Gary H. Stern
Federal Reserve Bank of Minneapolis
The 35th Annual Conference on Bank Structure and Competition
May 6, 1999
The process of reforming banking regulation and safety net policies has tested the endurance of many an analyst. Those at this conference know of, and in some cases authored, the continuous, 20-year flow of proposals on the subject of reform. This work has left few policy stones unturned, making it nearly impossible to advance new proposals. Yet, the actual regulatory and legal changes introduced over the periodalthough positive stepsare inadequate to address the safety net's perversion of the risk/return trade-off.
In fact, trends that have accelerated over the last decade, such as banking consolidation and increasing complexity of bank operations, make the reforms adopted appear even less equal to the task of addressing the major concernthe moral hazard problem that is inherent in providing a financial safety net. Because the government protects depositors and other bank creditors from losses, interest rates paid on bank deposits and other forms of bank debt do not fully incorporate the riskiness of the banks' activities. Banks then receive a price signal that is too low and end up financing higher-risk projects than they would otherwise. This behavior is not venal, despite its "moral hazard" designation; one need not conjure up images of bankers meeting to rip off taxpayers to understand how a firm would rationally react to the incentives created by mispricing.
But it now appears that a growing number of regulators and policymakers recognize the power of these price distortions (and Asia provided a powerful reminder). This shift in view has reduced but not eliminated the controversy over the need for regulatory reform, especially with respect to our largest banks. A growing consensus in favor of reform notwithstanding, there is no agreement about the appropriate measures to implement.
In my view, proposals that rely exclusively on unfettered markets to eliminate moral hazard are not credible, and those that rely exclusively on supervision and regulation are unlikely to effectively address the moral hazard problem. In contrast, the many years spent analyzing moral hazard have yielded promising reform proposals that make use of market signals to enhance discipline. Drawing on these plans, I propose combining market signals of risk with the best aspects of current regulation to help mitigate the moral hazard problem that is most acute with our largest banks. Let me make this last point clear: The reforms discussed would apply only to our largest banks, especially those that are considered too-big-to-fail (TBTF).
Anyone suggesting regulatory reforms for the largest banks must first confront the Federal Deposit Insurance Corp. Improvement Act of 1991 (FDICIA) and follow-up steps that regulators have taken. We find these steps to be positive and, in fact, they provide the framework for future reforms. Yet, by definition, our call for further reforms suggests that FDICIA and regulatory steps are incomplete in an important way. Moreover, we believe that problems with the regulatory approach have been exacerbated by increasing consolidation and complexity in the banking industry, which, by the way, may well be an appropriate and efficient response to changing market conditions and technological innovations.
The primary response, as exemplified by FDICIA and subsequent reforms, to the pernicious incentives created by the safety net is to regulate bank activities so as to limit risk taking. FDICIA was an important step forward in that it focuses supervisory resources on banks in poor financial condition. FDICIA, in theory, requires banks in relatively poor financial shape to pay higher deposit insurance premiums. Likewise, FDICIA creates a system wherein banks in deteriorating financial condition face restrictions on their activities and more frequent review than financially sound institutions. And for all banks, supervisors have implemented examination procedures which shift staff resources to review of higher-risk activities and of the banks' policies for controlling their risk exposure. Finally, regulators have hired more "rocket scientists," staff with the specialized human capital necessary to evaluate new forms of bank risk taking and management.
All of these steps appear beneficial. But even with them, we cannot rely solely on supervision and regulation to adequately contain moral hazard incentives. There are several factors underpinning this conclusion, including:
The ability of regulators to contain moral hazard directly is limited. Moral hazard results when economic agents do not bear the marginal costs of their actions. Regulatory reforms can alter marginal costs but they accomplish this task through very crude and often exploitable tactics.
