From the issue dated
In 1936, the Literary Digest confidently predicted that Alf Landon would defeat President Franklin D. Roosevelt by 54 percent to 40 percent. The poll had been fatally skewed toward upper-income Americans, having drawn names from sources like lists of automobile-license registrations.
Today most pollsters know how to devise truly random population samples. But public-opinion polls should still be approached with caution, says Adam J. Berinsky, an associate professor of political science at the Massachusetts Institute of Technology. In Silent Voices: Public Opinion and Political Participation in
Q. Why do you believe that "don't know" responses are sometimes biased in a particular direction, and aren't just random noise?
A. In the case of social-welfare policy, there's a lot of work in political science that suggests that this is a hard issue for people, and that folks who might tend to be more expansive or liberal or generous are more conflicted about these issues than conservatives are. ... These are the kinds of people who might remove themselves from opinion-poll questions because of their ambivalence. ...
Then I got to thinking about race. If you look at the 1989 mayoral election in New York City, polls suggested that David Dinkins would win by a large margin, but in fact he just squeaked by Giuliani. ... In the book, I find that it's older Jewish Democrats who were the people most likely to misrepresent their preferences -- most likely to say, "I don't know" when, in fact, they were planning to vote for Giuliani. Unlike the social-welfare-policy case, where people really don't know because they're conflicted, these are people who have a sense of where they stand but just don't want to tell you.
These older Jewish Democrats -- or, as I like to call them, Mom and Dad -- they didn't want to say that they weren't going to vote for Dinkins, because they're good Upper West Side New York Democrats, but at the same time there was an unease. It wasn't necessarily that they were racist in any way, but they didn't want to risk appearing racist in the context of a survey.
Q. Is this just a problem for professional pollsters? Or do academic political scientists face similar kinds of potential bias?
A. Some political scientists have wondered what would happen if a fully informed public was asked about its preferences. ... In some of their studies, low-income people who know a lot about politics are used as a proxy for poor people who don't know anything about politics. And there are obvious problems with that kind of method. The process of learning about politics is something that itself could change your interests. That sort of work makes me nervous.
Section: Research & Publishing
Volume 50, Issue 41, Page A14