Interest rate risk model assumptions are an integral part of the overall asset and liability management process. Managers and the Asset and Liability Management Committee (ALCO) review and develop model assumptions to ensure the results are reliable for business decisions and managing the balance sheet. Greater confidence in model results helps management develop business tactics and strategies to minimize risk and capitalize on opportunities.
Typically, asset managers use historical data to develop key model assumptions such as loan prepayment rates, deposit decay rates, repricing rates and rate betas, among others. However, the “pandemic years” (the time during the COVID-19 pandemic that occurred between March 2020 to around mid-2022, with lasting effects well afterwards) have made historical analysis incredibly challenging. Our goal here is to develop an approach to IRR model assumptions in the post pandemic environment.
In evaluating our situation, it looks like we have several options to consider. For any option, we should acknowledge the impact of the pandemic on historical data and document the selected course of action for management and ALCO.
Of course, our first option is to continue to use the assumptions we currently have in the model and complete additional analysis as conditions improve and warrant. For some assumptions, this might be just fine. For example, discounts rates used in present value analysis are often based on market rates and, while market rates are higher now than in the pandemic, the market rate approach seems appropriate. Loan prepayment rates, on the other hand, might need to be reviewed. Loan prepayment rates, for many institutions, declined dramatically during and after the pandemic years and most likely need to be reviewed. Consider a review of this and other main model assumptions and identify those that have been most impacted by the economic environment during the pandemic. Some assumptions most impacted by historical data during the pandemic seem to be loan prepayment rates, non-maturity deposit decay rates and rate betas (for both assets and liabilities).
For those assumptions where further review and analysis is warranted, options we might consider include data normalization or an adjustment to current assumptions based on management’s assessment. Normalizing historical data used in the analysis for a given assumption has some appeal. For example, we might adjust non-maturity deposits by the surge balance identified during the pandemic years when conducting a decay analysis. (As a note, most institutions experienced relatively significant deposit balance increases during the pandemic and those balances above a “reasonable” level were considered “surge” balances that were not likely to remain with the institution once the pandemic subsided.) In other cases, a management adjustment to an assumption value may be a reasonable approach. As an example, an historical analysis of loan rate betas over the past few years (given the movement in market rates) might result in abnormally low values that may not be realistic for the near term. Adjusting those values based on management expectations might provide improved assumptions for the interest rate risk model. Such adjustments are somewhat similar to a qualitative factor assigned by management in the allowance for credit losses model. Regardless of the option we use, it’s important to include documentation of the approach for management, ALCO and audit purposes.
Another approach for assumptions is to conduct sensitivity analysis for key variables. We can increase or decrease the value for each assumption by a percentage or factor and then evaluate the results on earnings and value. For example, we can increase or decrease non-maturity deposit decay rates by 50, 100 and 150 percent and process model results for each scenario. We can then compare the results to the base case to evaluate how sensitive the assumptions might be to significant changes. Typically, we would expand our analysis of assumptions with greater sensitivity to earnings or value at risk. Stress testing is one approach that can be completed regardless of the assumption and market environment. Market conditions are always subject to volatility, but the extremes felt during the pandemic years were far more challenging than the typical ups and downs. Over the next few years as we continue to adjust to the post-pandemic economy we will want to continue reviewing and assessing model assumptions. To prepare for future analysis, it may be beneficial to establish a process to capture core data and other information (such as rates, investment details, etc.) on a monthly basis, if not already in place. Historical data is often the most challenging part of conducting an analysis for model assumptions. Having a database of information provides a great foundation for success.