A Comparison Study On Forecasting Model Of System Risk Of Chinese Stock Market

Capital Asset Pricing Model(CAPM) was put forward by WilliamSharp, John Lintner and Jan Mossin respectively at 1964, 1965 and 1966according to Markowitz's choice thoughts of superior asset combination.It is an asset pricing model that maximizes the investors' effects inunfixed conditions. It not only theoretically resolves the problem of pricebalance in the condition of unfixed future returns and is called one of thethree basic theory together with time-value of money and venturemanagement, but also practically guides people's investment activitiesTherefore, capital asset pricing model has have been focusing much bythe investors from the beginning and has been well developed.As the basis of the price theory of financial market modernization,CAPM has been widely used in various fields such as the pricing ofsecurity fund and bond and investment decision making. The system riskcoefficientβin CAPM is used to measure the sensitivity of the excess profits of different assets and their portfolio to the market portfolio.Fixing the coefficientβcorrectly is of great importance.Coefficientβis an important index to reveal the system risk thelisted company faced, and it is also the necessary method that needed bthe portfolio management. How to estimate and forecast the Coefficientβof the stocks is important and of significant theoretical value.This paper gives three new mentalities to estimate the coefficientsβ:the CAPM based on the MRS-GARCH process, the one based on thestate space model and the one based on the nonparametric model.Utilizing the data of Chinese stock market from year 2001 to 2005, Weset up three forecasting models of the time-varyingβand then test them,the main conclusion are as follows: first, during the sample period theestimated value ofβshows no significant difference each year and mostof them tend to equal 1, but the derivation of which expands year by year.Second, the CU SUM SQ parameter stability test on the 21 industriesshows that, except for three industries, the coefficientsβof eighteenindustries are unstable, thus confirmed that the Chinese stock marketexists the time-varying coefficientβand therefore the variable parametermarket model should be introduced. Third, analyzing the mean absoluteerror (MAE) and the mean square predict error (MSE) of industry stockforecasted revenue, the results shows that there're many differencesbetween the three methods. Among which, the nonparametric one has the smallest error and the MRS-GARCH method has the largest one.According to the estimation result of the present market data, the abovethree methods are all better than the historical method.