An EBIT-based variant of the Duffie--Lando credit risk model

The Duffie--Lando (DL) credit risk model is based on the structural model developed by Leland and Toft, but modifies it by adding a noise term to the observation of firm value. This is done to reflect the fact that the asset process cannot be observed directly and thus that default times are basically unpredictable. Here we present a variant of the DL model that substitutes an earnings process for the asset value process used in the original version of the model. Using an earnings process as the state variable ties default to a piece of publicly available information widely considered to be a sign of a firm's financial well-being. It also allows for the precise specification of accounting noise, which is only sketched in the original DL model. In the last section, we specify a Bayesian methodology for applying the model in practice.