Ten Precepts for 21st Century Regulators

This post is a written version of remarks that one of us (Steve) gave at the OECD in Paris on 12 September 2023. It expands on the discussion in our earlier post Making Banking Safe.

The regulatory reforms that followed the financial crisis of 2007-09 created a financial system that is far more resilient than the one in place 15 years ago. Yet, the events of March 2023 make clear that the progress thus far is simply not enough. To ensure resilience, we need to do more.

To steer the process of further reform, we propose a set of 10 precepts that those who make the rules should keep in mind as they refine the prudential framework. These practical guidelines lead us to conclusions that mirror those in a recent post: regulation should be more rule-based (less reliant on supervisory insight or discretion); simpler and more transparent; stricter and more rigorous; and more efficient in its use of resources. Concretely, this approach means increasing capital and liquidity requirements; shifting to mark-to-market accounting; and improving the transparency, flexibility and severity of capital and liquidity stress tests.

With that as a brief introduction, we now focus on the precepts that should guide the reform of financial stability policy. Taken together, we think they form a coherent policy framework.

1.     An ounce of prevention is worth a ton of remediation.

The first and most important practical rule is that prevention is far better than mitigation or remediation.

Put differently, acting before things turn bad is much better than cleaning up the mess after they do. Letting financial distress become an imminent systemic threat and then trying to restore stability—which we think of as remediation or mitigation after the fact—adds to risk-taking incentives and makes future stress more likely. To be sure, there is overlap between crisis prevention and mitigation—for example, in limiting spillovers—but they have vastly different incentive effects.

Prevention means that regulators should limit incentives for risk-taking, while also requiring that institutions build in shock absorbers. While supervision is necessary, experience teaches us that it is a poor tool for prevention. In practice, the most effective prevention mechanisms are those that both have the desired impact on incentives and do not depend on exceptional supervisory insight: namely, high capital and liquidity requirements.

Before continuing, we note our view that the issuance of subordinated debt by banks is a poor substitute for common equity. In theory, supervisors could use a weak bank’s bonds to recapitalize it. Yet, as we highlight below, policymakers are very unlikely to use them when alternatives (such as bailout loans) are available. As a result, the buyers of bank subordinated debt will not require a debt market premium consistent with the risk of bank failure.

2.     You will always bail (even medium-size) intermediaries out.

To limit risk-taking incentives, resolution regimes must be credible. They must compel shareholders and managers to internalize the impact of losses, rather than to spread it to others.

Lawmakers and regulators have been trying for years to set up a system where they can credibly commit not to bail out intermediaries that pose a systemic threat. To see why this is likely to fail, consider what you would do if confronted with the choice between bailing out a bank (or a set of banks) and risking a crisis that threatens the integrity of the payments system or undermines the supply of credit to healthy borrowers. Our guess is that most people do not wish to be remembered as the parent of the next Lehman disaster.

This lack of tolerance for the risk of crisis is what we saw in March 2023 in the case of SVB and Signature. As medium-sized banks, they were not treated as systemic in life. But, to protect other regional banks, this supervisory complacency gave way as soon as they failed. In practice, it would take only a few medium-sized bank failures to mimic the impact of the demise of a global systemically important bank (G-SIB). So, in retrospect, the use of emergency systemic response tools to resolve these banks was less an issue of too-big-to-fail than of too-many-to-fail.

The case of Credit Suisse was obviously different. But again, authorities blinked. They chose not to implement an existing resolution plan, and instead first lent massive amounts to CS and then subsidized a takeover. (See, for example, the report of an independent Swiss expert group regarding the 2023 resolution of Credit Suisse.)

Based on this experience, authorities should acknowledge that—except for the idiosyncratic failures of a few small banks—they will err on the side of caution and bail out even medium-sized banks. To put it bluntly, when it really matters, we doubt that supervisors will implement living wills and resolution plans—even if banks have them.

To make the financial system resilient and to limit risk-taking incentives, regulators must take this virtually inevitable failure of supervisory commitment into account. That means minimizing the probability that we ever get to this point in the first place.

3.     Supervision is insufficient, so rigorous capital and liquidity requirements are critical.

We see this precept as one of the big lessons of the 2023 banking stress. It also is one of the big lessons that we take away from the Barr report. Supervision is important and, as a recent paper from the IMF emphasizes, it ought to be greatly improved. In the same way that we need enforcement of speed limits on our roads, we need someone to enforce financial regulations: institutions and markets won’t do it on their own.

But, in the end, even the best supervisors will never catch everything. Facing high information costs, they may not even be able to detect risk taking. And, even when they do observe risky behavior, authorities will be cautious about requiring timely remedies, let alone imposing penalties, as the burden of proof is normally very high. Finally, different supervisors will have different and changing tolerances for certain types of risk taking.

