CECL (Current Expected Credit Loss), the new GAAP allowance standard (ASC 326), went live for most lenders as of January 1, 2023. While most lenders won’t need to have full disclosures drafted until their fiscal year-end, the first quarter-end on March 31 means that many of these institutions will be reporting their first CECL number in their first quarter call reports. If your first CECL allowance result is due in a little over a month, it’s a good time to take stock of your process and make sure there isn’t anything you’re missing before that first official submission. The below checklist is our go-to when making sure that nothing’s been left out of the process.
1. Model Validation
It’s important to make sure that your CECL model is validated. As your model is new by definition of the changing standard, it will require a third-party validation. The requirements from SR Letter 11-7 are that the validation be performed periodically by someone independent from the model’s users/developers and who has the necessary expertise in CECL modeling to effectively challenge those users/developers. This effectively means selecting an independent team with experience in both building and utilizing these CECL models so that they can test for soundness and sensitivity. Experience here can be thin on the ground, so reach out to a third party if you haven’t found a validator yet – the year-end audit is not far off.
In terms of the periodic review requirement, we’ve seen institutions handle this in two ways. The first and most conservative method is to simply enact an annual validation requirement in the governance structure – this approach reassures stakeholders due to its regularity, but it can be too pricy to pass the cost/benefit test for many institutions.
In most cases, it is common to see validation upon implementation and then revalidation only when a significant change is made to the model. This could include a change in the historical data set, a change in the macroeconomic variables used for scenario conditioning, or general changes to inputs and calculations.
2. Prepayments
With the allowance calculation moving from a 12-month duration to a lifetime calculation, it’s important to understand how your chosen solution handles prepayment. While this may not have received much attention under Allowances for Loan and Lease Losses (ALLL), it can have a large impact on final CECL numbers. It’s important to hold reserve for your longer-term instruments, such as mortgages, to the behavioral lifetime rather than the contractual lifetime, otherwise you could be over-reserved. There’s a notable impact between holding reserve for the full duration of a 30-year mortgage and holding it for the expected behavioral life of 7-10 years (depending on your lending environment). This is one area where it makes sense to do a look-back analysis on your portfolio to make sure your portfolio’s prepayment experience is being reflected in the duration used in your CECL calculation.
3. Qualitative Adjustments
Many institutions are split on what to do with qualitative adjustments after adopting CECL. While some are dropping their frameworks because they have adopted more quantitative models for CECL, others are maintaining their frameworks on top of their modeled CECL results. The decision is largely driven by risk appetite / competitive pressure – more conservative institutions that dominate their geography tend to favor keeping their framework and stacking it onto their CECL results, whereas firms that need to stand out in crowded markets may need to be more liberal in their risk appetite.
When moving your framework forward to CECL, it makes sense to keep certain portions, such as lending experience or concentration risk. However, make sure that you do not include economic factors going forward. This was a common factor under ALLL, but under CECL economic forecasting is built into the model directly. Keeping an economic factor on top of that can open you to double-counting criticisms. Similarly, if you had a duration factor for your longer termed loans, that should also be dropped now because CECL is a lifetime measure.
4. Off-Balance Sheet Reserves
Calculating an off-balance sheet reserve is a requirement under CECL. Under ALLL, it may not have been done, or it may have been a small estimate (often a few hundred thousand dollars) that wasn’t truly quantitative and was rarely updated. Under the new standard, off-balance sheet allowance needs to be calculated when your institution has a commitment that must be funded on demand from the borrower against the amount of available draw that you as management reasonably expect to be drawn over the lifetime of the facility. Exceptions are when your commitments are unconditionally cancellable, meaning you have the option not to fund a draw request. The most common example here is credit cards, and sometimes unsecured consumer lines depending on how your institution writes the terms and conditions. Outside of credit cards, it’s worth doing a legal review to see which other line of credit products at your institution are unconditionally cancellable, thus lowering your off-balance sheet reserve.
As an important note, please remember that off-balance sheet reserve belongs under liabilities, not assets. As straightforward as that sounds, it’s common to present your on- and off-balance sheet CECL allowances together as a single total ACL when reporting on impact and quarter-over-quarter changes. That’s fine, but don’t forget to break them apart when booking the reserves.