There should be limited confidence that regulation and supervision will lead to bank closures before institutions become insolvent. In particular, reliance on lagging regulatory measures, restrictive regulatory and legal norms, and the ability of banks to quickly alter their risk profile have often resulted in costly failures. The existing literature suggests that the rules FDICIA established to ensure prompt closure of banks are unlikely to effectively reduce losses during a banking crisis, precisely because of the weaknesses identified above.
The ability of regulators to assess bank risk is limited. While regulators have access to inside information, there are still areas of profound informational asymmetry between regulators and banks. Moreover, the ability of regulators to directly rely on the output of internal bank models to overcome this asymmetry is currently limited. Problems include inaccuracies of the models as well as the difficulty regulators face in verifying the bona fide status and use of their output (i.e., does output represented as the one in 100 loss event really have that meaning).
Finally, even if regulators had the ability to accurately assess bank risk, they do not have a sound basis for determining how much risk in banking is too much or too little. Banking supervision and regulation, in other words, should not be expected to lead to something like an efficient amount of risk taking.
We also have strong doubts that FDICIA adequately addresses the problem of too-big-to-fail. As you know, policymakers and regulators indicated prior to FDICIA that they would rescue the liability holders of the largest banks because of their fear of contagion and systemic instability. Because of the TBTF assurance, liability holders did not have adequate incentive to charge large banks higher rates for higher-risk, and thus large banks took on more risk than they would have otherwise.
FDICIA establishes a process that subjects the bank-rescue decision to more sunlight, formal vote taking and consensus. However, it is not clear that these changes significantly altered the rescue process that was in place prior to FDICIA. Indeed the process codified by FDICIA in 1991 looks very much like the one leading to the rescue of Continental Illinois in 1984. Consider the following description of that process by the Comptroller of the Currency, Todd Conover:
"We debated at some length how to handle the Continental situation.... Participating in those debates were the directors of the FDIC, the Chairman of the Federal Reserve Board and the Secretary of the Treasury. In our collective judgment, had Continental failed and been treated in a way in which depositors and creditors were not made whole, we could very well have seen a national, if not an international, financial crisis the dimensions of which were difficult to imagine. None of us wanted to find out."
[Hearings Before the Subcommittee on Financial Institutions, Supervision, Regulation and Insurance, Sept. 18, 19 and Oct. 4, 1984, pp. 287-288]
These observations do not argue against any of the changes to regulation that have been made or that have been proposed. Indeed, that policymakers and supervisors have moved from a traditional command and control regime to a system where supervisory intensity and rules depend on the financial condition and management of the bank is a positive step. However, regulatory steps alone are unlikely to adequately address distortions created by the safety net and, especially, by the presumption of TBTF. Moreover, two trends that have accelerated since passage of FDICIA suggest that regulatory approaches may well have become even more unequal to the task of containing excessive risk taking by large banks than formerly.
Banking assets are increasingly controlled by the largest firms. In 1980, there were more than 12,000 banks in the country, with institutions with assets greater than $10 billion controlling 37 percent of total bank assets. These figures had barely changed by 1990 but, by 1998, there were far fewer banks (8,910), with the 64 over $10 billion in assets controlling a large share of total banks assets (63 percent). The accepted wisdom, buttressed by econometric models, forecasts a continuation of this trend. This rising concentration has almost certainly led to more TBTF banks.
Bank operations have become more complex. There are several reasons for this, including:
Increased bank size and geographic reach.
Increased scope of product offerings (e.g., insurance sales, underwriting
of securities, etc.).
Increased skills needed to offer some new products (e.g., derivative structuring) and implement new risk management systems that may differ radically from previous practices.
Such trends make it more difficult for bank supervisors to effectively monitor and respond to the risk taking of the largest and most complex banks for the general reasons enumerated above. How can regulators, for example, effectively alter the marginal costs of banks across so many activities? More generally, the increasing complexity of banks could raise the level of information asymmetry between the banks' managers and the regulator.
