Likelihood analysis of random effect stochastic frontier models with panel data

Abstract

The paper takes up posterior analysis of the stochastic frontier model with random effects when panel data is available. Available treatments of the model result in a likelihood function that is highly nonlinear and, as a result, applied researchers prefer to use fixed effect formulations when efficiency measurement is sought from panel data. The methodology is based on Gibbs sampling. It is shown how posterior distributions of parameters can be derived and how firm-specific efficiency measures can be computed.