With the economic downturn caused by the coronavirus pandemic, U.S. banks have had to prepare for the likelihood of elevated borrower defaults and have more than doubled their allowances for loan losses. This, on the back of about $115 billion of provisions in the first half of 2020, represented major progress toward absorbing the loan losses likely to result from the economic downturn.

Still, while the pace of provisioning may slow, we believe U.S. banks in aggregate are far from done with provisioning for pandemic-related losses. Our base case projection is that the pandemic will trigger 3% loan losses, forcing banks to provision more than $330 billion cumulatively. The fluidity of the current environment makes forecasting difficult. Still, if our base case proves roughly accurate--which will depend largely on the economy and government support measures--we would expect rated banks to report positive but muted earnings in the next four quarters on average because of the continued need to provision for loan losses.

The allowance levels of banks at the end of the second quarter may indicate that they are more sanguine in their loss expectations than we are, and performance will be highly sensitive to how those expectations evolve. Performance will also vary significantly from bank to bank, depending not only on the quality of their underwriting and loan losses, but also on the sufficiency of their second-quarter allowances and on their ability to absorb losses through prepovision net revenue (PPNR).

For instance, it appears larger banks generally have more aggressively built their allowances and have stronger PPNR than regional and smaller banks and are less likely to have bottom-line losses should loan losses approach our base-case estimate for the industry. We expect our ratings to hinge in part on such factors.

Our Expectations For Pandemic-Related Losses

We have set our base-case forecast for pandemic-related loan losses for Federal Deposit Insurance Corp. (FDIC)-insured banks in aggregate at about half of the 6.3% loss rate the Fed projects for the 33 banks part of the 2020 Dodd-Frank Act Stress Test (DFAST).

We consider this an adverse case and one that would be much less severe than the loan losses from the 2008-2009 financial crisis. Banks reported provisions in excess of 5.5% of their loans in 2008 and 2009 collectively and charged off more than 5% of loans in 2009 and 2010 and another 1.6% in 2011.

We consider our economists' forecast for 5.0% contraction in U.S. GDP this year and a 5.2% recovery next year. We balance that difficult outlook against the government support measures for individuals and businesses, such as the Paycheck Protection Program and expanded unemployment insurance, which have likely so far helped prevent outsized rises in loan defaults.

Chart 1

How Robust Are Allowances For Loan Losses?

In the first half of 2020, the ratio of allowances to loans for rated U.S. banks rose to a median 1.7% (from 1.0% at year-end 2019) and an aggregate 2.7% (from 1.3%). In other words rated banks increased their allowances by an amount equivalent to roughly 1.4% of loans (2.7% less the starting 1.3%).

The aggregate allowance far exceeds the median in part because of differences in loan mixes between some of the largest banks and the regional banks. Some of the large banks tend to have far greater proportions of credit card loans, which have required higher allowances than most other types of loans. For instance, Citigroup Inc. and JPMorgan Chase & Co., both of which have significant credit card loan balances, have some of the highest allowance-to-loan ratios.

Banks set their allowance under the current expected credit losses (CECL) accounting standard basically by estimating their lifetime loan losses under a base-case economic scenario--with projections for variables like GDP growth and unemployment. They then may make further adjustments to reflect the uncertainty of the base-case scenario or the likelihood of a more stressful scenario. The more stressful the projected scenario and the greater the bank's exposure to that scenario, the greater its allowance should be.

Provisions taken in the first half of the year drove about three-quarters of the increase in allowances. One-time adjustments to allowances made on Jan. 1, 2020, for the implementation of CECL --which requires banks to set allowances equal to expected lifetime losses--drove the remainder.

That implies that rated banks believe that lifetime losses on their loans have increased by about 1.1 percentage points because of the pandemic (equal to the 1.4 percentage point total increase in allowances less the portion due to CECL implementation on Jan. 1). (We refer specifically to allowances for funded loans because data on allowances for unfunded loans was not yet consistently available. We expect to gather that data once banks file consolidated regulatory financial reports. Including allowances on unfunded loans would push total allowances modestly higher.)

While allowances rose sharply, a 2.7% aggregate ratio and a 1.1 percentage point increase following the onset of the pandemic would still fall short, if our base-case loss projection of 3% were to be realized. That implies that banks may be expecting lower losses than our 3% base case. Therefore, to better gauge the robustness of allowances and make distinctions between banks, we look at not only allowances to total loans, but also the following measures:

• Allowance by type of loan
• For banks that participated in the Fed's 2020 DFAST, allowance to the Fed's projection of loan losses in its severely adverse scenario
• For banks that did not participate in DFAST, allowance to a projection of loan losses made by applying the Fed's aggregate loss rates in DFAST to each bank's loan mix
• Allowance to the company's projection of loan losses (for banks required to perform a company-run version of DFAST)
• Capital plus the allowance relative to regulatory risk-weighted assets

Looking at allowances relative to DFAST losses suggest large banks may have more aggressively added to their allowances than regional and small banks. Therefore, the 2.7% aggregate allowance for rated banks may exceed the 1.7% median not only because of loan mix (and greater credit card exposure), but also because of greater conservativeness. For instance, at the median, the allowances of DFAST banks (which are the country's largest banks) equated to 40% of DFAST loan losses. The allowance of other rated banks equated to just 30% of their implied DFAST losses (based on their loans mixes).

