Bias-correction in DEA efficiency scores using simulated beta samples: an alternative view of bootstrapping in DEA

Abstract

Bootstrapping of DEA efficiency scores came into being under the criticism that DEA input/output data may contain random error, and as a result the efficient frontier may be warped by statistical noise. Since the publication of the seminal paper by Simar and Wilson (1998), several researchers have carried out bootstrapping the DEA frontier, re-computing the efficiency scores after correcting the biases and developing confidence intervals for bias-corrected scores. We view bias-correction in DEA efficiency scores from a different perspective by randomising the efficiency scores that follow underlying beta distributions. In a step-by-step process, using the simulated beta samples, we show how to correct the biases of individual scores, construct confidence intervals for the bias-corrected mean scores and derive some statistical results for the estimators used in the process. Finally, we demonstrate this method by applying it to a set of banks.