Last Friday, the Federal Reserve Board released the results of its second stress tests for this year. The stress test results again demonstrated that large U.S. banks are highly resilient. Although the average maximum decline of regulatory capital ratios in these second stress tests exceeded the decline of capital ratios in the June exercise and two banks approached minimum risk-based requirements, we estimate that the average stress capital buffer would have remained unchanged if the Board had re-calculated SCBs based on these results.
Partly as a consequence of these strong results, as well as the improved economic outlook, the Board amended but did not eliminate its restrictions on capital distributions by all large U.S. banks. The legal justification that the Board gave for the restrictions is a provision of the capital plan rule that prohibits (without prior approval) a bank from making capital distributions “after the occurrence of an event requiring resubmission. . . .” The Board required a resubmission of large bank capital plans in June.
However, there does not appear to be any evidentiary or analytical basis to continue capital distribution restrictions beyond those contained in the Board’s recently developed stress capital buffer framework. The Board has indicated that it instituted these restrictions on capital distributions beyond the SCB, because the pandemic was perceived as a once-in-a-century event, creating unprecedented economic uncertainty and therefore requiring a special response. But the characterization of the economic situation and outlook released by the Federal Open Market Committee last week indicates that the current economic situation is not especially extraordinary and that the outlook is strong. Also, our prior research shows that the stress capital buffer framework is highly effective in reducing capital distributions in a stress event.
Moreover, our analysis of the stress test results indicates that the stress test projections understate the degree of resiliency of large U.S. banks in three ways. First, the adjustments made to supervisory models’ loss projections to capture some of the risks associated with the COVID event appear to understate the performance of banks if a severe drop in economic activity were to occur. Specifically, a recent BPI survey showed that a substantial majority (just over 60 percent) of loans that were active deferrals as of June 30 are now performing. The Federal Reserve did not disclose how much of the improvement in deferrals was incorporated in the stress test projections, especially for real estate loans. Second, banks with substantial trading and investment banking operations showed strong performance, and as a result saw lower peak-to-trough declines in their common equity Tier 1 capital ratios under stress. However, the diversification effect of trading revenue is underestimated in the stress tests, because of the double-counting of mark-to-market losses that banks’ trading assets experience when prices fall or rise. And third, the sharp increase in bank balance sheets driven by the influx of customer deposits as a result of the coronavirus event led the Federal Reserve’s projections of noninterest expense to increase significantly in the December stress tests, thereby artificially understating bank resilience.
In light of these results, it is important for the Board to make modifications to its SCB rule to help banks better anticipate changes in capital requirements and allocate capital more efficiently. Specifically, we recommend that the Board modify the provision of its capital plan rule that results in a prohibition on capital distributions in the event of a capital plan resubmission. The rule needs to offer a clear, transparent, and predictable rule-based test to govern capital distributions in the event of a required capital plan resubmission.
The current rule allows the Board to halt all capital distributions indefinitely by a substantial majority of the banking industry, simply by issuing a wholesale demand for capital plan resubmission and then not acting to approve any resubmitted plans. This possibility was not advertised or commonly understood as a possibility when the rule was adopted. The general assumption was that capital plan resubmissions would be idiosyncratic, based on bank-specific events, and promptly reviewed and approved—not that the rule would constitute effective suspension of the entire capital regulatory framework as applied to banks holding nearly 80 percent of the industry’s assets. That framework was in fact largely based on the use of restrictions on capital distributions as a means of regulating bank use of capital buffers; that function has been ended as the ad hoc restriction has been applied. The ad hoc restrictions have also effectively granted the Board new, entirely subjective power to replace bank boards of directors in making decisions about capital distributions. Previously, boards were entitled to make such decisions, so long as the end result was compliance with regulatory capital requirements.
