Abstract driving fast through city.

CECL Model Validations Are More Than a Regulatory Exercise

By Steve Picarillo, David W. Giesen

March 26, 2019

Now that your Current Expected Credit Loss (CECL) models are complete (or near completion), it is time to think carefully about the next steps in the transition process. While there is a natural inclination to think, “thank goodness that’s over,” and move to implementation, you are not quite there yet. These newly minted models, which will be used for financial reporting purposes, should undergo a comprehensive validation to independently review, check, and verify that the models function and perform as originally designed.

Model validation is a relatively new concept for financial institutions yet various forms of validations have been employed in other businesses for decades. For example, publishers utilize fact checkers and copy editors to validate the work of reporters to ensure the credibility of their articles. Likewise, for financial institutions, a proper CECL model validation helps ensure the credibility of financial statements, particularly the allowance for loan and lease losses (ALLL). The Model validation concept arises from regulators’ Model Risk Management (MRM) guidelines and policies that govern financial institutions’ use of tools that flow into critical financial and operational functions. While larger financial institutions validate their high-risk models each year, Ankura recommends that banks of all sizes adopt a comprehensive validation process for all high to moderate risk models, especially CECL models, as best practices for a sound MRM function.

Validation is the second line of defense in assuring model accuracy. A good validation goes beyond just simply checking the boxes and should identify correctable problems in advance of audit and regulatory review. Given the unique aspects of CECL, the prominence of loss reserving to the financial statements and the biting consequences of reporting errors, many entities are reaching out for guidance, assistance, and skilled resources to ensure proper, timely, and cost-effective CECL model validations.

Prior to commencing, it is important to understand what your firm wants to accomplish from the validation. At the simplest level, a model validation checks inputs, reviews outputs, challenges assumptions, variables, and modeling methodology, and ensures the proper evidential documentation and functionality of the model. However, there are significant benefits to be gained by designing a comprehensive CECL validation, that goes beyond the “check the box” approach. A proper validation scope will include a review of model risk and governance, model coding/programming, and a comprehensive review of the source and quality of data. If there are feeder or follow-on models involved, a proper validation will also evaluate those tools. Given the importance of the CECL transformation and properly estimated ALLLs, regulators, auditors, and investors will gain a significant level of comfort in firms that perform comprehensive and independent validations as part of their MRM.

CECL is vague on the “how to” aspect of determining the ALLL, and allows for sizable management judgements, interpretations, adjustments, and differing model methodologies. Opining on, challenging, and documenting the rationale of such determinations is an essential part of the validation. To this end, the model teams should be able to strike a balance between statistical accuracy and the accounting aspects of CECL, which may present certain challenges, as traditional modeling teams are typically inclined to quantitative analysis and prefer the statistical approach to modeling.

To address the specifics of CECL, your firm should look at the following topics when designing CECL model validations:

  1. The calculation of the life of the loan

    CECL requires ALLLs to be calculated based on the “life of the loan,” as such, financial institutions should take a granular approach to tracking prepayment trends within their portfolio. To accurately calculate the expected life of an asset, banks need to have detailed data of prepayment speeds and defaults for each loan product, market, and asset type. This is a mission-critical issue in both model builds and model validations, since the longer the credit’s term, the bigger the ALLL will be. A good validation will evaluate how loan life was estimated and how the life will change with shifting economic environments. This will be another area of regulatory and external auditor focus, especially in an increasing interest rate environment on the tail of an extended period of historic low interest rates.

  1. The defense of a “reasonable and supportable” forecast

    While most market participants agree that a two to three-year forecasting period is reasonable and supportable before reversion to the mean, the validator will need to look at the economic factors that govern your bank’s portfolio. Depending on the loan structures and the geographic location of the borrowers, more localized economic projections may be essential.

    Similarly, model validation should include developmental evidence and supporting documentation of the reversion approach. Since CECL allows for several mean reversion methods, a proper validation will need to address the selected approach(es) and the impact of each on the forecast.

  1. Judging management judgment

    CECL allows for significant management interpretation, some of which will be addressed in Q factors and data adjustments. While such modifications may be necessary, too many may prove the model ineffective. Model validators will look at the modelers’ rationale for each adjustment and will require that the ALLL impact of any adjustment be quantified and documented.

A CECL validation should strike a careful balance between raw statistical review, the accounting requirements, and an understanding of how banks lend money. This creates certain tensions between statistical, financial, and credit teams, who see lending differently. A proper CECL model validation can bridge these gaps and arbitrate critical differences before they impact your financial statements.

With quickly approaching implementation dates and the need to redesign the validation process, timely, accurate, and comprehensive CECL model validation has become a key concern for financial institutions. Many institutions, especially smaller banking entities, do not have the depth of skilled and independent resources available to handle all three lines of defense. With a small pool of qualified resources, CECL model validators are increasingly more difficult to find and, as the implementation dates approach, the pool of available resources will undoubtably evaporate.


Ankura’s seasoned professionals have hands-on experience in banking, modeling, risk, capital planning, consulting, regulatory compliance, and data analysis. Our team has the essential subject matter expertise to effectively validate CECL and other high-risk models. With deep experience in developing, validating, and auditing models for US and global banks, we tailor our solutions to our clients, recognizing one size or approach does not fit all.