CECL Implementation: Eight Takeaways
For several years, Accounting Standards Update 2016-13, Financial Instruments—Credit Losses (Topic 326), has been considered the most significant upcoming accounting standard for banks and other financial institutions. Topic 326, more commonly referred to as the CECL standard, was adopted on January 1, 2020, by more than 150 SEC issuers. BKD investigated adoption statistics for 116 financial institutions with less than $50 billion in assets that adopted CECL and identified certain trends that can assist your financial institution in its CECL adoption plan. Unless otherwise indicated, the total number of institutions evaluated within the steps below is 116. Eight relevant takeaways identified are described below.
The below graphs compile the loan loss reserve changes as a percentage of loans from December 31, 2019, to March 31, 2020, for CECL adopters between $5 billion and $50 billion in assets and less than $5 billion in assets. In the below charts, the x-axis represents each financial institution by asset size and the y-axis represents changes in loan loss reserves as percentage of loans from December 31, 2019, to March 31, 2020. This data illustrates a wider variety of results in larger and more complex institutions over the comparable period. This is primarily related to larger institutions being held to a higher level of scrutiny in the estimation process by auditors and regulators. In addition, based on our observations, larger institutions, specifically those with DFAST experience, tend to use statistical credit loss models that are less reliant on qualitative factor adjustments and can lead to more volatility and variation.
Based on our understanding of institutions that will adopt CECL on January 1, 2023, these institutions’ observations will align more closely with the group less than $5 billion in assets. In preparing for CECL adoption, understanding the effects experienced by the first adopters of CECL can be insightful and help inform how your institution moves forward with the adoption process, especially regarding the effect of acquired loans and economic trends.
1. Acquisitions Were a Significant Driver in Increased Reserves
When considering the 10 CECL adopters less than $50 billion in assets as of March 31, 2020, with the most significant increase in reserves as a percentage of loans, eight had an acquisition during 2019 and one had acquisitions in 2018 and 2017. For the majority of these institutions, acquired loans were a significant portion of their portfolio at CECL adoption. This effect is expected, as accounting for credit losses on acquired loans has materially changed as part of the CECL standard. Historically, purchased loans fell under separate guidance that didn’t allow for the recognition of an allowance for credit losses at acquisition. Under the CECL standard, an allowance for credit losses is to be recorded on purchased loans. As noted in the chart below, prior to CECL the banks with the top 10 allowance increases at adoption had a significantly lower allowance as a percentage of loans compared to the other banks that adopted due to the large amounts of purchased loans with no allowance. After adopting CECL, these 10 banks now have an allowance that’s comparable, and even slightly higher, than the other banks that adopted. For more information on accounting for purchased loans under the CECL standard, see this BKD webinar on the topic. If your institution is anticipating an acquisition in the coming years or you expect to have a large amount of acquired loans at the date of adoption, reach out to your BKD Trusted Advisor™ to help ensure you understand the accounting implications on day one of CECL adoption.
2. Unfunded Commitments Had a Significant Effect at Adoption
Another effect of adopting the CECL standard was an overall increase in allowance for unfunded commitments. With the adoption of CECL, increases on unfunded commitments were expected. Of the institutions considered, 21 percent actually experienced a more significant effect from unfunded commitments at adoption than from funded loans. Further, 43 percent of the institutions could attribute 20 percent or more of total CECL allowance increases at adoption to unfunded commitments. Much of this effect is due to the fact that many institutions didn’t previously record allowance for unfunded commitments. CECL more clearly defined the approach and requirements to record allowance for unfunded commitments that aren’t unconditionally cancelable. It’s important your institution understands the potential effect of unfunded commitments at the time of CECL adoption. See this BKD webinar outlining the effect of unfunded commitments on CECL adoption. See below for unfunded commitment information based on total assets of adopters.
