Speech

On the Need for an International Lender of Last Resort -- Stanley Fischer

January 3, 1999

The frequency, virulence, and global spread of financial crises in emerging market countries in the last five years -- Mexico in 1994, with the subsequent tequila contagion in Latin America and for a day or two in East Asia; East Asia in 1997 and 1998, with contagion spreading crisis within the region; Russia in 1998, to an extent itself affected by Asian contagion, with the Russian contagion spreading to Latin America in addition to eastern Europe and the rest of the former Soviet Union -- coupled with the Congressional debate in the United States over the increase in IMF quotas, has led to the most serious rethinking of the structure of the international financial system since the breakdown of the Bretton Woods system in 1971.

As a result, governments and the international institutions will in the coming months and years put in place a series of changes designed to strengthen the international financial system. In emerging market countries, the emphasis will be on improving economic policies, and strengthening banking and financial systems, corporate governance, and the ability to deal with reversals of capital flows. In the industrialized countries where the capital flows originate, measures are being considered to improve the regulation of and information about the activities of international investors; and no less important, in recent months the leading central banks, in recognition of the feedbacks between the emerging market and the industrialized economies, have taken actions in the interests of their own countries that stabilize the world economy. The international financial institutions are being asked to improve surveillance -- including of short-term capital flows; to encourage the adoption of banking and other international standards in emerging market countries, and monitor their implementation; to improve the information provided to markets and the public more generally; and to consider changes in their lending practices, including through the provision of guarantees² and possible precautionary or contingency lending.³

The vision that underlies most proposals for reform of the international financial system is that the international capital markets should operate at least as well as the better domestic capital markets. To say this is to drive home the point that volatility and contagion cannot be banished, for asset prices inevitably move sharply, and there are significant correlations among the prices of assets, particularly in related industries. But while volatility and contagion will always be with us, we can surely do better in reducing the frequency and intensity of emerging market financial crises, and the extent of contagion, than we have in the last five years.

It is natural at a time like this to seek guidance in the original debates on what later became the Bretton Woods system. However those debates do not shed much light on the present difficulties, because for the most part in their work on the IMF⁴ the Bretton Woods participants were dealing with current account problems,^5,6 whereas recent crises have been driven by the capital account, by sudden and massive reversals of capital flows. Rather, as we consider how to make the global capital markets operate better and how to reduce the frequency and virulence of financial crises, I would like today to revisit an older literature, one that emerged out of the financial crises of the last century, that on the lender of last resort.

We are all of course familiar with the classic contribution of Bagehot (1873). The most famous lesson from Bagehot is that in a crisis, the lender of last resort should lend freely, at a penalty rate, on good collateral. But thinking on this issue neither started with Bagehot -- for as many have pointed out, Henry Thornton's 1802 analysis of monetary policy and the role of lender of last resort is remarkably sophisticated⁷ -- nor did it end with him. In particular, contributions in the last two decades by among others, Benston et al (1986), Garcia and Plautz (1988), Goodhart (1995), Goodhart and Huang (1998), Holmstrom and Tirole (1998), Kindleberger (1978), Meltzer (1986), Mundell (1983), Schwartz (1988), and Solow (1982) have sought to build on and develop the analysis of the role of lender of last resort. In the international context, following Kindleberger's work two decades ago,⁸ recent contributions by Calomiris (1998), Calomiris and Metzler (1998), Capie (1998), and Giannini (1998), discuss the potential role of the IMF as a lender of last resort, with Capie arguing that its inability to create money makes it impossible for the Fund to operate as lender of last resort.

I will start this lecture by reviewing the case for a lender of last resort in the domestic economy, and the set of rules that the lender of last resort is supposed to follow. I will then discuss the moral hazard problem that is created by the existence of a lender of last resort,⁹ and measures to mitigate it. I then turn to the international system, and will argue that it too needs a lender of last resort. I will also argue that the International Monetary Fund, although it is not an international central bank, has undertaken certain important lender of last resort functions in the current system, generally acting in concert with other official agencies -- and that that role can be made more effective in a reformed international financial system.

The lender of last resort role of the central bank is associated with the prevention and mitigation of financial crises. A financial crisis is a typically sudden actual or potential breakdown of an important part of the credit system. Financial crises and panics have been taking place for centuries (Kindleberger (1978) and MacKay (1841)). They are associated with a loss of confidence in the standing of some financial institutions or assets, and because the chain of credit is based on tightly interlinked expectations of the ability of many different debtors to meet payments, can spread rapidly, contagiously, through the financial system, and if unchecked, have significant effects on the behavior of the real economy. In economic theory panics can be modelled as cases of multiple equilibria,¹⁰ possibly dependent on herd behavior.

