Overview
Model risk is risk that an institution suffers loss due to a incorrectly developed or utilized model. It can result from
- The model is Incorrectly Specified or Developed - this can result from mistakes in the understanding the math behind the model or mistakes in coding
- The Model is Applied Incorrectly - this can result from applying a model to situations where the assumptions are violated in a material way. All models involve assumptions. The decision makers that apply the results of a model need to be aware of the assumptions of the model, and have a good grasp of when those assumptions do not apply.
Model Validation and Monitoring
- Duplicative Review - models should have multiple sets of eyes reviewing the model. This typically means there is developer developing the model, a separate individual assigned to review and audit the model, and a user of the model. All three should review and sign off on the model. The model should be subjected to a review of the assumptions, the math, and the code implementing the model.
- Ongoing Monitoring - if the model is being used in production, the model should go through ongoing testing and review. In particular, model forecasts should be checked against actual results and the model assumptions should be validated periodically.