Skip to Main Content

Approaching CECL and the COVID-19 Crisis from a Model Governance Perspective

In addition to the primary challenge of responding to changes in loan performance stemming from the impact of COVID-19, banks are also faced with changes in how they employ models to estimate credit losses under the CECL standard. Against the backdrop of multi-billion dollar increases to loan loss reserves in recent quarters, it’s a great time to explore the background behind the CECL standard, how COVID-19 has disrupted its implementation, and the role sound model risk management and governance can play to minimize the impact of this systemic crisis on models.

What Are Current Expected Credit Loss (CECL) Models?

In June 2016, the Federal Accounting Standards Board (FASB) updated the accounting standards related to the estimation of credit losses. ASU 2016-13, commonly referred to as the Current Expected Credit Loss (CECL) standard, triggered the need to develop and validate new credit loss models.

CECL changes the allowance for loan losses from an incurred loss model, under which probable losses are incorporated into the allowance, to the expected loss model, a forward-looking approach in which the allowance includes an estimate of losses over the remaining life of the asset. Consequently, behavioral models that estimate prepayment, default, and recovery have become more important and impactful under the new accounting standard. Such models are sensitive to macroeconomic variables (such as interest rates, unemployment rates, etc.) and, of course, to drivers related to the borrower’s creditworthiness. Regarding the modeling approach, the new accounting standard is not prescriptive regarding the methodology and allows for a range of approaches.

In addition to the new standard, another complicating factor in the implementation of CECL is the range of model risk management practice and sophistication among banking institutions, which is mostly driven by the size and complexity of the institution. The larger institutions were the first ones that needed to comply with the new Supervisory Guidance on Model Risk Management (SR-11-7/OCC 2011-12). Later, as these best practices were applied by the FDIC to smaller institutions though FIL-22-2017, there was a narrowing of this wide range.

Broadly speaking, regulatory agencies have emphasized that banks should tailor the scope of their model risk management practices to the size and complexity of their businesses. As such, many differences in approach remain, leading to difficulties in implementing CECL. However, even in the face of these differences (some necessary, others not), the key to a successful implementation and ongoing support of a firm’s CECL models is a strong model risk management framework.  

How COVID-19 has Disrupted the Implementation of CECL Models

As alluded to earlier, implementing CECL across all credit-sensitive assets is a difficult exercise even at the largest institutions, as it requires melding risk modeling activities with economic forecasting groups and finance activities. This challenge is even more acute at smaller institutions where modeling resources are already stretched thin. Even employing vendor models at these institutions requires competent staff to incorporate the models into banking systems, modeling teams to review and in many cases maintain those implementations, and other personnel to take the time to review and understand the projected credit losses. COVID-19 has made all of this even more difficult.

The answer from policy makers under the Coronavirus Aid, Relief, and Economic Security (CARES) act signed into law on March 27 was to allow banks the option to delay the implementation of the CECL standards until December 31, 2020 or the end of the national emergency – whichever comes first. In addition, on March 27, 2020, the Office of the Comptroller of the Currency (OCC), Federal Reserve and Federal Deposit Insurance Corporation (FDIC) released a joint regulatory capital rule providing a new option for phasing in the impacts of CECL into regulatory capital over the next five years, including a full offset of two years to estimate the impact of CECL versus the incurred loss approach. This impact analysis is something many in the industry have been calling for since the new standard came into being in 2016.

COVID-19, CECL, and MRM: A Recipe for Better Model Governance

With CECL delayed but still a perennial priority for financial institutions, a well-developed model risk management (MRM) framework will play an integral part in coordinating efforts to minimize the impacts of the pandemic on new and existing models. In particular, the pandemic may have already “broken” some CECL models, while others may require significant adjustments. Conceptually sound fixes and adjustments will need to be rooted in sound model risk management practice, which should be embedded throughout the organization. This will make the development or redevelopment exercises less time consuming and the finished product more robust.

More than anything, the COVID-19 crisis poses a data availability challenge, as many creditworthiness variables, macroeconomic series, and risk factors have experienced what is statistically known as a structural break. That is, the trajectory of these variables, leading up to the crisis, are no longer useful (or as useful) in predicting the likely trajectory after the crisis. This has made reasonable scenario analysis and the assignment of probabilities to scenarios even more challenging.

As a simple example, in a pre-vaccine pandemic world, what is the most likely economic scenario over the reasonable and supportable forecast horizon for a given credit portfolio? An even more difficult question is, has the reasonable and supportable forecast horizon for all our credit portfolios changed? Another challenge presented by these structural breaks is the inability for many models to handle the current data. In many cases, the data causes an error. A more insidious problem is when models continue to run, but there are hidden problems caused by the data, such as forecasting using data beyond the range of calibration.  

A strong model risk management framework will help CECL model validators and developers, business leaders, and risk management utilize the delay to assess existing scenarios and make appropriate adjustments, likely employing some expert judgement to augment the development process.

Regarding the modeling approach, as was mentioned above, there is flexibility since the new CECL standards are not specific about which model methodology should be adopted. The delay will give institutions more time to review their current modeling approaches, analyze any deficiencies highlighted or brought about by the pandemic, and begin planning for CECL model enhancements or redevelopment. In the most difficult cases where a model has been deemed unfit for purpose due to one or more reasons, a strong model risk management framework is even more important, as it provides the exception process for ongoing model use, the oversight of any model overrides, and the ability to facilitate safe and timely rebuild and replacement.

It is also important to remember that models are used for specific portfolios and/or products. With that in mind, developers, validators, and users of the models need to understand the risk factors driving the model, their impact on those targeted portfolios/products, and whether the CECL model is performing as expected. The model monitoring and performance reporting framework provides an opportunity to review and challenge the current risk drivers and begin the exploration of new drivers. For models under development, this review process may also alleviate wasteful rebuilds.

Where to Turn for Model Validation and Model Risk Management Advice

No matter the size of the institution or the sophistication of its CECL model development efforts, a sound MRM framework can help guide and manage short-term strategies when models break due to the pandemic, including the use of model overlays and model exceptions. Furthermore, it will also guide longer term model (re)development, performance monitoring, data acquisition, and model validation strategies.  This does not make the implementation process easy, but it does make it easier and may spare the institution from costly and time-consuming model rebuilds as well as regulatory criticism.

Whether you need help building out your MRM framework or getting your models back up to speed through validation and testing, you can reach out to the RMA Model Validation Consortium (MVC) for help. Contact us today to schedule a discovery call to assess your institution’s challenges and needs. 

Kevin Oden

As the Managing Director of the RMA Model Validation Consortium, Kevin is passionate about providing high-quality model validation services at a competitive price point for RMA member banks. Kevin holds a Ph.D. in math from UCLA and was a leader in risk management and model validation for Wells Fargo Bank.



Kutteh, William J

Will Kutteh is the Senior Manager of Risk Products at RMA, overseeing the RMA Model Validation Consortium and Dual Risk Rating solution. Will has over five years of experience in the financial services industry, with a focus on Enterprise Risk and Model Risk Management. He has performed validations of models used in liquidity measurement, credit analysis, interest rate risk management, CCAR stress testing, and CECL. Will holds a B.A. in economics from Wake Forest University.