Many economic time series have features of nonlinearity and regime change. Hamilton's Markov-Switching (MS) model is useful to capture these characteristics. The purpose of this study is to apply Markov-Switching model to examine the behavior of stock returns and the joint behavior of stock returns and macroeconomic variables.
The analysis of univariate MS model, allowing both mean and variance to change, shows quite different results according to the time horizon. The peculiarity of stock returns from 1980 to 1994 is largely determined by changes in volatility, whereas the one from 1986 to 1994 determined by changes in mean. The results of bivariate MS model show that there exists a negative relation between stock returns and interest rates, a positive relation between stock returns and business cycle. We also draw a comparison between volatilities estimated by different models. MS model does not seem to capture enough variation in volatility compared to GARCH.