An Effective and Simple Tool for Measuring Loss Aversion

In prospect theory (PT) the loss aversion index, lambda, measures the size of the concave kink of the gain-loss utility function at the reference point. A truth-telling mechanism for assessing personal beliefs, the quadratic scoring rule, is extended to measure loss aversion. We control for the bias captured by decision weights in PT and quantify lambda efficiently with only three quadratic scores. In an experiment, we demonstrate these features for risk and extend the tool to ambiguity. We find median values of lambda = 1 at the aggregate level for both sources of uncertainty. Probability and event weighting are less pronounced but accord with earlier findings from the literature. These weights depend on the sign of the corresponding outcomes, which is implication of reference-dependent preferences. Event weighting is also observed at the individual level. After controlling for these weights, we find very few subjects who are loss averse or gain seeking.