Abstract:

Measuring risk preferences is challenging because people’s
choices over lotteries violate expected-utility axioms and sometimes
vary depending on how the lotteries are framed. We develop a two-stage
procedure to measure risk preferences, and we demonstrate it via a
survey about hypothetical retirement investment choices administered
to 601 Cornell students. The first stage is the standard method of
eliciting choices over risky lotteries. In the second stage, we
confront participants with their inconsistencies—their different
responses to choices framed differently that should be the same
according to expected-utility axioms—and allow them to update their
choices. Our key assumption is that individuals’ updated, “reasoned”
choices more closely reflect their preferences than their original,
“untutored” choices. We find that on average, participants update in
the direction of consistency with expected-utility axioms, and their
reasoned choices may exhibit less risk aversion than their untutored
choices. Our results suggest that deviations from the axioms may
typically reflect decision errors rather than non-expected-utility
preferences. Our two-stage procedure may hold promise as a way to
measure risk preferences for the purpose of setting optimal defaults
or giving advice about portfolio allocation.