KEY TAKEAWAYS

• U.S. banks reported significant improvements in third-quarter earnings, compared with the prior quarter, as they sharply reduced provisions for credit losses, held allowances roughly flat, and saw significant declines in consumer borrowers deferring loan payments.
• While we are maintaining our base-case forecast for 3% aggregate pandemic-related loans losses, taken mostly in 2020 and 2021--pending further clarity on the economy, additional government stimulus, and vaccine development and distribution--we believe the probability of losses reaching that level have dropped somewhat.
• At that loss rate, we would expect most rated banks to generate positive, but relatively weak, earnings over the next year, an important factor in the stable outlooks S&P Global Ratings has on many banks.
• While lower provisions drove higher earnings in the quarter, low interest rates are weighing on net interest income and preprovision net revenue. Absent a rise in rates, banks would likely exit the pandemic with weaker profitability than they entered it.

U.S. banks' asset quality showed important signs of stabilizing in the third quarter as the economy rebounded briskly. The proportion of consumer loans on deferral fell sharply and the sharp rises in allowances for credit losses that characterized the first half of the year for most banks, stabilized. That allowed banks to report substantial rebounds in earnings from the prior quarter.

Whether the asset quality stability lasts will likely depend on how durable the economic rebound will be through 2021, the extent of further government stimulus, and the success of vaccine development and distribution.

S&P Global Ratings believes there remains a high degree of uncertainty about the evolution of the coronavirus pandemic. Reports that at least two experimental vaccines are highly effective are promising, but this is merely the first step toward a return to social and economic normality; equally critical is the widespread availability of effective immunization, which could come by the middle of next year. We use this assumption in assessing the economic and credit implications associated with the pandemic (see our research here: www.spglobal.com/ratings). As the situation evolves, we will update our assumptions and estimates accordingly.

Until we gain some further clarity, we are maintaining our base-case forecast for the U.S. banking system of 3% pandemic-related credit losses (see "What Lies Ahead For U.S. Bank Provisions For Loan Losses," Aug. 12, 2020). We would expect banks to take those aggregate losses mostly in 2020 and 2021. At the same time, we believe that the probability of losses reaching or exceeding 3% has dropped somewhat. At a 3% loss rate, U.S. banks (rated and unrated) in aggregate would likely only be about 40% finished with pandemic-related provisions, which they would probably take over the next year.

In that scenario, we believe rated banks, many of which would have to at least double the provisions they took in the first three quarters of 2020, would generally remain profitable over the next year, even if profits are weak. We would expect most to report low- to mid-single digit returns on equity. However, an expectation of continued profitability along with restrained capital payouts and solid liquidity owing to continued inflow of deposits, supports our stable outlooks on many banks.

At the same time, banks' ability to absorb additional losses through earning has notably declined as ultralow interest rates drive net interest income down and net interest margins (NIMs) to multi-decade lows, pressuring preprovision net revenue. A loss rate greater than 3% is more likely to cause significant bottom-line losses now than it would have when NIMs were much higher.

At current NIMs, banks are also likely to exit the pandemic with weaker profitability ratios than they entered it. For instance, we estimate the third quarter returns on common equity of banks that experienced declines in net interest income would have been roughly 160 basis points (bps) higher at the median had their net interest income been flat with the year ago period.

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 Federal Reserve projects for the 33 banks included in the June 2020 Dodd-Frank Act Stress Test (DFAST).

A 3% loss rate 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 4.0% contraction in U.S. GDP this year and a 3.9% recovery next year. Whether the government extends further stimulus and the success in developing and distributing vaccines will also likely have material impacts on credit losses. Enhanced unemployment benefits and eviction protections have lapsed for many Americans. The stimulus provided to small- and medium-size businesses through the Paycheck Protection Program (PPP) has also ended. To date, Congressional Democrats and Republicans and President Donald Trump have not agreed on an additional stimulus bill.

We will update our 3% pandemic loss estimate as the path of economic growth and vaccine and stimulus efforts become clearer.

Chart 1

How Robust Are Allowances For Loan Losses?

The ratio of allowances for loan losses to loans for rated banks was 2.8% in aggregate and 1.7% at the median at the end of the third quarter. Those ratios did not change much from the prior quarter, but they were up sharply from 1.3% and 1.0% at yearend 2019.

