Time-Varying Copula Networks for Capturing Dynamic Default Correlations in Credit Portfolios

Abstract

Grasping the dynamic interdependence of credit defaults is important to successful portfolio risk management and systemic risk assessment. Though useful in modeling joint default behavior, traditional copula models may assume a static dependence structure that does not reflect the changing nature of credit market linkages. The paper proposes a Time-Varying Copula Network (TVCN) model that can be used to model and visualize time-varying default correlations in credit portfolios. The proposed framework enables the detection of changes in the systemic connectivity and contagion pathways by temporal shifts in dependence modeling, which can be combined with a network representation. To estimate the time-varying copula parameters, a state-space model is used and the parameters are updated in a recursive model. The empirical model based on credit default swap (CDS) data show that the TVCN approach is effective to observe non-linear and asymmetric effects, particularly when financial stress arises. Findings indicate that the correlation with default is not fixed, but varies in response to macroeconomic shocks, liquidity constraints, and interlinkages between sectors. This dynamic definition offers better understanding into portfolio diversification, stress testing, and risk concentration patterns than their static counterparts.