Research Of Financial Time Series Based On Markov Switching Cointegration Model

Vector error correction model (VECM) is an important expression of cointegration. It overcomes the false regression, and describes the long-term (static) and short-term (dy-namic) characteristics in the sequence of high-dimensional economic variable effectively. But VECM is based on the assumption that the component of time series vector is of unit root. However, most data in the financial markets have the characteristics of long memo-ry, and the existence of price adjustment,transaction costs and the change of economic policy in real economy make the long-term equilibrium relationship between variables not often occur in every period. Based on this, Markov Switching fractional Vector Error Correction Model(MSFVECM) is proposed in this paper. Due to the complexity of the model and thus difficulty in parameter estimation through usual likelihood method, we try the Bayesian method through Gibbs sampling in this article to estimate the parameters. The structure of this paper is as follows:Firstly, in Chapter 1, we give a brief description of the definition of cointegration. In Chapter 2, we describe the background of VECM and improved FVECM and MSVECM, in which we point out the the necessity of the MSFVECM. In Chapter 3, We try the Bayesian method through Gibbs sampling to estimate the six parameters of MSFVECM, and give the procedure to complete the estimates. Finally, we confirm the validity of the estimation method through Monte Carlo simulation. Then we use this method to analyze the financial market and find out that the new model is appropriate and the estimation method is efficient.