An experimental methodology testing for prudence and third-order preferences

We propose an experimental method to test individuals for prudence (i.e. downside risk aversion) outside the expected utility framework. Our method relies on a novel representation of compound lotteries which allows for a systematic parameterization that captures the full generality of prudence. Therefore, we develop a general technique for lottery calibration in experiments. Since we investigate a very subtle third-order property we test our method in the laboratory employing a factorial design. We find that it yields robust results and that prudence is observed on the aggregate as well as on the individual level. Further we show that preferences based on statistical moments, in particular skewness seeking, can at most approximately explain individuals' behavior in the experiment.