Further Improvements of Finite Sample Approximation of Central Limit Theorems for Envelopment Estimators

A simple yet easy to implement method is proposed to further improve the finite sample approximation of the recently developed central limit theorems for aggregates of envelopment estimators. Focusing on the simple mean efficiency, we propose using the bias-corrected individual efficiency estimate to improve the variance estimator. The extensive Monte-Carlo experiments confirm that, for relatively small sample sizes (≤ 100), with both low dimensions and especially for high dimensions, our new method combined with the data sharpening method generally provides better ‘coverage’ (of the true values by the estimated confidence intervals) than the previously developed approaches.