36th Annual Conference on Bank Structure and Competition
It's a pleasure to address this Conference once again. I say once again because I summarized a part of my Ph.D. thesis, which dealt with modeling individual bank behavior, at the 1972 Conference. If this pattern continues, I'll make my next appearance in 2028, when I'll be 88. Maybe by then I will have finally attained wisdom.
The theme of this year's Conference is the extraordinary current transformation of the financial services industry and the implications for financial services regulation. It's hard to imagine a more appropriate topic. From my perspective the transformation has been generally positive. With the elimination of interest rate ceilings in the early 1980s and the final removal of branching restrictions by the Riegle-Neal Act in June 1997 the industry is far more competitive than it was two decades or even one decade ago. Moreover, a case can certainly be made that competition, efficiency and customer convenience have all been enhanced by the steady erosion of Glass-Steagall product restrictions in the 1980s and ‘90s, and their final demise last year with Gramm-Leach-Bliley.
One aspect of the transformation, however, is challenging many bank customers and shareholders, and all affected bank regulators: the growth of very large — and, more importantly, complex — banking companies. The challenge for regulators, of course, is to supervise effectively the risk-taking activity of these companies. Given the inherent difficulties regulators face in monitoring large banking companies, several proposals have been made to enlist market discipline to supplement formal bank supervision. At this Conference last year, my colleague Gary Stern at the Minneapolis Fed proposed requiring coinsurance for deposits exceeding the coverage limits of regular deposit insurance. Others, including Fed Governor Larry Meyer, have endorsed requiring large banks to issue subordinated debt to the public. Changes in the relative yields on such debt presumably would at least in part reflect changes in the bank's risk posture.
Both of these proposals clearly represent steps in the right direction. Neither, however, confronts what I believe is a significant impediment to the efficient resolution of problem institutions: potentially inappropriate Fed lending to problem banks through the discount window. Such lending can systematically allow uninsured depositors and other uninsured creditors to escape losses and therefore reduce their incentive to apply market discipline. While FDICIA contains provisions intended to restrict Fed lending in problem situations, the restrictions may be ineffective in particular cases for reasons I will discuss in a few minutes.
First, let me briefly review the Minneapolis and subordinated debt proposals, since these are probably the most prominent efforts to date to deal with the broader risks presented by the huge banking organizations that now dominate much of the financial landscape. The essence of the Minneapolis proposal, again, is coinsurance. All uninsured depositors would be required to absorb some loss if their bank failed, even if the bank were very large, but the loss would be capped at, say, 20 percent of an individual deposit. Requiring all uninsured depositors to suffer some loss when a bank fails would reduce moral hazard. But limiting the loss at a fairly low proportionate level would diminish the risk of systemic problems and thus increase the likelihood that regulators would allow even a large insolvent bank to fail rather than propping it up at potentially substantial cost to taxpayers. In this way the Minneapolis proposal seeks — quite sensibly, in my opinion — to balance effectiveness and feasibility.
Subordinated debt proposals, which were discussed in sessions earlier this morning, would require larger banks to fund some minimum portion of their assets with subordinated debt. In the event of a bank failure, subordinated bondholders would be paid off only after all depositors — insured and uninsured — and perhaps other creditors as well, were paid. Consequently, they would be especially sensitive to bank risk-taking, since they, unlike equity shareholders, would not participate in any upside gains that risk-taking might produce. Given this, changes in the yield on a bank's subordinated debt would signal changes in how the market viewed the bank's health. The proposals typically require that large banks issue debt several times a year to provide frequent signals and to punish excessive risk-taking with an immediate increase in borrowing costs.
As I indicated earlier, in my view both of these proposed reforms are worthy ones and deserve a place in any serious consideration of safety net reform. Indeed, Congress recognized the potential usefulness of the subordinated debt discipline in writing the Gramm-Leach-Bliley Act, which requires that large banks wishing to conduct nonbanking activities in a bank subsidiary issue such debt and maintain top market ratings on it.
The main point I want to make today, however, is that in the absence of greater restraint, Fed discount window lending could materially undermine the potential benefits of either the Minneapolis or subordinated debt proposals. I recognize that this is hardly news to this audience, but it may not be widely appreciated beyond it.
The Fed's discount facility exists primarily to assist healthy banks facing temporary liquidity problems. It should not be used when a bank's solvency is in question, for obvious reasons. The managers of an insolvent bank have a strong incentive to use any funds received through the window in all-or-nothing gambles for recovery. The problem, of course, is that most such gambles fail, which increases the loss borne by the deposit insurance fund and possibly taxpayers. Ideally, the Fed would distinguish between solvent and insolvent banks when lending. In practice, however, it is often very difficult to distinguish between banks experiencing temporary liquidity problems and truly insolvent institutions. Therefore, in the past, the Fed typically has lent whenever a bank could provide acceptable collateral.
