How Do You Know Your Modeling Assumptions Are Right?
Interest rate risk modeling requires the user to make assumptions about member and management/board behavior. For example, some members will pay back their loans ahead of schedule and the rate of prepayment will increase if rates fall and decrease if rates go up, especially on mortgage-related products. Likewise, the incentive for members to move some of their deposits from lower-cost products, such as regular shares, to higher-yielding CDs increases when rates go up (note: static income simulation ignores this important aspect of IRR management). Assumptions must be made about the likelihood of these events occurring in the future.
Has your examiner asked “how do you know your modeling assumptions are right?” If so, you answered correctly if you said you don’t. Assumptions, as the name implies, are assumptions. They are not facts. A prepayment speed on a loan can be calculated with certainty when the prepayment occurs. To put that prepayment speed into a model assumes that behavior will continue, but it may not. How that behavior changes when interest rate changes is yet another assumption. Despite this, the assumptions are an important and necessary part of modeling a credit union’s IRR exposure.
The new IRR rule, 12 CFR Part 741, Interest Rate Risk Policy and Program, says assumptions should be “reasonable and supportable” and that credit unions should “assess the sensitivity of results relative to each key assumption.” We agree with these comments and work with our clients to develop reasonable assumptions, yet we realize there is no right assumption. Therefore, we conduct literally thousands of stress tests annually to help our clients understand the impact of their assumptions to their IRR exposure. For example, we often stress test early withdrawal speeds by making them 50% faster than what is in the base simulation. Whether you are a client of ours or not, a recommended approach is to stress test assumptions to see how they may impact results. Assumptions that have a material impact may require additional research and refinement whereas others may not.
The bottom line is stress testing and documenting—followed by assumption adjustments if appropriate—should help to satisfy the “reasonable and supportable” component of the new rule.
