Research On VaR Measure Of Portfolio Using High Frequency Data

Financial risk has far-reaching influence and long duration.The risk measurement of financial market has attracted much attention.On one hand,with the development of economic globalization and financial integration,the dependence of regions and markets are increasing.On the other hand,investors usually invest their assets in two or more projects in order to reduce risk.Therefore,the financial risk measurement for portfolio will be of great importance.Value-at-risk(VaR)is one of the traditional portfolio risk measure methods.The process is as follows.First,the GARCH class volatility model is build using the low frequency data.Then,Copula function is utilized to capture the dependence among the assets.Finally,the VaR of the portfolio is estimated.However,the low frequency data will lose a lot of information in frequent trading.High frequency model such as the Realized GARCH model and HAR model capture more information in intraday data,both of which have been extensively investigated.However,the employment of high frequency model in portfolio remains to be explored.In this paper,the Shanghai Composite Index and the Hang Seng Index are selected as examples to demonstrate the risk prediction capability of our model.The Realized GARCH and HAR-RV models were build base on the high frequency data.Copula function was utilized to predict Portfolio risk.In this paper,two high-frequency data modeling methods were compared.Furthermore,the high frequency models were also compared with the low frequency model.The results demonstrated that the Realized GARCH model has better precision in portfolio risk prediction.The arrangement of this work is as follows.First,we build the volatility models of the Shanghai Composite Index and the Hang Seng Index.We utilize five closing prices of the two indexes to calculate the rate of return and the realized volatility(RV)of the them.We utilize the daily closing price to calculate the rate of return.The results indicate that the Shanghai composite index shows an obvious volatility from the end of 2015 to the middle of 2016,which is a result of stock market crash.The edge distribution model of Realized GARCH and HAR-RV were established respectively.The GARCH model was also build using low frequency data as a control group.A new sequence with independent and identically distributed was obtained.Furthermore,the descriptive statistical analysis was carried out.Second,five Copula functions were used to capture the dependencies between the two markets.The result shows that the SJC Copula function is the best,indicating that the Shanghai Composite Index and the Hang Seng index have an asymmetric tail dependent structure.Further,the time-varying SJC Copula model was built to capture the dynamic changes of the dependence.Third,the dynamic VaR risk measurement of portfolio was carried out.The results in different models were compared.The VaR measure of portfolio with equal weights for Shanghai Composite Index and Hang Seng Index was carried out.Then,by utilizing the Monte Carlo simulation method with rolling window,the dynamic risk value VaR of the two indexes under the significant level of 1%,5% and 10% was obtained.After Backtracking the VaR results,we found that the SJC Copula-Realized GARCH model is the best one among the SJC Copula-GARCH model and the SJC Copula-HAR model at each significant level.The prediction of SJC Copula-GARCH model is the worst.Nowadays,the uncertainty in financial markets is increasing.In this paper,we utilized both low-frequency data and high-frequency data for the volatility research,and finally utilized the results for risk estimation of the Shanghai Composite Index and the Hang Seng Index.The study is of great significance.First,our work contributes to the academic research.By comprising the results of GARCH,Realized GARCH and HAR models,we build an optimal model for VaR of portfolios.(2)This work provides information about market risk for investors.The Shanghai Composite Index and the Hang Seng Index of the Chinese market were used as examples to demonstrate the capability of risk forecast.Theoretically,the methodology employed in this paper can also be extended to other markets.(3)This work also provides guidance for financial regulators on market risk,which has an important role in the sound development of the capital market.This article can be further improved in the following three aspects:Firstly,the forecasting mode can be extended.RV is used as an example to demonstrate the capability of risk forecasting.Other methods for Realized measures,such as the WRV(Weighted Realization)or RK(Nuclear Volatility)also can be used.Secondly,the mode for the high-frequency data volatility can also be extended.In this work,the Realized GARCH and Heterogeneous Autoregressive HAR-RV models were used.Other types of models such as various distributed Realized GARCH models,extended HAR-RV-J,HAR-RV-CJ can be used.Thirdly,the Copula model can be extended.The objects of this paper are the Shanghai Composite Index and the Hang Seng Index,so we use two-dimensional Copula function to capture the correlation between assets.When measuring risk for three or more portfolios,we can consider using vine Copula function to characterize dependency relationship.