During the 1980s, the Fed frequently lent for extended periods to banks that eventually failed. In a number of cases, this lending helped provide uninsured depositors and other creditors sufficient time to remove their funds from a troubled bank, which left the deposit insurance fund to foot the entire bill when the bank ultimately went under. It seems obvious that the benefits of proposals to increase market discipline and reduce moral hazard would be significantly limited if these lending practices persisted going forward.
An example. I can offer a very clear illustration of the problem I've just summarized from fairly recent experience in my own Federal Reserve District. The commercial real estate crisis in the late 1980s and early ‘90s was particularly severe in the Washington, D.C., metropolitan area. The $1.6 billion National Bank of Washington, which I'll call NBW, incurred heavy losses during the crisis, which in May of 1990 culminated in NBW's parent, Washington Bancorp, defaulting on maturing commercial paper. The default received conspicuous press coverage that sparked a substantial outflow from NBW's uninsured deposits. Even though uninsured deposits accounted for about one-third of NBW's deposit base, however, the outflow did not force NBW's closure because the funds were replaced by discount window loans. These loans sustained the bank until it was eventually closed in August, while approximately $310 million in uninsured negotiable CD's and foreign office deposits were drawn down. It is estimated that, ultimately, NBW's failure will cost the FDIC fund approximately $220 million.
Although the NBW experience was particularly prominent, it was far from unique. According to a 1991 House Banking Committee report, between 1985 and 1991 about 530 depository institutions, or about one-quarter of the number that eventually failed, received discount window loans prior to failure. Clearly, during at least the late 1980's, uninsured depositors and other creditors had good reason to expect that they would have time to withdraw their funds from troubled banks before incurring losses. Moral hazard, in other words, was high.
The effect of FDICIA. Looking forward, we all know that FDICIA included provisions aimed at restricting Fed lending to potentially insolvent banks. Won't these restrictions prevent Fed loans from bailing out uninsured depositors at troubled banks in the future?
In at least some cases they may not, in my view. To be sure, FDICIA, passed in 1991 in the wake of the banking industry's problems in the late 1980s, limits discount window lending. For 'undercapitalized' banks — banks with a capital ratio below 4 percent — and for banks with the lowest CAMEL rating, the limit is 60 days. For 'critically undercapitalized' banks — banks with a capital ratio below 2 percent — the limit is only 5 days. But the 60-day restriction for undercapitalized banks can be extended for additional 60-day periods if a bank's primary federal regulator or the Fed certifies that the bank is viable for the longer run. And the Fed can breach the 5-day limit for critically undercapitalized institutions if it is willing to risk incurring a small financial penalty. At most, the financial penalty amounts to any interest the Fed earns on increases in lending made after the 5-day period.
I would acknowledge that this penalty, while it would probably be small in financial terms in most cases, might loom larger from a reputational perspective, since the Fed must report such penalty payments to Congress. Still, it is unlikely that these limitations would have prevented or significantly reduced Fed lending to NBW, had the restrictions been in place at the time. In accounting terms — which is what matters for purposes of the FDICIA constraints — NBW was not recognized as undercapitalized until a few days before it was placed into a conservatorship. It was only in those last days that examiners required capital-reducing loan charge-offs. Consequently, even if FDICIA had been in place, in the absence of any additional constraints the Fed could well have continued to lend to NBW until the very last days. At worst, depending on NBW's CAMEL rating when discount window lending began, the Fed might have been restricted to lending for 60 days. But two-thirds of uninsured deposit withdrawals (and half of all large CD and foreign office deposits) occurred in the first 60 days of Fed lending.
Implications of the NBW experience. As I indicated earlier, the Minneapolis proposal seeks to balance effectiveness in reducing moral hazard and the feasibility of actually seeing the proposal adopted, and I salute this pragmatic approach. Our Bank's experience with NBW, however, does suggest how Fed lending could undermine the proposal in practice. While the plan would impose losses on remaining uninsured depositors after a bank is closed, Fed lending may well allow many depositors to withdraw their funds before closure, as indeed occurred in the NBW case.