Surprisingly, there is no accepted definition of the term lender of last resort,¹¹ and there are also important disagreements about what the lender of last resort should do. I will start with the more traditional Bagehot conception, as summarized and developed by Meltzer (1986, p83).

Meltzer states that

The central bank is called the lender of last resort because it is capable of lending--and to prevent failures of solvent banks must lend--in periods when no other lender is either capable of lending or willing to lend in sufficient volume to prevent or end a financial panic.

He lists (pp 83-4) five main points, the first four derived from Bagehot:

• The central bank is the only lender of last resort in a monetary system such as [that of the United States].

• To prevent illiquid banks from closing, the central bank should lend on any collateral that is marketable in the ordinary course of business when there is no panic [emphasis added]. It should not restrict lending to paper eligible for discount at the central bank in normal periods.

• Central bank loans, or advances, should be made in large amounts, on demand, at a rate of interest above the market rate. This discourages borrowing by those who can obtain accommodation in the market.

• The above three principles should be stated in advance and followed in a crisis.

• Insolvent financial institutions should be sold at the market price or liquidated if there are no bids for the firm as an integral unit. The losses should be borne by owners of equity, subordinated debentures, and debt, uninsured depositors, and the deposit insurance corporations, as in any bankruptcy proceeding.

Meltzer's statement for the most part agrees with other formulations, but does not emphasize the view, summarized for instance by Humphrey (1975) and attributed to Thornton, that the overriding objective of the lender of last resort should be to prevent panic-induced declines in the money stock,¹² and that there is thus no conflict between its monetary control and its duties as lender of last resort.¹³ In some more recent formulations this view has been extended to the precept that

• In the event of a panic, the central bank should provide liquidity to the market, but not to individual institutions.

I want now to take up six points: first, whether the central bank is or has been the only lender of last resort; second, whether the lender of last resort must have the ability to create high-powered money; third, the critical notion that the central bank should lend against collateral that is good during non-crisis periods; fourth, the notion of the penalty rate; fifth, the Humphrey et al argument that the lender of last resort should lend to the market but not to individual institutions;¹⁴ and sixth, the view that the principles of lender of last resort lending should be stated in advance.

First, is the central bank the only lender of last resort? As a historical matter, no. There are really two roles that are described as that of lender of last resort in the literature. There is first the crisis lender, the institution that provides the financing to deal with a crisis. This financing may be provided to the market, or to individual institutions. Second, there is the crisis manager, the institution that takes upon itself the responsibility for dealing with a crisis or potential crisis, whether or not it itself lends for that purpose. At times institutions other than the central bank, including in the United States the Treasury, and even private institutions, such as clearing houses, and in 1907 J.P. Morgan, have played one or both of these roles.¹⁵

Let me expand on the role of the crisis manager. In a panic, it is necessary to find some means for dealing with the collective action problem. A panic is the realization of a bad equilibrium when a good equilibrium is possible, and there is a need in such situations for some agency or group of institutions to take the lead in trying to steer the economy to the good equilibrium. This can be done in two ways. First, through its lending activities, the crisis lender can act as a leader, thereby inducing or allowing others to follow in a way that mitigates or prevents the crisis. Second, there is a managerial, or facilitating or coordinating role in which other agents or institutions may be encouraged to act in the right way, for instance by their extending a loan to an institution whose failure could have systemic consequences. If a certain authority, and access to resources, are necessary for taking this coordinating role, then a Treasury may be able to do it as well as a central bank. In many countries the central bank is formally charged with the responsibility for maintaining the stability of the financial system, and it therefore tends to act as crisis manager, as in the recent LTCM case. But while the central bank is generally both the crisis manager and the crisis lender, neither role has to be carried out by the central bank.¹⁶

The separation of the roles of crisis lender and crisis manager is likely to become more frequent as the task of supervision of the financial sector is separated from the central bank. For instance, in the United Kingdom, the responsibility for crisis lending now rests with the Bank of England, while the Financial Services Authority has responsibility for organizing banking rescues.