The Board should clarify when a capital plan resubmission will be required, provide what standards other than compliance with existing capital rules the Board will consider in deciding when to lift it, and set a limit on the maximum amount of time a bank can stay under those constraints. These changes should be implemented through a notice-and-comment rulemaking process to promote predictability and certainty to banks and their investors about when capital distributions can be paid under any circumstance, including upon resubmission of a capital plan.
Description of Second-Round Stress Test Results
In broad terms, the results of the second-round stress test released last Friday were in line with our expectations. The average stress capital buffer derived using the new results is almost identical to the average SCB generated using the June 2020 results. Specifically, our analysis estimates an average stress capital buffer of 3.3 percent in each of the two stress scenarios: severely adverse and alternative severe. The generated average SCB based on June’s stress test results was also 3.3 percent.
Although the average SCB was essentially unchanged between the June and the December stress tests, the decline in the average CET1 capital ratio was higher in the December test by about ½percentage point. This difference arises because the most recent test was generally more severe for banks subject to the 2.5-percent SCB floor. This muted the increase in the average SCB resulting from the decline in the average capital ratio. We believe this result reflects the unique characteristics of the pandemic, and the adoption of the current expected credit loss (CECL) methodology by banks. Under CECL, banks must reserve the full expected lifetime losses on loans and certain commitments at inception. Therefore, they do not have to wait to book a reserve until a loss on the loan is probable.
While the average SCB was little changed across the two stress tests conducted this year, we observed some material moves in the SCB of individual banks. These changes depend significantly on the banks’ business models, the composition of their loan portfolios, and their responses to the uncertainty associated with the COVID event, including the building of allowance for credit losses under CECL.
Exhibit 1 shows the change in the generated SCBs of individual banks derived using the December 2020 stress test results versus the June 2020 SCB published by the Federal Reserve. Each purple circle in the chart represents the difference between the SCB derived using the results from the December severely adverse scenario and the June SCB. Similarly, the open orange squares represent the difference between the SCB derived using the results from the December alternative severe scenario and the June SCB. The observations are ranked, starting from the banks with the largest decreases in their SCB (to the left) relative to those with the most material increases in their SCB (to the right).
Although more than one-third of the banks experienced a reduction in their derived SCBs and another one-third had SCBs that basically remained unchanged, the remaining one-third had higher SCBs. Three banks experienced significant increases in the maximum decline in their CET1 ratios under stress. Specifically, one bank would have experienced on average a 4.1-percentage point increase in its SCB between June and December, and two other banks would have registered an average increase in their SCBs of 2.6 percentage points each. In the rest of this section, we show that the variation in SCB moves for all banks reflect the level of allowance for credit losses at the start of the stress tests, differences in bank business models, and the expansion of bank balance sheets as a result of massive influx of deposits during the COVID event.
The sharp increase in allowances for credit losses in first half of 2020 was a key driver of stress test results . . .
The first important driver of the SCB is the level of allowances for credit losses banks had at the start of the stress tests. Although loan losses were higher under the most recent stress scenarios relative to the June stress test, provisions for loan losses fell. This was because allowances for credit losses rose sharply over the first half of this year, reducing the need for additional loss provisions. Indeed, Exhibit 2 shows a strong negative correlation between the level of allowance for credit losses at the start of the stress tests for each bank (represented on the x-axis) and the change in projected provisions between the June and December stress tests (represented on the y-axis).
The top-left corner of Exhibit 2 also shows three banks with a material increase in provisions for loan losses between the December and June stress tests. These same banks saw material increases in the December SCBs relative to June. Two of those banks have a reserve ratio near 1¾ percent. The three banks were projected to increase their provisions significantly over the planning horizon, because of the notable increase in commercial real estate loan losses in the most recent test. The Board noted in its description of model adjustments that it had increased the loss given default (LGD) for income-producing properties backed by hotel properties. Thus, the adjustment to the LGD of income-producing CRE loans resulted in a much higher path of provisions for loan losses in the most recent test. Loan losses rose significantly less across the other asset classes apart from first-lien mortgages.
