A simulation study of hypothesis tests for differences in efficiencies

This paper reports the results of a simulation study to evaluate the performance of Banker’s new asymptotic DEA tests for inefficiency differences between two groups. Their performance is evaluated relative to the performance of conventional parametric corrected ordinary least squares (COLS) tests and the Welch and Mann-Whitney tests used earlier in the DEA literature. The authors consider five different underlying production functions, four different inefficiency distributions and four different sample sizes in the present study. The results include the following: (1) The new asymptotic DEA tests outperform COLS-based tests even when the parametric functional form employed for COLS estimation is identical to the underlying production function. (2) The asymptotic DEA tests are robust in that they perform well for different underlying production functions and inefficiency distributions. (3) Welch’s mean and Mann-Whitney tests perform worse than both the new asymptotic DEA tests and the COLS-based tests. (4) The performance of COLS-based tests is improved when Banker’s new test statistics are employed using COLS estimates of production inefficiencies. (5) No marked differences in Type II error are found between the various tests with one-output, two-input production technologies. But, the Welch’s mean and Mann-Whitney tests and COLS-based tests perform slightly better than the asymptotic DEA tests in terms of proportion of Type II errors with one-output, five-input production technologies.