4.33 A Robust Test For Stochastic Dominance
Statistical tests are developed for the hypotheses
that first- or second-degree stochastic dominance
holds between some pair of random variables in a set.
The tests are applicable when the distributions of the
random variables are unknown, but samples from each
distribution are observed. These tests do not require
that the distributions be in parametric families, and
do permit some statistical dependence of the random
variables within an observation period, and across
periods. Thus, the tests are robust with respect to
functional form and some patterns of statistical
dependence. The paper applies the tests to data on
daily returns from closed-end mutual funds, and
concludes that some funds are second-degree
stochastically dominated.