Understanding the Implications of Fed Governor Tarullo’s Suggested Changes to Stress Testing

Understanding the Implications of Fed Governor Tarullo’s Suggested Changes to Stress Testing

Yesterday, Federal Reserve Board Governor Daniel Tarullo gave a speech, Next Steps in the Evolution of Stress Testing, at the Yale School of Management. Governor Tarullo’s remarks covered a wide range of issues surrounding stress testing, including a new way of conceptualizing the stress test using a Stress Capital Buffer. Most of these concepts were not included in the proposed rule issued by the Board on the same day, but his remarks merit serious study as the Board works to develop a future proposal. Set forth below are some initial impressions and areas for future analysis and research, though we caution the speech does not provide complete details for its various proposals, so our analysis is necessarily limited and preliminary.

Creation of a “Stress Capital Buffer” and Changes to Key Assumptions

The speech describes a “stress capital buffer” (SCB) approach to setting post-stress capital requirements, which would alter the overall structure of CCAR. Today, under CCAR, the Fed projects revenues, losses, capital distributions and balance sheet changes over a nine-quarter period of severely adverse stress, and banks must demonstrate that they would meet a series of minimum capital requirements – for example, a 4.5 percent minimum common equity Tier 1 (CET1) ratio – after the stress. To simplify this assessment, the Fed’s new approach would appear to convert the CCAR stress test into a new, additional risk-based capital requirement, in which the new SCB would substitute for the capital conservation buffer. Under that requirement, a bank would need to maintain CET1 capital equal to the sum of (i) 4.5 percent, (ii) its SCB, and (iii) its GSIB surcharge, if applicable. The SCB would be the greater of (i) 2.5 percent or (ii) the maximum decline in the bank’s CET1 ratio under the CCAR severely adverse scenario. (The speech is not clear on whether and how the stress capital buffer would apply to existing CCAR post-stress risk-based capital minimum requirements other than the CET1 requirement. Because the SCB is considered a substitute for the capital conservation buffer, we assume that it – like the capital conservation buffer currently – would not be combined with non-risk-based capital measures.)

Although we believe a GSIB surcharge component of the SCB is not appropriate, as we describe in detail below, the SCB approach itself contains several ideas that we believe merit further analysis and study:

  • The speech notes that the Federal Reserve may end its current practice of assuming that a bank would continue making dividends and share repurchases under stress that would be prohibited by operation of the capital conservation buffer rule. In particular, the speech notes that the Fed may instead only assume that a bank will maintain its dividends for a one-year period, which is a much more realistic approach, given experience during the financial crisis.
  • The speech notes that the Federal Reserve may eliminate its current soft limits on capital distributions (i.e., its enhanced scrutiny of proposed capital plans with a dividend payout ratio above 30 percent). These limits have not been well grounded in analysis, and have appeared superfluous, so their elimination appears warranted.
  • The speech notes that the Federal Reserve may end its current practice of assuming, counterfactually, that bank balance sheets will grow during times of severely adverse economic stress; instead, it may assume that bank balance sheets would remain static in such circumstances. This is certainly more realistic, but we believe further consideration of a more nuanced, line-of-business approach is warranted.
  • While details were not provided, it appears that the new approach would result in an ongoing capital requirement, rather than the current point-in-time approval of a capital plan (with deviations from that plan not permitted without a resubmission). This might mean that a bank that was above its minimum CCAR capital requirement could at any point during the year make an unplanned share repurchase or dividend payment without resubmitting its CCAR plan. On the other hand, a bank could be prohibited from making a pre-planned dividend payment or repurchase if it had suffered losses, and such actions would take it below its CCAR minimum requirement. If this is a correct reading, the costs and benefits of this resulting lack of predictability are worth further study.

Incorporating the GSIB Surcharge

In describing the potential introduction of an SCB approach, the speech notes that the Fed may add a bank’s GSIB surcharge to the effective CCAR post-stress minimum capital requirements. As we have described in previous research, there are significant analytical, policy and practical problems that strongly caution against such an approach, including significant limitations and weaknesses in the methodological framework used by the Federal Reserve to calibrate the GSIB surcharge itself. We further describe several of these concerns here.

While the speech does not consider any of the potential costs to markets or the economy of incorporating a GSIB surcharge into CCAR, we believe the Board should do so prior to advancing such a proposal. As an effective matter, the GSIB surcharge is a tax on capital markets activities.1  Thus, its addition to CCAR minimums will give banks a more powerful incentive to reduce the activities that produce it.

As a conceptual matter, the speech disputes the notion that inclusion of the GSIB surcharge duplicates the global market shock and counterparty default shock already imposed on GSIBs. It notes, correctly, that ”these additional scenario components capture direct losses to which any firm engaged in material trading activity is exposed” and, “[t]hese components of our stress test currently apply only to GSIBs because GSIBs are the only firms currently in CCAR for which these exposures are material.” However, if one considers a GSIB’s retail and wholesale businesses, it remains the case that CCAR currently stresses both of them; however, the addition of the GSIB surcharge would result in stressing the wholesale businesses twice, and the retail businesses only once. One can pose “macroprudential” justifications for doing so, but the effect is the same: still greater incentives for large U.S. banks to reduce their double-taxed support for capital market funding for U.S. businesses.

