| With the development of computing technology and information transmission speedsincrease, and the importance of speculative, more and more attention has taken intohigh-frequency data, scholars have begun to focus on high-frequency data on the stockmarket.Discussion of the extreme financial risk management for financial regulators and financialinstitutions, will improve the timeliness to prevent extreme events and enhance the accuracy ofthe measure of risk based on high frequency data.Based on November2011by-second data of theShanghai Composite Index and Shenzhen Composite Index, we consdierded high-frequency dataextremal analysis of China’s securities market, with the method of application of financial timeseries we discussed the extreme value of the high-frequency data’s equilibrium, the relationshipbetween volatility characteristics, distribution characteristics, and risk metrics of China’ssecurities market.Try to discuss the extreme value of high frequency data t is a new attempt ofresearch aboute microscopic results of and China’s securities market.In this paper we mianly discussed：（1）Statistical characteristics,the distribution ofcharacteristics, patterns of variation of various frequency data of intraday’s stock market（.2）withError correction model and GARCH model to reserch high-frequency data of China’s securities inthe equilibrium relationship and its volatility,then compared variation adjustment coefficient ofvarious frequency extreme value with the closing price, compared GARCH (1,1) of Shenzhencomposite index’s high-frequency extreme value with GARCH-M, observed spillover effects andthe interaction of extremum between Shanghai Composite Composite Index and ShenzhenComposite Index.（3） In the purpuse of satisfactory fitting distrbution of seconds ofhigh-frequency data, we use generalized extreme value distribution model to discuss intradayvarious frequency extreme data of Shanghai Composite Index, to find the distribution parameterswith the changes in the characteristics with different frequencies of the nature of the distributionof high-frequency data,at the end of discussion we computed VaR;at the same time,we made a riskmeasurement research of high-frequency data in the local extreme market conditions of China’sstock market by POT model and the Hill model.The thesis is divided into four parts:Chapter1： Introduction. This chapter discusses the background and significance of extremeevents in financial markets using high-frequency data as the breakthrough point, reviews andsummarizes the scope and methods of high-frequency data in the financial sector, development and acquisition of high frequency data in volatility, the results and current achievements ofhigh-frequency data extrema history.Chapter2:Typical characteristics of the Chinese stock market’s high-frequency data.With theaccumulation of deep reaserching of the high-frequency data, reaserchers has discovered sometypical characteristics of high-frequency financial data, this article summarizes the typicalcharacteristics of the high-frequency financial data, taking the Shanghai Composite Index data asan example to study different frequencies of high-frequency returns as statistical study, then findsthat the frequencies of returns less than day are stationary series, kurtosis is greater than3,skewness is less than0, shows a significant peak left tail "phenomenon,these characteristicsprepares the way for the following different models of various high-frequency returns.Chapter3: the associated characteristics of extreme value of high-frequency in the ShanghaiComposite and Shenzhen Composite Index.First this chapter discusses the equilibriumrelationship of various frequency of extreme value and the closing price in the ShanghaiComposite, details the short and long term equilibrium relationship of15seconds maximum,minimum and the closing price of the Shanghai Composite, Similarly other frequencies ofextreme value and the closing price equilibrium relationship then found that adjustment factor ofdifferent frequencies of extreme value to the closing price obeys the periodic motion, this canshows that the choice of the frequency will affect the investment decisions of investors, abalanced relationship of the maximum and minimum values reached the closing price isasymmetry, this verifies that China’s stock market presentes “falling speed is greater than risingspeed "phenomenon. Then this chapter reaserches the volatility of extremum data of the ShanghaiComposite and Shenzhen Composite Index by the GARCH model and GARCH (1,1),finding thatGARCH (1,1) is better to portray the volatility of the market extremum. Finally, this chapterreaserches "spillover effect" of the Shanghai Composite and Shenzhen Composite Index in thesense of second-order moment， found that Shenzhen Stock Exchange reflects rapidly whentheres is a change on Shanghai Stock Exchange and the impact is far-reaching,but Shanghai StockExchange reflects hysteretic when theres is a change on Shenzhen Stock Exchange.Chapter4, the risk reaserch on extreme value theory. In this chapter three methods ofextreme value theory are used: generalized extreme value distribution, the POT model and theHill estimator depicts the characteristics of distribution of different frequency of the extremereturn in the Shanghai Composite, making the risk measure of high-frequency data under extrememarket conditions. The results showed that the VaR measure under the generalized extreme valuedistribution is similar to the Hill estimatorAfter molding each frequency of extreme sequence of generalized extreme value distributionin the Shanghai Composite, finding the characteristics of the distribution parameters changes with the different frequencies, observing the best fitting of the distribution of of high-frequency data inseconds, surveing VaR of the different frequencies observed under the generalized extreme valuedistribution.Obtained sub-block distribution parameters by maximum likelihood estimation andfinding that the scale parameter and location parameter are gradually increased, the shapeparameter is around0.2and there is a little change of the shape parameter confidence interval.Further analysis of the distribution of residuals finding that5minutes and10minutes blocks ofgeneralized extreme value distribution fits better. With the possiblity of breaking the new recordswe found that reference criteria for the frequency of transactions in the stock trading process canbe considered30minutes or more. Finally we calculatesVaR0.99andVaR0.95of the generalizedextreme value distribution,finding that with the with the block sizes increasethe value at risk hasalso increased; analyzid15seconds of high-frequency closing price of the Shanghai CompositeIndex by POT model, gains threshold is0.03, in over-the mean figure, estimated the modelparameters by Generalized Pareto distribution and found that the shape parameter is very close to10minutes of the shape parameter of generalized extreme value distribution, GPD diagnosticdiagram illustrates the fit is better, finally calculated VaR and ES of15seconds of the ShanghaiComposite Index,and compared with VaR and ES of normal distribution, the ES were, resultsshowed normal distribution underestimated the risk; threshold is0.03of15seconds ofhigh-frequency return of the Shanghai Composite Index in the Hill estimator,99%quantile is near0.079by Hill estimate which is very close to10minutes of VaR of GEV distribution,95%quantile is maintained at0.0453by Hill estimator. |
