Comparisons of Stochastic Frontier Effect Models: Monte Carlo Simulation

This paper compares, by Monte Carlo simulations, the small sample properties of three different types (KGMHLBC, RSCFG and FE) of the stochastic frontier effect models which are designed for examining how observable firm characteristics influence on technical efficiency level. RSCFG has the scaling property that environmental factors affect the scale of technical inefficiency, but not the shape of the inefficiency distribution while KGMHLBC does not possess the property. Both RSCFG and KGMHLBC assume a specific distribution of technical inefficiency and are estimated by the maximum likelihood. On the other hand, FE does not impose any distributional assumption of inefficiency and statistical disturbance nor the uncorrelation assumption between inefficiency and input variables and is estimated by the fixed effect treatment. Our simulation results reveal that FE is robust and insensitive to various specifications in the estimation of production technology and the marginal effect of an environmental factor on efficiency while RSCFG and KGMHLBC are likely sensitive to a priori distribution of technical inefficiency. Comparing rank correlations between inefficiency estimates and true inefficiency, however, FE produces the least best estimate of inefficiency.