The Empirical Research Of Chinese Securities Market Based On The Family Of GARCH Model

Chinese securities market has past over twenty years, since the Shanghai Stock Exchange was officially inaugurated.However, no matter in the regulation or in the institution are not standardized perfectly. This makes the securities market is not very stable,with stong volatility. Unstable securities stock market, unscientific investment and some vicious financial events which occured occasionally makes the securities market full of high-risk, So the investment institutions and individuals are facing serious challenges. In this case, for scholars and investors, how to accurately describe the securities market price and identify market future yield are concerned about, Therefore, The study of volatility are particlarly important.This paper highlights the important value of VGARCH model and non-parametric GARCH model on the study of volatility. VGARCH model is based on GARCH model on which adds trading volume. It can eliminate some extreme and unreasonable values, reduce the expectations of volatility effectively, control the risk in a smaller range and make the forecast of volatility more stable and more accurate, it extends the class of GARCH modern, So VGARCH model has a good reference value in financial data analysis. Nonparametric GARCH model is under the condition that we don’t know the distribution function of error interference items, using local regression smooth technology and R software simulation to estimate the distribution function of error interference items, then GARCH. It has great flexibility and provides an important theoretical basis in the analysis of volatility. This paper put the yield of the shanghai real estate index and shenzheng composite index as sample, use GARCH models to fit and forecast the volatility, and compare the superiority between parameter method and nonparametric method, the following results:(1) the sequence of yield of the shanghai real estate index and shenzheng composite index are not normal, the volatilities are smooth, and they all past the ARCH-LM test, Therefore we can use GARCH model to fit them.(2) This paper adopts four prediction error metrics, for the mean square error, mean absolute error, gaussian quasi maximun likelihood loss error and logarithmic loss function error, the smaller these values,the closer the predicted values and the real values. In the comparative analysis of the two groups of data, we draw a conclusion that in five GARCH models, GARCH(1,1) is good for the forecast of the volatility of shanghai real estate index returns; VGARCH(1,1) is good for the forecast of the volatility of shenzheng composite index returns.(3) Comparing parameter method and nonparametric method for the two groups of data, we have a conclusion that nonparametric GARCH(1,1) model for the forecast of volatility of shanghai.real estate index yield is better than the GARCH(1,1) model, the prediction error isall. So nonparametric GARCH(1,1) model can suitably describe the characteristics of the volatility of shanghai real estate index yield; VGARCH(1,1) model for the forecast of volatility of shenzheng composite index yield is better than the nonparametric GARCH(1.1) model, the prediction error is small. So VGARCH(1,1) model can becomingly depict the features of the volatility of shenzheng composite estate index yield.This article mainly has two characteristics:(1) After GARACH model’s coming, more and more scholars extensively research on it.Now the model is on the development and, the parameter estimation method is been improved. A large number of empirical studies that GARCH model has been widely used, however, scholars either using the class of GARCH to model the data, or comparing with parameter estimation methods. On the basis of the predecessors, We comparative study of the suitable model in GARCH models and nonparametric GARCH model, then we get a model which greatly combining model with method.(2) This paper selects two representative index as sample, shanghai real estate index and shenzheng composite index, we get the above conclusion, So it states that we can not determine fitness only from the model or the method, we should choice model and method to analyze the data according to different purposes and different methods.