By Michael Gullette and Josh Stein
This increased procyclicality of CECL is due to the inability of economic forecasters to accurately predict turns in the economy. In fact, based on preliminary modeling of CECL estimates by large banks prior to CECL’s Jan. 1, 2020, effective date, CECL would have increased the procyclicality of loan loss reserves during the financial crisis time period of 2006-2010. In other words, the very “procyclicality problem” that was supposed to be fixed by CECL would likely be exacerbated by it.
The real story: countercyclical capital transition working well
With this in mind, ABA welcomes additional work to study the CECL accounting standard. A December FEDS Notes article takes the first crack at it. Unfortunately, the article—prepared by Federal Reserve staff members—has limitations that can be misinterpreted. Indeed, those who read only the headlines may come away with inappropriate conclusions. Most importantly, the unprecedented conditions of the pandemic made it an awful time to base any review of CECL, and the study’s authors acknowledge that.
So, while the Fed staff “find limited evidence that the impact of CECL on allowances is associated with decreased lending in the pandemic,” we question why someone would ever expect to find any such evidence under the circumstances, especially since interagency regulatory capital transition rules had been put into place precisely to address such concerns. As a reminder, an initial CECL regulatory transition rule allowed adopters effectively to defer for two years the initial regulatory capital impact of CECL allowances, then amortize those initial differences over the succeeding three years. A revised transition rule was then enacted in March 2020 also to allow 25 percent of incremental CECL allowances after the adoption date to likewise be deferred and amortized. This revision addressed the concerns that bankers had pointed out relating to the volatility and potential procyclicality of CECL.
Here, the Fed staff concluded that the transition rule successfully neutralized the aggregate capital impact of CECL on adopters. Neutralizing the capital impact was meant to neutralize the impact on lending, and it appeared to have worked—the Fed should take credit for that. With this in mind, ABA believes that the banking agencies should now consider how such a countercyclical measure could be made permanent. This should be the headlining conclusion of their paper. (As the Bank Policy Institute notes in a 2021 paper, the countercyclical impact of the initial deferral supports the notion that it should be made permanent.)
Provisioning responses to economic outlook: the wrong question
The Fed staff also concluded that loan loss provisioning under CECL “was noticeably more responsive to the dramatic changes in economic outlook that occurred in the COVID-19 pandemic” than incurred loss provisioning. This conclusion appears mainly based on the percentage changes in loan loss allowance coverage ratios (calculated by dividing the allowance related to the respective loan portfolio by the cost basis) recorded between Jan. 1, 2020, and June 30, 2020.
ABA disagrees with the notion that, during the pandemic, CECL caused credit loss estimates to be more responsive to the economic outlook than incurred loss estimates. More on this in a moment—but more importantly, a focus on CECL’s responsiveness to economic forecasts is misguided. Banker concerns related to CECL are not whether the economic outlook should be reflected in credit loss estimates; the concerns relate to whether banks can accurately forecast changes in the economy in the first place. The historical inability to do so among professional forecasters is what causes volatile credit loss provisions, resulting in procyclicality. A study purely of responsiveness to economic outlook misses the point.
Here, it is generally agreed that no one would have foreseen the pandemic nor the various government responses to it. The pandemic simply does not provide a good basis for determining whether banks can predict changes in the economy. However, even when looking at the sensitivity of credit loss estimates to economic outlooks, we question whether the first two quarters of 2020 actually do prove CECL estimates to be more sensitive. While the percentage increases in allowances between Jan. 1, 2020, and June 30, 2020, appear to be higher for CECL banks, it would be premature to conclude that CECL caused the difference.
The analysis omits key differences between CECL banks and non-adopters. Non-CECL adopters had higher overall coverage ratios prior to 2020 than their large bank counterparts, and this was despite historically lower charge-off experience (which would normally support lower allowances). Moreover, CECL allowances are product-sensitive, with consumer lending, which is dominated by the larger CECL banks, being significantly more volatile than commercial or real estate-based lending. In fact, large banks generally have only about a third of their loan portfolios in real estate-secured lending, whereas smaller non-CEC banks generally have two-thirds of their portfolios in real estate. In other words, the increases in non-CECL coverage ratios were based on higher starting points that suppress their percentage increases. With this in mind, the median Q1 2020 allowance coverage ratio for commercial real estate, the mainstay of most community bank businesses, was actually higher for non-CECL banks than for CECL banks, according to the acting chief accountant at the OCC.
