Keys to CECL Validation Success in 2019

CECL ValidationBY MARK JORDAN, FI CONSULTING

At present, financial institution SEC filers are busy building and testing their internal models and processes to calculate reserves under the new current expected credit loss (CECL) accounting standard. The industry is entering a critical phase for teams to validate loss models and aggregate, analyze, and share preliminary results with senior management.

Model validation is a critical activity to support this year’s parallel run. Model risk management teams must provide effective challenge, including CECL-specific requirements. These include reasonable and supportable (R&S) periods, macroeconomic factors, portfolio segmentation, and reversion to the mean. Without confidence in the defensibility of CECL results, the other components of end-to-end production will not succeed.

This post summarizes FI’s view of key model validation challenges in the lead-up to CECL parallel run and offers thoughts on how to best tackle these challenges.

Key Challenges to Validating Models for CECL

Lack of prescriptive methodology. The Financial Accounting Standards Board (FASB) left open for financial institutions themselves to determine how to meet the standard’s principles, such as defining the R&S period for their credit loss estimates. Banks are in the process of converting existing loss models, adding new ones, or outsourcing to third parties, but they do not have a clear gauge of what will meet auditors’ and regulators’ definition of “compliant.” For example, open-ended loans such as credit cards require well-reasoned judgments to determine the “life” of the loan and the allocation of future payments and charge-offs attributable to current balances.

Uncertain effective challenge. There are many stakeholders in the CECL process, including accounting, risk, finance, credit, lines of business, treasury, technology, and data. Assigning accountability to the appropriate risk oversight team to assess the quality of judgments for key requirements of the standard is therefore a consequential decision. Model validators may not be thinking about CECL-specific requirements such as R&S period, mean reversion, and macroeconomic inputs. The overall approach to qualitative adjustments under CECL also needs to be challenged for consistency, soundness, and degree.

Hard deadlines and competing needs. While 2018 brought some regulatory relief with respect to capital stress testing, many banks will continue with their stress testing efforts while also managing CECL parallel run. The challenge, then, is keeping CECL validation efforts on track while simultaneously attending to other validation demands. Regulators are scheduling CECL-specific exams during 2019 and are focused on fundamental questions including R&S period, mean reversion, and segmentation. The urgency and criticality of CECL implementation – on top of business-as-usual activities – create significant strain on the ability of institutions to stay on pace with validation plans.

Keys to CECL Validation Success

Building a defensible estimation of lifetime credit loss depends on a strong foundation of data and models underpinning a reasoned analysis of core drivers of credit risk. Validation activities – designed and executed following industry governance frameworks – will test that foundation. Keys to a successful CECL validation effort include:

Create a two-pronged defense strategy. The principles-based nature of the standard means there are no objective measures of compliance. A consensus for “reasonable and supportable” may emerge only after several cycles of internal audits and examinations. A strategy to gain the confidence of examiners should thus be two-fold. First, be proactive in your engagement with auditors and regulators. Take advantage of the relative openness of the current environment, as all parties want to succeed in meeting the new standard. Use interaction and dialogue to invite examiners and auditors into your thinking and give weight to their feedback in the design of your program. Second, when explaining your thinking, go beyond charts and spreadsheets with your supporting artifacts. Insist that your project teams fully document the rationale for their assumptions, the theoretical approaches, and the calculations used to derive estimates.

Engage an independent view. Your second-line enterprise risk management organization is on point to conduct CECL model validations. But overlying the models are the specific requirements of the standard. Your detailed execution plan for parallel run should spell out exactly who in the organization is responsible for challenging the decisions related to these requirements. We have seen organizations assign resources from other parts of the institution’s oversight function to provide credible challenge for CECL requirements, freeing validators to focus on model validation. Model risk management should adopt an iterative engagement approach with model teams to provide feedback throughout the model development process, which will minimize the risk of model findings pushing the CECL program off-track.

Leverage resources across the organization. Evaluate the feasibility of employing other teams within the bank with quantitative expertise to validate CECL models and credit judgments. Take advantage of the fact that this is a multidisciplinary effort that is already pulling in many key bank functions. This is an undertaking that will impact everyone across the bank, not an isolated “accounting project.” Use the “bully pulpit” to raise awareness of the stakes and enlist the resources you need to help you through the validation phase.

Parting Thoughts

For the longer term, use what you are learning from the validation process to improve your existing model governance framework. Convergence around CECL models and their usage lays the foundation for a bank-wide genealogy of common features, data, and assumptions shared across models. Parallel run testing offers an opportunity to fine-tune governance of CECL models for smooth quarterly disclosure runs with high confidence in estimations in the future.