EDITOR’S NOTE: This article is excerpted from a paper of the same title, published in Brookings Papers on Economic Activity in Sept. 2017.
IN THE WAKE of the global financial crisis of 2008–2009, financial regulation has undergone a dramatic overhaul, in the U.S. and elsewhere. There are many elements to the new regulatory regime, including enhanced capital requirements and stress testing, liquidity rules, resolution planning, margin and clearing requirements for derivatives transactions, and much more. With the bulk of the rule-making and implementation now nearly complete, it is a natural time to take stock of the changes: to ask whether the new regulations are working as hoped, how they are meshing with one another, and what the unintended consequences and other inefficiencies might be.
This article discusses three principles that can be used to assess the efficiency of those parts of the regulatory regime that are most directly tied to bank equity capital, including the standard risk-based Basel III capital requirements, the leverage ratio, and the Federal Reserve’s stress-testing process. Although these are far from being the whole regulatory toolkit, they are among the most important pieces of it, and these elements alone have become very complex.
PRINCIPLE 1: CONSOLIDATE CONSTRAINTS
With respect to capital regulation in particular, basic economic theory suggests that, in a “normal times” steady state, one can achieve a relatively efficient regulatory outcome that leads banks to properly internalize the social costs of their activities with a single capital requirement that forces each bank to maintain a sufficient ratio of equity to risk-weighted assets, provided the risk weights are chosen appropriately. Moreover, requiring different banks to maintain different ratios of equity to risk-weighted assets, as Basel III does with its capital-ratio surcharges for global significantly important banks (G-SIBs), also makes sense within the context of the same framework.
However, crucially, the same economic logic does not support having multiple independent constraints on bank equity ratios – as is the case when banks have to separately satisfy minimum values for their risk-based capital ratios, their leverage ratios, and their post-stress capital ratios. This is because when banks have heterogeneous business models, different constraints can bind in equilibrium for different banks. As a result, two different banks can face different relative risk weights when performing the same two activities, which distorts their behavior, just as would happen if different non-financial firms faced different relative marginal tax rates for the same two activities. Thus, a first core design principle is that wherever possible, multiple constraints on the minimum level of equity capital should be consolidated into a single risk-based constraint.
PRINCIPLE 2: DYNAMIC RESILIENCE
Next, it is important to think about what optimal regulation looks like away from the steady state, when the banking system has been hit with a negative shock that reduces its capital base below the natural long-run level. In this case, as long as there are flow costs to raising new external equity, ratio-based capital requirements are not sufficient to implement an efficient outcome. Rather, in addition to specifying capital ratios, the regulator must also compel banks to recapitalize – i.e., to raise new dollars of outside equity, above and beyond what they would voluntarily do on their own. Thus, a second design principle is an emphasis on what might be called dynamic resilience: After an adverse shock, the ability of regulators to implement a prompt recapitalization of the banking system is at least as important as setting the exact value of the capital ratio in normal times. In many ways, this is an obvious point, but it’s one that has been underappreciated in much of the work in this area, which has been more concerned with calibrating static optimal capital ratios.
PRINCIPLE 3: COMBAT REGULATORY ARBITRAGE BY FILLING IN CONTINGENCIES EX POST
Finally, one needs to ask how regulators can best respond to the inevitable gaming of any rules that they write down. A natural instinct when seeing that one particular rule (say, a risk-based capital requirement) has been arbitraged is to propose another rule that the historical data suggests would have worked better. This, in part, is the logic invoked by those arguing for a more prominent role for a non-risk-based leverage ratio requirement. But it is useful to bear in mind the wisdom in Goodhart’s law: “Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.” In other words, any rule, once codified ex ante, will tend to be arbitraged, and this problem cannot be easily addressed by proposing more rules. Rather, a third core principle is that regulators should explicitly aim to take an incomplete contracting approach, filling in certain contingencies ex post, once they have observed how banks are responding to the existing rules.
These three principles in turn lead to the following specific policy recommendations for updating and strengthening the current capital-regulation regime:
DIAL BACK THE SUPPLEMENTARY LEVERAGE RATIO (SLR): Having an SLR that is either binding or near-binding is counterproductive and distortionary. There are two broad ways that the SLR could be made to be less constraining on bank behavior. First, the minimum level of the ratio could be reduced – for example, from 5% to 3% for the G-SIBs. Alternatively, the denominator of the ratio could be adjusted to exclude the very safest assets, including central bank reserves, Treasury securities, and initial margin for centrally cleared derivatives. In principle, either approach could serve the desired purpose, though the latter has the drawback that it could create a sharp cliff between Treasuries, which would now have a zero risk weight, and near-riskless substitutes (for example, agency and highly rated corporate bonds). However, to the extent that either approach makes the SLR less likely to bind at all, this distinction will tend to matter less.
