Don’t Know Responses and Preference Gaps: An Assessment Using the Gilens Dataset

• In Event: Advances in Public Opinion Research Methods

Abstract

There is increasing interest in the inequality of political representation in policy, particularly the opinions of different income groups. Most research finds that preferences of the rich are over-represented, not only in the United States but across much of Europe and possibly elsewhere. But, there are substantial limits on this mis-representation, as scholars have demonstrated that the policy preferences of different income groups are strikingly similar. Moreover, even where the preferences do differ, the middle (and poor) have their way almost as often as do the rich. In this paper, we consider whether the observed preference gaps we do observe our real. Building on other previous research, we argue that differences in information account for a nontrivial portion of preference gaps. More specifically, we hypothesize that income is correlated with information, and that low information leads respondents to guess, effectively at random, leading to greater gaps in between groups as income differences increase. We then test the hypothesis using the dataset from Gilens’ now-classic studies of unequal representation in the US between 1982 and 2002. The results support our expectations, as we find that proportion of “don’t know” (DK) responses is negatively related to income levels, that the proportion is positively related to preference gaps, and that the proportion also predicts larger divided opinion distributions, as we would expect from guessing. These results imply that a portion of the differences in opinion that we observe is not real, which has fairly clear implications for our assessments of policy congruence that we contemplate in the concluding section.

Keywords: groups, policies, information, opinion, representation, congruence

Authors

• Mads Andreas Elkjaer, University of Copenhagen

• Christopher Wlezien, University of Texas at Austin