Why Directors Should Consider Updating CECL Models

Now that the current expected credit loss (CECL) accounting standard has been fully adopted across the industry, many banks are beginning to evaluate the performance of their models and identify enhancements that might be needed to confirm that their allowance for credit losses adequately reflects their current risk profile.

As they carry out this essential next phase of CECL implementation, boards and management teams can benefit by recognizing some of the common limitations and areas of concern that others in the industry are encountering. In addition, directors and senior executives should understand the critical principles and practices to apply when making changes to their CECL models.

Commonly Observed Modeling Changes
As they analyze the performance of their models, many banks are seeing some recurring patterns. In most cases, observed model changes have centered around enhancements to existing approaches rather than more-fundamental overhauls of the process.
Common areas of focus include:

• Segmentation. Most banks’ initial segmentation frameworks were based on call report codes, but many are now considering more-granular portfolio characteristics and moving to further stratification of existing segmentation.
• Historical data. As banks accumulate their own historical loss experience and compare peer data to their own internal findings, many are reevaluating the time periods for which loss data is used, incorporating data from recent years.
• Economic forecasts. Banks are reevaluating the continued applicability of the macroeconomic factors they have been using since the initial CECL adoption, as well as the forecasting approaches they apply in their models.
• Modeling approaches. Many banks are considering whether deployment of more-complex models could offer benefits by enabling them to quantitatively apply economic and loss information within the model. Banks continue to compare the costs and benefits of internally developed models against the purchase of third-party models.
• Qualitative factors. The appropriateness of qualitative factors continues to be a critical focus area for banks. More structured, scorecard-based approaches are becoming more prevalent. Banks must consider whether qualitative adjustments remain appropriate, if model enhancements duplicate factors that already were incorporated in quantitative models and if other factors are needed to account for modeling gaps or new industry risks.

Best Practices for Updating CECL Models
When making changes to CECL models, banks should follow a robust and structured process to evaluate the appropriateness and impact of the modifications. Key to this process is evaluation and documentation of fundamental change elements, including:

• Rationale and impact of changes. Modeling enhancements and changes are inherently judgmental. Banks should be prepared to explain the rationale and reasons driving the changes.
• Model enhancement versus error correction. Banks should be prepared to evaluate and document the impact arising from any changes, which includes determining whether the changes should have been made in an earlier period or are based on new information. For example, while shifting from the use of peer data to internal data could be considered an enhancement, the decision to exclude a prior vintage from the model forecast might need to be evaluated as a correction of an error in the existing approach.
• Compliance with model risk management policy. When evaluating current models or modeling enhancements, it is important to verify that all CECL model elements still comply with the bank’s broader model risk management policy.
• Testing of model changes. Before fully deploying changes or enhancements, parallel runs of previous and revised models are essential to assess the impact of changes and to identify and isolate the drivers of differences between the old and new approaches.
• Model governance. Modeling changes might necessitate enhancements to the key controls governing the model, including the completeness and accuracy of any new data being used. Banks also should revisit existing controls and model evaluation approaches to determine whether they continue to adequately address estimation risks.

By following a disciplined and well-documented CECL model evaluation and update process, banks can improve model performance to more accurately reflect their current risk environments while also maintaining compliance with applicable regulatory and financial reporting requirements. Enhanced and updated models strengthen a bank’s overall credit risk management framework by providing boards and management teams with more-accurate, insightful information for establishing an appropriate allowance for credit losses.