Abstract:
We investigate whether violations of canonical axioms of choice under risk are mistakes or a manifestation of true preferences. First, we elicit axiom and gamble preferences and then allow subjects to revise their potentially conflicting preferences. Among the behavioral patterns that allow for a clear-cut interpretation on the decision level, we find that roughly 70% of axiom violations are intentional whereas only 30% are mistakes. On the subject level we can clearly categorize almost half of our subjects. Among those, roughly 24% are rational expected utility maximizers, 24% make occasional mistakes, and 52% refute the normative value of these axioms.