Relatedly, the speech seems to dismiss the need for consideration of how other post-crisis regulations has reduced the “macroprudential” risks of wholesale businesses. Given that various liquidity, margin, funding and derivatives regulations were expressly designed by the Federal Reserve and other regulators to reduce the systemic risk posed by these businesses, this dismissal is hard to understand. After all, as the speech itself notes, it is the “adverse impact of a GSIB failure on the financial system as a whole” that is “the basis for the GSIB surcharge.” It would thus seem to follow that regulations enacted to reduce exactly that – the adverse impact of a GSIB on the financial system – would be quite relevant in calculating any GSIB surcharge that the Fed may integrate into CCAR. As noted, at the end of the day, a firm’s GSIB surcharge is effectively a simple function of the size of its securities portfolios, derivatives exposures, and market funding. Yet the existing GSIB surcharge equates the adverse systemic impact of a bank that complies with the new derivative stay protocols, collects and posts mandatory and substantial amounts of margin for its non-cleared swaps, complies with more stringent limits on its counterparty exposures, and maintains the proposed mandatory amounts of total loss absorbing capacity (TLAC) under the new legal resolution regime, as exactly the same as one that does not do any of these things. On this point, the speech concludes that “current work on orderly resolution is surely no argument against the promotion of systemic stability and macroprudential aims by integrating the surcharge into CCAR.” But it would seem that work that reduces the systemic impact of a GSIB’s failure certainly should be reflected should the Fed choose to incorporate into CCAR a surcharge that is intended to measure and tax exactly that thing.

The speech also notes that such measures are ongoing, rather than complete. However, by the Fed’s own account, these should be finalized within the next few months.

Changes to Scenario Design

The speech also notes that the Fed is considering making some revisions to the Board’s “Policy Statement on Scenario Design Framework for Stress Testing” – namely, making the severity of the change in the unemployment rate less severe during economic downturns. These changes have the potential to resolve concerns we raised in a TCH research note earlier this year. In particular, the CCAR 2016 severely adverse scenario assumed a recession that includes an increase in the unemployment rate that is more severe than prior years’ scenarios, and considerably more severe than the 2007-2009 financial crisis. In addition, the increase in the unemployment rate in the 2016 scenario was substantially more sudden than was experienced during the 2007-2009 financial crisis, which is likely to cause simulated losses to accumulate rapidly and in greater amounts over the stress period. By more severely stressing unemployment rate changes, the 2016 Federal Reserve scenarios tend to discourage lending to households and small businesses and, more broadly, create incentives for banks to shift away from loans whose performance is especially sensitive to unemployment rates.

The Incorporation of “More Macroprudential Elements”

The speech identifies the incorporation of “more macroprudential elements” into CCAR as a key policy goal of the Fed, but there is a serious question of whether that expansive goal can be achieved through further stress-based capital charges on eight U.S. banks, particularly as much of finance migrates away from them (for regulatory reasons, among others). There are innumerable examples. The largest Designated Market Maker on the NYSE is not a GSIB. A hedge fund is now the largest dealer in U.S. interest-rate swaps. According to a report, in the interdealer market for US Treasuries, eight of the top ten firms ranked by volume on BrokerTec over May and June of 2015 were non-banks; those non-bank firms accounted for 85 percent of the volume. Of course, none of these firms are subject to CCAR, and the addition of “macroprudential” elements to CCAR would have no effect on them – except to expand their market share by making GSIBs less competitive in those markets. Thus, paradoxically, as more “macorprudential” elemental are added to CCAR, its macroprudential effectiveness could actually decrease.

Furthermore, in other contexts, proponents of the GSIB surcharge’s incorporation into CCAR have argued that GSIBs need to emerge from stress with higher capital levels in order to face the market. We believe that this concern may be exactly backward: if a stress of the type specified in CCAR actually were to occur, the better question might be whether there is a market left to face the U.S. GSIBs. Thus, we strongly urge (and expect to undertake) research into this question: what would be the effect of a CCAR stress on the rest of the financial sector, and how would the resulting crisis be managed? It seems quite plausible that such a stress would cause the failure of hundreds of small banks, the collapse of numerous market-funded lenders and hedge funds, and other dislocations among non-banks. For GSIBs (and other large banks), it seems unlikely that the concern would be their lacking capital, and quite likely that the concern would be regulatory and legal incentives causing them to hoard their liquidity, turn away deposits, and refuse to purchase troubled non-banks and small banks.