Second, does the lender of last resort need the ability to create high-powered money? There is no question that it helps, most obviously in the case that a panic takes the form of a run from bank deposits into currency. Then the central bank can create the currency needed to deal with the panic, and at no first round cost to the taxpayer. However, as emphasized by Kaufman (1988), Schwartz (1988), and others, panics caused by a demand for currency are rare, and many banking problems result in shifts of deposits among banks, from those deemed unsound to those thought to be healthy.¹⁷

More generally, a panic may take the form of a run -- possibly enhanced by contagion --from the liabilities of one set of institutions to those of another. In these cases it may be possible to recirculate the liquidity from the institutions gaining high powered money back to those losing it. In principle that can be done by the market, if it is able to distinguish the merely illiquid from the insolvent companies.

Still -- and this is a critical point, because the line between solvency and liquidity is not determinate in a crisis -- it depends on how well the crisis is managed -- the task of distinguishing insolvent from illiquid becomes more difficult, and an official lender of last resort is likely to be needed to help restore normalcy.

The central bank is better equipped to deal with such a situation than an institution that cannot create high-powered money, not least because it is frequently necessary to act rapidly to stem a crisis, and the central bank is best placed to do that. But it is possible to set up an agency to deal with potential banking sector problems and endow it with sufficient funds -- perhaps from the Treasury -- to cover the anticipated costs of normal crises. In any case, the costs of major financial system difficulties will one way or another be borne by the fiscal authority, either explicitly, or implicitly, in the form of lower central bank profits over an extended period of time.¹⁸

We should also note that in dealing with banking crises, the lender of last resort has more often acted as crisis manager, as coordinator, without putting up its own funds, than as lender. Goodhart and Schoenmaker (1995) show that in the twenty-year period ending in 1993, taxpayer or deposit insurance money was used in over half the 120 banking rescue packages they study -- in part, because the central bank simply did not have the real resources that were required to deal with the banking problem.

All this is straightforward, provided the country has a floating exchange rate. However we should recall that at the time Lombard Street was written, the Bank of England was bound by gold standard (or currency board) rules, and the Bank of course did not have the ability to create gold. Nor, in the event of what was then called an external drain, a demand for gold to export, did it have sufficient gold to meet all the potential demands on Britain's gold reserve and it had frequently to borrow from the Banque de France (and vice versa). Nonetheless, Bagehot enjoined the Bank to act as lender of last resort. In practice, in the three crises preceding Lombard Street, the Bank of England was given permission to break the gold standard rule, and since Bank of England credit was in the event accepted as being as good as gold, it managed to stay the panics. The key was not the legal right to create high-powered money, but the effective ability to provide liquidity to those who wanted it. That function could also be exercised by other institutions whose credit was accepted in preference to the liabilities of institutions in trouble, for instance, in the United States, at times the Treasury.

The question of a lender of last resort arises in similar forms today, in countries with currency boards. If the question is how to deal with domestic financial institutions that may suffer liquidity problems, one solution, adopted in Bulgaria, is to set up an agency that is endowed with sufficient resources to lend in the event of a panic or banking sector problems.^19,20 If the problem is how to deal with a potential external shock that puts pressure on the domestic banking system, then the country may either hold excess foreign exchange reserves, or as in the case of Argentina, borrow from the markets and the official sector, and put in place international lines of credit. In these cases the private and public sector lenders to the central bank are acting as the country's lender of last resort.²¹ More generally, a country seeking to maintain a fixed exchange rate, will need foreign reserves or access to foreign financing to deal with potential external shocks.

In sum, in a modern system, with a flexible exchange rate, the central bank is best placed to operate as lender of last resort, because it can create liquidity as needed. But while desirable, that is not an essential prerequisite for a lender of last resort, particularly one acting as crisis manager. In a fixed rate system, the central bank is well placed to deal with what used to be called an internal drain, but not necessarily to deal with external shocks.

Third, why does Bagehot insist that the lender of last resort lend against collateral, and why the further requirement that the test be whether the collateral is good in normal times? The availability of collateral is a rough but robust test of whether the institution in trouble is likely to be solvent in normal times.²² By applying this test, the lender of last resort avoids the need to form a judgment on the solvency of the institution applying for liquidity, while retaining the capacity to operate at the speed necessary to stay a panic.²³ At the same time, by basing the decision to lend on the availability of acceptable collateral, the lender of last resort reduces the moral hazard that the potential borrower would take excessive risks in its portfolio by holding assets that would not be accepted as collateral.

