The Federal Reserve released last month the hypothetical economic scenarios for its annual stress test for large banks. While the Severely Adverse scenario incorporates significant economic stress, it fails to account for other risks, particularly interest rate risk. These gaps also reflect more fundamental weaknesses with the current supervisory stress testing framework.
Which Firms are Covered?
This year’s supervisory stress test will cover 32 bank holding companies (BHCs). The stress of a global economic recession will apply to all firms. Two will also face a counterparty default scenario. Nine will face three scenarios: global recession; counterparty default; and global market shock.
Last year’s stress test covered only 23 firms. Those between $100-250 billion in total assets are stress tested only in even-numbered years. A related stress testing process for banks (rather than BHCs) kicks in at $250 billion. Transition provisions mean that the stress test exercise won’t apply to some BHCs over the $100 billion threshold. Rapid growth has long been a red flag to supervisors, but current rules allow such fast growers to ease into stress testing requirements. Consider New York Community Bancorp (NYCB), which has grown by 95% over the past two years. NYCB’s recent stock performance is an even bigger red flag:
It can take time for banks to develop the reporting infrastructure necessary for the stress testing exercise. There is also nothing preventing supervisors from taking a close look at the bank, including its internal stress testing processes. Still, supervisors devote a lot of resources and agency expertise to the annual stress testing exercise and it’s hard to justify excluding the most troubled large bank from this process. We’ll have to wait until 2026.
The 2024 Scenarios
The chart below shows some key metrics for the Severely Adverse scenario. Low points are highlighted in yellow.
The scenario envisions a severe recession, but also a quick bounce back. Real GDP growth averages 5.6% during the last six quarters of the stress test horizon. Except for 2021’s post-COVID bounce, that’s higher than any quarter since the mid-1980s. Unemployment peaks at 10% but falls to 7.4% over the next six quarters. Unemployment) also peaked at 10% in October 2009, but it took 14 quarters to fall to 7.5%.
Projected interest rates stand out most prominently. The 3-month Treasury rate immediately falls by 340 basis points and remains low throughout the stress test horizon. Mortgage rates fall by 260 basis points in the first quarter and remain low. Interest rate risk played a major role in each of last year’s large bank failures.
The stress test scenarios don’t just ignore IRR, they reward it. As of December 31, 2023, Charles Schwab had unrealized losses on its investment portfolio of $21 billion, more than 50% of Tier 1 capital. BAC has unrealized losses of more than $100 billion, or more than 45% of Tier 1. The supposedly stressed rate scenarios would make these losses disappear. Most of the unrealized losses were in the HTM portfolios, so would not have an immediate capital impact, unless it was accompanied by a liquidity event. However, that’s exactly what happened with SVB, Signature, and First Republic. It’s as though the French defense ministry decided in 1946 to give that Maginot Line another try.
Fundamental Issues with the Current Stress Testing Regime
Gaps in the 2024 stress test scenarios reflect more fundamental limitations to the annual stress testing process. These limitations fit into two broad categories: reliance on a single scenario and adding undue complexity to make the scenarios appear more plausible.
Reliance on a Single Scenario
The 2024 stress test scenario may be more realistic and possibly more conservative than a rising rate scenario. The problem is that is the only scenario applied to most banks. Lots of bad things can happen in a $27 trillion economy and an effective stress testing program should look at vulnerabilities to a wide range of adverse conditions.
The Federal Reserve introduced two “exploratory” scenarios to try to address this issue. Scenario A combines a mild recession with a rise in market interest rates. Scenario B includes a severe recession and an even sharper increase in rates. The scenarios also assume that 20% of noninterest-bearing deposits shift into time deposits and an increased reliance on wholesale funding. There will also be two exploratory market shocks that apply to the eight G-SIBs.
Whether the results of the exploratory scenarios will have any real-life consequences is another story. The Fed describes the exploratory analysis as focusing on hypothetical risks to the broader banking system, rather than focusing on firm-specific results. Results from the exploratory analysis will not affect capital requirements or dividend distributions. The Fed will publish aggregate industry results in June. It remains unclear whether and to what extent the results of the exploratory scenarios will affect the supervision of individual firms. What happens to firms most vulnerable to the exploratory scenarios? Will supervisors turn down dividend requests on that basis? I have my doubts.
The Plausibility Trap
Stress scenarios should be at least somewhat plausible. That doesn’t mean likely. A true stress scenario should be like the Spanish Inquisition – nobody expects it. And you can take plausibility too far. Efforts to make the stress scenarios more plausible can muddy the waters and complicate the analysis. Recoveries follow recessions but the details are another matter. Supervisors could merely assess a bank’s vulnerability to a severe recession without pinpointing the timing and the extent of the economic recovery.
Perhaps the biggest offset in supervisory stress tests is pre-provision net revenue (PPNR). Losses don’t occur in a vacuum and strong revenues can offset credit, market, and operational losses. But can we trust stressed revenue projections? You must deal with a lot of moving parts. Asset and liability mixes and yields can and do change over time. Earnings at risk projections aren’t very reliable out more than about a year and they only measure interest rate risk. Looking at the potential impact of dozens of other macroeconomic variables further complicates the picture.
Other measures of loss-absorbing capacity, such as capital risk weights or loan loss reserves, don’t assume revenue offsets. Instead, they focus on loss exposure. Revenue potential is a separate, more subjective consideration. PPNR also tends to disappear when a bank fails. Strong core revenue may make a troubled bank more attractive to an acquirer but won’t help in a liquidation.
Supervisory stress testing is just one weapon in the regulators’ arsenal. Regulators can and should look at a bank’s internal stress testing processes and other risk measures to identify vulnerabilities. However, the high profile and public disclosure of supervisory stress testing can lead observers to read too much into favorable results. Opponents of the Basel Endgame Proposal cited favorable supervisory stress testing results as some kind of proof that further tightening of capital rules was unnecessary.[1] This response makes it more important than ever for supervisors to address some of the gaps in the current stress testing regime and acknowledge some of its limitations.
[1] The following comment from Morgan Stanley summarizes this position: “If regulatory capital levels as assessed under CCAR have been robust in recent years at the largest U.S. banking organizations—which the Board has repeatedly asserted publicly—then significant further RWA increases are unnecessary.”