Semiparametric GARCH Model Compared With GARCH, Nonparametric GARCH Model

With the rapid development of information technology and the globalization tendency of the money market, new financial products continually come forth so that the complexity of financial investment and bank operation increased. And territorial financial storm and banking crisis show that people should strengthen the study of financial steady and control money market. People starved for constructing different theoretical models, and need the demonstration research on the basis of microcosmic datum. One of the important resources of risk of financial system is the fluctuating price of financial assets. Thus constructing model by fluctuating is the core of the financial assets research study content. At present, the thoughts of parametric, nonparametric and semiparametric model are used in finance domain. A lot of demonstration research show that there are significantly volatility clustering and persistence and the tails of distribution of financial datum are heavier than a normal distribution. So, people couldn't use common time series to model the volatility of financial datum. GARCH model is the one of useful models. Parametric GARCH model is in common use but it is sensitive to model misspecification. People proposed nonparametric GARCH model to deal with the problem. But there was a problem of"dimensional disaster"in nonparametric GARCH model, and people were difficultly to explain the model. So, in this paper we propose a semiparametric GARCH model. The semiparametric GARCH model combines parametric model and nonparametric model. It can explain the model and decrease the error between estimation values and true values. The semiparametric GARCH model is one of the semiparametric models, but the approaches in existence are not used in semiparametric GARCH model. In this paper we proposed a approach to semiparametric GARCH model, i.e. two stages iterative algorithm of semiparametric GARCH model.In this paper, GARCH, nonparametric GARCH and semiparametric GARCH are used to fit the volatility of Chinese stock market. The demonstration research show that there are significantly heteroscedastic in return series. The parametric GARCH and semiparametric GARCH models are more accurate than a GARCH model. semiparametric GARCH models is more stronger than nonparametric GARCH model in explanation. And semiparametric GARCH model can settle the problem of"dimensional disaster".