Functional Coefficient Stochastic Volatility Model

To estimate assets' volatility utilizing time series models has important practical and academical meanings. Usually, there are two approaches of analyzing volatility, and one of them is utilizing the time series Econometrics model which we take in this paper. We proposed a generalized stochastic volatility model, the Functional Coefficient Stochastic Volatility (FCSV for short) model to formulate behavior of volatility. The FCSV does belong to the family of stochastic volatility model, an alternative to ARCH-GARCH type model, however, as being set up under nonparametric representation the FCSV generalized usual stochastic volatility model. Moreover, FCSV does not only include stochastic volatility model in the literature, but also enable researchers to deal with the structural changes phenomenon of volatility, which was not captured by past stochastic volatility models. By applying MCMC and Bayesian analysis, we are able to consistently estimate the FCSV. And the statistical inference is made based on Posterior Odds Ratio. Additionally, in the empirical testing, the FCSV successfully captures the structure reform of Chinese RMB exchange rate at July, 2005.