Although a reduced role for the SLR is desirable, many of the concerns that have motivated its advocates are absolutely legitimate, namely: the general potential for the current risk-based regime to be gamed; the particular vulnerability to such gaming of complex model-based approaches to setting risk weights; and the lack of any risk weight at all on even relatively risky sovereign securities. The remaining recommendations attempt to address some of these concerns.
INTEGRATE THE RISK-BASED CAPITAL REQUIREMENT AND THE COMPREHENSIVE CAPITAL ANALYSIS AND REVIEW (CCAR) INTO A SINGLE CONSTRAINT: One way to accomplish this integration is put forward in a recent proposal by the Fed. The idea is that there would just be a single overarching risk-based capital requirement. It would start with a baseline risk-based ratio. But then this ratio would be augmented with a “stress capital buffer” that incorporates estimates of net losses coming from the annual CCAR process. And this stress capital buffer would be subject to a 2.5% floor.
In other words, each bank would now face just a single constraint, and the effective risk weight for any asset i would be the sum of a statutory time-invariant Basel-style risk weight and a component that reflects asset i’s performance in the severely adverse stress scenario. Moreover, the latter piece would not be set in stone but could vary year-to-year. Although there would be time-variation in risk weights, the fact that there is only a single binding constraint at any point in time implies that all banks face the same cross-sectional tax rates on their activities, which is the key to minimizing the sorts of industry-level distortions described above.
DESIGN ANNUAL STRESS SCENARIOS WITH REGULATORY ARBITRAGE IN MIND: At first glance, one reaction to a consolidated constraint of the sort described just above might be that it is just a relabeling of the usual risk-based capital requirement. If so, one might ask what the independent role of the stress testing process is – i.e., why do we need the CCAR when it is just being folded into the conventional risk-based capital regime?
Again, the point to emphasize is that the component contributed by the CCAR to an asset’s effective capital charge, namely that coming from the stress capital buffer, is not a time-invariant constant based on a rule-making process like the usual risk weights, but rather it is free to vary with each year’s design of the stress scenario. To take maximum advantage of this flexibility, it should be used proactively to combat regulatory arbitrage. One way to do so would be to purposefully design each year’s CCAR stress scenarios to react to rapid growth or surprisingly high profitability in particular lines of activity. This could be done at quite a granular level. Indeed, a natural starting point for the exercise might be to have supervisors ask who, for example, the 20 most highly compensated line managers or traders are in each big bank each year, and then to think about stressing the exposures most closely associated with these employees. The underlying idea is to learn as much as possible about the incentives at play by observing the behavior of bank executives, and then to condition the CCAR design based on what is learned from this behavior. (NOTE: An approach of this sort may well have surfaced the “London Whale” risk exposures that lost JPMorgan over $6 billion in 2012, as managers and traders responsible for the risk were among the highest paid in the organization. See “JPMorgan Chase Whale Trades: A Case History of Derivatives Risks and Abuses, Majority and Minority Staff Report,” Permanent Subcommittee on Investigations, United States Senate, pp. 57-59.)
MAKE USE OF THE COUNTERCYCLICAL PROPERTIES OF THE STRESS CAPITAL BUFFER: There is one other way to take advantage of the potential for time variation in the stress capital buffer. In the wake of a negative shock to banking system capital, one part of the optimal response is to relax the required capital ratio requirement in a countercyclical manner. This will happen naturally – and in a broadly symmetric way across banks – if there is a single capital requirement that incorporates a stress capital buffer, provided that the underlying stress scenario envisions less further deterioration in the macro environment once the economy has already declined significantly. (NOTE: When the unemployment rate is 5%, the severely adverse scenario might contemplate unemployment rising by 5 percentage points, to 10%. But when the unemployment rate has already hit 8%, the further increase modeled in the severely adverse scenario might be only 4 percentage points, to 12%. This would tend to reduce the going-forward stress capital buffer, all else being equal.)
