Federal Reserve’s Off Cycle Stress Test Is Most Severe Stress Test Ever, Punishes Capital Market Activity

Key Takeaways

  • Despite positive economic trends noted in early June, the application of a rigorous stress test in February 2020, the limited time period before the 2021 stress test and the continued resilience of the banking sector including the industry’s voluntary refrain from stock buybacks early on in the crisis, the Federal Reserve conducted an Off Cycle Stress Test (OCST). Only the quantitative results of the OCST will be published December 18, 2020.
  • Since the assessment of resubmitted capital plans will not be published until sometime in 1Q21, it is very likely the prohibition of stock buybacks and dividend caps will be extended for the first quarter of 2021.
  • While the Fed has not stated whether they will use the OCST results to resize the Stress Capital Buffer (SCB), this stress test incorporated some of the most extreme macro scenario variables and Global Market Shock (GMS) factors ever. Resizing the SCB at this point would be unnecessarily punitive and potentially stymie economic recovery.
  • Substituting the Fed’s decision-making regarding buybacks and dividends undermines the regulatory capital framework and lessens the accountability of firms’ executive management and board of directors to their shareholders and regulators.
  • The OCST is one of the most severe stress tests ever administered as measured across macroeconomic variables and the GMS factors.

The quantitative results of the Off Cycle Stress Test (OCST) will be released this Friday, December 18. The release, however, will not be accompanied by the Federal Reserve’s assessments of banks’ resubmitted capital plans. Instead, the Fed has announced that they will publish their capital plan assessments sometime in the first quarter of 2021. The assessments are critical to removal of the prohibition of stock buybacks and dividend caps that were instituted in June 2020 and implemented even though the largest banks had already voluntarily foregone stock buybacks until there was a sustained economic recovery.

SIFMA believes the Fed should release this capital plan assessment in the very earliest part of the next quarter to remove ambiguity regarding dividends and stock buybacks. More importantly, the return to reliance on existing capital regulations, including the Stress Capital Buffer (SCB), to determine appropriateness of capital actions is essential to maintain credibility and integrity in the U.S. capital framework. Placing responsibility for capital management back into the hands of a firm and its board of directors is appropriate as it is a core obligation of both parties to which shareholders and the Federal Reserve hold them accountable. Moreover, the Fed regularly assesses the capital planning process including governance and decision making annually as part of the supervisory stress test and therefore should be comfortable with this outcome.

The Fed should not use the results of the OCST to resize the SCB because it sets a dangerous precedent of resizing capital requirements immediately following a crisis and far too soon after a rigorous stress test has been performed. Moreover, it implies a level of precision in stress testing outcomes which is unsupportable and results in highly procyclical outcomes.

The OCST is one of the most severe supervisory stress tests ever employed and assumes unprecedented economic deterioration, compounded by the lack of consideration of any additional government stimulus or the implementation of COVID-19 vaccines.

The OCST, like all supervisory stress tests, uses macro scenario variable to generate loss rates and pre-provision net revenue (PPNR). However, the variables used in this stress testing cycle were based on hypothetical scenarios that were unprecedented in the speed, level and duration of economic deterioration relative to previous CCAR cycles and historical observations. The variables and factor shocks were of significant magnitude and generally seem extreme given the evidence of nascent economic recovery. Furthermore, the use of such draconian assumption appears excessive given the Fed was likely designing the scenario for the 2021 supervisory stress test at the same time the OCST variables were announced.

The use of OCST would be more insightful and helpful if it was applied surgically as a tool to understand more vulnerable parts of the economic recovery instead of a blunt instrument. It would also be a compliment to the regular cycle stress test results and would be helpful in informing the next stress test cycle and more importantly supervisory assessments of capital and risk management.

