On February 11, the Federal Reserve released the global market shock (GMS) component of the 2020 severely adverse scenario after having released the macroeconomic scenarios last week. In this year’s stress tests, 11 of the 34 participating firms must incorporate an instantaneous shock to their trading exposures and losses associated with the default of their largest counterparty.
BACKGROUND
Banks with significant trading operations are subject to the GMS component in the stress tests. The GMS includes more than 35,000 large moves in market prices and rates generally calibrated to market moves from the second half of 2008. This year’s description of the scenario also indicated the calibration of shocks may include “multiple periods or hypothetical (but plausible) events based on salient risks.”[1] The Fed does not disclose precisely how the different sets of shocks included in the GMS are calibrated. This year’s description of the GMS includes slightly more information on the time horizons used by the Fed to calibrate the risk factor shocks. Specifically, the document states that the calibration of the shocks in more liquid markets—interest rates, foreign exchange, or public equities—is done using a shorter time horizon (such as 3 months). In contrast, shocks to assets classes traded in less liquid markets—private equities or non-agency securitized products—would be calibrated assuming a bank holds the asset for a longer period (such as 12 months). However, it offers no historical or empirical analysis to support a 1-year holding period, which appears ahistorical and counterfactual.
Banks are required to recognize the trading and counterparty mark-to-market losses from the GMS in the first quarter of the planning horizon of the stress tests. Because many banks subject to the GMS reach the trough in regulatory capital ratios well before the end of the planning horizon, the frontloading of stress losses via the GMS leads to a larger decline in regulatory capital under stress. This results in a higher required level of capital to pass the test than for loan losses, which are spread over the 9-quarter planning horizon.
GMS 2020
The Federal Reserve indicated that the 2020 GMS would be especially stressful for bank holdings of collateralized loan obligations (CLOs) and private equity investments; therefore, we first analyze the moves in the market shocks to these two asset classes. Next, we report the shocks to a set of risk factors analyzed in a study published by SIFMA (2019).[2] Similar to the conclusions presented in that study, the probability of the shocks occurring simultaneously remains quite remote in this year’s GMS. However, the size of risk factor shocks traded in more liquid markets is less extreme compared with prior years.
Figure 1 displays the shocks to AAA, AA, and A tranches of CLOs included in the last four stress testing cycles (the most relevant tranches for banks). The market value shocks included in this year’s GMS range from –23% for the AAA tranches of CLOs to –68% for the A tranches. This year’s shock is more severe than the 2019 shock; however, the 2017 and 2018 shocks to the AA and A tranches of CLOs were more extreme. Since the bulk of bank CLO exposures are to the AAA tranches, this year’s CLO shock will likely generate higher projected losses under the severely adverse scenario than in prior years.
The other asset class mentioned in the Fed’s descriptions of the scenarios was private equity investments. These include investments on a principal basis in standalone companies, real estate, general and limited partnerships interests, and hedge funds, including seed capital. Because there are so many different types of private equity investments, the GMS includes 240 specific shocks segmented across the main industry groups and geographical areas. Figure 2 displays the distribution of the shocks to private equity investments in GMS 2020 (purple), with GMS 2016 (orange) and GMS 2013 (yellow) for comparison. Until now, the assumed decline in the value of private equity investments had been the greatest in GMS 2016. Specifically, the median private equity shock in the 2016 scenario was –47% compared with –35% in the 2013 scenario. The chart shows that the private equity shocks included in this year’s test are slightly more severe than in 2016 (median is –50%), as indicated by the slightly higher median value of the shocks, but the distribution in 2016 had a fatter left tail. Thus, without knowing the holdings of bank private equity investments, it is difficult to draw strong conclusions about the severity of this year’s private equity shocks compared with 2016. As a reference, the private equity index declined about 55 percent during the second half of 2008. It therefore appears to be in line with a calibration of the shock to the crisis.
In the remainder of this note, we look at the set of shocks across various asset classes also included in the SIFMA (2019) paper. The SIFMA study concluded that most of the shocks analyzed had a very small probability of occurring and were even less plausible when assumed to occur instantaneously. Figure 3 displays the shocks to corporate bond spreads, credit card asset-backed securities (AA-rated), and the S&P 500 Index. The dashed purple/orange/yellow lines represent the largest historical adverse market moves over 10-day, 3-month, and 6-month horizons, respectively.[3] Except for the shock to the B-rated corporate bond spread, the remaining shocks shown in Figure 3 are within the most severe realization registered over a 6-month period. Moreover, the BBB- and A-rated corporate bond spreads do not exceed the largest historical adverse move observed over a 3-month period. Finally, the shocks to the S&P 500 and the credit card ABS are less severe than the largest historical 10-day adverse move, including a noticeable decrease in the severity of the shock to AA-rated credit card ABS.
Figure 4 plots the shocks to Treasury rates, MBS spreads in the to-be-announced market, and EUR/USD and GBP/USD exchange rates. As shown by the purple dashed line, the 2020 shocks are very close to the 10-day largest adverse move in the price or spread of the underlying asset. This suggests that the calibration horizon of the GMS was more specific this year relative to prior years. For instance, shocks to assets traded in more liquid markets, such as the moves in the EUR/USD and GBP/USD factors, are now much less severe than the historical 10-day largest adverse in those currency pairs.
Banks are also required to use the GMS to estimate losses in the counterparty default scenario. This scenario assumes that the largest counterparty of each bank with substantial trading or custodial operations defaults unexpectedly in the first quarter of the planning horizon. However, the application of the GMS in the largest counterparty default (LCD) scenario leads to a substantial overstatement of risk. This is because the GMS horizon remains well in excess of the time required to close counterparty default exposures, even during periods of severe market dislocations. For example, the Basel III counterparty capital charges assume considerably shorter closeout periods, ranging from 5 to 10 days, unless there is a collateral dispute. Therefore, much more work needs to be done to make the LCD component more realistic.
SUMMARY REMARKS
This analysis finds that some of the market shocks are more severe than those included in previous years, such as the decline in the value of private equity investments and that of collateralized loan obligations. But the overall aggregate level of GMS losses for each bank will depend on each bank’s trading positions on the as-of date specified in the scenario, which makes it difficult to assess the severity of the GMS over time.
Still, it appears the overall calibration of the GMS remains quite extreme. As a result, significant work needs to be done to make the calibration of the GMS more transparent and the probability of the shocks occurring simultaneously more plausible.
Disclaimer: The views expressed in this post are those of the author(s) and do not necessarily reflect the position of the Bank Policy Institute or its membership, and are not intended to be, and should not be construed as, legal advice of any kind.
[1] See, Federal Reserve Board, “2020 Supervisory Scenarios for Annual Stress Tests Required under the Doff-Frank Act Stress Testing Rules and the Capital Plan Rule,” February 20. Available at https://www.federalreserve.gov/newsevents/pressreleases/files/bcreg20200206a1.pdf
[2] See Securities Industry Financial Markets Association, “Global Market Shock and Large Counterparty Default Study: Recommendations for Reforms Based on a Statistical Analysis of Stress Testing Scenarios,” August 2019. Available at https://www.sifma.org/wp-content/uploads/2019/09/SIFMA-GMS-LCD-Study-FINAL.pdf
[3] These estimates are taken from SIFMA (2019).
