Portfolio VaR Modelling using EVT-Pair-Copulas Approach

The purpose of this research is to model Value-at-Risk (VaR) of portfolio with EVT-Pair-Copulas approach. In the financial literature, a significant amount of empirical studies have been done on the characteristics of financial assets returns and researchers have found a set of stylized facts about this subject. In this regard leptokurtic, left-skewed, weak autocorrelation, volatility clustering, and heteroscedasticity can be mentioned. Any estimation of risk without considering these characteristics or using unrealistic assumptions about financial assets returns increases the probability of failure in the risk management process. For this purpose, at first, the marginal distributions of returns are obtained using extreme value theory (EVT). Concerning characteristics of financial assets returns and also the primary filter to apply EVT, we use heteroscedasticity models for the marginal distributions of assets. Then the structure of the dependence between different stocks is estimated by using C-Vine, D-Vine, and R-Vine pair copula models. Afterward, the VaR of portfolio is estimated using the Monte Carlo simulation method. The final results show that the model with GARCH marginal distribution and R-Vine pair copula has been able to achieve the best performance among rival models at 95% confidence level.