Semiparametric Bayesian inference for stochastic frontier models

In this paper we propose a semiparametric Bayesian framework for the analysis of stochastic frontiers and efficiency measurement. The distribution of inefficiencies is modelled nonparametrically through a Dirichlet process prior. We suggest prior distributions and implement a Bayesian analysis through an efficient Markov chain Monte Carlo sampler, which allows us to deal with practically relevant sample sizes. We also consider the case where the efficiency distribution varies with firm characteristics. The methodology is applied to a cost frontier, estimated from a panel data set on 382 U.S. hospitals.