This sort of macro-sensitivity is already incorporated in the design of the underlying CCAR scenarios. But under the current regime with multiple constraints, it does not have as uniform a countercyclical effect on required capital because not all banks are equally bound by the post-stress capital requirement. Moreover, while there is also a separate formal countercyclical buffer built into the standard risk-based Basel regime, this buffer has not to date been deployed by U.S. regulators. This may in part reflect the political-economy challenges associated with varying an explicit and highly visible statutory requirement. By contrast, if the countercyclical variation is instead an implicit by-product of changes in the annual CCAR assumptions, it may be easier to implement on a semidiscretionary basis.
CONSIDER INCREASING G-SIB SURCHARGES: Although it is important to be cautious about the potential distortions associated with imposing different cross-sectional risk weights on different banks, it also the case that, holding the structure of these risk weights fixed, it can make good sense to set a higher overall minimum capital ratio on those banks whose failure creates larger social costs. So, something very much like the existing G-SIB surcharges should continue to play a role in any blended requirement – that is, the baseline capital-ratio requirement should be higher for the largest and most systemic firms.
Moreover, to the extent that multiple binding constraints such as the SLR have reflected a general desire to push more capital into the biggest banks, this objective can be accomplished more efficiently using the G-SIB surcharges. This would increase the overall amount of capital in the banking system and would do so without creating the sort of distortionary cross-bank activity-migration incentives discussed above. If anything, higher and more progressive G-SIB surcharges might have a beneficial incentive effect, by encouraging the largest banks to exit those lines of business where they do not create enough in synergies to outweigh the added social costs associated with their size and interconnectedness.
STRENGTHEN THE CCAR PROCESS AND INFRASTRUCTURE WITH AN EMPHASIS ON DYNAMIC RESILIENCE: The above recommendations for adapting the CCAR process all refer to how it should be used in normal times. In particular, these recommendations are all in the spirit of integrating the stress-testing process more tightly and efficiently into the normal-times regime of setting minimum risk-based capital ratios. But it would be a mistake to think of this as the only role for CCAR. Another vital aspect of stress testing – indeed, much of purpose of the original 2009 Supervisory Capital Assessment Program (SCAP) – is not to regulate capital ratios ex ante but, rather, to promote a rapid recapitalization of the banking system in the wake of a large negative shock.
To put this point into perspective, it is useful to think back to how events unfolded during the early stages of the financial crisis. Problems with subprime mortgages were already surfacing in late 2006. The first serious tremors associated with the crisis were felt in August 2007, with investor runs on multiple asset-backed commercial paper programs. At this point, there was no longer any real doubt about the nature of the shock confronting the financial system, even if its exact magnitude was yet to be determined. And yet during the interval from the start of 2007 through the third quarter of 2008, the largest U.S. financial firms – which, collectively would go on to charge off $375 billion of loans over the next 12 quarters – paid out almost $125 billion in cash to their shareholders via common dividends and share repurchases, while raising only $41 billion in new common equity. This all happened while there was a clear and growing market awareness of the solvency challenges they were facing. Indeed, the aggregate market capitalization of these firms fell by approximately 50% in the pre-Lehman period from the start of 2007 through the end of June 2008.
It seems indisputable that the severity of the crisis would have been mitigated if policymakers had clamped down on these payouts earlier and had compelled banks to raise substantial amounts of new equity. With this observation in mind, a central question to ask about CCAR is this: Suppose we were granted a do-over, and it was late 2007. If we had the current CCAR process in place, would things have turned out differently? Would we have seen significantly more equity issuance at this earlier date by the big banks, and hence a better outcome for the real economy?
The answers to these questions are not entirely clear. On the one hand, the rule underpinning the current CCAR framework gives the Federal Reserve the authority to curtail a bank’s payouts to shareholders in the event that its post-stress capital ratios fall below the specified minimum. There is somewhat more ambiguity as to whether the same rule also gives the Fed the authority to compel new equity issues, as opposed to letting a bank come into compliance with its required post-stress capital ratio by shrinking its assets – e.g., by slowing its loan growth or by selling assets. So, one useful direction for reform is to strengthen the CCAR rule so as to make it clear that the Fed does indeed have the authority to compel new equity issues when doing so is necessary to prevent an undesirable contraction in bank balance sheets at a time of macroeconomic stress.
At the same time, having the authority to do something is necessary, but it is not sufficient – there also needs to be the institutional resolve to follow through. And such resolve can be hard to come by at a time of systemwide stress, when banks can be expected to object strenuously to having to do what they perceive to be highly dilutive equity issues, and when regulators are likely to be skittish about further unsettling the market for bank stocks. Thus, in addition to rewriting the formal CCAR rule, another important aspect of the annual CCAR process should be an explicit form of war-gaming, whereby regulators rehearse the details – both among themselves and in cooperation with bank executives – of exactly how they would go about implementing a rapid recapitalization of the system in the face of large looming losses. The hope would be that repeated rounds of such war-gaming would help to build the institutional culture and muscle memory needed to go forward with an aggressive systemwide recapitalization plan when the time comes.
