This paper develops a statistically unbiased and simple method for
measuring the difference of independent empirical distributions estimated
by bootstrapping or other simulation approaches. This complete
combinatorial method is compared with other unbiased and biased methods
that have been suggested in the literature, first in Monte Carlo
simulations and then in a field test of external and internal scope
testing in contingent valuation. Tradeoffs between methods are discussed.
When the empirical distributions are not independent a straightforward
difference test is suggested.
