Glossary term
Model Risk
Model risk is the risk of loss, poor decisions, or compliance problems from using models that are wrong, misused, or poorly governed.
Updated
Read time
What Is Model Risk?
Model risk is the risk of loss, poor decisions, compliance problems, or misleading reporting from using models that are wrong, misused, poorly implemented, or poorly governed. It matters because financial institutions, investors, lenders, insurers, and businesses rely on models to estimate uncertain outcomes.
Models can support credit scoring, portfolio risk, capital planning, pricing, fraud detection, valuation, stress testing, underwriting, and forecasting. A model can be mathematically impressive and still create risk if the data are weak, assumptions are stale, users misunderstand it, or outputs are treated as more certain than they are.
Key Takeaways
- Model risk comes from flawed models, bad data, poor implementation, or misuse.
- It can affect lending, investing, valuation, capital, compliance, and customer decisions.
- Good governance includes validation, documentation, monitoring, and clear ownership.
- Model output should support judgment, not replace it blindly.
Where Model Risk Comes From
Model risk can enter at every stage of a model's life. The problem may be the design, the data, the code, the assumptions, the controls, or the way the business uses the result.
Risk Source | Example |
|---|---|
Bad assumptions | A model assumes relationships that no longer hold. |
Weak data | Inputs are incomplete, biased, stale, or mislabeled. |
Implementation error | Code, formulas, or system integration do not match the design. |
Misuse | Users apply the model outside its intended purpose. |
Governance and Validation
Model risk management usually includes inventorying models, assigning ownership, documenting assumptions, validating independently, monitoring performance, and setting limits on use. For banks, supervisory guidance emphasizes that model risk should be managed throughout the model life cycle.
Validation does not mean proving a model is perfect. It means testing whether the model is fit for purpose, whether limitations are understood, and whether users know how much confidence to place in the output.
Monitoring matters after a model is approved. A credit model, valuation model, or risk model can degrade when borrower behavior changes, markets become stressed, a product mix shifts, or historical relationships stop holding. Good governance treats model performance as something to watch over time, not a one-time approval box.
Consumer and Investor Exposure
Consumers may encounter model risk indirectly through credit approvals, insurance pricing, fraud flags, loan limits, and account decisions. Investors encounter it through portfolio risk systems, valuation models, factor models, and forecasts. In both cases, a model can turn data into decisions that have real financial consequences.
The risk is not only that a model is wrong. It is also that a decision maker may give a model more authority than it deserves, especially when the output looks precise.
The Bottom Line
Model risk is the gap between model output and reality. Strong controls, validation, and humility are necessary because models can be useful without being certain.