The requirement that the collateral be good in normal times is the critical insight. Bagehot's implicit view is that there is a good, normal, equilibrium towards which the lender of last resort is trying to steer the system. By urging lending on the basis of the value of collateral in normal times, he ensures that the lender of last resort does not allow the panic in the market to become self-fulfilling. The further suggestion that the rules of collateral be applied generously goes in the same direction, albeit at the cost of increasing moral hazard.²⁴

Fourth, the penalty rate. The penalty rate is intended to limit the demand for credit by institutions that are not in trouble -- it can be interpreted as one of the ways of dealing with borrower moral hazard, the risk that financial institutions will take excessive risks in normal times secure in the knowledge that they will be able to borrow cheaply in tough times.²⁵ But we need to ask relative to what rate is the penalty. Following the logic that the lender of last resort should try to achieve the good -- non-panic -- equilibrium, the penalty must be relative to the interest rate during normal times. It is not necessarily a penalty rate relative to the rate at which institutions would lend to each other in the market during a panic. In practice, the lender of last resort has frequently lent at a non-penal rate (Giannini, (1998)).

Fifth, should the lender of last resort lend only to the market, and not to individual institutions? This Thornton-Humphrey precept presumably is designed to help deal with the moral hazard of official lending to individual institutions, that might therefore be inappropriately saved from bankruptcy. On this view, given the provision of sufficient liquidity to the markets, the private sector will be able to decide which institutions should be saved. This is a worthy idea, one that should be followed when possible, for instance the day after the 1987 stock market crash, which at that point posed a generalized threat to the market rather than a specific threat to individual institutions. But given the externalities that can follow from the failure of a large institution, and the inherent uncertainties in the midst of a panic over market conditions following the panic, and thus over which institutions should survive, the precept cannot be accepted as a general rule of conduct for the lender of last resort.

Finally, Bagehot wanted the principles on which the lender of last resort would lend to be clearly stated in advance. The notion is that the knowledge of the availability of lender of last resort facilities in the event of a crisis would reduce the incentive to run on otherwise healthy institutions, just as the availability of deposit insurance should prevent runs. But another tradition, that of constructive ambiguity, has become more widely accepted. This view fears the moral hazard created by the existence of the lender of last resort, and therefore seeks to leave potential borrowers in doubt as to whether they will be able to borrow if in trouble. Given the uncertainty, they are more likely to exercise caution in their lending and portfolio decisions. But given the uncertainty, they are also more likely to pull their lines of credit sooner if they anticipate a crisis.

Which is the right view -- spelling out the rules, or constructive ambiguity? By this stage we are deep into issues of moral hazard, to which I will now turn.

"`Moral hazard' ... refers to the adverse effects, from the insurance company's point of view, that insurance may have on the insuree's behaviour."²⁶ The standard but extreme example is that of an individual with fire insurance who burns down the property; the less extreme example is of an insurance holder who takes less care than he otherwise would to prevent the event (e.g. a fire) against which he is insured.

It is important in considering moral hazard to recognize that in the presence of hidden actions, there is no perfect solution to the problem. It will not generally be optimal in a situation of risk to suppress moral hazard entirely. Rather, the optimal solution will generally combine the provision of insurance with measures to limit moral hazard. One aspect of the solution is "the use of incentives -- structuring the transaction so that the party that undertakes the actions will, in his own best interests, take actions that the second party would (relatively) prefer. For example, fire insurance is often only partial insurance so that the insuree has a financial interest in preventing a fire." (Kreps, (1990), p577). Partial insurance, risk sharing, is a quite general feature of the solutions to problems involving moral hazard. So is the use of regulations to try to reduce the risks, for instance by requiring buildings to be equipped with smoke detectors and water sprinklers.

In the case of the domestic lender of last resort, moral hazard problems could arise with respect to both the actions of managers of financial institutions who believe they would receive loans from the lender of last resort during a crisis, and the actions of investors in those financial institutions. However we have to be careful. If the lender of last resort were able to distinguish perfectly and intervene only to stop unwarranted panics, leaving institutions that would be insolvent in normal times to fail, the managers of these institutions and their investors would face the right incentives.²⁷ But since the lender of last resort is unlikely to be able to distinguish perfectly between warranted and unwarranted crises, and since in addition deposit insurance and the too big to fail doctrine create the presumption of moral hazard, measures that attempt to offset the moral hazard of both managers of and investors will be helpful.

