Minimax regret priors for efficiency estimation

Abstract

We propose a minimax regret empirical prior for inefficiencies in a stochastic frontier model and for its other parameters. The class of priors over which we consider minimax regret is given by DEA interval scores and, for the parameters, the class of priors induced by maximum likelihood estimates. The new techniques are shown to perform well in a Monte Carlo study as well as in real data for large U.S. data banks.