Notably, losses on first-lien mortgage loans increased 33 percent in the December stress tests relative to June. This increase is almost surely driven by undisclosed Board model adjustments, to account for deferral and forbearance programs extended by banks to help their customers overcome the economic distress brought on by the pandemic. In prior stress tests, losses on mortgage loans have been driven by the path of the house price index in the stress scenario. Since the peak-to-trough decline in the house price index was smaller in the December stress scenarios than in the June scenario, and the housing market has been performing well, the increase in losses on mortgage loans is most likely being driven by assumptions about the performance of loans under forbearance programs. Although the share of residential real estate still in deferral is higher than other loan categories, that higher share reflects the nature of the deferral tenor in residential real estate loans. This should not be interpreted as a sign that those loans have more difficulties in migrating from the deferral status back to the performing status. Indeed, the housing market is holding up well because of increased demand for residential real estate outside of major cities as a result of remote work.
Banks with large trading and investment banking operations showed strong performance . . .
An important driver of the decrease in the generated SCB of some banks was the strong performance of noninterest income, driven by robust trading revenues and investment banking fees. The December projections went up relative to June for two main reasons. First, the Board made a significant change in the stress scenarios by delaying the deterioration in economic conditions by one quarter. Second, forecasts are derived from models that use an autoregressive framework, and the strong performance seen in the second quarter was projected to continue over subsequent quarters.
Exhibit 3 shows that banks with substantial trading and investment banking operations—measured as the share of trading revenues and investment banking fees exceeding 20 percent of operating income in 2Q20—generally experienced an increase in their projections of operating income cumulatively over the nine quarters of the planning horizon. The boost in trading revenue and investment banking income for these banks supported noninterest income and helped offset the decline in net interest income and the increase in noninterest expense over the nine quarters of the projection horizon.
Notwithstanding the positive impact of trading revenues on noninterest income in the December stress tests, a recent BPI post shows that the projections of trading revenue are understated in the stress tests for the banks subject to the global market shock (GMS). This understatement happens because mark-to-market losses on trading assets are captured both by the GMS and in the trading revenue projections. In an ideal scenario, the trading revenue series would only include fees, commissions, and trading gains reflecting bid-ask spreads associated with new client activity. However, in the stress tests, the trading revenue series also includes both mark-to-market gains and losses and trading-activity-based revenue. As a result, these projections for trading revenue for banks subject to GMS double-count mark-to-market losses on bank trading inventories expected to be incurred under a stress event.
Balance sheet growth dampens bank performance through higher noninterest expense projections . . .
Finally, projections for noninterest expense (equal to the sum of compensation expense, fixed-asset expense, and all other noninterest expense) rose substantially in the December stress test relative to the June projections. Based on the Board’s description of its noninterest expense model, an increase in total assets across the two stress tests leads automatically to an increase in noninterest expense projections over the nine-quarter planning horizon. This usually is not an important shortcoming of the model. However, because the growth in bank assets was so large in the first half of 2020, the implications for the projections of noninterest expense are material.
According to the results shown in Exhibit 4, the correlation between asset growth and the change in noninterest expense projection between June and December is 0.65 across the 33 banks included in the stress tests. As a result, banks with faster balance sheet growth between 4Q19 and 2Q20 also had higher projections for noninterest expense in the December stress tests. In aggregate, total assets rose 11.6 percent across banks subject to the tests in the first half of 2020, and noninterest expenses rose 9.4 percent in aggregate.
However, since the increase in bank size was mainly driven by the influx of deposits and the corresponding increase in deposits at Federal Reserve Banks, there is no obvious reason for the supervisory models to be mapping banks’ increased balance sheets into higher projected noninterest expense during the nine-quarter planning horizon.
