We compare different ways of modeling real-world (as opposed to risk-neutral) probabilities of default over a fixed time horizon conditioned on a vector of explanatory variables like financial ratios or macroeconomic indicators. Besides a simple approach based on a logistic regression, we discuss a maximum expected utility approach, which chooses the model measure from a one-parameter family of pareto-optimal measures. These are defined in terms of consistency with the data and consistency with a prior measure. The consistency with the data is measured in terms of a feature vector and the consistency with the data in terms of relative entropy. We apply this setting to a very general class of utility functions, namely the class of hyperbolic absolute risk aversion (HARA) utility functions. The different models are compared with respect to different performance measures, power measures on the one hand and calibration measures on the other hand. The numerical comparison is based on Fitch Risk?s North American Loan Loss Database.     «

We compare different ways of modeling real-world (as opposed to risk-neutral) probabilities of default over a fixed time horizon conditioned on a vector of explanatory variables like financial ratios or macroeconomic indicators. Besides a simple approach based on a logistic regression, we discuss a maximum expected utility approach, which chooses the model measure from a one-parameter family of pareto-optimal measures. These are defined in terms of consistency with the data and consistency with...     »