Implicit in this critique of policies that rely solely on regulation to address moral hazard is the view that market forces can bring benefits that are not otherwise available. I am not, however, arguing for the other extreme. In my judgment, a policy of complete reliance on the market is not the answer either because such a policy is not crediblethe government will and should intervene when there is a negative shock to the entire system. Consequently, a policy that gives the pretense of relying solely on the market still creates moral hazard.
Several reform proposals look to unfettered markets to manage the risk taking incentives created by the safety net. One option, privatization, would eliminate the federal safety net and with it the need for regulators to monitor banks. Instead, private insurers would assume that role. Several large commercial banks are the proponents of these schemes, which vary from plan to plan but which often call for stripping deposit insurance of its federal features, such as the full faith guarantee of the U.S. government.
Narrow banking plans, whose features are well known, essentially take a similar approach. The uninsured, unregulated "wide banks" under this scheme are the same as banks under a privatized system and have the same economic justification for their creation. The difference is that the narrow banks would meet the residual demand for insured deposits. The safe, transparent assets of the narrow bank eliminate the need for existing safety and soundness examinations and, presumably, for federal insurance, although it could be retained if desired.
A serious problem with these laissez faire approaches is that they do not credibly address the potential for instability in the banking system nor the related TBTF problem. The complete absence of a federal safety net creates the potential for banking panics which could have substantial financial and real costs, but one need not hold this view to conclude that laissez faire reforms will fail to reduce mispricing of risk. This is because policymakers and regulators have often indicated through deed and word that they will rescue liability holders of the largest banks because of their fear of contagion and systemic instability. These practices raise doubts about the credibility of the "no government support" pledge central to privatization and narrow banking plans. As such, narrow banking and privatization are not likely to reduce materially market expectations of rescues and hence the accompanying under pricing of risk. And as the discussion suggests, the problem is particularly acute for large institutions.
In our view, incorporating market signals into the regulatory process is an essential step in addressing moral hazard. However, incorporating market signals from creditors who are put at risk requires a credible policy framework, one which accounts for TBTF and potential systemic instability, and which supervisors and legislators will embrace. The idea is to require, by law, that depositors and other creditors at all banks, not just small institutions but even those considered TBTF, bear some risk of loss in the event of the failure of the institution, and to incorporate the market signals that this policy generates into the current regulatory regime.
A wide body of empirical research suggests that bank creditors, credibly put at risk, do assess and act upon the risk taking of banks. That is, these investors alter their pricing to reflect changes in the riskiness of the bank. The incorporation of new information by creditors means that market prices for bank funding will directly affect the bank's marginal cost of taking risks with a precision that regulation cannot achieve. In particular, this means that greater reliance on market signals is potentially capable of reducing the mispricing of risk taking which occurs today.
Moreover, market participants have proven reliable in incorporating the implications of new bank products and services in their risk assessments, so further financial innovation can be expected to be accommodated within this approach. Finally, market participants have an option for addressing information asymmetry not available to regulators. If market participants do not understand a bank's strategy or operations, they can charge a risk premium to account for that weakness, thereby providing banks with a signal to provide more information.
Development of a policy framework for the use of market signals requires two steps. First, policymakers must credibly put creditors and others capable of providing market discipline at risk of loss, so that market signals are generated. To accomplish this successfully, the reform must address TBTF and instability. Second, bank regulators must explicitly and systematically incorporate market signals into the supervisory process.
Creating Credible Market Signals Analysts have suggested several credible methods for developing market signals. The Federal Reserve Bank of Minneapolis has called for amending FDICIA so that uninsured depositors and other creditors at TBTF institutions would lose a modest but meaningful percentage (e.g., 20 percent) of funds if their bank fails. This plan takes advantage of the vast experience of private insurers who have long used coinsurance to address moral hazard. In the banking context, such coinsurance is credible because the limitation on loss size should preclude substantial financial spillovers and thus should reduce the likelihood of contagion. As a result, Congress and bank supervisors would have less incentive to bailout TBTF institutions or worry excessively about the threat of panics. Certainly, though, there is a tradeoff between market discipline and banking stability. This reform would make banking panics a possibility, something that is highly unlikely today. We assert, however, that no effective reform can avoid this tradeoff.