This indicates that loss reserving practices at smaller institutions are normally robust, often informally reflecting lifetime loss estimates. Such robust allowances during the pandemic should not be a surprise, however, as banking examiners were quick and persistent in communicating to bankers throughout the pandemic period that credit loss allowances, whether on a CECL or an incurred loss basis, should reflect all risk and all loss content.
As a result, even if a level of responsiveness of allowances to economic outlooks is indicated through the numbers, it may be due more to sophistication and advanced forecasting techniques rather than any steps to comply with a new accounting standard. The high level of qualitative allowances (those not directly based on quantitative modeling) applied during the pandemic by all banks and the resulting wide ranges of allowance coverage ratios between companies make it difficult to assess how economic outlooks actually end up in numerical estimates.
Regardless, when the new accounting standard was being debated by the Financial Accounting Standards Board, the objective was to provide countercyclical credit loss provisioning through allowances built one to two years earlier than incurred loss allowances, not the one to two quarters that the Fed staff is citing. Therefore, a proper test of CECL remains to be performed.
Throughout those FASB debates, FASB members and others acknowledged that no one understood their borrowers better than community bankers. Given the high cost of CECL compliance, therefore, it remains questionable whether the CECL results noted above represent an adequate improvement over current incurred loss practices, especially for those community banks who must bear the cost of reengineering their credit loss estimation systems and processes for 2023 adoption. While the banking agencies have endorsed the Federal Reserve’s “SCALE” method that is meant to streamline CECL practice for community banks, the robust allowances recorded since 2020—most of which are for credit losses that still have yet to materialize—suggest that lifetime loss estimation is the current objective in a practical sense. Prior to the 2023 effective date, banking regulators and auditing firms should publish specific guidance on how community banks can easily comply with the standard while avoiding the costs just cited.
Permanent regulatory capital mitigation should be pursued
ABA’s 2019 discussion paper on the need for a CECL Quantitative Impact study noted two high level objectives a QIS must address:
- Examining how CECL forecasts impact lending throughout typical economic cycles. In addition to testing for overall procyclicality, the study would evaluate: the possible migration of lending from the regulated banking industry to nonbank institutions; the ability of community banks to compete in their markets when subject to CECL; and the effects on consumer lending and lending to small businesses, specifically to low-to-moderate-income borrowers.
- Offering practices and guidance that can potentially mitigate harmful effects of CECL.
Related to the second objective, the Fed study appears to support the notion that a countercyclical regulatory capital mechanism can be effective in avoiding adverse effects on lending. As we’ve discussed above, the banking agencies should pursue a permanent mechanism.
Related to the first objective, testing throughout the economic cycle is needed; we hope such work continues. With that in mind, however, the study’s analysis of the impact of CECL on lending during the pandemic noted a statistically significant decrease in “Other Consumer” loans at CECL banks compared to non-CECL banks. It was the only line of business that exhibited this trend. “Other Consumer” primarily consists of auto, student and installment loans—those often issued to LMI borrowers. Further study is needed, therefore, to assess CECL’s impact on lending to this segment.
Credit loss estimation is complicated. CECL’s lifetime loss objective makes it even more so. Overall, it is good to see the Federal Reserve staff devote time to understanding the effects of CECL. While the staff acknowledge the challenges of studying CECL during the pandemic, readers who go beyond the headlines will understand that this study only creates more questions than it answers. More work is needed.
Michael Gullette is SVP for tax accounting policy at ABA. Josh Stein is VP for tax and accounting policy at ABA.
ABA Viewpoint is the source for analysis, commentary and perspective from the American Bankers Association on the policy issues shaping banking today and into the future. Click here to view all posts in this series.