Capital Distribution Restrictions
Based on the stress tests results as well as the improved health and economic outlook, the Board had a strong basis for allowing banks to exit the ad hoc prohibition on share repurchases and limitations on dividends resulting from the resubmission process. While the banking sector and the economy more broadly will benefit from the Board’s decision to convert the wholesale prohibition on repurchases to a new restriction, the Board should have allowed the normal regulatory capital framework of the stress capital buffer to exclusively govern capital distributions in the first quarter of next year, for the reasons we will describe.
The Board has a sophisticated, elaborate, and conservative framework it developed over the past 10 years to evaluate whether a bank holding company should be allowed to distribute funds to its investors and to encourage capital conservation during periods of economic stress. The framework combines minimum capital standards; extra capital buffers required of the largest banks; and additional capital buffers based on results of a stress test, assuming a severe deterioration in the economy. Even though the economic downturn caused by the coronavirus pandemic was the first real test of that framework, the Board abandoned its automatic buffers governing capital distributions. Instead, it placed ad hoc prohibitions and limits on distributions to shareholders of all large banks subject to the framework. We believe the Board has done so because the pandemic was likely perceived as a once-a-century event generating unprecedented uncertainty. This event thereby required it to replace bank boards of directors in determining capital distributions, even when such distributions were in compliance with the Board’s regulatory framework.
Although that characterization of the economic situation seemed accurate in March, it is clearly not so now. The FOMC reported last week that the unemployment rate, at 6.7 percent, is 2.6 percentage points above its longer-run rate. By contrast, at the end of 2009, at the beginning of the recovery from the Great Recession, the FOMC judged the unemployment rate to be 4.9 percentage points above its longer-run rate. That gap fell to 4.1 percentage points at the end of 2010 and to 3.6 in 2011. It was only at the end of 2012, three years into the expansion, that the gap fell below where it is now.
Moreover, the FOMC expects the economy to grow at more than twice its trend growth rate next year, and nearly twice its trend growth rate in 2022, closing the unemployment gap by the end of that period. Nor do FOMC participants see the outlook as uniquely risky. As the FOMC reported last week, 11 participants see the risks to growth as broadly balanced or weighted to the upside, while only six consider the risks as weighted to the downside. Participants do judge the outlook to be highly uncertain, but they saw similarly elevated uncertainty in 2008, 2009, 2010, 2011, and 2012.
At a minimum, an improved economic outlook and understated stress test results raise questions about the need for continued ad hoc capital distribution limits. For these reasons, and to foster predictability around capital distributions, we continue to recommend that the Board seek public comment on a proposal on changes to the provision of the capital plan rule that results in a prohibition on capital distributions in the event of a resubmission of a capital plan. These changes should include a provision that firms should be allowed to make capital distributions subject to a clearly specified rule-based test, and not a subjective process, in the event of a resubmission of their capital plans. There should also be a clear temporal limit in this provision of the rule, including clarifying when exactly the process and resulting limits come to an end. More broadly and generally, we believe any limits on capital distributions should be rule-based and subject to notice and comment.
Even with the modifications we have suggested to the capital plan rule, a challenge remains: the capital plan rule is unclear about the exact circumstances that trigger a capital plan resubmission. The broad grant of authority to require resubmissions, and the resulting prohibition on capital distributions, is all a process that could be misused. A better course would be for the Board to establish clear, predictable, and transparent rule-based limits on capital distributions. Such limits would clarify under what future conditions a large bank may or may not make a capital distribution on a capital plan resubmission, and the exact conditions under which such resubmissions would be required.This would help regulated institutions as well as market participants to better predict a bank’s ability to make capital distributions in the event of a capital plan resubmission.
The Board will let firms know by March 31, 2021, whether their stress capital buffer requirements will be recalculated. See Federal Reserve Board, December 2020 Stress Test Results, Appendix D, at 144, available at https://www.federalreserve.gov/publications/files/2020-dec-stress-test-results-20201218.pdf.