Some of the large banks were able to absorb major provisions in the first half of the year and remain profitable in part because of especially strong revenue generation in segments of their capital markets businesses.

Some regional and small banks with lower proportional allowance levels likely simply have lower-risk loans, such as First Republic Bank. Others hold significant levels of loans that had been marked as part of acquisitions, meaning those loans may require no allowance. Therefore, the ratios of allowances to total loans (including acquired loans) for those banks, such as TCF Financial, First Horizon, and Truist, may be stronger than they appear.

Table 1

Table 2

Table 3

What Do These Allowance Levels Say About Future Provisions?

As the extent of loan losses becomes clearer, banks will need to set provisions roughly equal to the sum of:

• The excess of pandemic-related losses over current allowances; and
• An amount to replenish their allowance--likely to a ratio of allowances to loans around where it was at Jan. 1, 2020--as losses are absorbed. (See the appendix for a simple calculation to explain this further.)

If we assume that the 22 banks part of DFAST (excluding 11 others with foreign parents) ultimately have losses equal to 50% of the Fed's projection in its severely adverse scenario, those banks would be just shy of halfway finished with their provisions. In fact, these banks reported provisions and Day 1 CECL adjustments in the first half of 2020 equal to about 25% of the Fed's projected DFAST provisions (or half of those provisions assuming 50% of the Fed's losses). Assuming a loss rate equal to 25% of the Fed's projection would mean these banks are essentially finished with provisions.

For banks not part of DFAST--who tend to have lower allowance levels--we believe they have further to go. Applying 50% of DFAST loan loss rates by type of loan to their portfolios implies that they are only about 32% of the way done with provisioning. If we apply 25% of DFAST loan loss rates, these banks would be about 60% complete with provisioning.

We reflect this general weakness compared with larger banks in our ratings. For instance, more than one-third of our ratings on U.S. banks have a negative outlook, and most of these are regional banks.

Table 4

Chart 2

How Well Positioned Are Banks To Absorb Future Provisions?

We believe that most rated banks could generate enough preprovision net revenue to remain profitable over the next four quarters assuming:

• Banks report enough additional provisions to cover 50% of the loan losses in DFAST (half of the Fed's projected amount for DFAST banks and half of the amounts implied by the Fed's aggregate loan loss rates for non-DFAST banks)
• PPNR remains steady at the second-quarter rate over the next four quarters

We use a four-quarter period (through the second quarter of 2021) because CECL should force banks to take provisions as their loss expectations evolve. However, there is a possibility banks could take pandemic-related provisions more gradually and beyond the next four quarters, making losses easier to absorb with earnings. That would occur if their expectations of rising losses build only gradually. Also, a significant portion of the actual charge-offs could occur later than the next four quarters. While we believe most banks would remain profitable for the period as a whole, they could take losses in any given quarter that has a disproportionate share of the remaining needed provisions.

Banks that have already built the largest allowances relative to potential losses will need the least PPNR to cover additional provisions. Banks with the strongest PPNR generation will be able to cover the greatest amount of additional provisions needed.

Although most banks would remain profitable in this scenario, many would have very thin earnings. Furthermore, any dividends they pay and balance sheet growth they experience would weigh on their capital ratios. The Fed's limitation on dividends for banks that were part of the stress test provides some support for capital ratios. The Fed mandated that dividend payouts are not to exceed the average of a bank's net income for the four preceding quarters.

Table 5

Table 6

Underwriting And Loan Exposures Will Be Key Factors

The macroeconomic and bank-specific factors discussed above--mainly economic performance, effectiveness of government support, bank allowances, and PPNR--will not be the only factors that determine performance. Each bank's underwriting quality and exposure to industries and individual borrowers most directly affected by the pandemic will also play major roles.

We believe our ratings reflect distinctions in such factors. In fact, the negative rating actions we have taken since the start of the pandemic have focused largely on companies with significant exposures to pandemic-related industries, such as energy. We expect a combination of company-specific attributes and macroeconomic factors to drive ratings in the coming quarters.

Appendix

The following simple mathematical example explains how a hypothetical bank with the following characteristics would need to take a $60 provision:

• An allowance of $80 at June 30, 2020, that compares to an eventual $100 of pandemic-related losses (80% allowance to pandemic losses)
• An allowance of 1.5% of loans ($40/$2,667) as of Jan. 1, 2019 (after the CECL Day 1 adjustment)
• An allowance of 3.0% of loans ($80/$2,667) as of June 30, 2020, after $40 of provisions are taken in the first half of the year

Absorbing the $100 in losses would push the bank's allowance to negative $20. Therefore, it would likely look to replenish that allowance to $40 (the starting point on Jan. 1), costing it $60 in provisions, to maintain an allowance to loans ratio of 1.5%. In that scenario, the bank would still have 60% of its provisions remaining ($40 taken in the first half of the year and $60 remaining).

Table 7