The best way to counter these shortfalls of supervision is to make capital and liquidity requirements sufficiently rigorous.

4.     Early resolution reduces social costs and depends on proper measurement of capital.

Timely resolution is key both for incentives and efficiency. It reduces the ability of bank managers to extract from the resolution authority an expected rent (the value of a put option) that increases as it approaches default. It also reduces social costs.

To make timely resolution feasible, supervisors need to measure capital properly. This lesson is another outcome of the SVB episode. As interest rates rose, the value of SVB’s held-to-maturity securities fell. But these unrealized losses did not alter their measured regulatory capital. On a mark-to-market basis, the bank’s disclosures indicate that it was close to being insolvent by mid-2022. On this basis, it should have been shut down nearly a year before it failed. (See our discussion here.)

To improve the regulatory measure of capital, supervisors should shift toward mark-to-market accounting for the marketable assets that banks hold. They would then use that capital calculation to put a bank into resolution when the leverage ratio falls below the regulatory minimum (say, 2 percent), shutting them down pre-emptively.

(For options on reforming the role of deposit insurance in resolution, see here.)

5.     The social cost of bank equity capital is far lower than the private cost, so capital requirements should be higher than you think.

The Modigliani-Miller theorem teaches us that the value of a firm is independent of its capital structure. Put differently, the net present value of a firm’s cash flow does not change with its financing method. When the theorem holds, the marginal cost of equity funding is the same as it is for any other source of funding.

In practice, however, the Modigliani-Miller theorem fails for a variety of reasons. The two most important are the distortions arising from taxation and bankruptcy.

The differential tax treatment of debt and equity finance leads to a preference for debt. Importantly, however, this tax-induced preference creates a private cost of issuing equity, not a social cost.

Bankruptcy is a bit different, and far more pernicious. Because of limited liability, owners get the upside of any bet, but the extreme downside belongs to a bank’s counterparties and to the government. Again, this distortion creates a preference for debt financing, as it increases the value of the put option mentioned earlier. The closer the put is to being at the money—the more leveraged the bank—the more valuable is this option. In this case, requiring greater equity finance reduces social cost by countering the distortion: it increases the distance to default and lowers the value of the put. Put slightly differently, with a higher fraction of equity financing, the owners face more of the downside risk from the bank’s activities.

So, while bankers are correct to say that equity financing is privately expensive, social costs are clearly far lower.

In further support of increased capital requirements, we note the evidence from the BIS showing that better capitalized banks lend more, and that they lend to higher-quality borrowers (see our earlier post). In short, capital is the foundation for healthy lending. The opposite also is true: insufficient capital fosters zombie lending—to postpone the recognition of bank losses when borrowers fail.

How high should capital requirements be? For the largest banks, we suggest aiming for a leverage ratio of 15% (omitting reserves from the computation of total exposure). This ratio is similar to “The Minneapolis Plan” proposal for ending too-big-to-fail. For a typical bank, a 15% leverage ratio implies a risk-weighted capital requirement in the range of 30%. For comparison, the current risk-weighted requirement for the 34 largest U.S. banks ranges from 7.0 to 13.6 percent. But these organizations have a similar volume of long-term subordinated debt outstanding. As this debt approaches maturity, regulators should require SIFIs to issue equity to replace it.

6.     Those creating negative externalities do not voluntarily bear the cost. If they don’t complain loudly about the regulatory burden, regulators aren’t being tough enough.

Financial stability suffers from a problem of “the commons.” The incentive to take financial risk is analogous to the motivation for grazing on public lands or for fishing in public waters. Just as a farmer has the incentive to let their cows overgraze, leading to the starvation of others’ herds, an actor in the financial system has incentives to take risks that, because of spillovers, put others at risk by depleting systemic resilience.

Surely, no one who is depleting systemic resilience will voluntarily pay the price. When regulators force them to internalize these costs, they will complain. So, if financial entities are not complaining loudly, regulators probably are not doing enough to ensure financial resilience. Put differently, regulators and supervisors need very thick skins. They also need credible legal and political backing when complaints about supervision inevitably arise.

7.     To be effective, stress tests must be stressful, flexible, and (to some extent) transparent.

Earlier we highlighted how supervision typically falls short. Properly developed and executed stress tests can help fill this gap. In fact, stress tests may be the most important tool available to supervisors.

