Bayesian synthesis of portfolio credit risk with missing ratings

Dror Parnes

Abstract

ABSTRACT

In this study we develop a universal Bayesian model that accounts for the time frames until future credit rating downgrades, the time lengths of no credit coverage, and the time intervals until default of individual debt issuances. We then use a maximum likelihood estimation to assess the projected average default rates (over diverse horizons) of debt portfolios that exhibit similar credit characteristics. We demonstrate the scheme over multiple large homogeneous portfolios and further authenticate the validity of the derived credit profiles by appointing to each uniform portfolio a descriptive credit grade. These combined credit ratings tend to be higher than their respective individual bonds. This reconfirms the universal advantage of portfolio diversification.