Research On Realized Volatility Forecasting Of CSI 300ETF And Its Application In Option Pricing

With the rapid development of financial markets,relative to the familiar stocks,funds and other types of investment products,futures,options and other types of derivatives have slowly come into the public eye,and the concern about the ups and downs in the financial markets has increased significantly.The volatility of asset returns is an important indicator of risk,and accurate prediction of volatility plays a crucial role in portfolio allocation,risk management and derivatives pricing,and the study of capital market volatility has become an important direction for people to prevent financial risks.Therefore,this paper conducts modeling and forecasting research on the realized volatility of CSI 300 ETF,and applies the results of each model prediction to option pricing,which is of great practical significance to promote the healthy development of financial markets.In recent years,scholars have mostly used 5-minute high-frequency data to study volatility because 5-minute high-frequency data has less microstructural noise compared with higher-frequency data and has more intra-day information than low-frequency data.Therefore,this paper selects 5-minute high-frequency data of CSI 300 ETF from May29,2012 to May 31,2022 as the research sample and calculates the daily Therefore,this paper selects the 5-minute high-frequency data of the CSI 300 ETF from May 29,2012 to May 31,2022 as the research sample and calculates the daily,weekly and monthly realized volatility,jump volatility,positive and negative realized semi-variance and sign jump variance,which are variables obtained from the historical data,and the descriptive statistics of the variables show that the volatility of the CSI 300 ETF has obvious long memory characteristics,and combined with the heterogeneous characteristics of the financial market,this paper selects the HAR model for modeling research,and there are studies showing that macroeconomic policies will have an impact on the stock market Therefore,this paper constructs 12 volatility forecasting models based on the classical HAR family models,combining structural mutation and economic policy uncertainty index,and uses these 12 HAR models to estimate the sample of realized volatility of CSI 300 ETF,so as to explore the effect of each explanatory variable on volatility forecasting Then,the out-of-sample volatility forecasts are conducted using a rolling time window approach,and the forecasting accuracy of each model is compared using a loss function-based MCS test.Finally,the paper also applies the prediction results of the constructed HAR class model to the option pricing of CSI 300 ETF options to explore the practical application value of the HAR class model.The empirical results show that: first,from the sample estimation results of each model,the significance of jump volatility,sign jump variance,positive realized semi-variance,economic policy uncertainty index and structural mutation coefficients is better,which indicates that they have stronger forecasting ability for CSI 300 ETF volatility,while the significance of negative realized semi-variance coefficients is worse,which indicates that negative realized semi-variance has weaker forecasting ability for CSI 300 ETF volatility.The coefficients of negative realized semivariance are less significant,indicating that negative realized semivariance has weaker predictive power for the volatility of CSI 300 ETF.The inclusion of either structural mutation or economic policy uncertainty index in the HAR model leads to better model fit,indicating that the inclusion of these two variables in the HAR model is effective for volatility prediction.Second,this paper performs MCS tests on the predicted volatility based on three loss functions,MAE,MSE and MAPE,and finds that the sign jump variance improves the prediction accuracy of the volatility model most significantly,and the addition of structural mutation also improves the prediction accuracy of the volatility model,and the prediction accuracy of the model considering the economic policy uncertainty index is higher than that of the model not considering the economic policy uncertainty index.The HAR-RV-SB-SJV-CEPU model with simultaneous consideration of sign jump variance,structural mutation and economic policy uncertainty index has the best prediction accuracy.Third,after constructing the pricing models under the generalized geometric Brownian motion framework and using the Monte Carlo simulation method for option pricing,a comparison of the option prices calculated by each model using two loss functions,MAE and MSE,shows that the option pricing effect of the HAR-type model is significantly better than that of the B-S option pricing model,and the HAR considering the economic policy uncertainty index and the HAR considering the structural mutation class models both improve the option pricing effect,and the HAR-RV-SB-SJV-CEPU model with the best volatility forecasting ability exhibits the best pricing effect in option pricing.