Discover more from Without Warning
Peacetime Bank Stress Tests: Who's That For?
"The scenarios are not forecasts"
Last week, the Fed announced the design of this year’s iteration of its stress tests for large banks. In the US, the stress test drives large banks’ risk-based capital requirements.
The 34 largest banks face a 4.5% minimum risk-based capital charge plus a “stress capital buffer” calculated from their losses in the stress test—the latter of which can’t be lower than 2.5%. Moreover, the eight GSIBs are subject to the GSIB surcharge, which, as of Oct 1, 2022, ranges from 1% (State Street) to 3.5% (JPM):
While the Fed reiterates that stress test design decisions are not forecasts, the “severely adverse” scenario tested by the Fed is clearly informed by a general macro risk outlook.1 For instance, the 2023 stress test scenario’s developments in foreign economies focus on greater stress in advanced economies; 2022’s test, meanwhile, had “greater stress in emerging market economies, partly driven by building risks in the Chinese economy.” The 2023 “global market shock,” which applies to firms with the largest trading activity, examines stress from falling prices, while last year’s focused on inflationary risks. Other such contrasts exist.
Stress Test Target (Audience)
Germaneness of scenario be damned, it’s not clear the benefits of stress tests in peacetime—that is, outside of a financial crisis—come from the tests’ tether to the macro risk reality. Given that the tests drive banks’ risk-based capital, what’s important is just that they’re tough—so banks have “enough” capital—and that they’re variable so that banks can’t unduly smooth their scores.
The tough macro scenario is never really going to look like the next financial crisis. Look at covid. Look at the test-tube financial crisis playing out at Credit Suisse right now; indeed, the main reason CS is even still alive at all is that its capital position was strong and that it was able to source new equity (but even that is looking increasingly like it might not be enough—see: the foundering share price).
The success of the US stress test in 2009 came from the fact that it publicly stress-tested a worsening of the situation that the financial system was already facing—and had a fiscal backstop. The covid-era “sensitivity analysis,” which came earlier in the crisis timeline, was similarly helpful in that it showed the system to be decently well-capitalized against further deterioration of the pandemic (which the Fed knew before it announced it would release the analysis)—and obscured individual bank names while doing so. This latter decision to not disclose individual bank names made sense since it was still a fragile juncture of the crisis—indeed, the Fed even already knew not all banks had “passed”—and since there was no fiscal backstop for the results. (←While these are the main “good” reasons for not disclosing individual names, then-Fed Vice Chair for Supervision Randy Quarles really only cited the analysis’s data/design roughness as the reason for this choice.)
In a crisis, the banking sector is plagued by a perception of undercapitalization relative to the amount needed to continue credibly providing balance sheet capacity to the system. A well-designed stress test can thus serve as a way for regulators to use their confidential supervisory information to reveal to the market that the system is fine and business should resume. There is of course a key balance to be struck here of credibility (toughness), reassurance (being able to say everything is fine), and government support (demonstrating enough tools/funds to address any possible extent to which things are not fine).
Outside of crisis-time however, the public stress test process isn’t really for anyone except maybe bank equity investors. They benefit from some scenario analysis being applied to banks and get a little more of a window into banks’ balance sheets. The tests have little to do with the genesis of stress tests in the US. For instance, in 2020 (!!), Goldman was over a percentage point short from its new, stress-test driven capital ratio. While it had to eat a dip in its share price, there were no reports of Goldman losing business following this result; I would suspect precisely zero counterparties abandoned Goldman on account of these results. And thus, importantly, while bank equity investors may “benefit” from the additional info, it’s not clear the peacetime public stress tests (unlike their crisis-time counterparts) have any impact on the amount of capital allocated to the sector as a whole.2
Capital: Enough & Ungamed
We don’t want banks to “teach to the test”—which they could do if stress tests were unchanged from year to year. Of course this separate point is also ultimately a question of banks having “enough” capital: unvarying, game-able stress test scenarios can allow a bank to run down its capital while still reporting high modeled levels of capital.
For instance, in the pre-GFC decades, as regulators shifted capital regulation toward a greater emphasis on banks’ internal risk models—particularly in Europe—banks, whether intentionally or not, gamed them. Here, for instance, is a graphic from the IMF’s Tam Bayoumi charting large European banks’ risk-based capital against their simple leverage ratios. You can see the correlation break down over time and even turn negative by 2008—suggesting maximal gaming:
Thus, it’s important to keep any modeled component of capital requirements variable, and thus less game-able. Fed Vice Chair for Supervision Barr’s stated desire to contemplate additional stress scenarios seems to serve this aim. The Fed started down this path last week:
For the first time, this year's stress test will feature an additional exploratory market shock to the trading books of the largest and most complex banks, with firm-specific results released. This exploratory market shock will not contribute to the capital requirements set by this year's stress test and will be used to expand the Board's understanding of the largest banks' resilience by considering more than a single hypothetical stress event. The Board also will use the results of the exploratory market shock to assess the potential of multiple scenarios to capture a wider array of risks in future stress test exercises.
As this bonus scenario gets “assessed” for “future stress test exercises”—aka, eventually contributes to capital requirements—this will clearly add to the test’s variability aspect, making the tests less preemptively game-able. For better and for worse, an “additional exploratory market shock to the trading book” could be designed in such a way that it’s particularly hard on just JP Morgan, or just Goldman, or just… who knows.
Capital Is Capital
Okay, so we want banks to have “enough” capital. And, to make sure that any model-driven reported amount of capital is actually “enough,” we want to keep the tests changing so banks can’t stage their balance sheets to manipulate (and misrepresent) their results.
This will help ward off financial crisis by keeping banks better capitalized ahead of whatever crisis risk may come. Getting the specific scenarios “right” outside of crisis-time is much less important.
Aside from this, the Fed’s stress test design framework structurally builds in some countercyclicalities. For instance, this year’s unemployment shock is greater (to get to the 10% minimum), as is the shock to residential real estate to reflect the year’s increase (though a non-monotonic one) in house prices.
A counterargument here is that the banking sector actually attracts additional equity capital on account of the Fed effectively stamping approval on banks’ balance sheets. This may suggest to investors that the Fed would be unlikely to let a bank get wiped out—as there would be serious reputational costs for a Fed that had recently signed off the bank’s resilience. However, it’s not clear a public stress test is required for the sector to receive this benefit. Indeed, the reason for the GSIB surcharge is precisely to offset this perception which long predates the stress testing process. Rating agencies have long explicitly applied too-big-to-fail premiums to banks. And irrespective of a public stress test and its design, the Fed supervisory process is always ongoing—and manifesting in capital payouts, fines, business restrictions, etc.—and thus its credibility always in play. It should thus be clear that this perception is not the result of the performative aspect of a peacetime stress test with specific bank results published.