The FDICIA limitations on the window could prevent Fed lending from facilitating withdrawals from uninsured accounts in one case: where a decline in recorded capital is the first signal that a bank's condition is deteriorating, and the decline is sufficient to cause the bank to become critically undercapitalized. In such a case, the Fed can lend for only 5 days — perhaps not long enough for a significant portion of depositors to escape. Typically, though, recorded capital declines only after examiners force banks to write down questionable loans. To the extent that such write-downs lag other evident signs of weakness, such as press reports of declining profits or management shake-ups, uninsured depositors have an opportunity to flee before closure. If reasonably clear signals of a bank's deterioration can be generally expected even shortly before recorded capital declines, and the Fed routinely behaves as it did in the NBW case, uninsured depositors and other creditors will have little incentive to constrain the bank's risk-taking. Moral hazard would be high and the purpose of the Minneapolis proposal would be frustrated.
The NBW case raises questions regarding subordinated debt proposals as well, although perhaps somewhat less pressing ones. While discount window loans usually prevent a bank from defaulting on deposit contracts, they could just as well prevent default on subordinated debt contracts, thereby undermining efforts to impose market discipline through a subordinated debt requirement. Since default on a debt contract could force a bank to close as quickly and surely as default on a deposit, Fed lending officers may find themselves forced to choose between making loans to a bank that may be used to redeem maturing bonds or allowing a bank to fail. If past practices continue, such loans might well be made in many cases as long as acceptable collateral were available.
The expectation that loans of this sort might be forthcoming would undermine efforts to impose market discipline through a subordinated debt requirement. Subordinated debt proposals might offer greater prospects for success than the Minneapolis plan in this regard, however, since the proposals call for relatively long maturity debt, which implies that a sizable portion of debt holders would be unable to rely on Fed lending for protection. While the Fed has lent for extended periods in some cases in the past, it seems reasonable to presume that in most instances the lending period would not be long enough for all bonds to mature. Some bondholders, then, would lose. So while the possibility of Fed lending might reduce the incentive of bondholders to monitor their respective banks' risk-taking at the margin, the incentive for holders as a group could remain fairly strong.
Destructive Ambiguity. As NBW began to encounter difficulties, some of the bank's uninsured depositors — those, at least, who were familiar with the Fed's behavior in earlier problem situations — may have wondered if NBW would be propped up with Fed loans. The bank was fairly small, which presumably reduced the probability of assistance. On the other hand, NBW was the fourth largest bank in an important city, which suggested that some Fed assistance might be made available. In short, creditors probably considered the prospects for Fed assistance to be uncertain. Maybe the Fed would lend; maybe it would not.
Some observers might view this uncertainty positively as 'constructive ambiguity.' In their view, because a troubled bank's uninsured depositors and other uninsured creditors are not certain Fed assistance will be forthcoming, they have reason to monitor bank risk-taking.
But if these depositors and creditors are unsure about the prospects for Fed lending, the implication is that they see at least some chance of assistance, which would reduce the incentive to monitor risk. This might just as reasonably be called 'destructive ambiguity.' Beyond its impact on uninsured creditor behavior, such ambiguity may expose the Fed to political pressure in problem bank situations, since it implies the lack of a firm and consistent rule on procedure when these situations arise. Indeed, ambiguity may invite creditors who stand to lose money to try to bring political pressure to bear on the Fed.
A conundrum. To summarize, we face a conundrum. Worthy proposals for dealing with Too-Big-To-Fail, like the Minneapolis and subordinated debt plans, attempt to contain moral hazard in large banking organizations by reinforcing the regular supervision of these companies by bank regulators with market discipline. But these plans and others like them can be undercut if Fed lending to troubled institutions permits uninsured depositors and other creditors to withdraw their funds before closure. My NBW example and similar instances in the 1980's demonstrate that, historically at least, the Fed, for whatever reason, has behaved in this fashion in some cases with predictable results. Nothing in the reform proposals I've discussed would seem to preclude similar behavior in other cases going forward. While FDICIA addressed the problem, its focus on accounting capital makes it likely that the conundrum will persist unless something else changes.
What is the something else that needs to change? The answer is pretty obvious. The Fed should adopt a more restrained approach to lending — shifting away from a lending posture that historically has appeared bent on reducing the risk of near-term systemic financial instability to something approaching zero. What may be less obvious is that there is at least some basis for believing the Fed can actually make such a shift if it approaches the challenge in a longer-term, strategic context.
A shift to greater lending restraint — striking a more appropriate balance. The Fed's strong bias toward preventing or at least minimizing banking or other financial market disruptions is hardly surprising. After all, this was a principal objective of the framers of the Federal Reserve Act. Lending to troubled banks can forestall financial disruption in the short run, and in some cases may actually prevent disruption by, for example, facilitating the acquisition of a troubled bank by a healthy one.
