As CECL adoption ramps up and nears completion, many banks—both 2020 and 2023 filers—are looking to validate CECL estimates, ensure that models remain predictive, and address auditor concerns. And the volatility of the macroeconomic environment doesn’t help, changing a bank’s portfolio mix and adding additional stress on CECL models.
At the same time, many institutions' loan books are changing, meaning that historical loss rates are not a reliable comparison point for future-looking loss forecasts. Further, there is no one-size-fits-all approach to CECL, so finance and risk departments must consider all elements related to portfolio credit characteristics and the macroeconomic environment to determine an appropriate approach. The wide variety in approaches to CECL further muddies the waters making it more difficult to compare results to peers and more difficult for management to gain comfort over the functioning of CECL models.
Independent third-party validation and calculation review can play an important role, but with the complexity of these calculations, they may not be enough—that’s where benchmarking comes in. Model benchmarking compares model outputs with results generated using a different methodology as a validation tool. A benchmark model may help users gain greater confidence in their production of a “champion” model, or alternatively, highlight the sensitivity of model outcomes to the selected methodology and assumptions.