How do and should we deal with the potential moral hazard problems in the domestic economy that arise from both deposit insurance and the presence of a lender of last resort? First, official regulation;²⁸ second, encouragement for private sector monitoring and self-regulation; third, by seeking to impose costs on those who make mistakes, including when appropriate by enforcing bankruptcies.

To be eligible for the use of the discount window, banks are expected to obey restrictions on their activities, designed to limit the riskiness of their portfolios.²⁹ Although some banks have been allowed to apply their own risk-control systems, there is no disagreement on the necessity for controls on their risk-taking activities, and little disagreement on the need for official sector monitoring in the form of bank supervision. While regulation extends across the board in the financial markets, as institutions become less critical to the operation of the payments system, the relative reliance on the provision of information to investors increases. These regulations are intended to limit the likelihood of panics and the need for a lender of last resort, while not preventing well-informed risk-taking by investors.

Second, the system seeks to encourage private sector monitoring, particularly by sophisticated investors. The limit on the size of bank accounts covered by deposit insurance is one such device, albeit one that -- because of the practice of too big to fail -- rarely operates when large institutions get into trouble. In addition, when the lender of last resort as crisis manager arranges a rescue package financed by the private sector, it encourages more careful monitoring by the financing institutions and similar firms in future.

Third, the lender of last resort should seek to limit moral hazard by imposing costs on those who have made mistakes. Lending at a penalty rate is one way to impose such costs. Short of bankruptcy, losses incurred should be borne by equity holders and holders of subordinated claims on the firm, and in general management should change.³⁰ Insolvent institutions should be sold or liquidated, with losses borne in the first instance by owners, and then in reverse order of legal precedence. All this of course is made much easier by the existence of bankruptcy laws, which help ensure that workouts for insolvent firms are carried out in an orderly way. In general, the crisis manager is expected to take account of the moral hazard impact of its actions, and is frequently judged on those grounds.

How well do these devices to limit moral hazard work? A first judgment, based on the frequency of domestic financial crises around the world during the last two decades, is not very well. But as we consider this question more carefully, we should remind ourselves that moral hazard is something to be lived with and controlled rather than fully eliminated; that some crises are bound to happen in any system that provides appropriate scope for private-sector risk-taking; that -- as proved by the long history of financial crises well before there were lenders of last resort and deposit insurance -- not all financial crises are caused by moral hazard rather than simple errors of judgment and waves of euphoria and depression; and that the right question is how well the financial system and the economy operate in such a regime compared with a relevant alternative, for instance one in which there is neither a lender of last resort nor deposit insurance. Although I am not aware of careful studies that have attempted to make this more sophisticated judgment, it is fair to guess that such a study would conclude that we need in future to do a better job of controlling moral hazard in the domestic financial system.

We return now to the issue of constructive ambiguity versus spelling out the rules for the operation of the lender of last resort. Constructive ambiguity is one way of dealing with moral hazard, by leaving private sector agents to consider the risk that they will not be covered by the actions of the lender of last resort in a crisis. The main benefit of spelling out the rules --provided they are good and implementable rules, which are understood to be implementable --should be a reduction in the frequency of panics. This benefit has to be compared with the lesser risks that are likely to be taken by investors and the possibly greater frequency of crises under constructive ambiguity. Further, in a rational expectations equilibrium with constructive ambiguity, there will be occasions when the private sector is disappointed because the putative lender of last resort fails to deliver -- and ex post, the economic costs of any given crisis are likely to be greater when the potential lender of last resort fails to act.³¹

Although the central bank or lender of last resort will never be able to spell out precisely the circumstances under which it would act as either a crisis lender or crisis manager, it should specify the general principles on which it will act. For instance, it could announce that it will be willing to act as crisis lender in the event of a systemic crisis, and the rules that it will follow -- for instance some of the Bagehot rules -- as both crisis lender and crisis manager. There are three reasons to spell out the rules: first, and this is Bagehot's justification, by specifying a good set of rules, the central bank reduces the likelihood of unnecessary self-justifying crises; second, by announcing and implementing a particular set of rules it provides incentives for other stabilizing private sector behavior, for instance in the holding of assets good for collateral; and third, by spelling out the rules in advance, it somewhat limits its own freedom of action after the event.³²

Now to the point. Is there likely to be a need in the reformed international financial system for a lender of last resort, for an institution that takes the responsibility to deal with potential and actual crises, either as a crisis lender, or as a crisis manager, or both?

