The Application Study of Financial Risk Measurement Based on GARCH-Copula Model
|School||Tianjin University of Finance and Economics|
|Keywords||Copula Function GARCH Model portfolio VaR GED Distribution|
Risk Measurement of finance has always been an important part of the modern financial risk management. And as to the financial risk control, it is very important to quantify the risk of finance assets. Usually, the data of finance assets with spikes, thick tail and volatility clustering features, its modeling to quantify risk assets should be based on asset data characteristics selected an appropriate model and distribution assumptions. At present, the method of VaR is the most important method to quantify the value of finance assets or asset portfolio risk, the model used mostly are GARCH model, copula functions, and a combination of both, in different distributions assumed to get VaR values. Especially in the distribution assumption, scholars has already been conducting in-depth study on the basis of the fat-tail t-distribution, the GED distribution and extreme value distribution, to better fit the data and build the model. In the portfolio, each asset data may have different characteristics, and therefore its need to choose better distributions to fit, and ultimately by the copula functions to get the joint distribution function. It can be getting the GARCH-Copula Model, which has fully considered the characteristics individual assets, and then get the value at risk VaR will be more effective.Considering the characteristic of each assets, this paper have used the t distribution and GED distribution to fit the data and construct the GARCH-Copula model to get the VaR of the portfolios, at the assumption of mixed distribution of t distribution and GED distribution. In contrast, the paper using the GARCH-Copula model based on the t-distribution and GED distribution respectively to measure the VaR of the portfolio.Also we have used the GARCH-VaR model and other simply method to measure the portfolio’s VaR.Under the condition that we have already fully considered the individual data’s characteristics, choosing the suitable distribution an d GARCH-Copula model, to describe the volatility clustering of data, construct a robust model and to estimate the VaR of the portfolio.