Abstract
This article attempts to simultaneously investigate different regimes in both mean and volatility of post-war US GDP growth using a four-regime Bayesian Markov switching model. Bayesian approach suffers from the label switching problem that leads to the failure of identifying regimes. We introduce two methods to deal with the label switching problem in posterior simulations of parameters. The four regimes identified by either of the two methods capture different characteristics in mean and volatility of US GDP growth.
Funding
This research is supported by the National Natural Science Foundation of China [grant number 71301072] and the Doctoral Fund of Ministry of Education of China [grant number 20120091120001].
Notes
1 For simplifications of the Bayesian inference, we use precision instead of variance.
2 Symmetric priors are chosen for the Markov switching model since the use of asymmetric priors would reduce the marginal likelihood of the model (Geweke and Jiang, Citation2011).