Bayesian analysis of unemployment dynamics in Japan

Abstract

In this paper, we investigate the unemployment dynamics in Japan within the framework of Bayesian modelling. To consider structural changes in a model for the matching function specified in Cobb-Douglas form, we regard not only the matching efficiency but also the elasticities of new hiring with respect to unemployment and with respect to vacancies as time-varying parameters. Then, from a Bayesian perspective, these are treated as random variables and smoothness priors are introduced. In addition, a set of models for the matching function and the smoothness priors is described in a state space representation. The parameter estimation is carried out using Kalman filter and fixed-interval smoothing. The average for the period between January 2009 and December 2010 suggests that 60% of the total unemployment rate was a result of structural and frictional factors and that 40% was attributable to a labour demand deficiency. Further, in terms of matching efficiency, the Japanese labour market is not viewed as functioning effectively even in the late 2000s.