Combining DEA and stochastic frontier models: An empirical Bayes approach

The paper proposes to combine stochastic frontier models and linear programming methods by using DEA measures as priors of efficiency in the stochastically frontier model and Monte Carlo methods are developed to perform empirical Bayes inference in the new model.

Abstract

The paper proposes to combine stochastic frontier models and linear programming methods by using DEA measures as priors of efficiency in the stochastic frontier model. These prior measures are revised to obtain posterior measures using Bayes’ theorem. Monte Carlo methods are developed to perform empirical Bayes inference in the new model. The methods are organized around Gibbs sampling with data augmentation. The new techniques are illustrated in the context of efficiency measurement in US airlines.