Don't Bother Polling People on Why They Vote, Researchers Say

As we mentioned last week, a recent Washington Post/ABC News poll asked voters whether they were more or less likely to vote for a candidate with various personal traits commonly identified with one of the 2008 presidential hopefuls (though no names were used). The poll found that voters were most likely to be turned off by a candidate over the age of 72 (58 percent said they were less likely to vote for such a candidate), while only 6 percent said they were less likely to vote for a black candidate. Full results here.

But as one reader in Virginia argued to us, one shouldn't assume that this has much bearing on reality. We decided to investigate. While the pollster who conducted this survey, ABC's Gary Langer, ardently defended its utility, some researchers agreed with the reader, saying the entire premise is flawed.

The question comes down to this: Are people able to accurately describe in the abstract how they vote? Stanford University's Jon Krosnick, a leading expert on the psychology of political behavior, says no.

"People just simply cannot answer these questions accurately," Krosnick tells News Desk. "When we study, as academics, the factors that cause vote choice, we do it statistically. We don't rely on people's explanations."

Langer counters: "Each of these candidates has attributes that either hurt them or help them," he says. "Evaluating individual attributes ... is a real part of the dynamic."

The poll, Langer says, is not about predicting votes, but about identifying the important dynamics of the campaign. "Clearly, these results to work in that respect," he says.

Another public opinion expert, Robert Shapiro of Columbia University, raised a second qualm with the poll: Because each trait was closely identified with one candidate, well-informed respondents may have evaluated the quality based on the candidate, not the other way around. But the degree to which respondents were thinking of a particular candidate is not known, further muddying the waters.

"I think in the abstract, these things are not very useful," Shapiro says. He adds that the only really important demographic for these types of questions are political independents without strong party attachments.

Both Krosnick and Shapiro argue for a different approach: instead of asking people to describe their preferences, they say behavior should be studied experimentally. One way, they said, would be to create hypothetical candidates with certain qualifications, and manipulate the gender, race, or age among respondents while keeping the political resume the same, and measuring the change in preferences.

Etc.: Politics, Post Civil Rights, on USNews.com