Extreme risk modeling: An EVT–pair-copulas approach for financial stress tests

This paper presents a semi-parametric copula-GARCH risk model for financial return series with a stress testing perspective. The marginal distributions of the returns are specified using the Extreme Value Theory (EVT), putting a specific emphasis on extreme returns. The joint distribution is then built up using the pair-copulas theorem, based on the marginal distributions and the pair dependence structures. The model performance is assessed for three sets of assets, namely equity indices, exchange rates, and commodity prices. The empirical results support a better static and dynamic properties of the presented model compared to most common specifications used in practice. The proposed model and the alternative specifications are then carried out to perform stress testing exercises on hypothetical portfolios, where financial returns are considered as risk factors. The results show that the use of a wide range of risk models produce significantly different results, in terms of the corresponding stress scenario and in the corresponding impact on the portfolios. Hence, considering flexible and consistent specifications, as in the proposed model, allows ensuring a better credibility of the stress scenario and enhances the usefulness of the stress testing results.