To understand the magnitude of the macro variables, the severely adverse scenario assumed that the unemployment rate would increase steadily from 9.5% in 3Q20 to a high of 12.5% during 4Q21 and only decline to 7.6% by the end of the nine quarters. These assumptions are considerably onerous as the U.S. unemployment rate had already demonstrated significant and speedy improvement at the time the scenarios were published.

The OCST incorporated equally severe assumptions regarding Gross Domestic Product (GDP) growth.  The Fed’s stress assumption for 3Q20 GDP growth underestimated actual GDP growth by nearly 40%. Moreover, consensus projections for 4Q20 GDP growth is approximately 3% while the OCST assumes a 5.9% decline. The negative trend continues through most of the nine quarter period.

The same patterns are evident across other domestic variables such as the House Price Index, and the Market Volatility Index where the trajectory of the nine-quarter estimates reflects continued significant economic turbulence at unprecedented levels and duration. Even as a stress test, employing such extreme variables in the wake of the rigorous 2020 regular cycle stress test, the “real time” stress test of COVID-19, and the short timeframe before the next regular cycle stress test is unnecessary.

Certain GMS factor shocks were made considerably more stringent while the GMS factor shocks that were left unchanged were already extremely severe.

Shifting to the OCST’s GMS factor shocks, we see the same pattern of significantly increased severity of stress test assumptions well beyond any historical precedents.

For some asset classes such as corporate credit, private equity and securitized products, the OCST relied on 2020 regular cycle stress test GMS factors.  We believe this is because those GMS factor shocks were already extremely severe when compared to historical observation even including the COVID-19 experience. SIFMA wrote on the severity of the 2020 GMS factor shocks and the assessment of the GMS factors in comparison to the COVID-19 experience in June. These blogs can be accessed here and here.

GMS factor shocks for equity, energy, metals,  interest rates and certain FX currency pairs were modified substantially from the 2020 regular cycle stress test despite that only some of those asset classes demonstrated deterioration in excess of the 2020 GMS factor shock.

As mentioned above, other OCST GMS factor shocks, such as those for equity, energy, metals, interest rates and certain FX currency pairs were revised likely because the performance for some of those asset classes during the COVID-19 experience  was close to or exceeded the GMS factor shocks used in the regular cycle supervisory stress test. Importantly, however, none of the single day price moves for these asset classes during COVID-19 dislocation exceeded the 2020 regular cycle supervisory stress test GMS factor shock. For some product classes even using an extended calibration, the asset class price performance during the COVID-19 dislocation did not exceed the regular cycle stress test GMS factors shock. Nevertheless, the Fed increased the GMS factors which exemplifies the exaggerated severity of market deterioration built into the OCST.

To provide some perspective on the magnitude of adjustments that were made to the OCST GMS factor shock, we have highlighted some of the impacted asset classes. For example, the OCST GMS factor shock for U.S. equity markets as expressed by the S&P 500 was a 34.2% spot decline, which is 32% higher than the GMS factor used in the 2020 regular cycle stress test and still higher than any historical observation. During the COVID-19 dislocation, the S&P 500 index only declined 12% on its worse day and 23% if calibrated on a 10-day basis. The Fed also significantly increased the GMS factor shock along the U.S. equity volatility curve where the nearer-term volatility points were increased by as much 200%. Generally, the same degree of severity was applied to material foreign equity markets in similar ranges. It is unclear how the magnitude of the upward adjustments was sized, or the plausibility of such movements were assessed.

The OCST GMS factor shock for energy showed similar increases in price shocks from the 2020 regular cycle stress test. For example, the GMS factor shocks across all types of crude oil were increased close to 50% in the short end of the curve with the size of the shocks declining slightly along the curve.  The same modifications are seen in the diesel, heating oil and gasoline GMS factor shocks. While the prices of some of these products were impacted by the COVID-19 dislocation due to an episodic declining appetite stemming from global lockdowns, the driving factor in the price decline was the Saudi-Russian price war. Moreover, and important to the timing of the OCST, oil prices had stabilized to more normal historical levels by mid-summer. Other energy products such as natural gas, which were not impacted by price wars were nevertheless materially modified. The OCST roughly doubled the GMS factor shocks from the 2020 regular cycle stress test. The magnitude of the adjustments is hard to support particularly given many natural gas product prices were even higher than pre-pandemic price levels by August.   

