On the Realized β Coefficient Long Memory Testing and Prediction Modeling

This paper estimates one day β coefficient of shares in shanghai stock exchange, tests realized β coefficient long memory characteristics and makes a model to predict further with long memory stochastic volatility model, based on a wealth of highfrequency transaction data which have more information, and using the new calculated application of time-varying coefficient β. Shanghai Stock index shows that both realized covariance and the realized covariance between individual stock and shanghai stock index have significant similar long memory characteristics, but the β coefficient of individual stock does not have significant similar long memory characteristics generally. The predicting results of sub-scale and sub-sector show that ARMA (2,1)model has the advantages in forecasting with model; prediciting accuracy will decrease when the group (or scale)grows; but in general for conservative industries, predicting accuracy is better than some high-tech industry.