This is not the Kindleberger question, of whether there is a leading central bank that accepts some responsibility for the performance of the global economy. For instance, when Kindleberger (1986) blames the Great Depression on the absence of an international lender of last resort, he means that no agency -- and the natural candidates were the Bank of England and the Federal Reserve -- pursued a monetary policy that took account of the international dimensions of the crisis in which it found itself. Kindleberger would probably say, approvingly, that in the present crisis, the Fed has acted as international lender of last resort in that sense, even though it was taking actions in the interests of the United States.

Rather the question is whether there is a need for an agency that will act as lender of last resort for countries facing a crisis. There is such a need: it arises both because international capital flows are not only extremely volatile but also contagious, exhibiting the classic signs of financial panics,³³ and because an international lender of last resort can help mitigate the effects of this instability, and perhaps the instability itself. This applies particularly to emerging market countries, where the crises of the last five years have been concentrated.

This conclusion also rests on the view that international capital mobility is potentially beneficial for the world economy, including for the emerging market and developing countries. But this potential can only be realized if the frequency and scale of capital account crises can be reduced. The founders of the Bretton Woods system provided for the use of capital controls to deal with capital flows. Some controls -- particularly market-based controls that seek to limit short-term capital inflows -- can be envisaged as a useful part of a transitional regime while the macroeconomic framework and financial structure of the economy are strengthened. The use of controls to limit capital outflows has been advocated in this crisis by several academics³⁴ and adopted by Malaysia. But it is surprising how few countries have taken that route. Indeed, policy makers in Latin American countries that generally had such controls in the 1980s have rejected them this time around, emphasizing that controls were inefficient, widely avoided, and had cost them dearly in terms of capital market access. It remains an open question whether this crisis will result in the more frequent use of controls on capital outflows, either in normal times, or in the midst of crises.³⁵ The answer will depend to an important extent on the success of the reforms to be implemented in the next few years.

I will argue not only that the international system needs a lender of last resort, but also that the IMF is increasingly playing that role,³⁶ and that changes in the international system now under consideration will make it possible for it to exercise that function more effectively.

In focusing on the Fund's potential lender of last resort role, I leave aside its other important functions, among them: "[t]o promote international monetary cooperation through a permanent institution which provides the machinery for consultation and collaboration on international monetary problems"³⁷; lending for current account purposes to countries that lack market access; surveillance and the associated provision of information; and technical assistance including policy advice and monitoring.

Let me immediately dispose of the argument that the IMF cannot act as a lender of last resort because it is not an international central bank and cannot create international reserves. As already discussed, the domestic lender of last resort -- whether as crisis lender or as crisis manager -- is not necessarily the central bank. As crisis lender, the IMF, whose financial structure is close to that of a credit union, has access to a pool of resources, which it can onlend to member countries. As crisis manager, it has been assigned the lead in negotiating with member countries in a crisis, and it cooperates in arranging financing packages.

The question arises whether the IMF, as crisis lender, has sufficient resources to do the job. The Fund has reached its present size as a result of a series of quota increases, approximately once every five years. Relative to the size of the world economy it has shrunk significantly since 1945. If the IMF were today the same size relative to the output of its member states as it was in 1945, it would be more than three times larger than it will be when the present quota increase is completed;³⁸ if the quota formula applied in 1945 were used to calculate actual quotas today, the Fund would be five times its size; and if the size of the Fund had been maintained relative to the volume of world trade, it would be more than nine times larger -- that is, the size of the Fund would be over 2.5 trillion dollars.³⁹ However the decline relative to the likely borrower base is smaller, for whereas all members but the United States and probably Canada were potential borrowers in 1945, most of the rich countries are at present unlikely to borrow from the Fund.

Despite this significant shrinkage relative to the original conception, the Fund as lender of last resort is still able to assemble a sizeable financial package in response to a crisis. In case of systemic problems, the Fund can borrow from the New or General Arrangements to Borrow. Further, as demonstrated in the recent Brazilian and East Asian packages, member governments and other international financial institutions may add significantly to these packages in cases they deem to be of systemic importance or particular importance to them. Whether the Fund will in future be large enough relative to the scale of problems, will depend on the future scale and volatility of international capital flows, which will in turn depend on the effectiveness of reforms, including measures to deal with problems of moral hazard.

