Research Proposal for SFB guest research program “Effective Forecasting of Default Probabilities”

Empirical analysis of credit risk employs various combinations of financial ratios for modeling and estimation of default probabilities. Hereby most studies concentrate on short-term default risk, considering at most a one-year default horizon. However prior findings on the dynamics of risk drivers that have been established in the Corporate-Finance and Asset-Pricing literature are mostly ignored. In this research project we will empirically study the impact of forecasting and incorporating dynamics of covariates on accuracy of default prediction. The focus here is manly on long-term default risk, considering horizons up to five years. The study can potentially reveal shortfalls of established and widely used measures of default risk, such as Altman’s Z- score, distance to default or ratings. Consequently, the results might indicate how these can be improved. Plan of action If default of a firm within a timeframe [t, T] is indicated by a dummy variable TtY, = 1, then default probabilities can be modeled by)'() ( , tTt XFYP βα + = , (1) where tX denotes a vector of risk drivers measured at time t and F is the cumulativ