Varieties of risk elicitation

We explore a variety of risk preference elicitation procedures that involve direct choice from a set of lotteries, including budget lines (BL) and binary choice lists (HL). We find statistically significant violations of the expected utility hypothesis (EUH) consistent with disappointment aversion, and also find violations of first order stochastic dominance, but both sorts of violations are mostly small and only slightly impair the predictive power of a parametric implementation of EUH. The estimated coefficient of relative risk aversion, gamma, varies widely across individual subjects (consistent with EUH) and also across elicitation tasks (inconsistent with direct implementation of EUH). An alternative nonparametric measure of risk preferences displays similar patterns. The two risk preference measures are highly correlated with each other for each elicitation task. Each separate measure varies widely across individual subjects and across elicitation tasks, with low to nil correlation between BL tasks and HL tasks. Some of the variation across tasks can be explained by attributes such as graphical vs text representation that have no role in decision theory.