The aggregate allowance far exceeds the median, in part, because of differences in loan mixes between the largest banks and the regional banks. Some of the large banks tend to have far greater proportions of credit card loans, which require higher allowances than most other types of loans.

While allowances have risen sharply this year, ratios of 2.8% and 1.7% at the aggregate and median would still not be high enough if our base-case loss projection of 3% were realized. That implies banks may be expecting lower losses than our 3% base case. 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); and
• 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 2.8% aggregate allowance for rated banks may exceed the 1.7% median not only because of loan mix (and greater credit card exposure), but because of greater conservatism. At the median, the allowances of DFAST banks (which are the country's largest banks) equated to 40% of DFAST loan losses in the third quarter, little changed from the prior quarter. The allowance of other rated banks equated to just 30% of their implied DFAST losses (based on their loans mixes).

Some regional and small banks with lower proportional allowance levels likely have lower-risk loans, such as First Republic Bank. Others hold a significant amount of loans that were marked as part of acquisitions, meaning those loans may require no allowance. 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.

Chart 2

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 Jan. 1, 2020--as losses are absorbed.

If we assume 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 slightly more than halfway finished with their provisions. (At the end of the second quarter, they were less than halfway finished by this calculation). 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 allowances--we believe they have further to go. Applying 50% of DFAST loan loss rates by type of loan to their portfolios implies they are less than 40% of the way done with provisioning (up from less than a third in the prior quarter). If we apply 25% of DFAST loan loss rates, these banks would be largely done with provisioning. Again, these numbers imply banks generally expect lower losses than we have incorporated in our base-case scenario. A handful of these banks released reserves in the third quarter.

Table 4

How Well-Positioned Are Banks To Absorb Future Provisions?

We believe 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), and
• PPNR remains steady at the third-quarter rate over the next four quarters.

We use a four-quarter period (through the third quarter of 2021) because the current expected credit losses accounting standard (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, 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. Any shareholder payouts they make and balance sheet growth they experience would weigh on their capital ratios. The Fed's prohibition on repurchases and dividend increases for banks that were part of the stress test provides some support for capital ratios, at least through the end of the year. It also mandates that dividend payouts do no exceed the average of a bank's net income for the four preceding quarters.

Table 5

Table 6

Table 7

Low Rates Have Hurt The Earnings Capacity of U.S. Banks

We estimate that net interest income fell in the third quarter by a median 3% for rated U.S. banks compared with the same period last year, helping drive down preprovision net revenue (PPNR) by a median 5%. (Coupled with a 3% median rise in noninterest expenses, the fall in net interest income more than offset a 5% rise in noninterest income.)

A median drop in net interest margins of more than 50 bps from the prior year outweighed the positive effect of 8% loan growth and 13% asset growth on net interest income. Without such growth, net interest income would likely have fallen substantially further. In fact, banks that had little or negative loan growth and sharp declines in NIMs experienced double-digit declines in net interest income. The minority of banks that grew net interest income generally benefited from strong loan growth or were asset sensitive (probably with a significant portion of fixed rate assets).

The loans originated as part of the Paycheck Protection Program, which were material for some banks, appear to have had a small positive impact on net interest income, limited by the 1% yield on such loans.

For banks that heavily rely on net interest income, particularly regional banks, the drop in NIM can be especially painful. Most rated banks generate about two-thirds of their revenue from net interest income, although the largest banks, the trust banks, and some others have much lower reliance.

We believe the fall in net interest income and PPNR means the earnings capacity of U.S. banks has dropped relative to before the pandemic, even when neutralizing for differences in provisioning.

For instance, the following measure helps illustrate the impact of lower rates on banks' returns on average common equity (ROCE):

• Year-over-year change in net interest income divided by average common equity, and

We annualize and tax adjust that ratio at a 20% rate.

For banks that had net interest income declines, the ratio implies the declines cost those banks roughly 160 bps of ROCE at the median. If their net interest income had held flat, their ROCEs would have been 160 bps higher.

This excludes the roughly 35% of rated banks that reported higher net interest income related primarily to strong asset growth or acquisitions. However, even most of those banks have been weakened by lower NIMs.

Table 8

Table 9