Experiments in filtering political opinions

Changes in the wording of core measures of political attitudes in the American National Election Studies have generated a good deal of controversy about artifactual sources of change in the U.S. electorate. This research, based on several field experiments and replications, investigates the effects of using or not using various types of opinion filter questions that have appeared in the SRC/CPS series over the years. The analysis shows that the use of a filter interacts significantly with a respondent's level of education and interest in politics, particularly the latter, in determining whether a respondent will offer an opinion on a given public policy issue. But the study also demonstrates that such an interaction occurs primarily when the question about interest in politics is askedafter, rather thanbefore, a set of political issue items. In interpreting these order-and-context effects the authors develop a self-perception model of how respondents infer their interest in politics from information that isavailable in memory about their own behavior in the survey interview.