On the Meaning of Survey Reports of Roll‐Call “Votes”

Contemporary efforts to evaluate representation often compare survey measures of how citizens say they would vote on legislation to what elected officials do in office. These comparisons generally suggest poor representation. We argue here that this common design is unlikely to effectively evaluate representation because responses to survey questions differ in important aspects from voting in legislatures. Measurement error and construct validity undermine the comparison. Three survey experiments show that providing partisan and nonpartisan information readily available to legislators materially changes respondents' expressed preferences on roll‐call votes. With information, expressed policy positions are both less centrist and more closely matched to legislator behavior in their preferred party. Respondents also appear aware of their own lack of knowledge in evaluating roll‐call policy votes. The treatment effect of information decreases in confidence judging policy in that area. We show similar patterns for respondent opinions on Supreme Court decisions.