Model Risk Management
Model risk involves the use of financial models, and the potential that errors in setup, input, or interpretation of results can lead to material misstatement of results. Model risk can be present in internally created financial models or in vendor-supplied financial models and/or results.
Having a model’s mathematics validated periodically (important only if there is a change to the underlying software and/or application of the software) can be one way to reduce model risk. However, what about errors in the setup, data input, or interpretation of results? Beyond periodic, formal written model validations, how can decision-makers limit their exposure to model risk? Below are some key items that those involved in financial modeling should consider to aid in reducing model risk:
- Ensure a clean audit trail for model assumptions. Whether modeling internally or engaging an outside third party to conduct modeling, key assumptions utilized (e.g., deposit pricing, prepayment speed, principal cash flows) should be easily available for review by decision-makers. Key assumptions regarding credit/default risk, interest rate environments modeled, and market rates that drive model results should be well-documented and available for review.
- Automated does not mean “error free.” Even though many aspects of financial modeling now include a high degree of automation, ensuring that the automated inputs for financial models are being appropriately captured by the model is the first step in automation being a benefit, and not a drawback, in the financial modeling process. Additionally, if there are any new products added or changes made to existing products, those responsible for the modeling/report generation should be made aware of the changes.
- Separation of duties. Is there sufficient separation between those taking risks (e.g., treasury, investment department or advisor, loan department, etc.) and those reporting risks? An independent third party unaffiliated with investment activity for the credit union (whether both functions occur within or outside the credit union) can be a strong control in the modeling process. The modeler should be encouraged to independently test the risk, including stressing vulnerability to expectations not coming true.
- Modeling risk vs. plans. Planning what you expect to happen is typically opposite of the risk if things go wrong. As a result, a key part of avoiding modeling risk is ensuring that the modeling is addressing things that can go wrong, such as deposits migrating from low-cost accounts, loan balances decreasing, and credit risk increasing.
- Apply common sense. Step back and evaluate whether the results make sense. In our experience of performing model validations, it is common to see extremely optimistic deposit values that are far beyond any price an institution would actually pay to acquire the deposits. Another example is the modeling showing huge gains on the portfolio, while the institution states that its loan rates are very competitive and the loan growth is above market. Results along these lines should raise flags about the exposure to the risk results being wrong, or if the institution feels the results are correct, additional explanation should be available.
Financial modeling should provide decision-makers with useful and relevant information to aid in the decision-making process. Ensuring that a robust system for mitigating model risk exists can help ensure the reliability of that information.
