Bayesian Methods for TFP Analysis of a Multi-Region Economy with Dynamic Structure and Application to Japan

Abstract

In this paper, we present the Bayesian methods for total factor productivity (TFP) analysis of a multi-region economy with dynamic structure. Conventional approaches to the empirical analysis of regional economic growth are incapable of providing an appropriate description of the behavior of TFP. Accordingly, we introduce a set of Bayesian statistical models based on regional production functions. In our framework, TFP is regarded as a time-varying parameter, and is estimated using a Bayesian smoothness priors approach. Compared with conventional approaches, this application of a Bayesian method makes it possible to analyze the behavior of TFP in detail. In addition, we take into account the fact that elasticity of output with respect to factors of production may vary across regions, implying a structural heterogeneity among the regions. Further, the possibility of structural changes in regional economies is incorporated. As an empirical application, we present the results for an analysis of Japanese prefectural data.