Font Size: a A A

Frequency Financial Time Series Volatility

Posted on:2010-05-27Degree:MasterType:Thesis
Country:ChinaCandidate:J ZhangFull Text:PDF
GTID:2199360275483127Subject:Operational Research and Cybernetics
Abstract/Summary:PDF Full Text Request
High-frequency financial data has large sample size and short sampling cycle. It contains abundant information of market. It is the best representation of the financial market. In order to research deeply on the microstructure of market, analyzing and modeling the volatility of high-frequency financial time series become the hotspots of domestic and foreign econometric researchers. In 1998,Andersen & Bollerslev proposed a new method of estimation, the Realized Volatility method. Another method—Realized Range-based Volatility was brought forward by Christensen and Podolskij in 2005. Then the domestic scholars put forward effective methods by extending the two methods.In this paper, we analyze the statistical characteristics of high-frequency data in the Shanghai Stock Market, which is sampled with the frequency of 1 minute. The result verifies that the typical characteristics of high-frequency time series are reflected in China's stock market. For example, high kurtosis, heavy tail, non-normal distribution, congregated volatility, long memory process and intraday"U"-shaped trend.Based on the research results, a new estimating method, Adjusted Weighted Realized Volatility, is proposed as the improvement of Weighted Realized Volatility and proven to be unbiased and more efficient. According to these estimating methods, we work out the estimation values of Integrated Volatility and prove the effectiveness of the Adjusted Weighted Realized Volatility. Then logarithmic Weighted Realized Range-based Volatility is chosen to fit data owing to its best statistic property. Afterwards the AR (i)-FIGARCH (p,d,q) model is established because the return rate series has the characteristics of stable and long memory. The parameters of the model are estimated by the method of Polymerization sequence analysis and maximum likelihood estimation.Finally, the comparison between logarithmic Weighted Realized Range-based Volatility and its fitted value shows that the model has favorable fitting result. Then we calculate the return rate of the sample. By comparing it with the predictive value, the AR(i)-FIGARCH(p,d,q) model is proved to be valuable in estimating and forecasting volatility.
Keywords/Search Tags:high-frequency financial time series, volatility, Adjusted Weighted Realized Volatility, AR(i)-FIGARCH(p,d,q) model
PDF Full Text Request
Related items