Bayesian Regime-switching Analysis via Stochastic Approximation Monte Carlo

Abstract

Monte Carlo methods have received much attention in the recent literature of the regime- switching analysis. However, the conventional Markov chain Monte Carlo (MCMC) algorithms, such as Metropolis-Hastings, tend to get trapped in a local mode in simulating from the posterior distribution of regime-switching time-series, rendering the inference ineffective. In this paper, we focus on the finding the best likelihood value in Bayesian nonlinear time-series model (Kim and Cheon, 2010) and the detection of multiple regime-switching in monthly simple returns and quarterly unemployment rate via the stochastic approximation Monte Carlo algorithm (Liang et al., 2007). The numerical results indicate that our method outperforms MCMC significantly for the regime-switching identification, and provide 3 and 5 regime switchings in monthly simple returns and quarterly unemployment rate, respectively.