Turning Point Detection Using Markov Switching Models with Latent Information
A crucial task in the study of business cycle is the detection of the turning points, indicating the beginning (end) of a phase of growth (recession) of the economy. The dating proposed by experts are generally evaluated with the support of some empirical procedure. Many efforts have been devoted to propose statistical models able to detect the turning points and to forecast them. A class of models largely used for these purposes is the Markov Switching one. In this work we use a new version of the Markov Switching model, named Markov Switching model with latent information, which seems particularly able to detect the turning points in real time and to forecast them. In particular this model uses a latent variable, representing the cycle of the economy, as information to drive the transition probabilities from a state to another. Two examples are provided: in the first one we use Japanese GDP data, where our model and the classical Markov Switching model have a similar performance, but the first provides more information to forecast the turning points; in the second one we analyze USA GDP data, where the classical Markov Switching model fails in the turning points detection, whereas our model fits adequately the data.
KeywordsGross Domestic Product Business Cycle Turning Point Extended Kalman Filter Markov Switching
Financial support from Italian MIUR under grant 2006137221_001 is gratefully acknowledged.
- Altissimo, F., Marchetti, D. J., & Oneto, G. P. (2000). The Italian business cycle: coincident and leading indicators ans some stylized facts. Temi di Discussione del Servizio Studi–Banca d’Italia, 377.Google Scholar
- Bry, G., & Boschan. C. (1971). Cyclical analysis of time series: selected procedures and computer programs. NBER Technical Paper, 20.Google Scholar
- Diebold, F. X., Lee, J. H., & Weinbach, G. C. (1994). Regime switching with time-varying transition probabilities. In P. Hargreaves, (ed.), Nonstationary Time Series Analysis and Cointegration (pp. 283–302). Oxford: Oxford University Press.Google Scholar
- Diebold, F. X., & Rudebusch, G. D. (1989). Scoring the leading indicators. Journal of Business,60, 369–391.Google Scholar
- Filardo A. J. (1998). Choosing information variables for transition probabilities in a time-varying transition probability Markov switching model. Federal Reserve Bank of Kansas City, RWP 98–109.Google Scholar
- Harvey, A. C., & Shephard, N. (1993). Structural time series models. In: G. S. Maddala, C. R. Rao, & H. D. Vinod (Eds.), Handbook of statistics Vol. 11 (pp. 261–302). Amsterdam: Elsevier Science Publishers B.V.Google Scholar
- Otranto, E. (2001). The Stock and Watson model with Markov switching dynamics: an application to the Italian business cycle. Statistica Applicata,13, 413–429.Google Scholar