12 CFR 225.8(k)(1).
See, e.g., Vice Chair for Supervision Randal K. Quarles, Federal Reserve Board, “Supervision and Regulation Report,” Testimony Before the Committee on Banking, Housing, and Urban Affairs, U.S. Senate (November 10, 2020), https://www.federalreserve.gov/newsevents/testimony/quarles20201110a.htm(referring to the pandemic as a “once-in-a-century shock”); see also Federal Reserve Board, Assessment of Bank Capital during
the Recent Coronavirus Eventat 1 (June 2020), available at https://www.federalreserve.gov/publications/files/2020-sensitivity-analysis-20200625.pdf(“While all large banks are sufficiently capitalized at present, the Board is taking additional action to preserve capital and to assess banks’ capital plans more frequently during this time of uncertainty.”)
FOMC, “Summary of Economic Projections” (Dec. 16, 2020), available at https://www.federalreserve.gov/monetarypolicy/files/fomcprojtabl20201216.pdf.
The Federal Reserve continues to use the incurred methodology in the stress tests to calculate the allowance for credit losses. This reduces uncertainty and allows for better capital planning at banks.
There is no obvious pattern to relative bank performance under the two tests, but the alternative severe scenario appears to generate more severe outcomes for the three banks with the largest increases in their derived SCBs.
See, e.g., Brenda Richardson, “Experts Predict What The Housing Market Will Be Like In 2021,” Forbes, Dec. 16, 2020, available at https://www.forbes.com/sites/brendarichardson/2020/12/16/experts-predict-what-the-housing-market-will-be-like-in-2021/?sh=57391bc536dc (saying “The housing market has been on fire this year with record-low mortgage rates and a sudden wave of relocations made possible by remote work.”); see also Issi Romem, “Explaining the Frenzy in the Housing Market,” The New York Times, Dec. 3, 2020, available at https://www.nytimes.com/2020/12/03/upshot/home-prices-explaining-frenzy.html.
According to the Board, the model used to generate the trading revenue projection of banks subject to the GMS does not have an autoregressive term. The improvement was driven by the delay in the deterioration of economic conditions in the two stress scenarios by one quarter.
12 CFR 225.8(k)(1).
We recommended these changes in a recent letter to the Board. See BPI Letter, “Amendments to Capital Planning and Stress Testing Requirements for Large Bank Holding Companies, Intermediate Holding Companies and Savings and Loan Holding Companies (RIN7100–AF95)” at 16–19 (November 20, 2020), available at https://www.federalreserve.gov/SECRS/2020/November/20201130/R-1724/R-1724_112020_137375_504270313357_1.pdf(hereinafter BPI Letter) (recommending certain changes to 12 CFR 225.8(k)(1)).
12 CFR 225.8(e)(4). For example, in our letter we recommended the Board propose and seek comment on clear criteria and standards to be used for purposes of requiring a capital plan resubmission. See BPI Letter at 18–19.
We believe these changes would be consistent with the principles articulated by the Board’s Vice Chair for Supervision. SeeRandal K. Quarles, Vice Chairman for Supervision, “Early Observations on Improving the Effectiveness of Post-Crisis Regulation,” American Bar Association Banking Law Committee Annual Meeting (January 19, 2018), at 2 (suggesting “we have an opportunity to improve the efficiency, transparency, and simplicity of regulation”), available at https://www.federalreserve.gov/newsevents/speech/quarles20180119a.htm; see also Randal K. Quarles, Vice Chairman for Supervision, “The Eye of Providence: Thoughts on the Evolution of Bank Supervision,” Federal Reserve Board, Harvard Law School, and Wharton School Conference: Bank Supervision: Past, Present, and Future (December 11, 2020), https://www.federalreserve.gov/newsevents/speech/quarles20201211a.htm(saying that “[t]ransparency helps guard against regulation that is arbitrary or capricious. Transparency is enshrined in statutory process protections such as those in the Administrative Procedure Act, which facilitates understanding and communication.”)