Institutions with < $5B Total Assets
Institutions with $5B–$50B Total Assets
3. Smaller Institutions Were More Likely to Delay CECL Implementation
The Coronavirus Aid, Relief, and Economic Security Act allowed for a delay of CECL implementation through the earlier of the termination of the COVID-19 national emergency or December 31, 2020, with the 2021 Consolidated Appropriations Act extending the allowance of a delay through the end of the emergency or January 1, 2022. Of the publicly traded institutions with less than $50 billion in assets we reviewed, 27 percent elected to delay CECL adoption in Q1 2020. However, smaller institutions were significantly more likely to delay. Of the 45 institutions that delayed the adoption of the standard, 42 were less than $10 billion in assets on the date of adoption. In total, 43 percent of institutions at $10 billion or less delayed adoption of the standard. Just 4 percent of institutions between $10 billion and $50 billion delayed CECL adoption, and no institutions larger than $20 billion delayed implementation. There are several factors surrounding the decision on whether or not to delay. BKD surveyed several public institutions surrounding their rationale for delaying or not delaying. Common responses from those that delayed included lack of comfort regarding the forecasting component specifically with COVID-19 and the desire to perform more stress testing in addition to general time restraints. Those that opted to adopt CECL pointed to additional costs of running parallel CECL calculations and the ability to separately quantify CECL’s adoption effect from COVID-19 effects. Larger institutions with DFAST experience tended to have more comfort with forecasting techniques for credit losses. A common takeaway from both adopters and those that delayed was CECL adoption is a challenging process.
4. CECL Adopters Increased Reserves in Q1 2020 More Than Those That Delayed Filing
As noted above, there were more than 40 institutions less than $10 billion in assets that delayed CECL adoption and, therefore, were still using the incurred loss model as of March 31, 2020. In comparing institutions under the incurred loss model and those that adopted CECL for the three-month period ended March 31, 2020, CECL adopters had larger increases in reserves (in comparison to the three-month period ended December 31, 2019). For CECL adopters, reserves as a percentage of loans increased on average by approximately 61 percent over the same period. In comparison, institutions that delayed CECL implementation saw loan loss reserves increase on average by approximately 16 percent. The chart below outlines the change in the allowance for loan losses as a percentage of loans from December 31, 2019, to March 31, 2020, for public filers less than $10 billion that adopted the CECL standard in comparison to public filers less than $10 billion that delayed adoption of the CECL standard.
As previously noted, a portion of this relates to the effect of transitioning from an incurred probable loss model to a lifetime loss model. Of the 116 banks we reviewed, only nine of them had a decrease in the allowance for credit losses related to loans at adoption. Through review of Form 10-Q filings, many banks cited duration as a driver of effect at adoption. Typically, the longer the duration of portfolio, the larger the potential for increases in allowance at adoption. In addition, as noted in item 2 above, the change in accounting for acquired loans played a significant part in increasing allowance for many CECL adopters. However, adoption date increases didn’t tell the whole story of Q1 increases in allowance for loan and lease losses. Of the 61 percent increase noted above, approximately 30 percent of it related to adoption date and the remainder related to COVID-19 effects. See item 8 below for a further discussion on COVID-19’s effect on CECL.
Institutions Sorted Ascending by Percent Difference
5. Obtaining Relevant & Accurate Loss History Data Was a Challenge During CECL Adoption
Based on discussions and consultations with various CECL adopters, a significant barrier for many institutions during the CECL adoption process was obtaining accurate historical loss data. Although not required, most CECL adopters felt they needed loss history over an entire business/credit cycle, and for many institutions, that dated back to 2007. The ability to have lifetime losses in both expansions and retractions in a business cycle is a beneficial input to understanding the long-term, or cyclical, historical loss experience. However, most adopters found it challenging to obtain that level of loss history.
This led to many institutions using peer data, as a proxy, as part of their CECL model. Use of peer data, as long as appropriate documentation is maintained, can be a viable option to overcoming a lack of historical data until a more complete loss history is compiled.
When using peer data, financial institutions should consider the need for a qualitative factor to account for the risk of using such data in their calculation. Navigating a lack of historical data is another hurdle in the CECL implementation process. Serving more than 1,800 financial institutions nationally, BKD can help.
6. PD/LGD Methodologies Were Most Common in CECL Adopters Below $50 Billion in Assets
When comparing methodology types, BKD used Q2 2020 data, as more robust model disclosure data was available in Q2 in comparison to Q1. There are several acceptable methodologies in estimating CECL reserves, and there’s no specific methodology required by the CECL standard. Of the 83 banks with assets less than $50 billion that disclosed their methodologies as of June 30, 2020, the probability of default/loss given default (PD/LGD) was the most common methodology used, and 47 of the 83 institutions disclosed the use of PD/LGD in some fashion. Of those 47 institutions, 27 disclosed only PD/LGD as their CECL methodology. The next most popular methodology selected was discounted cash flow (DCF), which was used by 29 institutions. Of the 83 institutions, 58 disclosed one methodology, while 25 disclosed a combination.
