A nonparametric test on mean difference of Data Envelopment Analysis efficiency estimates – bootstrapping approach

This paper presents a nonparametric method to test if the mean difference of DEA efficiency estimates between two groups statistically exists. A proposed method employs a bootstrapping approach to generating BCC efficiency estimates through Monte Carlo simulation resampling process. For the purpose of demonstration, we empirically apply the proposed method to the Korean bank industry and compare its result with the result by the traditional deterministic DEA method. The nonparametric statistical hypothesis testing procedure in this study, which considers not only stochastic variability of the DEA data, but also random radial deviations off the efficient frontier, serves as a useful tool for objectively evaluating whether the mean difference of DEA efficiency estimates between groups is statistically significant.