Experiments on portfolio selection: A comparison between quantile preferences and expected utility decision models

This paper conducts a laboratory experiment to assess the optimal portfolio allocation under quantile preferences (QP) and compares the model predictions with those of a mean-variance (MV) utility function. We estimate the risk aversion coefficients associated to the individuals' empirical portfolio choices under the QP and MV theories, and evaluate the relative predictive performance of each theory. The experiment assesses individuals' prefer-ences through a portfolio choice experiment constructed from two assets that may include a risk-free asset. The results of the experiment confirm the suitability of both theories to predict individuals' optimal choices. Furthermore, the aggregation of results by individual choices offers support to the MV theory. However, the aggregation of results by task, which is more informative, provides more support to the QP theory. The overall message that emerges from this experiment is that individuals' behavior is better predicted by the MV model when it is difficult to assess the differences in the lotteries' payoff distributions but better described as QP maximizers, otherwise.