Apparently not only were the survey writers sexist, they didn't understand that you don't ask leading questions. Unfortunately, I didn't write down the current wording that CBS uses, but the basic approach is to ask respondents if they would ever vote for a candidate who was a man/woman/black/Hispanic/atheist/etc., if most people they know would ever vote for such a candidate, and if America is ready for such a candidate. (Eighty-one percent said they would vote for a woman candidate, but only 56% said most people they knew would and only 60% said America was ready for a female President.)

For more from Kathleen, check out some of her articles for CBSNews.com:

• How Does Age Affect Polls? Older Respondents May Process Questions Differently
• Why Do Polls Yield Different Results? A Look Inside The Numbers Shows Why Similar Polls Can Seem So Different
• Do VP Choices Matter? CBS News' Kathy Frankovic Breaks Down Polling Evidence From Past Presidential Elections