EFFECTIVE IN 2020 for Securities and Exchange Commission registrant banks (and 2022 for private banks), U.S. generally accepted accounting principles (GAAP) will require banks to accrue for credit losses on loans and other financial assets using the current expected credit loss (CECL) approach. Under CECL, banks that make loans of any type will record provisions for loan losses (PLLs) at loan inception that equal the current expected credit losses on the loans over their entire lives. Subsequently, banks will record PLLs only when the current expected credit losses on loans change. Banks will estimate current expected credit losses based on historical loan performance (for example, loss rates), current economic conditions, and reasonable and supportable forecasts of future economic conditions, when they are able to develop such forecasts.
Banks will record considerably higher PLLs at loan inception under CECL than under the current incurred loss model (ILM), particularly for certain loan types. Under the ILM, a bank that makes a loan records a PLL at inception only for losses that are incurred, probable, and capable of reasonable estimation. At the inception of heterogeneous loan types (for example, commercial loans), which banks evaluate primarily on a loan-by-loan basis, banks typically record minimal if any PLLs under the ILM because of the difficulty of meeting its three conditions for loss accrual at the individual loan level. At the inception of homogeneous loan types, which banks primarily evaluate at the portfolio level, banks typically record PLLs under the ILM and bank regulatory guidance equal to the expected charge-offs of the loans over the next 12 months. This amount can be thought of as bank regulators’ rough proxy for credit losses that currently meet the ILM’s three conditions for loan loss accrual. Depending on the type of homogeneous loan, 12 months may be similar to (for example, credit card loans), somewhat less than (auto loans), or much less than (residential mortgages) the remaining lifetime of the loan. Hence, under CECL, banks will record far larger PLLs at inception for both heterogeneous and long-lived homogeneous loan types than they accrued under the ILM.
Similar in some respects to fair-value accounting, CECL has considerable conceptual attractiveness for both the management and the evaluation of banks. In particular, CECL requires banks to devote attention and resources to estimate their current expected credit losses and to ex post evaluate (for example, back-test) any forecasts they use to make these estimates. These estimates will change as economic conditions change, thereby making banks, bank regulators, and market participants better aware of changing economic conditions and their implications for banks’ loan losses and overall solvency. Such awareness should have beneficial effects on financial stability.
On the other hand, CECL fits uncomfortably in the existing GAAP model for gross loans outstanding and interest revenue, which are based on promised (not expected) principal and interest payments. In this essay, I explain how this uncomfortable fit yields PLLs that are more poorly matched to interest revenue under CECL than under the ILM. I explain how this worse income statement matching yields more procyclical effects on banks’ income and capital under CECL than under the ILM. I describe the conceptually correct approach that eliminates these procyclical effects while keeping CECL. Because accounting standard setters are unlikely to make these changes, I explain an approach that bank regulators could use to approximate the changes.
I also explain how CECL likely will yield procyclical PLLs after loan inception by expanding the set of loans for which banks accrue loan losses over the entire remaining lives of the loans. As a consequence of this expansion, banks likely will record larger PLLs when bad times occur after loan inception under CECL than under the ILM, despite the fact that CECL requires banks to record larger PLLs at loan inception. While there are no obvious conceptually correct ways to modify CECL or other accounting requirements to solve this problem, I provide some guidance regarding how bank regulators can best address it.
Finally, I explain how the larger PLLs recorded at inception and perhaps subsequently under CECL than under the ILM require bank regulators to rethink regulatory capital requirements.
THE ILM VS. CECL
I begin with a brief discussion of the ILM’s three conditions for loan loss accrual. First, loan losses must be “incurred,” which means rooted in the present. Conceptually, a loss is incurred when the borrower’s ability to pay deteriorates unexpectedly after the inception of the loan, so that the deterioration could not have been factored into the loan’s contractual interest rate and other terms. Second, loan losses must be “probable,” which in practice typically is a high probability threshold, such as 70% or 80%. This condition is difficult to satisfy at the individual loan level (that is, for heterogeneous loans) until loans are close to default but is easy to meet at the portfolio level (for homogeneous loans), even at loan inception. Third, loan losses must be capable of reasonable estimation, which means the loss can be estimated with adequate accuracy based on historical data or other evidence. This condition is also much easier to satisfy at the portfolio level than at the individual loan level.
