Research On Chinese Stock Market Volatility Forecasting And Option Pricing Based On Realized Volatility Model

At present,the financial market develops rapidly.In the research of modern financial theory and practice,the rise and fall of the stock market has been the focus of attention of enterprises and individuals,it becomes very necessary to measure the volatility of the market.The volatility of asset returns is an important factor in measuring and managing financial risks,and it is also closely related to the pricing of derivatives,the best choice of investment portfolio and the hedging of assets.In the past,domestic and foreign scholars and financial institutions have fully studied the volatility model,and most of them used the daily rate of return to analyze the volatility,but a lot of intrad-day trading information would be lost in this way.With the application and development of information technology,the fitting of volatility forecasting accuracy requirements gradually improve,using intra-day high frequency data,calculate the estimated realized volatility can approximate to the actual,the researchers also gradually realized volatility model as a measure of volatility of common tools,further promoting the research direction of updating the volatility model.In this dissertation,we select the high-frequency data of SSE 50 ETF every 5 minutes from December 4,2017 to December 2,2020 as the sample,and calculates the data series of three different frequencies,namely daily realized volatility,weekly realized volatility and monthly realized volatility.Through the test,we find that the realized volatility sequence has the characteristics of lognormality and long memory.Based on the Heterogeneity Market Hypothesis,three kinds of realized volatility sequences are used to construct the realized volatility model(HAR-RV model)and carry out the fitting prediction research on the realized volatility of the SSE 50 ETF.Adding new related explanatory variables into the volatility model has always been an important direction of model expansion.Considering the impact of market liquidity on stock market volatility,in this dissertation,we added liquidity indicators--turnover rate and Amihud indicators--into the model as new explanatory variables,and constructed the HAR-RV-A model and the HAR-RV-Tmodel.Through testing,it is found that the residual of the model has heterosceasticity.In order to better improve the prediction effect of the volatility model,GARCH model is introduced to eliminate heterosceasticity of the volatility model,and the HAR-RV-T-GARCH model is constructed.Through in-sample fitting and out-of-sample testing.Then two kinds of loss functions are selected to compare and analyze the prediction results of the above extended model.The results of in-sample fitting and out-of-sample testing both show that the four realized volatility models constructed have good prediction effect.Among them,the HARRV-T model and HAR-RV-A model with liquidity index added have better prediction effect than the traditional HAR-RV model,and the HAR-RV-T-GARCH model with residual variance elimination has the best prediction effect.When investors trade options,they are most concerned about the issue of option pricing,because whether the pricing is accurate or not will affect the investors' subsequent trading strategy construction and portfolio hedging.The volatility of underlying financial assets is a key parameter of option pricing,so it is very important to correctly estimate and predict the volatility of underlying financial assets.In this dissertation,the realized volatility expansion model constructed is applied to the option pricing of SSE 50 ETF.Through empirical analysis,it is found that the realized volatility model can better capture the volatility characteristics of the underlying asset sample,and the option pricing effect of the realized volatility model is obviously better than that of the B-S model.The extended realized volatility model has better volatility fitting effect and better pricing effect on SSE 50 ETF options,which proves that the extended realized volatility model constructed in this dissertation has good practicability.