CMStatistics 2017: Start Registration
View Submission - CMStatistics
B1617
Title: Spatial dependence and unobserved heterogeneity in stochastic frontier models: A Bayesian approach Authors:  Antonio Carvalho - Heriot-Watt University (United Kingdom) [presenting]
Abstract: The aim is to contribute to the literature of panel data Stochastic Frontier modelling and efficiency measurement in production and cost functions by discussing estimation of efficiency and other parameters of interest in the context of spatial dependence and unobserved heterogeneity. The proposed Bayesian approach allows for the decomposition of efficiency into a time-varying component and a persistent component in a random effects model with spatial spillovers. The measurement of efficiency with spatial spillovers is more complex than in the non-spatial case and measurement methods are discussed in relative and absolute scales. A Bayesian approach with a standard assumption of half-normal distribution for efficiency is outlined. Small sample performance of the model is deeply related to the underlying signal-to-noise ratios with good performance for larger samples and encouraging results for applied researchers. The approach is applied to aggregate productivity in European countries between 1992 and 2005, with a particular focus on transition economies and the role of spatial dependence and unobserved heterogeneity in the production frontier. Results show a large amount of persistent inefficiency which would be ignored under less complex estimation methods, and also a non-negligible amount of spatial dependence.