Structural Breaks and Regime Shifts in Financial and Macroeconomic Sectors
- Author(s): Su, Yanpin;
- Advisor(s): Chauvet, Marcelle;
- et al.
This dissertation focuses on the extensions of the Markov switching model (both univariate and multivariate time series) with applications in financial and macroeconomic sectors. Chapter one provides an overview for this dissertation. Chapter two (joint with Professor Marcelle Chauvet) proposes a flexible univaritate Markov switching model that allows for recent changes observed in the U.S. business cycle in the last six decades. The Markov switching model allows three Markov processes to characterize the dynamics of U.S. output fluctuations. I consider the possibility that both the mean and the variance of growth rates of real GDP can have short run fluctuations in addition to the possibility of a long run permanent break. We find that, differently from several alternative specifications in the literature, the proposed flexible framework successfully represents all business cycle phases, including the Great Recession. In addition, we find that the volatility of U.S. output fluctuations has both a long run pattern, characterized by a structural break in 1984, as well as business cycle dynamics, in which periods of high uncertainty are associated with NBER recessions.
Chapter three (joint with Professor Marcelle Chauvet) proposes a dynamic joint bi-factor model to study the nonlinear dynamic relationship between stock and bond markets via the covariance structure of the two latent factors and Markov processes. One latent factor representing stock market is extracted from three excess stock returns and the other unobserved factor displaying the bond market is extracted from 10-year excess Treasury bond returns. Each of these factors follows distinct two-state Markov process representing phases of these markets. The Markov process for the stock factor represents bear and bull stock markets, whereas the Markov process for the bond factor represents low and high Treasury bond return phases. The empirical results show that there is a mild positive correlation between stock and bond factors over the whole sample, which is consistent with most related literature. Surprisingly, the comovement between stock and bond markets is the strongest during the NBER-dated recessions which implies that investors should not use bonds to hedge against stock market risk at that time. The comovement between stock and bond markets is the weakest during pre-NBER and post-NBER recessions, which shows that the stock and bond portfolio risk is well diversified then.
Chapter four proposes a generalized impulse response function for joint bi-factor model which can be extended to the Bivariate Two-Markov-Chain VAR model, while the concept of the generalized impulse response is introduced by Koop et al. (1996) which can be used for both linear and nonlinear multivariate models. Compared with current popular generalized impulse response function in Markov-switching vector autoregressive models by Karame (2010, 2012), the joint bi-factor model contains two first-order Markov chains which can capture two different but related markets well, for instance, stock market and bond market. The empirical illustration of this generalized impulse response function makes use of estimation of the joint bi-factor model from chapter three. The generalized impulse response graphs show that both stock and bond markets react most strongly in the Bear market and react to the weakest in the Bull market when a unit stock shock occurs; both stock and bond markets react most strongly in the high bond return phases and react to the weakest in the low bond return phases when a unit bond shock occurs.