A percolation approach to modeling credit loss distribution under contagion

ABSTRACT In this paper, we suggest that, in addition to common exogenous factors, the structure of links in the economy might have a bearing on credit loss distributions. Insight gained from considering the structure of links would help in understanding not only how risky obligors might get but also how deeply a crisis might affect credit portfolios as a function of the internal structure of the economy. The theory of multiple interacting agents can provide both the conceptual framework and statistical tools for modeling the structure of links in the economy. The theory allows one to consider a connectivity parameter of the economy, bringing important insight to credit risk correlations. Using percolation theory, we offer an explanation as to why á-stable distributions might serve as a model for credit risk. We also demonstrate how the model presented in this paper can be naturally integrated into the framework of reduced-form models of credit risk. Correlations due to common exogenous factors are integrated with correlations due to contagion effects.