Modeling Dependencies Between Asian Sovereign Debt Markets With Vine Copulas

Abstract

This thesis evaluates the effectiveness of vine copula models, also known as pair- copula constructions, in modeling the dependencies between Asian sovereign debt markets. We analyze the dependence structure between 5-year credit default swap spreads on sovereign debt for Indonesia, Malaysia, China, South Korea, Thailand, and the Philippines. Marginal distributions are estimated using ARMA-GARCH models. Two vine copula models are constructed: a regular-vine (R-vine) cop- ula model and a canonical-vine (C-vine) copula model. These are compared to the conventional copula used in modeling dependencies in credit portfolios, the multivariate Student’s-T copula model. We show that vine copula models outper- form multivariate Student’s-T copula models in capturing dependencies between countries’ sovereign debt markets and reveal insights into the dependence struc- ture between the chosen markets. All three models accurately estimated VaR for Malaysia, China, South Korea, Thailand, and the Philippines when backtesting over the 2007-2008 financial crisis at various confidence levels, but extreme outliers still affected model accuracy in the case of Indonesia. Results point to the R-vine model as the most accurate in capturing both dependencies and market risk in Asian sovereign debt markets.