Contagion network, portfolio credit risk, and financial crisis

We develop a network-based nonlinear dynamical system model to study credit exposure risk at the portfolio level. This model captures the complex characteristics of contagion arising from the microstructural interdependencies among firms, especially looping effects. We find that when contagion features are not adequately modeled, portfolio credit risk is generally underestimated, but becomes overestimated during a crisis. This can partly explain the outbreak and subsequent intensification of a crisis such as the 2008 financial crisis. We also derive an expression for the portfolio loss of a regular network, which has implications for measurement and pricing of portfolio credit risk.