Current Expected Credit Loss (CECL) Model Validation

Darling Consulting Group (DCG) is a recognized leader and premier provider of independent risk management consulting and regulatory-compliant model validation services. Our validations align with regulatory model risk management guidance (OCC Bulletin 2011-12, Federal Reserve Board SR 11-7, and FDIC FIL 22-2017).

DCG’s Quantitative Risk Analysis & Strategy (QRAS) team has conducted over 500 credit-related model validations, including dozens of ALLL model validations. Since FASB released the CECL guidance in 2016, DCG has been transitioning our credit and ALLL expertise into CECL validations. We have collaborated with accounting firms, regulators, model vendors, and practitioners. We developed several CECL test models using a variety of methodologies to evaluate and master the nuances of each approach.

DCG understands that CECL represents one of the most significant changes to financial institution accounting in decades. As a firm who has been closely working with banks for four decades, we understand the challenges that come with such a drastic regulatory and accounting shift. DCG’s approach to CECL validation is that we should not focus on just one aspect of the modeling. A model designed only based on the CECL accounting rules may be too rigid for practical business use. A model with the best credit risk modeling may fall short when macroeconomic and qualitative factors are applied in an inappropriate way.

CECL models come is a variety of shapes, sizes, and levels of complexity. DCG’s CECL Model Validation service is tailored to assess models ranging from simple spreadsheet-based constructs to more sophisticated software implementations. DCG utilizes a 12-part CECL Model Validation scope based on our research and experience conducting trial validations on CECL test models. This 12-part validation scope was developed after careful consideration by DCG’s validation team and is designed to cover all of the core components of a comprehensive CECL validation. Some highlights from our 12-part validation scope are provided below:

Model Governance

Compared to the old incurred loss models (ALLL), the new CECL model requires even more expert judgment to calculate a defensible expected loan loss amount. Even quantitative components that were previously “taken for granted” will require more support than ever before. No matter how simple or complex the methodology, it is imperative that the model is built on a foundation of sound governance. DCG’s governance validation includes an evaluation of policies, internal controls, change control and model documentation.

Data Inputs

CECL requires the integration of data from disparate data sources. Beyond historical loss information, CECL also requires models to incorporate new data items such as life of loan and macroeconomic scenarios. The overarching goal of evaluating the data inputs is to ensure that the data itself and all data treatments are appropriate for the intended use of each data field. To that end, DCG has a specific data input evaluation framework that will ensure your data is unbiased and fit for use.

Model Methodology

As previously mentioned, DCG has developed a series of CECL test models using each of the common CECL methodologies for the purpose of mastering the nuances of each modeling approach, especially as they pertain to compliance with CECL accounting principles. As such, we are uniquely experienced to evaluate the mathematical and computational accuracy of each CECL model type. In our experience, the common CECL approaches can be organized by the categories shown below:

 

DCG understands the complexities and challenges related to each approach. Regardless of the combination of methodologies that your CECL model employs, our validation will provide you with thorough examination of the conceptual soundness of your methodology and the accuracy of your implementation.

DCG is familiar with most of the third-party CECL solutions, and has worked successfully together with several prominent third-party providers in past validations. For models where there is limited code disclosure, DCG also has several approaches to addressing the validation of model methodology in lieu of a full code review.

Model Testing

Model performance evaluation is one of the most challenging parts of the CECL validation, because the only way to truly test a CECL model is to run it over a recession and compare the model projections with what actually happened. Absent an actual recession, DCG has developed a three-pronged approach for tackling model testing, including unique sensitivity tests, benchmark tests, and challenger models.


Contact us to learn more or schedule a meeting with DCG’s experts. We look forward to the opportunity to provide you with our best-in-class CECL model validation services.