Bootstrapping the Malmquist productivity index - a simulation study

Abstract

This paper presents a Monte Carlo simulation study of the bootstrap algorithm proposed by Löthgren and Tambour for calculation of bootstrap confidence intervals for the firm-specific Data Envelopment Analysis (DEA) Malmquist productivity index. The simulation results indicate that the coverage accuracy of bootstrap confidence intervals are near the nominal confidence level for small to moderate sized samples. For larger sample sizes, a clear convergence of empirical to nominal level is found.