Stress tests lack COVID-scale stress
“Transparency surrounding the results of the tests is a bedrock principle of the stress testing framework, dating back to the Supervisory Capital Assessment Program (SCAP) that was conducted during the 2007-2008 financial crisis.” Letter from U.S. Senators Schatz, Brown and Warren to Federal Reserve Chair Powell and Vice Chair Quarles, June 23, 2020.
“It’s very worrisome. I think the stress test has become something of a compliance exercise.” Former FRB Governor Jeremy Stein, Brookings-Michigan Panel, “Has Dodd-Frank Reined In Systemic Risk?” June 30, 2020.
In recent months, the Federal Reserve acted aggressively to support nearly all parts of the U.S. economy. Unprecedented monetary policy actions, both in size and scope, served to maintain market function and the flow of credit. And, while we have misgivings about the Fed’s CARES Act-driven moves to support the nonfinancial sector, we applaud Chair Powell and his colleagues for their quick and decisive actions (see our previous posts here, here and here). This, together with fiscal policy support for individual households and small firms, has kept an awful situation from becoming far worse—at least for now.
But, the Fed’s responsibility extends beyond monetary policy to the regulatory and supervisory arenas: it is obliged to maintain the safety and soundness of the banking system (and, to some extent, of the broader financial system). On this score, and in stark contrast to its actions in 2009, the Board of Governors has come up significantly short. As we have noted before, the 2009 stress test (SCAP) was one of the most important tools for restoring confidence in the financial system. And, as the three U.S. Senators point out in the letter from which we quote, transparency is one of the key features of a stress test. Without full disclosure, suspicions will linger about the ability of the largest banks to provide credit to healthy borrowers if the COVID recovery falters. (See our earlier post for details.)
In this post, we examine the results from the Fed’s 2020 assessment of bank capital adequacy published on June 25. Based on the COVID-related sensitivity analysis—for which individual results are unavailable—one-quarter of the 33 banks tested fall below the regulatory minimum in the worst of the three cases. The fact that we can only guess which banks those might be creates suspicion regarding many banks.
The Fed’s report provides results for four scenarios. The first, labelled “severely adverse,” was constructed before the COVID downturn and announced on February 6, 2020. As we note in our previous post, by early May we knew that the economic reality was already worse. Since stress tests are a forward-looking capital planning exercise, it made sense to add scenarios that are truly severe from the current vantage point.
With this in mind, the Fed created the “V-shaped”, the “U-shaped” and the “W-shaped” COVID sensitivity scenarios. Recovery is fast in the first case, slow in the second, and in the third, there is a second wave. None of these appears to be particularly “severe” under present circumstances. Current forecasts suggest that the V-shaped scenario is very optimistic, that the U-shaped one is the most likely, and that the W-shaped path is a real possibility. Unfortunately, things could plausibly get worse than any of these.
Before getting to the details, we need to explain the calculations that form the basis for the capital requirements. While there are several ways to measure capital adequacy, we focus on two types. Both divide common equity (the accounting value of the firm to shareholders) by a measure of assets. The first, which the Fed calls the common equity tier 1 (CET1) capital ratio, weights assets by their risk. The second, what the Fed calls the supplementary leverage ratio (SLR), weights assets equally. Furthermore, since the purpose of capital requirements is to ensure banks have sufficient buffers to withstand adverse events, they take account of off-balance sheet exposures (usually in the form of derivatives) that can generate losses as well. That is, measures of assets for the computation of both the CET1 capital ratio and the SLR include both on- and off-balance sheet items. For the largest banks, the CET1 requirement varies, but is generally over 10%; and the SLR requirement is 5%.
To give some sense of the scale of the numbers involved, we can look at information for the six largest U.S. banks ranked by end-2019 assets—JPMorgan Chase, Bank of America, Citigroup, Wells Fargo, Goldman Sachs, and Morgan Stanley. Their assets range from about $900 billion to $2.7 trillion―risk-weighted assets are on average about half the total. Off-balance sheet exposures increase total assets by roughly 20%. Furthermore, at the end of last year, these six banks’ CET1 ratios were between 11.1% and 13.3%, and their SLRs ranged from 5.7% to 7.1%. (See our earlier discussion here, as well as the Kansas City Fed’s Capital Analysis here.)
Turning to the June 25 report, the figure below (from page 14) displays the minimum CET1 capital ratio in each of the four scenarios. The colored bars show the interquartile range. That is, in each of the cases, one quarter of the 33 banks subject to the 2020 exercise have capital ratios that fall below the bottom of the bar at some point over the three-year horizon of the stress scenario. As we look at these, it is worth keeping in mind that the minimum requirement for the CET1 ratio is the sum of a 4.5% base, a 2.5% conservation buffer, a systemic risk buffer for large banks, and a discretionary buffer added by supervisors. Any bank that breaches the 7% (4.5% + 2.5%) threshold is subject to restrictions on their activities. For a bank that falls below 4.5%, the supervisors issue a warning, replace management, and ultimately place the bank in receivership or conservatorship. Looking at the figure, it appears that 8 of 33 banks would fall below this critical threshold in the W-shaped recovery. Indeed, it is possible that one or more banks would sink below this threshold in every scenario: we just don’t know.
