Correlation and asset correlation in the structural portfolio model

Jon Frye

Abstract

ABSTRACT

To forecast the default distribution of a credit portfolio, a risk manager often relies on a structural model that contains a measure of correlation. Quite frequently, the manager sets correlation in the model equal to asset correlation. This common practice, it is argued here, may be an error. Correlation in the portfolio model would be identical to asset correlation under a set of assumptions. This study examines each assumption in detail. Relaxing each assumption creates a potential for difference. The potential is realized in studies that estimate correlation based on default data; these estimates tend to be less than the asset correlation. The estimation presented here rejects uniform values of correlation greater than 10.8%. It is shown by example that the difference in correlation values appears to be great enough to produce misleading statements of risk.