Pricing models can either be purchased from vendors or developed internally, and they can be mainframe- or PC-based. Internally developed models are either built from scratch or developed using existing customized models that traders modify and manipulate to incorporate the specific characteristics of a transaction.

The use of pricing models introduces the potential for model risk into the valuation process. Model risk arises when an institution uses mathematical models to value and hedge complex financial securities that are in relatively illiquid markets and for which price-discovery mechanisms are inefficient. In these circumstances, the models an institution uses may rely on assumptions that are inconsistent with market realities; employ erroneous input parameters; or be calibrated, applied, or implemented incorrectly. Accordingly, effective policies and procedures related to model development, model validation, and model control are necessary to limit model risk. At a minimum, policies for controlling model risk should address the institution’s process for developing, implementing, and revising pricing models. The responsibilities of staff involved in the model-development and model-validation process should be clearly defined.

In some institutions, only one department or group may be authorized to develop pricing models. In others, model development may be initiated in any of several areas related to trading. Regardless of the bank function responsible for model development and control, institutions should ensure that modeling techniques and assumptions are consistent with widely acceptable financial theories and market practices. When modeling activities are conducted in business-unit policies governing model development and use should be consistent with overall corporate policies on model-risk management. As part of these policies, institutions should ensure that models are properly documented. Documentation should be created and maintained for all models used, and a model inventory database should be maintained on a corporate-wide or business-line basis.

Before models are authorized for use, they should be validated by individuals who are not directly involved in the development process or do not have methodological input to the model. A sound model-validation process rigorously and comprehensively evaluates the sensitivity of models to material sources of model risk and identifies, reviews, and approves new models or enhancements to existing models. Ideally, models should be validated by an independent financial-control or risk-management function. Independent model validation is a key control in the model-development process and should be specifically addressed in a firm’s policies. Management should be satisfied that the underlying methodologies for all models are conceptually sound, mathematically and statistically correct, and appropriate for the model’s purpose. A model should have the same basic mathematical properties as the instrument being modeled. Pricing methodologies should be consistent across business lines. In addition, the technical expertise of the model validators should be sufficient to ensure that the basic approach of the model is appropriate.