Modeling And Forecasting Realized Volatility Based On HAR-RV Model

How to measure and model volatility is an important issue in finance. As an important characteristic,The volatility of financial assets is the core of asset allocation,asset pricing and risk management.With the rapid development of science and technology,all kinds of financial products trading more frequently,and the frequent intraday data contains more and more information.Compared to the other volatility,such as implied volatility and stochastic volatility,realized volatility has many better characteristics.It contains powerful information which comes of high-frequency intraday data and only requires a small amount of calculation.Some research shows that realized volatility has some characters such as fat tail、skewed and autocorrelation.Now there have some reasons to explain these characters.One of them comes from the heterogeneous market hypothesis,it insists that in financial market,there are many heterogeneous traders.Basis this theory,a new model is raised.This model is called Heterogeneous Autoregressive model of the Realized Volatility (HAR-RV).This model suggests that because of the different actions of heterogeneous traders,there generates different volatility components.This model is a one lag auto regression that to the different time volatility.Although this model is simple,it can well describes the empirical features of realized volatility,such as long memory、autocorrelation and fat tail.This paper chooses HAR-RV model in modeling and forecasting the realized volatility.Many methods have been proposed for modeling and forecasting financial market volatility,and we compare our HAR-RV models to GARCH models which generated by the most widespread procedure in academic applications.So compared with GARCH models,it can improve the credibility of the results of empirical studies which set up based on the HAR-RV model.Through empirical research in the paper,we find that our simple HAR-RV model produce superior performance.Both in Rsquared and in AIC,HAR-RV model is better than GARCH models.In the forecast results, particularly in dynamic prediction,it is obvious that HAR-RV model is better than GARCH models.The paper proves that the HAR-RV model have the advantages of practicality, simplicity in forms and of excellence in Description and prediction.