Confidence Intervals for Corporate Default Rates

Abstract

Rating agency default studies provide estimates of mean default rates over multiple time horizons but have never included estimates of the standard errors of the estimates. This is due at least in part to the challenge of accounting for the high degree of correlation induced by their cohort-based methodologies. In this paper, we present a method for estimating confidence intervals for corporate default rates derived through a bootstrapping approach. The work extends research in the academic literature on one-year default rates [Hanson and Schuermann (2006)] to the multi-year horizon case. Our results indicate that historical mean speculative-grade default rates are generally measured fairly precisely, with standard errors less the 10% of the estimated means. Investment-grade default rates, however, are measured much less precisely, particularly for issuers rated single A or above. Precision increases at longer horizons. Of practical importance, the results indicate that Moody's long-term ratings satisfy the Basel II criteria for effectively distinguishing relative credit risk. This is true even for "low-default portfolio" portion of the rating scale - letter ratings Aaa, Aa, and single A - because the default rates associated with these rating categories are significantly different from one another at the two-year and longer investment horizons.

Keywords: default probability, bootstrap, risk management

JEL Classification: C14, G21, G28