Consistent tests for risk seeking behavior: A stochastic dominance approach

We develop non-parametric tests for prospect stochastic dominance Efficiency (PSDE) and Markowitz stochastic dominance efficiency (MSDE) with rejection regions determined by block bootstrap resampling techniques. Under the appropriate conditions we show that they are asymptotically conservative and consistent. We engage into Monte Carlo experiments to assess the nite sample size and power of the tests allowing for the presence of numerical errors. We use them to empirically analyze investor preferences and beliefs by testing whether the value-weighted market portfolio can be considered as efficient according to prospect and Markowitz stochastic dominance criteria when confronted to diversi cation principles made of risky assets. Our results indicate that we cannot reject the hypothesis of prospect stochastic dominance efficiency for the market portfolio. This is supportive of the claim that the particular portfolio can be rationalized as the optimal choice for any S-shaped utility function. Instead, we reject the hypothesis for Markowitz stochastic dominance, which could imply that there exist reverse S-shaped utility functions that do not rationalize the market portfolio.