The speech also provides a lengthy discussion of potential features that might be introduced to make CCAR “more macroprudential.” In particular, the speech describes potential changes to address funding shocks, liquidity shocks, fire sale dynamics, common counterparty defaults and central counterparty reactions. Taking advantage of the horizontal nature of CCAR, as well as other horizontal supervisory tests, to assess interlinkages and concentrations that give rise to macroprudential risks seems highly worthwhile. But it will be critical when making such an assessment to take into consideration all of the wide range of existing, other macroprudential regulations and horizontal supervisory exercises that already require banks to take numerous steps to mitigate exactly these risks (Basel III liquidity rules, single-counterparty credit limits, etc.).


The speech includes a lengthy discussion of transparency in CCAR, and notes that the Fed does not intend to make the supervisory models by which is projects stressed revenues and losses public. In doing so it argues:

Full disclosure would permit firms to game the system–that is, to optimize portfolio characteristics based on the parameters of the model and take risks in areas not well-captured by the stress test just to minimize the estimated stress losses. In part for this reason, full disclosure could promote a “model monoculture” in which both supervisors and firms use the same models to evaluate risks. A critical part of the DFAST/CCAR process is that firms use their own models to assess their risks. We do not want them simply to copy the supervisory model and thereby increase the vulnerability of the financial system to the inevitable blind spots in even the best models.

We have some sympathy for this notion, for reasons worth exploring. CCAR establishes extremely high capital requirements, and does so on a very granular basis. Banks submit data differentiating among assets at a high level of detail – for example, distinguishing loans to purchase an automobile based on thousands of different criteria. Thus, if the results of the CCAR auto loan models were known, CCAR could take on high capital allocation powers: for example, banks would vary credit availability and pricing based on CCAR results for a given type of auto loan. We would not refer to this possibility as banks “gaming the system” but rather banks rationally responding to government-set hurdle rates – that is, not treachery, but obedience. Of course, to the extent that those hurdle rates are driving banks away from making independent credit decisions based on economic capital analysis, one could simply lower them.

Another alternative to mitigate this effect would be to reduce the number of capital measures for which post-stress minimums are identified – ideally, to one: the CET1 to risk-weighted measures used by Governor Tarullo in presenting how the new SCB regime would function. Clearly, a post-stress minimum leverage ratio has the greatest potential to produce capital misallocations, and as we have demonstrated, the CCAR leverage ratio was the binding constraint for 14 (of which 6 are GSIBs) of 33 banks in the last CCAR round.

Still, any benefits of regulatory opacity need to be considered against significant competing concerns. First, the concern that banks would or could “game the system” would also seem to warrant a range of other salient considerations, including (i) the extent to which bank and public awareness of the Fed’s own perceptions of risk might actually be beneficial to financial stability; (ii) the extent to which academic and practitioner analysis could result in the Fed’s models – which are becoming vitally important to U.S. economic growth – being improved; and (ii) the extent to which the Fed could develop, as an alternative to regulatory opacity, other means by which it might monitor and restrict the “portfolio optimization” with which it is concerned. Second, transparent regulatory assumptions have proven manageable in the actual risk-based capital regime, where regulators have also established a comprehensive range of risk assumptions in the form of risk weights – which have always been made public. Ultimately, if the goal of CCAR is to ensure that banks have sufficient capital to withstand a severely adverse stress scenario, why is it appropriate to maintain secrecy around the models that used to determine whether they could do so?

The speech also notes that “[e]ffective prudential regulation must be equally dynamic and should try to avoid pushing major financial firms toward measuring all of their risk positions in exactly the same way.” While we have sympathy for this notion as well, we do note that it is difficult to reconcile with the overwhelming preference shown by the Fed and other regulators for the use of standardized, government-established measures of risk in capital rules, liquidity rules, counterparty credit limits, incentive compensation, and other prudential regulation.

Other Key Issues

Governor Tarullo notes, and the proposed rule issued yesterday provides, that the Federal Reserve (i) will be eliminating the qualitative CCAR exercise for smaller financial institutions, highlighting the “unnecessary burden” caused by this process and (ii) is “progressively integrating the qualitative elements of CCAR into our year-round supervisory program for the largest banking organizations”; we commend each of these changes. Once the integration is complete however, we would recommend that the Board consider eliminating the annual qualitative exercise altogether. This is not because the qualitative components of stress testing are unimportant. Rather, the existing bank examination process is considered the best means of supervising credit underwriting, cyber security, consumer protection and numerous other aspects of banking. We do not see any legitimate basis for concluding that only capital planning must be overseen through a public up-or-down, binary regulatory judgment at the end of each year.

1 Consider the five factors that determine a bank’s GSIB surcharge:

  • The complexity factor includes almost exclusively securities and derivatives assets held in market making (as opposed to loans held as part of commercial banking);
  • The inter-connectedness factor includes almost exclusively dealer-to-dealer trading assets held in order to hedge customer positions held in market making;
  • The cross-jurisdiction factor includes almost exclusively cross-border dealer-to-dealer trading of the type captured by the interconnectedness factor;
  • The short-term wholesale funding factor includes almost exclusively the funding of securities positions; and
  • The size factor is not so exclusively focused on securities activities, but for the largest banks still comprises those assets as a large percentage of a firm’s total assets.