Model Risk

Financial models are simplifications of reality. They can be wrong because of mathematical or coding errors, incorrect data, unrealistic assumptions, or applying a model outside the conditions it was built for.

For example, risk models used by many banks before 2008 assumed that house prices across the United States would not fall sharply at the same time, based on historical data. When they did, losses far exceeded the models' estimates. A spreadsheet error has also contributed to large trading losses at major institutions.

Model risk is managed through model governance: documenting how models work and their limitations, independent validation and testing, controls over changes, regular reviews of performance against actual outcomes, and challenging key assumptions.

Banking regulators require banks to manage model risk formally, with guidance such as that from the US Federal Reserve and the Office of the Comptroller of the Currency. Even well-built models should be used alongside judgment, since no model can capture every possibility.