Jeff Prelle, Head of Risk Modeling at MountainView, a Situs company, served as conference chair at the CECL 2018 Congress Conference.
When it comes to Current Expected Credit Loss (CECL) planning, modeling and implementation, the consensus among financial institutions is that there is no consensus, because no two financial institutions are alike. At Situs, we often reinforce the importance of taking a strategic approach to CECL modeling and implementation. After all, CECL’s tentacles touch a wide range of functions within a financial institution, from model and credit risk to accounting and information technology (IT). But when it comes to achieving CECL milestones and tactics, what will work for one bank, credit union or lender may not work for another.
This was the resounding theme of the 2018 Center for Financial Professionals (CFP) CECL 2018 Congress Conference that Situs sponsored in October. As a sponsor, conference chair and panel moderator, I had the unique opportunity to listen to a range of perspectives on key obstacles facing institutions regarding CECL. Reflecting on the event, it occurred to me just how difficult it was for institutions to discuss CECL in a way that resulted in fervent agreement and nodding. It was not that the information and content lacked value or substance; it was simply that the insights provided by its participants offered a constant stream of new information and unique perspectives. Participant enthusiasm was found in notetaking rather than nodding.
While broad, common themes and goals, such as regulatory compliance and accounting accuracy, were discussed during the conference, these broad themes took a back seat to the more nuanced and institution-specific obstacles. Those obstacles, when overlooked, may leave an institution vulnerable to time-consuming setbacks or costly errors. In response to those obstacles, I have outlined a few key recommendations to help institutions avoid turning their CECL road map into a CECL roadblock:
(1) Get your data and systems assessed
One thing was clear during CECL 2018 Congress Conference: data is still a concern for many institutions, most notably smaller institutions. For some institutions, it is not just about whether they have “enough” data; it is whether they have enough quality data and a strong enough infrastructure to access, update and manage the data. During the conference, several institutions noted that the lack of confidence in data and data management is leading to model indecision. If you are an institution that can’t seem to make modeling decisions, a data assessment will help dissect your data challenges and give you practical next steps to address those challenges and comply, including best-fit models based on your current data.
(2) Learn from IFRS 9 implementation pitfalls: Parallel-test your models
Institutions overseas have already implemented IFRS 9, which means there are case studies to reference. While IFRS 9 is not a perfect replica of CECL, it offers US institutions valuable lessons from its implementation. For example, some institutions scrambled to get their models in place and did not take appropriate precautions to ensure the models were both reasonable and supportable across various scenarios (up/down scenarios, for instance). For both models and scenarios, implement a full parallel test that will run for at least 6-12 months prior to your “go live” date. Stress your models and scenarios during that time to better understand the results; use this period to establish benchmarks for your losses; determine your model attribution plan for Allowance for Loan and Lease Losses (ALLL); establish your audit plan so that it addresses any questions that arise; and vet your approaches with management. Simply put, make sure everything works. It is worth the extra time and effort.
(3) Make sure your model developers don’t get lost in thought
One can imagine it could be easy for a model developer to get deep into his or her data, models, and scenarios. This is great! However, it would be prudent to ensure your model developer has come up for air a few times to mingle with the various business lines. Here’s an easy way to find out: Does your model developer know the name of the manager for each of the relevant business lines? By consulting with the various business lines, your model developer will have a better understanding of loan and portfolio idiosyncrasies and will have likely ensured that any economic factors applied are relevant to the specific loan type or portfolio. The best line of defense, however, is to get your models validated, which will enable your institution to capture model risk earlier in the process.
(4) Create multiple scenarios
Some financial institutions are getting hung up on the likelihood that a certain scenario will occur, but that is not the best question to ask. When evaluating a scenario, it is not necessarily about the probability that a scenario will occur (there are infinite scenarios), it is the probability of a scenario being better or worse than another scenario. Financial institutions need to stretch their models to get a more reasonable and supportable estimate of loss by creating multiple scenarios. It may seem like a hassle at first, but when it comes time to justify your model decision, you will have hard data to show.
If you find yourself experiencing CECL roadblocks, it is likely due to an institution-specific challenge. To solve for institutions-specific CECL challenges, assess/audit your current approach, identify the trouble spots, deconstruct the problem and break it down into practical steps that can be addressed and logged in your documentation. Always keep in mind: If you do not document the process, these discussions get lost in institutional knowledge. To learn more about assessing, validating and implementing CECL, email Jeff Prelle, Managing Director and Head of Risk Modeling at Situs at firstname.lastname@example.org.