In this paper we outline an approach to make use of accounting and market based information to forecast corporate default. We evaluate a wide set of default forecasting models that make varying use of accounting and market based information.
We find that modified structural model approaches of the type used by Moody’s/KMV are best able to forecast bankruptcies out-of-sample for a set of 1,797 bankruptcies over the 1980 to 2010 period. We then find that these superior default forecasts are also able to explain relatively more of the cross-sectional variation in credit spreads for a sample of around 2,000 corporate bonds over the 1997 to 2010 period, and 453 credit-default-swap contracts over the 2005 to 2010 period.
The most interesting result that we document is the predictive information content of these default forecasts relative to the physical default forecast implicit in actual credit spreads. We find a robust positive association between the differences in actual credit spreads and implied credit spreads based on our best default forecast models and future credit returns.
This relation is robust to (i) industry controls, (ii) inclusion of known equity risk factors, and (iii) alternative weighting schemes. This positive relation is suggestive of a role for structured use of accounting and equity market information to serve as an anchor for evaluating actual credit market data.