On the Ability to Disentangle the Two Errors in the Normal/Half-Normal Stochastic Frontier Model

Abstract

In this paper, a simulation experiment is carried out in the framework of the normal/half-normal stochastic frontier model in order to analyse its ability to disentangle the two types of errors that form the composite error. According to the results obtained through the mean bias and the mean squared error of the parameters and efficiencies, and via Spearman rank correlation between actual and estimated efficiencies, a good performance of the model is only obtained when considering medium-sized or large samples and the variance of the inefficiencies highly contributes to that of the composite error. The problems of wrong skewness and absence of random error are also addressed. The influence on the results of selecting a wrong distribution for the inefficiency term is also analysed.