CREDIT RISK MODELS: AN ANALYSIS OF DEFAULT CORRELATION

This paper examines one of the major problems in credit risk models widely used in the financial industry to forecast future defaults and bankruptcies. We find that even after proper calibration, a representative credit risk model can severely underestimate default correlation. We further find that a likely reason for the underestimation of default correlation is the problematic common practice in the financial industry of using observable equity correlation as a proxy for unobservable asset correlation when the model is applied to predict default correlation. However, our results show that this proxy in common practice is not valid.