Specifically, these tests help us assess the capacity of a capital or liquidity buffer to cushion shocks. How risky are assets? What are the consequences of off-balance sheet exposures? What portion of capital is truly loss-absorbing when a severely adverse shock hits? What will happen to capital levels in the event of a severe recession that causes widespread distress? What are the spillovers from the dysfunctionality of key intermediaries? How can a bank or other intermediary meet liquidity needs when normal funding mechanisms dry up or become too costly?

A stress testing regime can be assessed based on three attributes: severity, flexibility and transparency. The mix of these characteristics determines the regime’s effectiveness.

If the scenarios are insufficiently dire, there is no point to the test. Importantly, the scale of the adverse stress must take account of the fact that we do not know how to conduct dynamic stress tests with multiple rounds in which institutions first respond to an initial shock and then react to the responses of others. To mimic this amplification mechanism, a crude, one-shot test—of the kind that we can implement—must impose initial shocks that are larger than anything in the historical record.

Second, flexibility is essential. Supervisors must change the scenarios aggressively and unexpectedly, and they should disclose them only after the date on which institutions fix their balance sheet (and off-balance sheet) exposures. Third, and closely related, there is considerable room for transparency, but there are limits. If professors were to publish tests before students take them, they wouldn’t be tests.

8.     Follow the money. What is most profitable (or growing fastest or done in the largest quantity) is often the biggest source of systemic risk.

Supervisors should always use their scarce resources to focus on what’s most profitable, what’s growing the fastest, and on what’s done in the largest amounts. These are likely to be the things that pose the greatest risk to the system.

To be clear, when you see a financial firm that is highly profitable, there are only three explanations for their apparent success. They could be geniuses and have made a unique discovery. They could just be taking more risk. Or they could be doing something illegal. Assuming we adequately police illicit activity, we can rule out the last of these. And, while there are some geniuses out there, there are likely only a few. This leaves risk taking as the most likely and most frequent source of exceptional profitability. So, if you follow the money—and exclude a few geniuses as well as the crooks—you will find the risk.

9.     Regulation is an arms race between public and private entities that authorities typically lose.

Everyone appreciates that the financial system is constantly evolving—sometimes very quickly. Some aspects of financial evolution are good, some are not. The good ones, like improvements in access to basic financial services, can create large social and economic benefits.

But not all innovation improves welfare. Financial institutions devote substantial resources to circumventing regulatory rules through legal and accounting schemes that sustain risk-taking without improving the allocation of resources. They pay handsomely for expertise that helps them meet the letter of new regulations while minimizing their functional impact. In the process, their large and costly legal, risk management and financial engineering operations both influence and exploit the official rules and regulations.

Years ago we proposed the creation of a regulatory counterintelligence agency that would design regulation to anticipate gaming and arbitrage by the industry. We now doubt the practicality of this proposal, as it would be both costly and require extraordinary cooperation across jurisdictions.

The realistic alternative is higher capital and liquidity requirements, which naturally limit the incentives for gaming. As we mentioned at the outset, an ounce of prevention is worth a ton of remediation.

To be sure, raising capital and liquidity requirements solely on banks would fuel the long-run shift of financial activity beyond the regulatory perimeter to nonbank financial intermediaries (NBFIs). Confronting this nonbank problem requires an enormous shift from regulation by legal form of organization to regulation by function. While this change is obviously difficult, it is essential for financial resilience. (See our earlier post here.)

10.  Other jurisdictions will always do what’s best for them regardless of what is good for you.

Because of spillovers, ideally both macroprudential and resolution policy would be coordinated across jurisdictions (see here). In practice, the scope for cooperation is very limited. Each of you has a mandate to contribute to the well-being of the people in your country. This mandate does not include sacrificing the welfare of your citizens for those abroad.

So, when you and your colleagues make decisions, you may take account of the impact elsewhere, but only insofar as that feeds back to you. While such noncooperative behavior is likely to prove suboptimal in the world of global finance, there is no general remedy. As one example, recall the famous admonition regarding spillovers from multi-jurisdictional bankruptcies: “financial institutions may be global in life, but they are national in death.”

To encourage awareness of the limits to cross-border cooperation and curb any resulting loss of trust, it helps to make regulatory goals and constraints as transparent as possible. It also makes sense to develop detailed cooperative agreements where they can be most beneficial (for example, in the resolution of global banks and nonbanks in the largest financial centers).

Conclusion

To conclude, we reiterate the inferences that we draw from these precepts: regulation should be more rule-based (less reliant on supervisory insight and discretion); simpler and more transparent; stricter and more rigorous; and more efficient in its use of resources. This means substantially increasing capital and liquidity requirements; shifting to mark-to-market accounting; and improving the severity, flexibility and ex-post transparency of capital and liquidity stress tests. Furthermore we need to ensure that resolution regimes are pre-emptive not reactive.