To get some sense of the current state of the large U.S. banks—and what a real stress test might look like if it we were to do it now (end-June 2020)—we turn to the Stern Volatility Lab’s measure, SRISK. This is a forward-looking, high-frequency measure of systemic risk computed using market (rather than accounting) values of bank equity. In our earlier post, we note that aggregate SRISK is at levels similar to those in late 2008 and early 2009 (see here).
We use the data on SRISK to compute a stressed risk measure that we will call an SRISK leverage ratio. This differs from the Fed’s stress test version of the SLR in several important respects. First, rather than rely on a specific economic scenario, the SRISK-based measure assumes that, over the coming six months, equity markets fall by 40% from their current level. Second, while supervisory measures of capital (including those in stress tests) base their calculations on accounting measures, SRISK uses the market value of equity. Third, the supervisors explicitly include off-balance sheet exposures, while our SRISK leverage ratio does not. To be sure, market prices of bank equity should reflect the value of off-balance sheet exposures, so our measure should capture some of this. Nevertheless, the official SLR (because it adds off-balance-sheet exposure to the denominator) tends to be lower than our SRISK leverage ratio.
If stress tests have a similar impact on CET1 capital ratios and on leverage ratios, we can use the latter to judge which of the banks might be doing badly. With that in mind, we compare our SRISK leverage ratio and the SLR from the Fed’s stress tests for the 14 publicly traded domestic banks that report the SLR: generally, these are the largest domestic banks in the Fed’s sample.
Since the Fed bases their stress tests on end-2019 balance sheets, we first compare the Fed’s results to SRISK information on that date. For each of three series, we plot each bank’s stressed leverage ratio (either using the SLR from the Fed’s analysis or the SRISK leverage measure for different dates) on the x-axis. On the vertical axis, we plot current leverage (June 26, 2020) based on the market value of bank equity: this is the ratio of the bank’s market equity to the sum of its fixed liabilities and market equity. The black dots show the end-December SRISK leverage ratio; the red dots are the stress-test SLR published by the Fed; and the blue dots are the current (end-June) SRISK leverage ratio.
Bank leverage, current and under stress
Note: The end-June 2020 leverage ratio on the vertical axis is computed as the current market value of bank equity divided by the sum of market value of equity and book value of fixed liabilities.
Source: Board of Governors of the Federal Reserve, Table 4; NYU Stern V-Lab; and authors’ calculations.
Looking at the chart, we start by noting the close relationship between the end-December SRISK leverage ratio and the end-June actual leverage ratio (based on market value of equity): the correlation is 0.76. This pattern suggests that SRISK at end-December did a reasonable job of anticipating current circumstances.
Second, reflecting the inclusion of exposure in the denominator, the Fed’s stress test SLRs (red dots) generally lie to the left of the end-2019 black dots. Importantly, the rank correlation between the red dots and the black dots is 0.78, again highlighting the utility of SRISK as an early-warning device.
Third, the end-June SRISK leverage ratios (blue dots) typically lie well to the left of the red dots. Because stress test results that include off-balance-sheet exposures typically would be even lower, the implication is rather dire: were the Fed to do a proper stress test today, most of the banks would have minimum SLRs of 3% or lower, well below the 5% level required of the largest banks, and even below the regulatory minimum for all banks.
All this suggests that we can use the current SRISK leverage ratio to get some idea of where individual banks stand. To do this, we examine the end-June level for 21 of the 33 U.S. banks subject to the stress test. (Note that there is no measure of the market equity value for the 10 subsidiaries of foreign banks in the Fed’s sample, so these are omitted.) The results are in the following chart. The question naturally poses itself: are the banks on the far right of this chart the ones that are in the bottom quartile in the figure from the Fed’s report? Unfortunately, we can’t know, because the Fed has not published the full results of its analysis.
SRISK stressed leverage ratios (as a percent of fixed liabilities plus market equity), June 26, 2020
Note: We exclude both TD Ameritrade and American Express.
Source: NYU Stern V-Lab; and authors’ calculations. Labels are the stock symbols for each bank.
To conclude, while we applaud the Fed for assessing the impact of the COVID crisis on banks, the decision to report only the aggregate distributions of the COVID sensitivity analyses leaves us guessing about the condition of the individual banks. We expect that some of the banks would fail a realistic stress test were it conducted today.
Had the supervisors required banks to build additional capital buffers in the last few years of the boom that ended in February―something that would have been relatively inexpensive given then-prevailing equity prices―it is possible that there would be little need to rebuild capital buffers now. Instead, we find it difficult to fathom why the Fed continues to permit the largest U.S. banks to pay dividends following the deepest U.S. economic downturn since the Great Depression. That just increases the likelihood that some banks will need to increase capital in ways that could hamper the recovery.
It did not have to be this way.
Acknowledgments: We thank the NYU Stern Volatility Lab for providing data that allows us to calculate the stressed SRISK leverage ratios for individual banks.