Read this if you are at a financial institution.
This article is part of our series on CECL implementation. You can read previous articles in the CECL series here.
Segments, sub-segments, pools, cohorts—by whatever name you call it, grouping loans (and other financial instruments) for CECL1 is kind of a big deal. Like choosing an inner circle of friends, creating effective loan pools can have a lot of influence over your CECL experience, from methodology decisions to your allowance estimates. As a CECL adopter, you are expected to evaluate, support, and document segmentation choices (no such requirement for your inner circle of friends!), even if you plan to use the same segmentation in place today. To do so successfully, consider these segmentation ABCD’s:
A: Accuracy and completeness of data
The accuracy and completeness of data used to determine the most appropriate segmentation under CECL covers a lot of ground—everything from what information you considered to be relevant and why, to where the data came from and how it was determined to be valid (aka accurate and complete). CECL requires loans sharing similar risk characteristics2 to be pooled together for “collective evaluation”; examples include loans with similar terms and structures, lien position on collateral (e.g. first, or junior lien), or collateral use (e.g. owner-occupied or investment real estate). As a result, “accuracy and completeness” applies not only to the data you rely on to pool loans, but also to what you determined the common risk characteristics to be, why those, what others you identified but ultimately didn’t use, and why. Read our earlier article, CECL Adoption: The five W's of data, for more information on data considerations.
B: Balance between granularity and significance
Striking a balance between how many segments you create and the significance of doing so can be a little like trying to achieve the “just right” goal of Goldilocks. For example, is pooling all your consumer loans together most aligned with your past loss experience, or does the type of collateral also influence your risk of loss? How far is too far (real estate, cars, boats, RV’s, tractors)? At what point does it become difficult to consistently demonstrate or predict meaningful differences in risk of loss for each? Several sections of the standard address this need to balance detail with what is useful3. In this way pools should be small enough that the risk characteristics they share are relevant to estimating inherent risk, but not so small as to be confusing, misleading, or not able to be modeled consistently over time. Being aware of how small a pool is in terms of the number of loans it consistently contains may be one consideration for whether or not the segmentation is too granular.
C: Controls over the selection of risk characteristics
Your segmentation choices will likely have far-reaching effects on other key decisions in your CECL methodology. Model selection, qualitative adjustments, and even if/what/how external or peer data may apply are examples of what could be impacted by your segmentation selection. As a result, and in addition to the above, your auditors and regulators will want to see evidence the risk characteristics driving your segmentation choices were robustly reviewed, challenged, tested, and documented. Further, they will want to see that you have a similar systematic approach in place, and ongoing, to identify when a loan no longer shares the defined risk characteristics of its segment, resulting in its removal from the pool to be assessed individually.4
D: Documentation tips
Documentation is like exercise—you know you should do it, but sometimes you don’t make it a priority. CECL opens the door for all kinds of documentation expectations, so coming up with a way to do this as you work through implementation can save you a lot of headache later. For segmentation, setting up a simple spreadsheet with the ABC’s to the left and columns to the right to list data, testing, key considerations, decisions, approvers, and even links to supporting evidence (data files, governance memos, etc.) is but one example of how you might keep track of these items as you work. Be sure to include any assumptions you had to make along the way (e.g. how you handled missing information on old or purchased loans), or aggregations (larger-level pools than you might have preferred) you accepted and why.
Finally, while you may be checking out what segmentation others in the industry are using—which will vary as it does today—what you’ll want to document most is why the choices you made are right for your institution.
For more tips on documenting your CECL adoption, stay tuned for our next article in the series on documentation. You can also follow Susan Weber on LinkedIn.
No matter what stage of CECL readiness you are in, our Financial Institutions team is here to help you navigate the requirements as efficiently and effectively as possible. If you would like specific answers to questions about your CECL implementation, please visit our Ask the Advisor page to submit your questions.
1Current Expected Credit Loss (CECL) methodology as provided for in the Financial Accounting Standards Board (FASB) Accounting Standards Update (ASU) Financial Instruments-Credit Losses Topic 326, commonly referred to as FASB ASU 326. A copy of the standard is available for download from the FASB website.
2Refer to FASB ASU 326-20-55-5
3Examples include FASB ASU 326-20-50-3, 326-20-55-10, and 326-20-55-11 (for financing receivables)
4Refer to FASB ASU 326-20-30-2