DCG has performed model validations for banks and credit unions of all sizes and balance sheet complexities for well over a decade. Over that time, the asset size of our clients has ranged from $50 million to nearly $50 billion. As you might expect, we typically see a correlation between the size and complexity of the institution and the sophistication of the ALM modeling software employed by that institution.
For example, larger institutions may designate a team of employees whose sole responsibility is to maintain a highly customized ALM model that is used to provide management with ongoing risk measurement results, whether they be NII, EVE, cash flow/liquidity, capital, what-ifs, or stress-testing. Most community institutions appoint a member of the treasury or accounting department to be tasked with running an in-house standardized vendor-provided model, usually in addition to myriad other responsibilities. We also see many organizations that outsource the ALM modeling function to a third party provider who performs the ALM modeling function and sometimes even prepares ALM reporting for that institution. Regardless of asset size, the appropriateness of an ALM model is dependent on its ability to capture the institution's risk profile.
While the complexity of ALM modeling software and expertise of the model owners and support staff (including outside 3rd party vendors) can vary greatly, regulatory expectation in regard to model governance remains fundamentally the same for all institutions. No matter how big or how small your institution is, the guiding principles for a solid framework of model oversight are consistent. However, the manner in which and to what extent they will be applied will vary depending on the size and complexity of the individual institution.
Below we will address several key practices management should consider incorporating into its ALM monitoring process.
Stress Testing: Sensitivity Testing Regulatory guidance proposes that institutions should
model alternative key assumptions to allow decision makers to see the impact each has on model results. Periodically testing various key assumptions and quantifying the impact to NII and EVE results should help ALCO understand how each key assumption impacts the overall risk profile.
This practice will enhance ALCO's confidence in model results, leading to more informed strategy formulation. For example, assumptions regarding deposit decay or average lives are perhaps the single most important determinant of economic value of non-maturity deposits. By modeling alternative assumptions (more/less conservative), management can gain a better understanding of the potential effect on earnings and capital.
Stress Testing: Scenario Testing In addition to testing key assumptions, institutions should
conduct scenario testing to simulate a range of future interest rate scenarios and potential impact on the institution's exposure to basis risk, yield curve risk, and option risk. These scenarios could include non-parallel rate changes (flattening or steeping yield curve twists), gradual rate ramps, and extreme rate change scenarios (i.e., +400bp) as conditions warrant. Results that provide a range of estimates for different scenarios and assumptions can help decision makers gain a better understanding of the impact these factors have on the overall interest rate risk profile.
Outcome Analysis: Model Back-Testing The Supervisory Guidance on Model Risk
Management stresses the importance of outcomes analysis and, specifically, model back-testing as a key way to gauge model performance. It states that when back-testing outcomes fall outside expected ranges, institutions should analyze the discrepancies and investigate possible causes. Differences may occur from errors in model set-up, model assumptions, or other factors such as market changes. While the most common type of back-test involves a comparison of projects to actual NII, model owners should also consider other analyses, such as a review of projected cash flows to those actually experienced. Comparing model outputs to the corresponding actual outcomes should help give management a comfort level with the model's accuracy or help to identify assumptions that may need tweaking.
Outcome Analysis: Benchmarking As a logical extension of NII back-testing, management
should also ensure the conceptual soundness and effectiveness of the ALM model through review, analysis, and benchmarking. Comparing actual versus forecasted activity can be helpful when validating cash flow projections, estimated prepayment speeds, growth projections, etc.
For example, management should review:
Cash flow reports to ensure that calls (investments/borrowings) are being evaluated and cash flows change as might be expected in the different rate environments
Rate reports to see that current rates are accurate and replacement rates are as intended
Compare market values calculated by the model to compare against pricing obtained from brokers for investments, residential mortgages, and FHLB advances (when possible)
Ongoing monitoring is not a one-time event, nor should it be done just to appease your regulator. Ongoing monitoring should be an integral part of management's ALM process, which, in turn, should promote greater confidence in the model results and, ideally, lead to better decision making.
About the Author
Since joining DCG in 2005, Karen French has worked in a number of capacities assisting clients in all aspects of asset/liability management. These include building ALM models and providing balance sheet management education, guidance, and support for DCG advisory clients, as well as performing independent process/policy reviews and both IRR and liquidity model validations for banks and credit unions across the county. She is also currently actively involved in DCG's capital planning products and services.
Contact DCG today to speak with one of our balance sheet experts, schedule a software demo, or discuss your institution’s needs and goals. We look forward to hearing from you.
© 2023 Darling Consulting Group, Inc.
