From CECL to capital stress testing: What you should know about model validation
As more complex models are developed to support financial reporting and key business decisions, regulators and audit committees alike are taking a closer look at institutions’ model risk management (MRM) framework. With an ineffective MRM framework, relying on a model may lead to financial reporting errors or operational losses.
One component to a sound MRM framework is model validation. In the face of increasing audit and regulatory scrutiny, many institutions are validating their new models prior to formal adoption. In addition to getting ahead of the regulators, their early moves are paying off through better risk management and improved business outcomes.
Allowance for current expected credit loss (CECL) models will soon be more prevalent across the entire financial institution sector. Vendors and in-house developers are currently updating and tweaking their CECL models, and some institutions have begun to run their model parallel to the current tool. Given the reliance on its output and the related impact to financial reporting, an institution’s CECL model will likely be rated high in the MRM framework, requiring model validation on at least a periodic basis. For more complex models, we recommend institutions plan to get their CECL models validated prior to the year
of implementation. (Implementation will be Q1 2020 for SEC registrants, Q1 2021 for public business entities, and Q1 2022 for all others.)
Capital stress testing
Capital stress testing, which focuses on loan portfolios after a two-year stress event, became an annual requirement for large banks after the 2008 financial crisis. While not a requirement for smaller institutions, the regulatory expectation of validating a capital stress testing model is now being pushed down to smaller institutions with over $1 billion in assets, especially for those experiencing high growth or with concentrated exposures.
Getting ahead of the regulators
Many small and midsized institutions are voluntarily validating their capital stress testing and CECL in advance of adoption. They’re doing it, not only to get ahead of the regulators, but also because of the devastating impact a defective model and model misuse may have on the organization.
Should your institution validate capital stress testing models now?
Here are some questions to ask:
- Is your institution basing strategic business decisions on the output of model?
- Do you understand the impact of all key assumptions used in the model and the impact it may have on your capital plans?
- Do you understand the limitations to the model being relied upon?
Depending on how you answered these questions, you should consider having your model validated as part of a sound MRM framework.
Benefits of validation
Early validation of CECL models will position your institution well in the face of pending regulatory
and audit scrutiny around the new rules.
Similarly, early and regular validation of both CECL and capital stress testing models will ensure your institution is making the best possible business decisions based on its use of its models. We find that clients who adopt and regularly validate capital stress testing models are making better lending decisions and experiencing improved management of assets and liabilities. In addition, we’re finding institutions that adopt a culture of innovation and constant improvement are consistently more profitable than their counterparts.
Another key benefit of early adoption is an improved ability to manage risk. Capital stress testing model validation helps ensure your lending practices won’t leave your organization undercapitalized if the economy turns sour. And in some cases, it can lead to changing business direction before risk turns into losses. For example, if it looks like a certain lending path could lead to too much exposure, your organization may make the business decision to get into a different type of lending.
Former statistics professor George Box once quipped, “Essentially all models are wrong, but some are useful.” His statement is a witty reminder that models, while useful, are also subject to error and misuse. Give us a call to find out how we can help you maximize the usefulness of all your models, reduce risk, and be more profitable.