Predicting technical effciency in stochastic production frontier models in the presence of misspecification: a Monte-Carlo analysis

Abstract

This paper provides a theoretical explanation for the sensitivity of technical efficiency measures to the choice of functional specification in stochastic production frontier models. It is shown that inappropriate functional specifications translate into a misspecification in the conditional mean of the stochastic frontier regression model. This misspecification, in turn, results in estimates of technical efficiency, confidence intervals and production elasticities being biased, even asymptotically. Monte-Carlo simulations reveal that the severity of the bias depends on the functional specification and the percentage contribution of the variance of technical inefficiency to the total variance of the composed errors.