Once at risk, uninsured creditors would demand additional information about their banks to the degree that current disclosures do not allow for a clear assessment of risk taking, and so we recommend mandating additional disclosure requirements only after assessing the data provided voluntarily under the new regime. With increased incentive for large depositors to monitor the quality of banks, the risk premia found in the rates charged by such depositors and other creditors put at risk should provide a more accurate reading on bank risk than currently exists.
A similar type of plan would, alternatively, require large banks to issue subordinated debt equal to some (small) percentage of the bank's assets. Subordinated debt holders come only before equity holders in making a claim on failed bank assets. Thus, they could lose a significant investment if their bank becomes insolvent. Most of the time, this gives subordinated debt holders incentives that regulators should desire. The debt holders want banks to take enough risk to generate profits, but they should demand bank closure before the institution becomes more than minimally insolvent. The Federal Reserve Banks of Chicago and Atlanta have been leaders in developing these plans and, more recently, members of the Board of Governors have expressed strong interest in such a proposal.
Although subordinated debt plans vary by author, their general structure has generated more support than proposals that rely on depositor discipline. In particular, holders of subordinated debt cannot demand immediate repayment as can some depositors, and therefore such plans are seen as potentially less destabilizing. Moreover, the subordination of debt could reduce the likelihood that holders would receive coverage during a bank bailout.
Subordinated debt plans, however, are not the proverbial free lunch. On the one hand, if issuance of long-term subordinated debt is mandated such plans may at best provide low-quality pricing data to regulators. In this case, information from the primary market may be limited because issuance is infrequent, and the secondary market for long-term subordinated debt may well be illiquid, especially if regulators require the issuance of such securities in greater amounts than currently demanded or supplied. On the other hand, if banks are required to issue short-term subordinated debt, they could be subject to panic-like drop-offs in funding. Some subordinated debt plans try to address this concern by requiring a mix of short-term and long-term maturities but there is no clear basis on which to divine the right combination.
Private insurance offers a third source for generating market signals. For example, Congress could require large banks to purchase a small amount of private deposit insurance, covering, say, 5 percent of deposits. In fact, in 1991 Congress mandated that the FDIC study the establishment of a private reinsurance system. Financial engineers could also design instruments to allow private investors to bear some of the cost of covering insured depositors. Such instruments already exist for natural disasters.
Private insurance proposals would require market participants to assess and segment banks by their riskiness. The effect of these arrangements on stability depends on the confidence of the insured creditors. If they do not believe private insurance capable of honoring their claims, insured creditors will run banks, although the limited amount of exposure they bear should limit the problem. In addition, private insurance arrangements do not address TBTF per se. But they could be structured such that the bailout by the government of the liability holders of a large bank with private insurance does not eliminate the insurers' requirement to make a payout. As such, pricing should still reflect the risk of a claim against the private insurer.
All of these proposals to put bank creditors at risk have the virtue of gradual implementation, if needed. The coinsurance rate, for example, need not rise to 20 percent at once (or ever), nor does the required amount of subordinated debt have to quickly meet its maximum level. This flexibility should reduce concerns about instability.
Incorporation of Market Signals Putting bank creditors at risk, however, is not a replacement for supervision. Some portion of large bank assets, for example, will remain opaque, leading to incomplete information for market participants and, presumably, to imperfect market signals. The supervisory process can generate and act on valuable information about such assets and therefore contribute to restriction of bank risk taking.
More generally, the market discipline provided by creditors may not be enough on its own to address excessive bank risk taking. As long as some bank creditors receive one hundred percent protection, the prices that banks face for raising money will be distorted. And none of the options for increasing market discipline would alter our current full protection for deposits under $100,000. Moreover, market discipline may be avoided, at least to some extent, if banks rely increasingly on insured funds in times of stress.