Finally, and also in the spirit of buttressing institutional resolve, whenever the Fed designs a CCAR stress scenario, it should be publicly accountable after the fact to explain how its assumptions for loan losses and other outcomes can be reconciled with the information in bank stock prices and credit default swap (CDS) spreads – particularly at times when these market prices are sending pessimistic signals.
Think again of the period from early 2007 to mid-2008, when bank stocks fell by about 50%. If a CCAR adverse scenario is being drawn up in a mid-2008-like environment, it seems hard to argue that it shouldn’t take on board the growing market skepticism about the state of bank balance sheets. Moreover, doing so should serve to heighten the pressure on regulators to push for a rapid recapitalization of the banking system. Of course, any market indicator can be driven by noise as well as news, and so it probably doesn’t make sense to have a mechanical rule tying market prices either to CCAR assumptions or to recapitalization requirements. But the current system, which has no real role for market-based information, is also far from optimal in this regard.
In closing, it is useful to note some caveats and qualifications. Perhaps the most important of these has to do with the limits of discretion in regulatory practice. A central theme of this article has been that it would be beneficial to rely less on multiple overlapping rules (such as risk-based capital ratios and leverage ratios) to deal with the challenging problem of regulatory arbitrage, and to instead give regulators more flexibility to respond to such behavior ex post, most importantly in the design of CCAR stress scenarios. In a similar vein, it has also been argued that CCAR stress scenarios should be responsive to movements in bank stock prices and CDS spreads, without necessarily writing these variables into a rule ex ante.
However, such a discretionary approach might not work quite as well as hoped. First, and most simply, the regulatory process might not be as nimble and flexible as it needs to be to create the desired benefits. For example, consider the suggestion that regulators look for areas in a bank where growth and profits are unexpectedly strong, or where compensation is unusually high, as clues to pockets of emerging risk and/or gaming of the rules. But what kinds of activities would actually be singled out in the course of such an exercise, and how useful would the information turn out to be? Absent any concrete evidence, it is hard to be fully confident. While this is not a good reason to dismiss a more discretionary approach out of hand, it may suggest that the most constructive first step would be for Fed officials to conduct some in-house trial-run testing of the approach before implementing it in practice.
Another potential concern for a more discretionary approach is that it can invite complaints from regulated banks about the CCAR process being nontransparent, arbitrary, and lacking in due process. Consider how a bank might respond if it is told that it is facing tougher assumptions on loss rates in a given year simply because it has been particularly profitable in some areas or is paying some of its employees in these areas generously. At the extreme, such complaints could manifest in legal challenges under the Administrative Procedure Act. And even if they did not, the associated pushback and political pressure might ultimately weaken regulators’ hands to the point where the discretionary approach becomes ineffective.
These are difficult issues and should not be minimized. Yet it may be possible to make some progress on them by taking the transparency bull more firmly by the horns. That is, the Fed should be very explicit about its theory of the case with respect to any aspect of the CCAR process that can be seen as less than completely transparent, and it should be committed to full transparency in those cases not covered by the theory. One distinction that may be helpful here is that between ex ante versus ex post transparency. There are good reasons why complete ex ante transparency – in the sense of telling the banks ahead of time what all the modeling parameters for the CCAR stress scenario for a given round will be – is undesirable. If one takes this form of ex ante transparency to an extreme, the CCAR degenerates into just another hard-coded capital rule, with all the associated vulnerability to regulatory arbitrage.
On the other hand, this argument does not imply similar costs to ex post transparency. So, absent a fundamentally different theory of the case, the Fed should be expected to disclose in significant detail after each year’s CCAR the specific analysis and evidence that led it to vary – e.g., the modeled loss rates for individual bank-by-asset-type categories relative to prior rounds.
All of this is in the same broad spirit as the Fed chair regularly testifying before Congress to explain monetary policy ex post, without necessarily committing to a monetary policy rule ex ante. And the hope would be that, over time, such ex post disclosure would enhance the Fed’s credibility with respect to how it handles its regulatory discretion and would therefore make a regime that relies on such discretion more politically resilient and ultimately more durable.