Implementation of stochastic dominance: a nonparametric kernel approach

Abstract

The empirical application of stochastic dominance in the selection of crop alternatives under risk has typically been based on a step function specification of the empirical cumulative distribution function (cdf). This specification has questionable consequences for the analysis, particularly with regards to the tails of the distribution. This article proposes an alternative nonparametric approach based on a nonparametric regression of the cdf. The effect of this alternative specification is then demonstrated using data from Northern Florida. Results are more realistic. The first-degree stochastic dominance results are less discriminating under the new specification while the second-degree results are more discriminating.