But this is not the end of the story. Whatever its effect on the short-run risk of financial disruption, Fed lending to troubled institutions can increase the risk these banks pose for taxpayers in two distinct ways. First, by allowing the removal of uninsured funds prior to closure, Fed lending shifts the burden of any losses ultimately incurred to the insurance fund, which increases the risk that taxpayers may eventually have to share the loss. Second — and perhaps more relevant to many of the points I've tried to make in this paper — frequent Fed lending to troubled banks weakens the incentive uninsured depositors and creditors have to monitor and discipline a bank's risk-taking. This, in turn, also raises the probability that taxpayers will have to share losses.
Given this, I believe we at the Fed need to strike a more even balance between working to reduce the risk of financial disruption, on the one hand, and reducing the risk to the insurance fund and taxpayers on the other. In the past, the Fed typically has lent to any bank with acceptable collateral, including many troubled banks that soon failed. For market discipline to be effective, we will need to give greater weight to taxpayer risk in making individual loan decisions, even in — indeed, particularly in — cases where a bank is both troubled and very large.
Implications for discount window administration. It's one thing to point out the need to achieve better balance in lending decisions generally. At the end of the day, however, the transition will have to be made on the ground, so to speak, one decision at a time. This change will not be an easy one. Fed lending officers will need to consider a borrowing bank's true financial health more carefully than at times in the past. If there is good reason to believe that a bank is insolvent in real terms, even if it is not yet undercapitalized in the usual accounting terms, the Fed should deny the loan. In practical terms, transitioning to a tougher stance may require us to take account of additional, non-traditional information in individual decisions to ensure that our assessment of a bank's health is as complete and accurate as possible.
The payoff. Again, this necessary transition will not be easy. As I noted earlier, differentiating solvent from insolvent banks, in real terms, is often difficult and rarely straightforward, especially if the lending decision needs to be made quickly. Moreover, given the Fed's past behavior, the initial decisions in a new regime will likely surprise uninsured depositors and creditors and disappoint their expectations, and this obviously could be disruptive for a time as uninsured depositors in other banks assess the health of their respective banks more searchingly and withdraw funds from institutions whose health is questionable. The transition could even provoke the failure of some weak banks.
However disruptive the transition may be, though, the short-term pain it causes will be justified in my view by the even greater pain and disruption it is likely to prevent over the longer term. Additionally — and this is a very important point — this change in the Fed's lending behavior will provide a solid foundation for proposals such as the Minneapolis and subordinated debt plans, and raise the chances that, if put in place, they would achieve the objectives their authors envision.
At this point, many of you in the audience are probably thinking, 'This all sounds nice — calling on the Fed to do the right thing. But everybody knows that when push comes to shove and a threat of real financial disruption emerges, the Fed will lend. The pressure on the Fed in these situations is just too great to resist. The only way out would be to legislate truly binding restrictions on the Fed's ability to lend — maybe a stiffer financial penalty if a bank actually fails.' Maybe so. But while one can imagine such legislation, even if restrictions were imposed, realistically the Fed would be left with some discretion so that the buck most likely will continue to stop with us. With this in mind, I believe that restraint on Fed lending to troubled banks, in order that plans to impose healthy market discipline on large banks can work, will have to be home grown. It is a matter, essentially, of establishing and maintaining credibility.
Are there any grounds for believing the Fed could achieve such credibility? I don't want to be unduly optimistic, but I believe there are grounds: specifically, in the Fed's success in building a high level of credibility for its anti-inflationary monetary policy strategy over the last two decades. After many years of stop-go monetary policy with disastrous results in the late 1970s and early 1980s, we bit the bullet, tightened monetary policy aggressively and persistently, and accepted the short-term economic consequences, public criticism and pressure that went with it. This sea change in monetary policy, in my opinion, is one of the principal reasons for the U.S. economy's extraordinary performance over the last four years.
I recognize that on closer inspection the analogy I've drawn may not be terribly comforting. After all, in the monetary policy area we acted only when our backs were against the wall, the need for painful action was understood, if grudgingly, by a significant number of ordinary Americans, and there really was no evident alternative. In contrast, confronted with the likely failure of a large bank, we could in many cases probably put off paying the piper to another time. Still, our earlier experience with inflation taught us that the short-term consequences of a policy transition can be endured, and that the ultimate payoff in better economic performance can be great indeed. I'm optimistic enough to believe quite firmly that we can do it again, and I look forward to celebrating our success when I return in 2028.