The Pre-Provision Net Revenue (PPNR) approach for trading revenue has a suboptimal outcome for those firms which are also required to perform the GMS.

A subset of firms is subject to the GMS that applies an instantaneous shock to a firm’s capital market exposures. This subset of firms is also required, like all firms subject to supervisory stress testing, to estimate PPNR using the macro scenario variables. We know that there is a weakness in the PPNR trading revenue model which results in an underestimation of trading revenues and consequently an underestimation of capital.

The combination of the GMS and the Fed’s PPNR model results in double counting of mark-to-market position losses because both approaches capture losses on trading positions.  

SIFMA understands that the Fed’s PPNR trading revenue model demonstrates an inverse relationship between a firm’s trading revenue and market volatility, as measured through the Chicago Board of Options Exchange Volatility Index (VIX). This inverse correlation however is not supported by historical experience, including the market patterns observed during the COVID-19 pandemic. Instead, historical analysis demonstrates that activity-based trading revenue is, in fact, positively correlated with the VIX.  During the COVID-19-related market dislocation, for example, the activity-based trading revenues of large trading firms increased significantly due to the volume and scale of market activity

SIFMA believes that the counterfactual relationship between the trading revenue model and real-world outcomes arises because the trading revenue model was trained on data drawn from the Fed’s FR Y-9C regulatory reporting. Specifically, the data was likely drawn from the FR Y-9C report’s trading revenue line item, which importantly combines activity-based trading revenue with mark-to-market position losses.  As noted, the GMS already captures extreme mark-to-market losses. Consequently, the PPNR trading revenue model estimates trading revenue using the combined trading revenue and mark-to-market losses. Thus, it underestimates trading revenue and replicates the positional losses that are generated by the GMS component.  Of note, the use of comingled activity-based trading revenue data and mark-to-market position losses is inconsistent with the Board’s approach to stress testing banking book credit product where the  Fed separates revenues derived from banking book products from credit losses.[1]  This inconsistent treatment for capital market product  is unwarranted and punitive to capital markets products.

The Fed should modify the PPNR trading revenue model and refine the data inputs to more closely  reflect only “true” trading revenue.

The Fed’s trading revenue model should be modified to remove mark to market losses and instead rely on data that is” fit for purpose” to ensure the integrity of the outcome.  Using the correct data series, the Federal Reserve’s trading revenue model would more accurately reflect the relationship between trading revenues and volatility, resulting in an approach that would complement rather than contradict actual experience and would be consistent with how firms assess risk and manage capital in connection with trading activities.


SIFMA is fully supportive of supervisory stress testing as a tool to assess capital adequacy including during periods of stress. Nonetheless, care must be taken to ensure that its assumptions are not only sufficiently stringent, but also plausible and avoid procyclicality, which would harm the ability of banks to support the recovery of the real economy. The Federal Reserve also needs to address the double counting issues inherent in the PPNR trading revenue model to ensure that capital markets activity is treated fairly. Treating capital markets activity appropriately in the stress testing process is an important element in promoting healthy and functioning capital markets and, in turn, the U.S. economic recovery from the COVID-19 event.

Coryann Stefansson is Managing Director and head of prudential capital and liquidity policy at SIFMA.

[1]           We further note that the Board has taken steps to avoid the double counting of losses in other contexts within the stress test.  See, e.g., Board of Governors of the Federal Reserve System, Dodd-Frank Act Stress Test 2020: Supervisory Stress Test Methodology at 19 n.31 (March 2020), https://www.federalreserve.gov/publications/files/2020-march-supervisory-stress-test-methodology.pdf.