A State-Space Stochastic Frontier Panel Data Model

In this paper we introduce a state-space approach to the econometric modelling of cross-sectional specific trends (temporal variation in individual heterogeneity) and time varying slopes in the context of panel data regressions. We show that our state-space panel stochastic frontier model nests some of the popular models proposed in the literature on stochastic frontier to accommodate time varying inefficiency and its dynamic version (productivity). A detailed discussion of alternative model specifications is provided and estimation (along with testing procedures for model selection) is presented. The empirical application uses the EU-KLEMS dataset which provides data in the period 1977-2007 for 13 countries and 20 sectors of each economy. Our main empirical interest is centered on productivity analysis and thus we focus on the stochastic frontier interpretation of this cross-sectional specific temporal variation. A post-estimation growth accounting is introduced in order to provide a quantitative assessment of the main factors behind sectoral labour productivity growth for each country.