A comparison of efficiency estimation methods via Monte Carlo analysis

In this paper we investigate the performance of the five efficiency estimation methods which include the stochastic frontier model estimated by maximum likelihood (SFML), the stochastic frontier model estimated by corrected ordinary least squares (SFCOLS), the data envelopment analysis (DEA) model, the combined estimation of SFML and DEA (SFML + DEA), and the combined estimation of SFCOLS and DEA (SFCOLS + DEA) using Monte Carlo analysis. The results include: 1) SFML provides most accurate efficiency estimates for the sample size 150 or over, 2) SFML + DEA or SFCOLS + DEA perform better for the cases with sample size 25, 50, and low random errors, 3) SFCOLS performs better for the case with sample size 25, 50, and very high random errors.