CECL was motivated by two related allegations,1 neither of which should be viewed as a fact. First, it is alleged that the ILM’s three conditions for loan loss accrual delay banks’ recording of PLLs in good times when realized credit losses (for example, loan charge-offs) are low. This allegation makes more sense for heterogeneous loans, for which the three conditions are hard to meet, than for homogeneous loans, for which the conditions are easier to meet. Second, it is alleged that this delay requires banks to accrue excessively large and unexpected PLLs in bad times when realized credit losses rise. This allegation ignores the fact that the ILM’s three conditions also apply in bad times, particularly for loans that are not yet severely delinquent. It also ignores the fact that a lot of information about banks’ loan performance other than PLLs (for example, delinquencies) is available to banks, bank regulators, and market participants.
Regardless of their correctness, CECL addresses these allegations in two primary ways. First, it eliminates the ILM’s probable condition. Under CECL, banks are required to accrue for future loan losses that they expect to occur with any probability, no matter how low. Second, CECL substantially weakens the ILM’s conditions both that losses are incurred and that losses can be reasonably estimated, in part by requiring banks to incorporate reasonable and supportable forecasts of future economic conditions into their estimates of expected credit losses, and in part by requiring the accrual of expected lifetime credit losses for all loans at inception.
As described earlier, under CECL, banks will record much larger PLLs at inception for both heterogeneous loan types and long-lived homogeneous loan types than they accrued under the ILM. While this is the case in both good and bad economic times, it will yield procyclical effects on bank lending primarily in bad times, when banks will be less willing and able to record such PLLs and thus less willing and able to lend.
In addition, I predict that banks likely will record larger PLLs when bad times occur after loan inception under CECL than under the ILM, even though CECL requires banks to record larger PLLs at loan inception. The primary reason is that the lifetime loss rates used under CECL likely will rise more in bad times than the 12-month (for homogeneous loans) or other (for heterogeneous loans) loss rates used for loans that are not currently severely delinquent under the ILM.
Because of the effects on banks’ PLLs both at and after loan initiation, bank regulators and other policymakers will have increased incentives to provide inducements for banks to lend in bad times under CECL than under the ILM. Such inducements could include suspending or weakening CECL, reducing regulatory capital requirements, and injecting capital, among other possibilities.
THE PROBLEM AT LOAN INITIATION
Under U.S. GAAP, banks account for gross loans outstanding and interest revenue as follows. At loan inception, gross loans outstanding equals the amount lent, and the effective interest rate is calculated as the internal rate of return that equates the amount lent to the present value of the promised principal and interest payments. The effective interest rate is higher for credit-riskier loans, holding the promised principal and interest payments constant, because the amount lent is smaller. The potential benefits of writing credit-riskier loans – for example, the (typically quite high) possibility of receiving the promised payments despite the credit risk – is reflected in higher interest revenue over the life of the loans.
Under CECL, the expected costs of writing credit-risky loans are reflected in their entirety in the PLL recorded at loan inception. In contrast, under the ILM, only a portion of these expected costs are reflected in the PLL recorded at loan inception. Hence, there will be worse matching of the interest revenue and PLL under CECL than there is currently under the ILM, not that the matching under the ILM is good.
Relatedly, because gross loans outstanding equals the amount lent at loan inception, under CECL net loans outstanding at loan inception will equal the amount lent less the PLL at inception. Banks thus will write down loans at inception more under CECL than under the ILM, as if they made worse lending decisions. These write-downs will be larger the higher the credit risk of the loans.
THE CONCEPTUALLY CORRECT APPROACH
The conceptually correct solution to the income-matching problem is not to jettison CECL; it is instead to change the accounting for gross loans outstanding and interest revenue. Gross loans outstanding should equal the amount lent plus the PLL at inception, so that net loans outstanding equals the amount lent at inception. This approach would enable banks to record the increased PLL at inception under CECL without reducing their income or their incentive to lend. The effective interest rate should be calculated as the internal rate of return that equates the amount lent to the expected, not promised, principal and interest payments. Interest revenue thus would reflect only the higher expected return on credit-riskier loans, not the benefits of possibly receiving the promised payments. Therefore, there would be no potential for mismatching the costs of making credit-risky loans to these benefits. This approach is similar to that currently required under GAAP for acquired credit-impaired loans (although there is no difference between gross and net loans outstanding for these loans at acquisition).
A MORE-FEASIBLE (BAND-AID) APPROACH
Despite the conceptual correctness of the approach described above, the Financial Accounting Standards Board (FASB) has shown no inclination to change the accounting for gross loans outstanding and interest revenue. Assuming that this remains the case, bank regulators could approximate the conceptually correct approach and eliminate the disincentive for banks to lend in bad times by adding the (after-tax) PLL recorded at loan inception to bank capital. This increment would need to be amortized over the life of loans to match the interest revenue recorded using the effective interest rate calculated based on the promised payments on the loan. Bank regulators would also need to modify capital requirements in the fashion described below.
