Resum:
We analyze the use of discrete choice models for the estimation of risk
aversion and show a fundamental flaw in the standard random utility model which
is commonly used in the literature. Specifically, we find that given two gambles, the
probability of selecting the riskier gamble may be larger for larger levels of risk aversion.
We characterize when this occurs. By contrast, we show that the alternative
random preference approach is free of such problems.