Duration dependence of the business cycle in Japan: A Bayesian analysis of extended Markov switching model

For the purpose of analyzing cyclical stochastic processes in business cycles, we propose a regime-switching model with duration dependence that makes use of the Weibull model. The advantage of this model is that it relaxes a constraint of the Markov-switching model in favor of time-varying transition probabilities, and investigates the property of duration dependence in business states. We employ Bayesian inference via MCMC to overcome the drawbacks of maximum likelihood estimation. Estimation using the composite index of coincident indicator shows that both the contractions and the expansions of the Japanese business cycle exhibited positive duration dependence during the last two decades.