THE PROBLEM AFTER LOAN INITIATION
The effect of banks’ use of CECL rather than the ILM on their PLLs after loan initiation reflects the net result of three effects. First, CECL requires banks to accrue for lifetime credit losses on all loans, whereas the ILM requires them to accrue for lifetime credit losses only on severely delinquent or otherwise impaired loans. Because fewer loans become severely impaired when moving from bad times to good times, this effect will tend to require larger PLLs under CECL than under the ILM in good times. Likewise, the reverse is true: Because more loans become severely impaired when moving from good times to bad times, this effect will tend to require smaller PLLs under CECL than under the ILM in bad times. This is the effect that critics of the ILM emphasize.
Second, loss rates rise when moving from bad times to good times (and fall when moving from good times to bad times), both for loans that are currently severely delinquent and for loans that are not. These changes in loss rates naturally are larger for losses over the remaining life of loans than for losses over any shorter period; for example, the 12-month horizon over which banks typically accrue for loan losses on homogeneous loans that are not yet severely delinquent is shorter than the remaining life of most types of these loans. Hence, banks’ PLLs will reflect larger decreases in loss rates when moving from bad times to good times (and increases when moving from good times to bad times) under CECL than under the ILM. This effect works in the opposite direction of the first effect. Moreover, this effect is likely to dominate the first effect unless the third effect described below is sufficiently strong.
Third, under CECL, banks are required to incorporate reasonable and supportable forecasts of economic conditions into their PLLs. Historically, the macroeconomic cycle has been moderately predictable, with good times followed by bad times and vice versa, although the periodicity of the cycle has varied. Most other economic cycles that affect loan credit losses, such as for real estate and other asset prices, have been considerably less predictable. Hence, banks have some ability to forecast future good times when times are currently bad, and future bad times when times are currently good. Were this ability sufficiently strong, it would substantially weaken the second effect and render the first effect dominant.
There is no evidence of which I am aware that banks (or other parties, such as bank regulators) have the ability to forecast cycle turns over periods of appreciable length. Even if banks were certain the cycle will turn (a perhaps rosy premise in today’s world), their ability to develop reasonable and supportable forecasts of the timing and strength of these turns is doubtful. Hence, I predict that the second effect will dominate, and banks’ PLLs recorded after loan initiations will be more procyclical under CECL than under the ILM.
It is not clear that there are any good solutions to this problem, which results from the unalterable fact that the world is an uncertain place. Bank regulators and other policymakers can address this problem in better and worse fashions, however. If bank regulators attempt to mitigate the procyclical effects of CECL by suspending or weakening its requirements during bad times, they will increase bank opacity and thereby reduce the confidence of market participants in the stability of the banking system. Preferable approaches would be for bank regulators or the federal government to systematically or in another transparent fashion reduce regulatory capital requirements or inject capital into banks as early as feasible in bad times.
RETHINKING BANK CAPITAL REQUIREMENTS
Bank regulators should adjust regulatory capital requirements to reflect the accounting approaches used, all else being equal. Capital requirements currently are based on the idea that banks reduce capital for recognized credit losses and hold capital against unrecognized credit losses. Assuming we retain the current accounting approach for gross loans outstanding and interest revenue, more credit losses will be recognized cumulatively under CECL than under the ILM, thereby reducing bank capital. Therefore, bank regulators should reduce capital requirements upon banks’ adoption of CECL, unless bank regulators’ goal is to exploit this adoption to increase capital requirements.
On the other hand, in the sadly unlikely event that the FASB adopted the conceptually correct approach to accounting for gross loans outstanding and interest revenue described above, banks’ PLLs recorded at loan inception under CECL or any other approach would have no effect on their capital. Only PLLs recorded after loan inception would affect bank capital. In this event, bank regulators should increase capital requirements.
1 Remarks by John C. Dugan, Comptroller of the Currency before the Institute of International Bankers, “Loan Loss Provisioning and Pro-cyclicality,” March 2, 2009.
About the Author:
Stephen G. Ryan is Professor of Accounting and the Charlotte Lindner MacDowell Faculty Fellow at the Stern School of Business, New York University. His research examines financial reporting by financial institutions and for financial instruments, including banks’ loan-loss reserving and the effects on banks’ financial reporting on financial stability. Ryan served on the Financial Accounting Standards Advisory Council for the period 2000–2003 and on the Federal Reserve Bank of New York’s Financial Advisory Roundtable for the period 2012–2018.