Font Size: a A A

The Construction Of Heterogenous Autoregressive Model With Time-Varying Parameters And Research On Risk Predition

Posted on:2020-10-22Degree:DoctorType:Dissertation
Country:ChinaCandidate:G Q LiuFull Text:PDF
GTID:1360330599975620Subject:Business management
Abstract/Summary:PDF Full Text Request
The volatility is the basis of further research on investment portfolio and risk management.How to improve volatility prediction in financial markets is a topic widely discussed by scholars and industrialists at home and abroad.Compared with low-frequency data such as daily data,intraday high-frequency trading data contains more market information and reflects the real volatility of asset prices to a greater extent.With the development of computer technology,the ability of storing and using high-frequency trading data in financial market is improving day by day,and more high-frequency trading information can be used in research and practice.And this provides convenience for research on volatility using high frequency data.In recent years,Realized Volatility(RV)is a representative theory of how to make full use of high-frequency trading information of financial assets in order to fully reveal the impact of intraday trading information,which avoids the shortage of traditional GARCH model and SV model in the use of information,and has the advantages of simple calculation and unbiased estimation.Under this background,the theises studies the Chinese stock market with the help of the theory of realized volatility from the perspective of high frequency data.through fully excavate the "jump","asymmetric" volatility and other intraday trading information to fit the realized volatility in Chinese stock market,and bring the time-varying characteristics of parameter estimation into the research scope.Thus,a novel prediction model of realized volatility with time-varying characteristics is constructed.On the other hand,the application of volatility in risk forecasting is further discussed,and the empirical analysis of risk prediction in Chinese stock market is carried out with the Extreme Value Theory.The Hang Seng Index of Hong Kong and the Shanghai and Shenzhen 300 Index are taken as the research samples to represent the mature market and the emerging market respectively.In order to analyze the stability and universality of the research results,this theisis focuses on the 5-minute trading data of Shanghai and Shenzhen stock markets and Hong Kong stock markets from January 1,2005 to December 31,2017.This sample range includes different stock price changes such as sharp rise,decline and slow change of stock price,and the conclusions obtained are more robust.The main contents of this theisis are as follows:Firstly,a HAR-CSJ model with take the "jump" and "asymmetric" volatility into account is constructed.As an emerging market,Chinese stock market is characterized by more frequent and violent fluctuations,and a strict statistical method is used to identify and measure the "jump" of the stock market in our country.Based on the semi-variance theory,the HAR-AJ,HAR-SJ,HAR-CSJ model with more intra-day trading information is constructed by taking the "asymmetric" volatility characteristics into account.At the same time,The HAR model constructed in this theisis and the existing classical HAR family model,that is,the basic HAR model,the HAR-C model of continuous component,the HAR-S model that models the decomposition of RV into positive and negative parts,and the HAR-SA model based on the total decomposition of day,week and month RV is compared and analyzed.Therefore,the traditional models discussed in this theisis have eight models,such as HAR,HAR-J,HAR-C,HAR-S,HAR-SA,HAR-AJ,HAR-SJ,HAR-CSJ and so on.The empirical results show that this theisis fully mining "jump","asymmetric" volatility information constructed by the HAR-CSJ model has an important advantage over other traditional models.In the process of RV modeling and analysis of Chinese stock market,the HAR-CSJ model of jump symbol is better than the HAR,HAR-J,HAR-C,HAR-S,HAR-SA,HAR-AJ,HAR-SJ model compared with this theisis.This shows that the Jump composition and symbol jump after the test is the most important characterization of RV.Secondly,a novel HARQ family model is constructed to relax the fixed parameter constraints in the traditional HAR family model.The fixed parameter constraints in the traditional HAR model are relaxed,and the time-varying characteristic of the parameter estimation variance is given.The traditional high-frequency volatility model is improved by introducing the RV fitting method of the parameter time-varying characteristic.A novel HARQ theory is used to upgrade the HAR model,which is based on the abundant intraday trading information,such as "jump" and "asymmetric" volatility,and the existing classical HAR family model.The HARQ family model with time-varying characteristics is constructed.In this theisis,eight models of HAR,HAR-J,HAR-C,HAR-S,HAR-SA,HAR-AJ,HAR-SJ,HAR-CSJ are upgraded to HARQ,H-SQ,H-SAQ,H-CQ,H-JQ,H-AJQ,H-SJQ and H-CSJQ models with time-varying parameters.The empirical results show that the time-varying parameters of both the Shanghai and Shenzhen stock market and the more mature Hong Kong stock market are significant,and the fitting effect of HARQ family model is better than that of HAR family model.At the same time,the logarithmic form of Log(RQ)has the best effect on the characterization of time-varying parameters,and the H-CSJQ model is the best one among the various models compared.Finally,the dynamic VaR prediction model of Chinese stock market is constructed based on the HARQ method and Extreme Value Theory(EVT).On the basis of HARQ model which has realized volatility fitting to high frequency data,combining with the EVT which can depict extreme volatility,this theisis constructs HARQ-EVT-VaR model,and carries out empirical research on VaR in Shanghai and Shenzhen stock market and Hong Kong stock market.The Conditional Coverage test(CC)method with robust analytical ability is applied to analyze the accuracy of the VaR predicted by each model,so as to judge the advantages and disadvantages of each model.The empirical results show that,on the basis of HARQ model,the EVT,which describes the extreme volatility,has a robust ability to predict the dynamic risk of the Chinese stock market.Among them,HARQ-EVT-VaR model is the best model in the comparative analysis of VaR prediction in Chinese stock market.This theisis discusses the realized volatility of Chinese stock market from the perspective of high frequency data,and analyzes it in detail from the theoretical and empirical perspectives.First of all,excavating the "jump" and "asymmetric" fluctuation information of the Chinese stock market to construct the HAR family model and compare it with the classical HAR family model.Secondly,the new HARQ family model is further used to upgrade it to relax the constraint of fixed parameters in the traditional model,so that it has time-varying parameter characteristics and improves the fitting effect of the realized volatility.Finally,based on the HARQ model and the EVT,the HARQ-EVT-VaR model is constructed and the dynamic risk of the Chinese stock market is predicted.In this theisis,a series of useful conclusions have been obtained,and the relevant research methods and conclusions have important reference value for investors and relevant management departments participating in the Chinese stock market.
Keywords/Search Tags:Realized Volatility, HAR, HARQ, Extreme Value Theory, Value at Risk
PDF Full Text Request
Related items