Bayesian analysis of a vector autoregressive model with multiple structural breaks

This paper develops a Bayesian approach for analyzing a vector autoregressive model with multiple structural breaks based on MCMC simulation methods, extending a method developed for the univariate case by Wang and Zivot (2000). It derives the conditional posterior densities using an independent Normal-Wishart prior. The number of structural breaks is chosen by the posterior model probability based on the marginal likelihood, calculated here by the method of Chib (1995) rather than the Gelfand-Dey (1994) method used by Wang and Zivot. Monte Carlo simulations demonstrate that the approach provides generally accurate estimation for the number of structural breaks as well as their locations