Eliciting Expectations

Abstract

We document how variation in the question format used to elicit belief distributions of percentage changes, such as return expectations, influences reported beliefs. We induce distributions to serve as a ground truth and vary how we ask about these distributions. Participants effectively report differences in means and volatilities irrespective of elicitation method. However, elicited levels vary as a function of question format. Reporting full distributions yields fairly accurate responses. In contrast, directly asking for means positively biases responses. Further, volatilities from confidence intervals are not dependable.  The evidence demonstrates complexities in survey design and helps researchers understand how to design and interpret surveys for meaningful results.

Keywords: Expectations, survey design, expected returns, recall, asset pricing, behavioral finance

JEL Classification: G02, G12, C90, D03, E03