Refining the bootstrap method of stochastic dominance analysis: The case of the January effect

This study investigates the effect of sample size and population distribution on the bootstrap estimated sampling distributions for stochastic dominance (SD) test statistics. Bootstrap critical values for Whitmore's (1978) second- and third-degree stochastic dominance test statistics are found to vary with both data sample size and variance of the population distribution. The results indicate the parametric nature of the statistics and suggest that the bootstrap method should be used to estimate a sampling distribution each time a new data sample is drawn. As an application of the bootstrap method, the January small firm effect is examined. The results conflict with the SD results of others, and indicate that not all investors would prefer to hold just a portfolio of small capitalization firms in January.