Empirical Studies on Default Contagion

One of the main challenges for the risk management of banks’ portfolios remains to correctly assess the correlation of corporate defaults in order to allocate the right amount of economic capital to portfolio risk. Many banks apply credit risk models that rely on some form of the conditional independence assumption, under which default correlation is assumed to be captured by the dependence of all firms in the portfolio on some common underlying risk factors. Well known examples of this approach include the Asymptotic Single Risk Factor (ASRF) model as developed by the Basel Committee on Banking Supervision (BCBS) and [72], or applications of the structural Merton model [105] like the KMV or the CreditMetrics model. Also reduced-form or mixture models, like the CreditRisk⁺ model, rely on the conditional independence framework. In the context of reduced-form models the conditional independence assumption is also often referred to as the doubly stochastic property. Essentially it says that, conditional on the paths of some common risk factors determining firms’ default intensities, the default events of the individual firms are independent Poisson arrivals with (conditionally deterministic) intensities.