Financial Institution Advantage and the Optimization of Information Processing

Chapter 4Model Risk Measurement

As discussed in Chapter 3, model outputs become truly informative when they are communicated along with characterizations of their perceived accuracy or inaccuracy, as well as with other limitations of the model itself. Such characterizations are often highly technical, can be challenging to produce and communicate, and will vary across model types and classes. An institution can overcome some of these challenges and complexities and boost its information flow by developing a model risk measurement framework that can simply summarize the strengths and weaknesses of each model in a way that's easy for decision makers to understand. Reducing complex and heterogeneous model diagnostics into a simple and uniform index is not easy and many trade-offs and simplifications will be required to make such a system practical. But the payoffs should be great. Not only are decision makers better informed as a result, but resources may be more effectively allocated when the effect of relative and absolute improvement in model performance can be easily understood by management.

Certainly, financial institutions are being pushed toward greater capability in model risk assessment. A 2014 survey by a major consulting firm identified model risk measurement as a hot topic for banks and insurers. As we have already noted, this reflects a mixture of firms' interest in mitigating risk, particularly the type of headline risks that can curtail senior executive careers ...