We present Bayesian statistics and Gibbs sampling, an MCMC simulation technique, as tools for making inferences in stochastic frontier models for panel data from the banking sector. In our empirical example, the Bayesian approach is applied to estimate a short-run frontier cost function for N = 58 branches of a Polish commercial bank, observed over T = 4 quarters of one year. We use a translog cost function (with regularity conditions imposed for an 'average' branch) and treat inefficiency as a random individual effect, assuming a varying efficiency distribution (VED) specification proposed by Koop, Osiewalski and Steel (1997).

Keywords: Bayesian econometrics, panel data, cost models, microeconomics of bank.
JEL: C11, C23, D24, G21

  Jerzy Marzec, Jacek Osiewalski - Bayesian Inference on Technology and Cost Efficiency of Bank Branches - plik pdf; (1,55 MB)