Research Of The Realized Volatility For Securities High Frequency Time Series

In the financial markets, the information is to influence the securities market price. Discrete data acquisition will definitely lead to varying degrees of lack of information, which we have good reason to believe that the higher the frequency of data collection, the less information loss.High-frequency time series is usually refers to the hours, minutes or even seconds for the data collected by frequency. Therefore, in order to more profound understanding of the financial markets, there is a need for high-frequency financial data to study, and as computer and communications technology advances, the high-frequency financial data research has become possible. Financial market risk is the focus which global financial institutions and regulatory authorities pay attention to. With the response, the rate of market volatility is measured accurately measure the value of the core issue of risk. Corresponding to, the market volatility is measured risk and the value of the core problem.Low-frequency fluctuations has two successful method in the time series of measurements field: The first category is ARCH models and the expansion of its form which 2003 Nobel Laureate in Economics owner Engle (1982) introduced; another is SV econometric model and the field of expansion of the form which Taylor (1986) introduced. As ARCH model and SV model in the low-frequency financial time-series modeling has the successful performance of the field, people will naturally think first ARCH models and SV models can direct high-frequency time-series modeling?Andersen and Bollerslev (1997), Wright and Bollerslev (1999) Empirical research shows that on the one hand, ARCH Model and SV class model can not explain the drivers of volatility, can not explain the reasons for the continued volatility; on the other hand their depicts for high Frequency time series is limited. Therefore, we must consider the different characteristics of the high-frequency time series from low-frequency of the time series to established the actual model.The last 10 years, Andersen and other scholars give a method using high-frequency data to estimate the volatility, the method is realized volatility. This method can be more accurate estimate of the volatility rate. "Realized" volatility is calculated by the square of the day returns rate, although the calculation is simple, but significance is profound. As long as the frequency of day yield is large enough, the realized volatility approach the integral volatility.On this basis, many scholars study deeply in both the characteristics and predict of the volatility, greatly expand this research field. We use three kinds frequency of high-frequency of Chinese stock market to calculate the realized volatility, we study the characteristics of the volatility and simulation, based on the realized volatility research at the Shanghai stock market volatility VaR.First of all , in the first chapter we introduce the research background, significance and Research at home and abroad. At present foreign scholars in the realized volatility application have the following studies: Integrated Volatility in Hull-White's (1987) stochastic volatility rate in the option pricing was direct applied ; Andersen and Bollerslev and others (2002) in the framework of Eigenfunction SV used realized volatility fluctuations to forecast integral volatility, and gave the empirical analysis; Andersen and Bollerslev, etc. (2003) research the forecast for realized volatility, and applied it in value at risk (VaR ) calculations. As the domestic securities market is not perfect enough, there are shortcomings in many aspects, on the domestic research in this area also less.in this paper,We use three kinds of high-frequency frequency of Chinese stock market to calculate time realized volatility , do a research for characteristics of the volatility , volatility imulation ,at the same time based on the volatility of the Shanghai Stock Market VaR . Our Research methods combine the theoretical analysis and empirical analysis, combine financial theory and quantitative methods ,using Eviews, Matlab, OxMetrics softwares, and through programming completed study.The second chapter describes the realized volatility basic theory, and select the three-minute, 5 minutes, 10 minutes time of high-frequency data of Shangzheng index and shencheng in 2007 as sample and estimates the Chinese stock market fluctuation use realized volatility, and the results is as same as the Western countries reached, that is, the returns of Chinese stock market and the realized volatility is not normal, showing obvious peak and thick tail, but standardized yield and realized volatility closer to normal, and decreased the peak and thick tail characteristics.In the third chapter ,we study the long memory of the realized volatility of the Chinese stock market, using R/S method to do the long-term memory test, auto-correlation function of fluctuations rate decay slowly along the time lag lengthen. We found that the realized volatility has long-term memory, these results are same with other researchers. Based on these, we use ARFIMA model to model the realized volatility of Shanghai Composite Index. We use MPL to estimate the parameters of the model. In order to make real objective results, we choose six instances ,they are ARFIMA(0,d,1),ARFIMA(0,d,2),ARFIMA(1,d,0),ARFIMA(2,d,0),ARFIMA(1,d,1),ARFIMA(2,d,2),to simulate the realized volatility which the time interval is 3 minutes, 5 minutes, 10minutes, and then by testing the prominence of Simulation model and parameters, we get the model for RV(3) is ARFIMA(1,d,1), for RV(5) is ARFIMA(0,d,1)and for RV(10) is ARFIMA(4,d,2).We use information guidelines AIC to optimize these models, choose the model made the AIC value of the function to a minimum. We found the model for RV(5) ARFIMA(0,d,1) is the optimization. This involved a lot of econometrics, time series and statistical modeling, and other aspects of the theory and method, as well as some computer software programming and use is the difficulty of this study. Finally, realized volatility will be applied to VaR calculation, taking into account returns distribution is not normal, with a clear peak thick tail, we estimated returns distribution parameters on the basis of the three assumptions that is the normal distribution, T students distribution and generalize error distribution, and obtained the VaR under the confidence level of 95%, 97.5% and 99%. Finally, LR test showed that t-distribution assumptions in a variety of circumstances passed the test, the assumption of GED distribution of VaR model based on RV(3) under the significant level 1%, has not yet passed LR test, underestimated the risk, and under other all circumstances, GED passed LR test. Normal at 2.5%, 1% level were not significant by LR test, underestimated the risk. Compared the failure days and expect days, t distribution fitting The Shanghai Composite Index sequence best, followed by the GED distribution, normal. It shows that the peak and thick tail. From the volatility model, the accuracy of the VaR model based on the RV (5) is the best, followed by the RV (10), RV (3).High-frequency data research is a new field, and high-frequency data has been based on realized volatility is one of the important contents of the financial study. Its calculation is simple, and has accurate forecasting without model restrictions, realized volatility has become a popular research tools. At present, the applications of realized volatility are only in the stock, the exchange rate or options, and many of them have directly aimed at one of the three laws of empirical framework to construct its analysis. Therefore, future studies should expand the scope of targets, attention to other financial assets and derivative products in the study. And future studies should research deeply the realized volatility for modeling, optimal frequency of sampling, and error of micro-structure.