A Bayesian approach is considered for identifying sources of nonstationarity for models with a unit root and breaks. Different types of multiple breaks are allowed through crash models, changing growth models, and mixed models. All possible nonstationary models are represented by combinations of zero or nonzero parameters associated with time trends, dummy for breaks, or previous levels, for which Bayesian posterior probabilities are computed. Multiple tests based on Markov chain Monte Carlo procedure sare implemented. The proposed method is applied to a real data set, the Korean GDP data set, showing a strong evidence for two breaks rather than the usual unit root or one break.
Bibliographical noteFunding Information:
The authors are grateful for the valuable comments of two referees. This work was supported by the Korea Research Foundation Grant funded by the Korean Government (MOEHRD, Basic Research Promotion Fund) (KRF-2008-C00017).
- Markov chain Monte Carlo
- Multiple breaks
- Unit root test