When Do Sensitive Survey Questions Elicit Truthful Answers? Theory and Evidence with Application to the RRT and the List Experiment

Corruption, vote buying, and other sensitive topics are difficult to study because people tend to under-report them in surveys. The degree of under-reporting bias has been shown to vary across studies, contexts, and question structures, but no systematic explanation for the variation has been advanced. I provide a simple theory that describes conditions under which an individual is more - or less - likely to respond truthfully to a sensitive question. The theory is based on the intuition that respondents lie to avoid looking bad in the eyes of interviewers. The main implication is that a respondent's second-order beliefs about the interviewer's priors are a key determinant of truthfulness. Empirical analysis of original data supports this claim: respondent's second-order beliefs correlate strongly with self-reported nonvoting and cheating. I show how second-order beliefs can be used to adjust for under-reporting bias.