Introduction to the Barra Default Probability Model

Nov 1, 2004

At Barra, we have developed a credit risk model for corprate issuers of debt.  One basic goal of our model is to forecast the default probability for firms that have publicly traded equity in addition to debt.  Initial statistical tests indicate that the Barra Default Probability (BDP) model performs significantly better than an agency-ratings based system.  In this article, we will give a brief outline of the model and the statistical testing.  Until recently, there have been two standard approaches to default probability structural models and reduced form models.  Structural models attempt to relate the default probability to the capital structure of the firm - the mix of equity and debt or, equivalently, the relative valuation of assets and liabilities.  Reduced-form models remain silent about the cause of default and model the statistical properties of the default event itself via Poisson-type distributions.  A new class of recently developed models introduces the notion of incomplete information [1, 2]. This class of models contains both traditional structural and reduced-form models as special cases [2, 3].  The Barra Default Probability model is a specific instance of such a model.  We turn to a brief description of structural models in order to motivate the discussion and introduce several useful concepts 1)Structural Models, 2)Shareholders' Equity = Assets - Liabilities, 3)First Passage Models, 4)The BDP Model, 5)Implementation of the Model