Best Practice Suggestions for Back-Testing Asset-Liability Management (ALM) Models for Accuracy

8/10/2022 - By Terry Treadwell, CPA

Back-testing has been challenging during this COVID period due to large volume variances and asset yield variances caused by high loan and deposit growth coupled with forecasts not incorporating PPP loan-related fees. The back-testing exercise helps Financial Institutions (“FI”) determine the overall reasonableness of the Net Interest Income (“NII”) base case forecast and finetune underlying assumptions in an ALM model. Throughout 2022, back-testing has become more challenging as market interest rates rise. There are various things to consider with back-testing:

Most models provide both static and dynamic balance sheet forecast ability. The static balance sheet model forecast is typically the most helpful to determine “base case” model NII accuracy (and what regulators expect to see tested). A static balance sheet forecast incorporates a cash flow runoff assuming the various asset and liability categories are reinvested back into the same categories with no new asset or liability growth.
Market rates for forecasts use either flat or forward rates. We find vendors modeling static balance sheet forecasts using flat, forward and sometimes both interest rate approaches. A static balance sheet forecast using a flat market rate assumption typically results in NII variance within 5% of actual results. With more significant rate increases now being built into forward rate forecasts, the base case NII may be hard to compare with actual historical results. Ask your vendor if they can provide both a flat and forward rate forecast to help you further evaluate these differences.
Period for back-testing. Examiners typically expect to see a 12-month back-test period applied which we encourage you to evaluate annually. However, for more accurate assumptions evaluation your FI can consider applying a shorter back-test period. Comparing a recent month actual vs. forecast using a rate/volume approach and performing these tests more frequently can help further isolate unusual trends or results that suggest possible model adjustments. Quarterly is typically the frequency for ALM model updating, and most model reporting includes monthly period forecasts.

Questions?

If you have any questions about Asset Liability Management (ALM), don't hesitate to contact our Financial Institutions team.

About the Author | Terry Treadwell, CPA

Terry is a consultant in the Financial Institution Advisory Group at Saltmarsh, Cleaveland & Gund with more than 25 years of experience serving financial institutions. Her primary areas of concentration are asset/liability management including interest rate risk modeling, strategic balance sheet planning and liquidity management.