Risk Management Models: Construction, Testing, Usage

Abstract

All models are wrong but some are still useful. Jarrow explores the implications of this truth for the development and proper use of risk management models. All models are based on assumptions that oversimplify conditions but are, hopefully, basically correct. In some cases, such assumptions are “robust,” meaning that the model is not much affected if they are a little off. Other assumptions are “critical” in that the model is seriously degraded if they fail at all. Another dimension on which assumptions can be separated is whether an assumption is capable of being tested against observable data. Jarrow expands on how these ideas apply in developing, testing, and using models. He reviews a variety of common practices, such as vega hedging using the Black-Scholes model, that make no sense in the theory of model building he offers.