CECL – Not Your Mom’s ALLL Calculation
CECL (Current Expected Credit Loss) is an ongoing, pervasive topic when I sit down and talk with credit unions. I believe this is partly because many are nervous and unsure of its potential impact, and how to start preparing for it. There is a ton of information floating around on the subject; some accurate, some inaccurate, and some completely erroneous. There are those in the financial industry that are under the false impression that this is only a marginal change and that 2021 is plenty of time to become compliant with the guidance. However, the consensus among the experts is that transitioning from the incurred loss to life of loan estimates is a significant change, requiring financial institutions to have robust, clean and complete data as well as tools and techniques ready to perform complex calculations and do sophisticated analysis.
While the Financial Accounting Standards Board (FASB) is not prescriptive in its guidance as to which methodology should be employed, many credit unions have already decided to use one of three — or a combination of these three — methodologies: Discounted Cash Flows, Default Probability/Loss Given Default or Historic Loss Rate Methodology. Each of these methods require significantly more data than most credit unions have in a centralized location today.
Discounted Cash Flows Methodology
This month, I’ll focus on the discounted cash flows methodology. When looking to apply this method to a mortgage loan portfolio, for example, you need to consider how difficult it would be to assemble the necessary data under your current data architecture. Under discounted cash flows, lenders are expected to forecast future cash flows, then discount the cash flow by estimated losses and adjust the resulting estimate for its present value, as well as for forecasted economic conditions. The output of this will be the contribution to the reserve. Let’s take a closer look at what data is needed to perform this analysis.
At a minimum we would need:
The exclusion of any one of these will result in a flawed expected loss output. Additionally, depending on how you estimate future losses, it might be advantageous to further disaggregate the loan records by more than loan type.
Let me give you a quick example illustrating this:
If in years during the recession, my loan production was heavy in super-prime paper, but since that time I have loosened the underwriting guidelines for less than prime, the expectation would be that my historic loss rates would be low, because the majority of my portfolio is comprised of low risk loans. By projecting a simple historic loss rate onto the current blended portfolio, I am likely grossly understating my estimated losses because my newer originations that have not matured through their peak loss period are much riskier than the previous loans that I have originated. As a result, my reserve could be insufficient to sustain actual future losses. In this case, credit quality might be an appropriate additional level of segmentation to do an effective analysis.
This brings me to a crucial subject area: pool ages.
Question: Which loan do you think has a higher probability of default in the next 30 days – a loan that is only 3 months matured or a loan that is 18 months matured?
Hopefully you answered the loan that is 18 months old. Loan defaults follow a pretty consistent pattern of low, high, low. Here’s an example: 5-year auto loans tend to experience very few losses within the first 12 months of their lives, and experience their highest losses in years 1 and 2 of their lifecycle, then the losses trend down through years 3, 4 and 5. If I’m forecasting losses over the next 12 months, should I apply the same estimated loss rate across the entire portfolio or should my loss rates be sensitive to the varying ages of the loans in the portfolio? This is where static pool analysis becomes essential. In the final guidance, static pool or vintage analysis must be included in your loss estimate formula.
Not Quite Done Yet
Now that the more basic steps of the discounted cash flows method have been completed, you may find that the expected loss output does not agree with what you think your real risk exposure is. You may find it necessary to make further adjustments using additional attributes. Based on an analysis that CU Direct performed in 2015, using aggregated data we’ve collected, it was discovered that loans without co-borrowers were up to 3 times more likely to default. If these results are true at your credit union, you may consider adjusting your estimated losses to include co-borrower as a consideration. How about updated credit scores? Does the positive or negative movement in borrower credit scores change default probabilities at various points in the lifecycle of the loan? These are just a few examples of the types of adjustments one might have to consider, but there are many other loan attributes that can potentially influence loss predictions — and by extension — impact the reserve amount. This is why the more data you have the better. However, data volume by itself is not enough. You need to be able to access it in one central source and most importantly, you must be able to trust its accuracy.
One Last Thing – Documentation
There seems to be a prevalent false sense of security among credit unions that because there is vagueness in some of the language in the final CECL guidance, their documentation for the reserve can also be vague. This simply, is not true. In fact, it is the complete opposite. FASB relaxed the language in the final guidance intentionally to provide financial institutions the flexibility to incorporate modeling that makes sense for their unique circumstances. With this flexibility comes a more onerous requirement for financial institutions to thoroughly document, defend, validate and support their CECL approach. This includes both the quantitative and qualitative components. As long as the math jive, the quantitative component of the estimate will be the easiest to document. The qualitative elements will be significantly more difficult. Consider the macroeconomic conditions in the estimate for real estate loans. Do you expect collateral values to increase or decrease over the next 24 months? Regardless of how you answer that question, if it is a consideration in the estimate, you have to document it and be able to withstand regulatory scrutiny and criticism.
The Good News
Those credit unions using CU Direct’s Lending Insights portfolio analytics are already half-way there. Many of you have gone through the data validation process and are confident that your data is accurate, complete and centralized. You have access to reports that provide loan performance analytics, historic losses by static pools, loss migration analysis, collateral valuation estimates, credit score migration, and more. If you haven’t been in the Lending Insights system for a while, this would be the perfect time to re-familiarize yourself with the tool. Now is the time to start identifying any data or loan types that you are not currently sending us but would like to. Contact firstname.lastname@example.org to set up a quick individual or group training.
Lending Insights will unveil its new CECL suite of products later this year that will meet the needs of credit unions, while helping achieve CECL compliance on day one!