A Research Of Realized Volatility On Hushen300Index Based On The Detection Of Structural Breaks

This article describes the theory of realized volatility obtained by the sum of the square of high-frequency yield during a day, such volatility used not to be directly observed becomes observable and is treated as a consistent estimate of the real volatility. Two new estimators are introduced to replace the bipower variation to seperate continuous volatility and jump volatility more effective according to asymptotic theory. Based on the separation work of continuous sample path variance and discrete jump variance, heterogeneous autoregressive (HAR) model is used here for volatility modeling and forecasting. Counting for the severe fluctuations in the Chinese stock market which may easily be easily caused by external factors because of its immaturity, a revised iteration cumulative sum of squares algorithm (ICSS) is also introduced to calculate the structural break point in the volatility sequence. According to these structural break points, mutations is divided into sub-samples and modeled to consider its predictive ability of the model simulation.The empirical results show that, realized volatility in its standard form is showing a distribution skewed to the right and a fat tail, and its logarithmic form has been very close to the standard normal distribution. In addition to the sequence of correlation, the volatility also seems to be strong clustering. Both the jump component and non-jump component sequences are autocorrelated, but autocorrelation of the jump component is weaker than non-jumping component. The HAR modeling regression suggests that model with jump and non-jump components concluded as the explanatory variable shows better predictive ability one day ahead than others, while the sub-sample regression shows the same result. Another point should pay attention to is that the ability of explanatory of jump component of the model is not significant, so model with only non-jump component may be better, and it is found that there are economies of scale in which the sequence of the volatility and leverage effect.