A Spatial Stochastic Frontier Model with Spillovers: Evidence for Italian Regions
| Author | Efthymios G. Tsionas,Panayotis G. Michaelides |
| Published date | 01 July 2016 |
| Date | 01 July 2016 |
| DOI | http://doi.org/10.1111/sjpe.12081 |
A SPATIAL STOCHASTIC FRONTIER
MODEL WITH SPILLOVERS: EVIDENCE
FOR ITALIAN REGIONS
Efthymios G. Tsionas* and Panayotis G. Michaelides**
ABSTRACT
Efficiency measurement using stochastic frontier models is well established in
applied econometrics. However, no published work seems to be available on effi-
ciency analysis using spatial data dealing with possible spatial dependence
between regions. This article considers a stochastic frontier model with decompo-
sition of inefficiency into an idiosyncratic and a spatial, spillover component.
Exact posterior distributions of parameters are derived, and computational
schemes based on Gibbs sampling with data augmentation are proposed to con-
duct simulation-based inference and efficiency measurement. The new method is
illustrated using production data for Italian regions (1970–1993). Clearly, fur-
ther theoretical and empirical research on the subject would be of great interest.
II
NTRODUCTION
Efficiency measurement using stochastic frontier models is well established in
applied econometrics and numerous applications have seen the light of day in
agricultural economics, health, marketing etc. Stochastic frontiers have been
introduced by Aigner, Lovell and Schmidt (1977) and Meeusen and van den
Broeck (1977) and have found numerous applications, so far. For a compre-
hensive review see Bauer (1990) as well as the excellent reviews by Greene
(1993, 2001). Extensions of stochastic frontier models have been considered,
among others, by Greene (1990), Stevenson (1990), and Tsionas (2000, 2002).
Although much work has been done in connection with stochastic frontier
models using cross-section or panel data, no published work seems to be
available on efficiency analysis using spatial data. In fact, countless studies
have been conducted applying stochastic frontier models to regional data, but
none –to the best of our knowledge –deals with possible spatial dependence
between regions. This has been the point of departure of our investigation.
Analytically, the article develops a model that allows for spatial correlation
and spillover effects in inefficiency. The starting point is the usual linear model
*Lancaster University
**National Technical University of Athens
Scottish Journal of Political Economy, DOI: 10.1111/sjpe.12081, Vol. 63, No. 3, July 2016
©2015 Scottish Economic Society.
243
To continue reading
Request your trial