The Analysis of Survey Data with Framing Effects

A well-known difficulty in survey research is that respondents' answers to questions can depend on arbitrary features of a survey's design, such as the wording of questions or the ordering of answers. We develop an empirical framework for analyzing survey data characterized by such framing effects. We show that the conventional approach to analyzing data with framing effects – randomizing across frames and pooling the data for estimation – does not generally identify a useful parameter. We consider a range of alternatives, and describe the conditions under which each yields a parameter of interest. One approach attempts to isolate the answers of respondents that are unaffected by framing. A second approach attempts to recover information on the distribution of “true answers” in the population – when true answers exist – using a technique analogous to post-survey re-weighting for non-response bias