Estimating stochastic production frontiers: A one-stage multivariate semiparametric Bayesian concave regression method

This paper describes a nonparametric Bayesian estimator for production frontiers that satisfies the axioms of monotonicity and concavity. An inefficiency term that allows for a departure from the homoscedastic prior distributional assumption is jointly estimated in a single stage with cross-sectional data. Our Monte Carlo simulation experiments demonstrate that the frontier and efficiency estimations are computationally competitive, align well with economic theory, and allow for the analysis of larger data sets than existing nonparametric methods. We use the proposed method to investigate Japan's concrete industry, an important component of the nation's construction sector, from 2007 to 2010. Our finding of a significant size-weighted inefficiency supports the argument that economic stimuli given to Japan's concrete industry may result in large losses due to inefficiency. (C) 2020 Published by Elsevier B.V.