ABSTRACT

In this chapter, the authors provide the utility-based model performance measures and model building approaches. They examine the performance of probabilistic models for investors in incomplete markets. The authors consider probabilistic models for the random variable, which is the loss relative to the par amount of a defaultable debt security. However, many models that practitioners use do not provide probabilities, but rather relate some characteristics of a certain variable, such as its conditional expectation, to a vector of explanatory variables. The authors discuss the performance measures for a few specific, commonly used, regression models, under the assumption that the investor’s utility function is from the generalized logarithmic family. They demonstrate how the different pieces of information that are provided by the specific regression models create value for an investor.