Thus, the supervisory process will continue to play an important role. In this environment, to rein in moral hazard effectively, the signals created by the market need to be incorporated in two general areas of regulation: deposit insurance assessments and the supervisory process.
Assessments: Under any credible reform of the safety net, the government will continue to provide a base level of deposit insurance. To limit moral hazard then, the government must charge deposit insurance assessments that vary with the riskiness of the bank (as signaled by the market). The means of incorporating market signals into the assessment depends on the source. Private insurance premiums would offer the most straightforward source of incorporation, but other market signals would work as well. The risk premia evident from the prices charged by depositors or other creditors put at risk of loss suggest a way to differentiate deposit insurance pricing. In order to reduce administrative costs, the deposit insurer would most likely group institutions with fairly similar risk premia into assessment categories. And the insurer would surely use market premia as only one factor in the assessment setting process, along with supervisory and other information that add predictive value.
Supervisory Process: The existing regulatory system relies in part on "triggers" which require risky institutions, as evaluated by bank supervisors, to face more supervisory scrutiny, higher insurance costs and more restrictions on activities than do sound institutions. Under Prompt Correction Action (PCA), for example, banks with declining capital ratios face increasing restrictions on activities until the point that regulators close them. There are also links between the frequency of bank examinations and capital ratios and supervisory ratings of management.
The point here is not to endorse the particular triggers used currently. (In fact, some have argued persuasively that PCA would be more effective if it were based on market measures of capital rather than book measures.) Rather, I want to emphasize the importance of incorporating market signals into supervisory triggers more generally, in order to enhance the ability of supervisors to address moral hazard.
How might this incorporation actually work? In a fairly simple approach, supervisors could make use of the new, risk-based deposit insurance premium groupings to sort banks into various trigger categories. Alternatively, supervisors could put banks into regulatory categories based on their funding costs relative to some benchmark rate (e.g., the difference of a particular institution's debt yield over the relevant Treasury rate or the average for bank debt). Under the combined regime, regulatory action would be triggered when banks moved from one market signal group to another and/or when they move from one regulatory category to another.
Of course, these comments gloss over many hard to resolve details. Indeed, it must be acknowledged that distinguishing between the noise inherent in rates on bank liabilities and meaningful measures of bank risk taking is a challenge. With this difficulty in mind, consider the pristine performance of a system incorporating market signals during a period of intense instability in fixed-income debt markets, as occurred in the latter part of 1998. For example, should large banks face restrictions on their activities because yields on their liabilities suddenly spike up sharply only to return to average levels? To the degree that the rise in yields reflects an overall increase in spreads between a risky rate and the risk-free rate, a system incorporating market signals need not trigger supervisory action. Supervisors could also filter out shocks in rates by using some sort of averaging procedure, but averaging has the serious disadvantage of masking price signals.
But focusing the difficulty of using market signals in banking regulation, especially during extreme circumstances, obscures the relevant comparison. Would the current regulatory-based system achieve more desirable outcomes than a system which incorporates regulation and market signals? Do we think that regulation alone would react with greater efficacy to periods of instability in credit markets and in the banking system more generally? For all of the systematic differences between regulation and market signals described above, I believe the burden of evidence rests on those who would argue that the current system is affirmatively superior to one making use of market signals.
Despite the call for change, the reforms suggested here are modest in many ways. They maintain deposit insurance, the existing supervisory structure, and phase in change gradually. At the same time, the reforms create the infrastructure for future efforts. They do so by explicitly recognizing and starting to address directly the fundamental and interrelated problems of moral hazard and banking instability and by taking steps toward explicit incorporation of market signals in the regulatory process. They also give regulators and policymakers time to learn and gain experience, thereby likely enhancing the quality of additional, future reforms.
Why bring up future reforms on the heels of this proposal? One answer is that if banks continue to become more like other financial services firmsa trend that seems firmly in placethen the justification for regulation and the safety net changes. These reforms provide a framework for transitioning from existing regulation, to regulation and market signals, to a system even more reliant on the market.