Overall, a common theme is much more diversity in methodology in CECL compared to incurred. For 2023 adopters, we expect a shift to more simplistic loss methods but still expect larger diversity in methods, including use of PD/LGD and DCF. For smaller and less complex institutions, the weighted-average remaining maturity (WARM) model has been identified by regulators as a simpler option. Four of the early CECL adopters used the WARM model as part of their overall CECL calculation, and one used the WARM model as its sole CECL model. See this BKD Thoughtware® article where we examine the pros and cons of the WARM model.
7. Most Forecast Periods Used Were Either One Year or Two Years
The CECL standard requires a “reasonable and supportable forecast” and is one of the most significant assumptions included within the standard. Based on disclosure information as of June 30, 2020 (most recently available at the time this article was drafted), for CECL adopters with less than $50 billion in assets, 68 used either one year (39 filers) or two years (29 filers) as the length of time the institution could support a reasonable and supportable forecast. The below chart outlines the forecast periods identified by public filers with less than $50 billion in assets. The “other” category included several lengths of time, including the shortest period noted (6 to 12 months), the longest period noted (two to five years), and several ranges spanning a number of years in the one-to-five-year range. In our experience, selecting a forecast length isn’t the real challenge. Instead, determining the factors management considers in setting the length and potentially updating their forecast period poses the bigger challenge.
8. The CECL Adoption Resulted in Increases to Reserves & Provision Expense, but COVID-19’s Economic Effect Also Played a Significant Role
As illustrated above, CECL adoption resulted in significant increases to allowance reserves at adoption. As expected post-adoption, provision expense has increased for community banks under both the CECL model and the incurred loss model as a result of COVID-19’s economic effects. However, as noted below, the effect for CECL adopters was more significant on average than for incurred loss banks. In our experience, a portion of this effect related to numerous CECL adopters using statistical regression-based models that incorporate economic variable forecasts into the quantitative components. Given forecasted economic variables were being significantly affected by COVID-19, this caused quantitative model outputs to increase. On the other hand, their incurred loss counterparts quantitative model inputs weren’t being significantly affected, i.e., charge-offs and risk grading, which caused them to be more reliant on Q factor adjustments.
Using data from 339 community banks (both public and nonpublic) between $2 billion and $50 billion in assets, we noted the following:
- For the 132 (including nonpublic) banks that adopted CECL at January 1, 2020, reserves as a percentage of loans increased from an average of 78 basis points as of December 31, 2019, to an average of 126 basis points at March 31, 2020, and ended with an average of 136 basis points for the quarter ended June 30, 2020. This calculates to an overall increase of approximately 74 percent when comparing reserves as a percentage of loans at June 30, 2020, to December 31, 2019, for CECL adopters, both public and nonpublic.
- For those using the incurred loss methodology, reserves as a percentage of gross loans went from an average of 104 basis points to an average of 113 basis points, ending with an average of 115 basis points over the same periods noted above, with an overall increase of an average of approximately 11 percent. Of the 339 community banks identified above, there were 207 banks within this group considered that calculated reserves under incurred methodology throughout these six months.
- Similar trends were noted in provision expense as a percentage of loans. For CECL adopters, the average provision as a percentage of loans was four basis points for the quarter ended December 31, 2019, 32 basis points for the quarter ended March 31, 2020, and 26 basis points for the quarter ended June 30, 2020. The institutions under the incurred loss model experienced increased provisions of four, 14, and 16 basis points on average throughout the referenced period.
This information is illustrated in the chart below.
There are many relevant trends and conclusions of early CECL adopters thus far, and the standard’s landscape continues to evolve. Each financial institution is unique and must consider its individual characteristics when developing and refining its CECL model. Consult with your trusted advisor to formulate the model that fits your institution best. For more information on CECL, please visit BKD’s CECL Resource Center or submit the Contact Us form below.