How to validate a stress test model
**Please check our most recent blog post regarding the latest changes to the FASB deadlines.**
The prevalence of stress testing within banks and credit unions has risen considerably in recent periods thanks to increased regulatory attention and the benefit of greater insight into financial institutions’ portfolios. While there aren’t set requirements in place for those under $10 billion in assets, interagency expectations have been set for stress testing by community banks.
Many banks, even following the recent economic crisis, maintain high concentrations of commercial real estate (CRE) or are targeting the segment for growth. To defend and support a CRE concentration, stress testing is a key component.
To provide institutions tips and best practices, Sageworks and CEIS Review recently hosted a webinar, Stress Testing: Drafting a Battle Plan for the CRE Portfolio. Liz Williams, managing director at CEIS Review, reviewed the types of stress testing and timeline of regulatory expectations. Williams also provided six main requirements for bottom-up stress testing:
1. Define the objective
2. Create process for gathering data that is accurate, meaningful and is updated regularly
3. Develop and document appropriate and relevant scenarios or “what-ifs”
4. Analyze the results and draw conclusions
5. Establish management oversight, review and action
6. Obtain independent validation of the methodology
Gain meaningful insight into portfolio performance
Learn More
At the conclusion of the webinar, an audience member posed a question in response to the sixth requirement mentioned above: “How do you validate your stress test model if the various scenarios don’t actually take place in real life?”
Williams offered some guidance, stating that validation includes looking at the process in a variety of ways. It starts with validating that the information you’re gathering and using is accurate and meaningful. It would also be important to review the accuracy of the calculations – that they’re doing what is intended – and then review the results. “It is admittedly a hypothetical exercise and it is certainly not known whether any of these scenarios would ever transpire, but it gives you some information about how your portfolio could perform,” Williams noted.
Williams provided two examples:
A bank that did fairly well in the last credit cycle and is looking to grow their commercial real estate portfolio could easily receive pushback from regulators that the concentration is increasing. Using actual loan-level information to look at how the portfolio might perform under stress is very powerful in terms of supporting and maintaining that level of concentration.
On the other side, a bank that didn’t do so well in the last credit cycle may have made some adjustments to underwriting standards and processes. Using the same loan level data in the stress test analysis can really show the impact of those changes and how the portfolio might perform better under stress in the future.
Stress testing isn’t an attempt to predict a future outcome, according to Williams. Rather, it is part of the analysis process to show how the portfolio might hold up in different scenarios.