5. Available-for-Sale Securities
While not a part of the CECL standard itself, the Financial Accounting Standards Board also updated its rules for available-for-sale (AFS) securities. A credit loss calculation is now required not just for any held-to-maturity (HTM) securities but also for AFS securities to determine if part of an instrument’s loss of value is credit related. This can be tricky, as it will require a comparison between an allowance calculation and market pricing, and the standard calls for the use of a discounted cash flow for the calculation. This means that on top of the new reserve calculation for your securities, you need to source probability of default (PD) and loss given default (LGD) risk metrics for each investment security. This can be a headache, especially since oftentimes the calculation method and source of risk metrics will not align with your loan portfolio process.
The good news is that there is no requirement for this calculation on securities that are backed by the U.S. government. For treasuries, Fannie/Freddie bonds, or other asset-backed securities linked to the federal government, you can skip this step and say that the allowance is zero. This can be a time saver, especially as several institutions hold only U.S.-backed securities in their investment portfolio. As a word of caution here, it’s important to talk through this with your treasury team in the likely event that they have not been involved with your CECL process. Any future positions they trade into without government backing will require this calculation – including municipals and highly rated corporates.
6. Data Cleanup
As we all know by now, the most painful part of the CECL process has been getting data ready for calculation. The more complex modeling and calculations have required new electronic data on a per-loan basis, some of which have not traditionally received high scrutiny during the data entry process. While getting data cleaned up and ready was part of everyone’s project to set up their CECL process, the truth is that some of these issues will require ongoing attention.
The most common issues have cropped up around maturity dates. We all know CECL is a lifetime measure, but what happens to a loan’s calculation when the maturity date is in the past? Whether these loans are just waiting on close-out procedures or actively being worked out, your chosen model may cause an error when trying to determine remaining life with past dated maturities. It’s always a good idea to make a manual check of maturities part of your quarterly process. Not only can you avoid a foreseeable error here, but you can also check for overly long termed loans. It’s a less common issue, but sometimes data errors can push your loans into comical territory. For instance, evergreen lines of credit, with no contractual maturity, may have had their maturity set to 100 years as a core system workaround, which doesn’t play nice with a CECL calculation.
Other common issues involve loan characteristics that haven’t traditionally been checked before they’re entered in the core system. Industry for commercial loans is one to watch for, as it may be a segmentation factor under your CECL framework, or it may drive your model’s results. Between standard industrial classification (SIC) codes and three different North American Industry Classification System (NAICS) code lists (2012, 2017, 2022), this is a common area of confusion.
Loan-to-value (LTV) percentages often drive commercial real estate results, and there is typically a great deal of grief for institutions here. Standards in data entry may not be the same between different commercial real estate (CRE) segments, and it is common to see current and original LTV values kept in the same core system field depending on which group of lenders has originated the loan. Also, due to the difficulty in parsing multiple types of collateral for a given loan, the LTV and collateral values for each loan may only show the primary property. In cases where additional properties or non-real-estate collateral was integral to the credit decision, you can find yourself over-reserving by only using the primary collateral.
Additionally, FICO has presented similar confusion on the consumer side – core systems will often have the same “current versus origination” FICO value issues. It’s also common to see a wide variation in FICO data quality between different consumer products – each lending team will often have its own procedures that may not align with the data needed for CECL calculation. These loan characteristic issues have been prevalent, and it’s important to add a check on their ongoing data quality. We’ve seen repeatedly that even after a data cleanup project, data entry on new loans can fall into old habits and cause continuing issues. The best practice will be to not only amend your data entry procedures (which no doubt you already have) but to also verify that the new originations are meeting the new quality standards going forward.
7. Governance/Documentation
It’s also important to ensure that your documentation is up to date. In addition to changing from an ALLL policy to a CECL policy, you’ll want to make sure that your procedures are updated as well. Whether it’s data preparation, calculation performance, or testing, auditors will be looking to see documentation and support of the new processes and models. Bear in mind that they will pay particular attention to committee minutes. Whether you use an allowance committee, credit committee, or other function for your quarterly allowance, stakeholders will be looking to see that there is broad, minuted agreement from senior management about the assumptions and model used to calculate your CECL reserve. Finally, at year-end you’ll need to update your disclosures – make sure that you plan and draft in advance and can explain the modeling and economic decisions you made as an institution.
While the above points aren’t exhaustive, we’ve found over the past several years that they drive the right kind of questions, making sure that your institution has taken a holistic CECL approach and hasn’t left anything out. Even though we’re at the “eleventh hour” for producing the results themselves, many firms still need to select an independent validator or need help drafting disclosures and documentation before year-end. The right advisor will have a wide range of experience with CECL and will be able to solve any issues that arise in order to get you across the finish line.