Research In Models Of Long-memory Volatility And Empirical Analysis

Since half of the century, China’s economy has made a rapid development. Especiallysince China formally became a WTO member in2001, in face of the complex and changingworld economic environment, China’s economy will inevitably suffer from kinds ofchallenges which come from the external environment. As the lifeblood of the economicdevelopment, the development of financial markets has become the focus of researchers andmanagers. Researching on fluctuations of financial assets in the financial sector has been oneof the focus of attention of many economic researchers, especially when the rapiddevelopment of computer technology, greatly facilitate acquisition and storage of a variety offinancial data, making the acquisition and storage of high-frequency or ultra-high frequencydata become a reality. Classic volatility model is mainly based on low-frequency data, such asday, week, and month. Not only the amount of the data is small, but also need longer cycle toobtain sufficient data. Use the financial high-frequency data for research, can obtain sufficientdata in a very short period of time, and shorten the data acquisition cycle greatly, at the sametime, saving the research time. Because the high frequency data contains richer price changesand long-term day phenomenon relative to the low frequency data, also plays a very importantrole in the study of market microstructure theory. Therefore, different from volatility modelbased on low-frequency data in the past, this paper is to study the volatility and its longmemory on the basis of high-frequency data. Since Engle (1951) found this feature of longmemory time series, the financial sector has put widespread concern in the long memory oftime series. Because the long memory means that the state now will continue to affect thefuture, which plays a negligible role in financial risk management. This article studied twovery important long memory time series models: the FDN model and ARFIMA model,followed by the ARFIMA-RV realized volatility model.Since the proposed of realized volatility, Zhang Shiying and other scholars expanded andimproved it, getting the Weighted realized volatility(WRV). Access to a large number ofdocuments founded that few studies have mentioned Weighted realized volatility model.Although many scholars believed that the ARFIMA model has been able to forecast thevolatility very well. But how to achieve an effective combination of Weighted realizedvolatility and ARFIMA, to suggest a line with the characteristics of high frequency data volatility estimation model is a relatively new research direction and difficult, the research ofthis article is expanded in the background of this issue. Through the analysis of the empiricalresults found Weighted realized volatility sequence a long memory, Weighted realizedvolatility has a greater normality, based on this feature and its long memory builded theARFIMA-ARIMA-lnWRV model, determined the values of the parameters，and then test thefitting effect of the model，proved that the model is good，finally introduced the developmentof the VaR model and application of WRV in the VaR.