CECL model validation: data and documentation to play an important role

While financial institutions are still in the early phases of current expected credit losses (CECL) planning, implementation and monitoring, others have already selected and simulated their model or model software and are ready for the next phase of the process: model validation.

Since CECL requires significantly more data, including historic, current and forecast assumptions, as well as more sophisticated analysis, validating a CECL model will likely require additional considerations beyond the current expected approach. We highlight a few reasons why internal and third-party model validators will place more focus on data and documentation.

Data Reliability: The importance of data is paramount and its quality is at the cornerstone of CECL. Data plays a significant role in the methodology you choose, the accuracy of your estimate, and the impact of the other integrated loss reserve analyses. When validating data reliability, an analyst or third-party model validation partner will need to consider the quality and completeness of data over the life of the loan, how data is stored, managed and scrubbed, and the quality and relevance of the forecast inputs and assumptions used in a model.

Governance and Documentation: While FASB did not prescribe a specific methodology, the standard states that the methodology must be “reasonable and supportable,” which means that modeling methodologies and processes should be well-documented. Thorough model governance and documentation will ensure that your organization can track changes to the models and movel processes, and will give auditors and regulators confidence in your approach. Documentation should extend beyond methodology to include model processes, model use, model maintenance, data and assumption selection, and portfolio segmentation processes. Since these model outcomes and process integrate with other models and inform quarterly accounting reports, it is important to ensure this information is accurate and can be relied upon for both business and regulatory reasons.

In future articles, we will focus on other key aspects of CECL model validation, including model testing/outcomes analysis, conceptual model, model use, and reporting.