We noted in the domestic case that while it is not essential that the lender of last resort be the central bank, it is helpful. Would it be useful for the IMF to be able to create reserves? Under Article XVIII of the Articles of Agreement, the Fund can by an 85 percent majority allocate SDRs "to meet the long-term global need, as and when it arises, to supplement existing reserve assets". It is possible -- indeed it was very much possible a few months ago -- to envisage circumstances under which a general increase in reserves would be useful, for instance, at a time of a seizing up of flows of credit in the world economy. However, a general allocation of SDRs has to be made in proportion to quota holdings, and would not in its current form be well suited to dealing with a problem that affects a specific group of countries, such as emerging market countries.⁴⁰

The IMF thus has the capacity to act as crisis lender to individual countries, and in specified circumstances, through an issue of SDRs, could lend more broadly. It also acts as crisis manager. Kindleberger (1978, p226) complains that the Fund is too slow in emergencies, but it has in recent years demonstrated the ability to move very rapidly, using the Emergency Financing Mechanism introduced after the Mexican crisis. The main constraint on the IMF's ability to act in time is that governments delay too long in approaching it, in part because excessive delay is a characteristic of governments that get into crises, also because they hope somehow to avoid taking the actions that would be needed in a Fund program.

The IMF, although it is not an international central bank, thus acts in important respects as international lender of last resort. But the job can surely be done better.

Before addressing that issue directly, I will discuss four central elements in the reform of the international system: exchange rate systems; reserve holdings; measures to bail in the private sector; and international standards.

Over a century of controversy has produced no clear answer to the question of which exchange rate system or monetary regime is best. A country's history, particularly its history of inflation, is a critical consideration in determining its choice of exchange rate regime. Nonetheless it is striking that the major external crises of the last two years -- in Thailand, Korea, Indonesia, and Russia -- have affected countries with more or less pegged exchange rates. Further, the assumption that the exchange rate was stable profoundly affected economic behavior in those countries, especially in the banking system, and contributed to the severity of the post-devaluation crises.

At the same time, we should not forget that several countries with fixed rates, particularly Argentina and Hong Kong, have succeeded in holding the line, and that some with flexible rates, among them Mexico, South Africa and Turkey, have been severely affected by the global economic crisis. Nor should we forget that many countries benefitted from using the exchange rate as a nominal anchor in disinflating, and that the fear of devaluation is often the best discipline on weak governments.

There is a tradeoff between the greater short-run volatility of the real exchange rate in a flexible rate regime versus the greater probability of a clearly defined external financial crisis when the exchange rate is pegged. The virulence of the recent crises is likely to shift the balance towards the choice of more flexible exchange rate systems, including crawling pegs with wide bands. If countries choose to fix, they may well want to do so definitively, through a currency board, rather than a less credible normal peg, assuming of course that the necessary preconditions are in place. But while we will probably see fewer nominal pegs in the coming years, we are unlikely to move to a system in which exchange rates for all countries float.⁴¹

A shift towards floating rates is likely to reduce the frequency of sharply defined foreign exchange crises. But because some countries will continue to peg their rates, and because sharp shifts in international investor sentiment regarding even a country with a floating rate can set off a panic and contagion, there will still be a need for an international lender of last resort.

Second, international reserves. The first line of defense in dealing with capital flow reversals, aside from macroeconomic policy and exchange rate responses, is to use the foreign exchange reserves. There has been surprisingly little emphasis in discussions of the present crisis on the fact that the countries with very large reserves have done better in dealing with the crisis than those with small reserves. But that is a fact, and it is very likely that countries seeking to draw the lessons of the present crisis will decide they should hold much larger reserves than before. This is already happening in the case of Korea -- and it will not be the only country to move in that direction.

Should countries hold larger reserves and rely less on the lender of last resort?^42,43 In many models, and in practice, the ratio of reserves to short-term external liabilities is an important factor determining the likelihood of a financial crisis (Calvo, 1995). This emphasizes that countries need to set their reserve holdings on the basis of capital, as well as current, account variables.⁴⁴ It is thus likely that the demand for reserves will increase as capital accounts become more open and international capital flows more readily. We should recognize that a general desire by emerging market countries to build up reserves by running current account surpluses in the next few years will impart a deflationary impact to the world economy.

Those reserves could be obtained not only through a current account surplus,⁴⁵ but also by international agreement on, for example, an issue of SDRs. Both aggregate demand and the distribution of seigniorage are different in these two approaches; there are also legitimate concerns that recipients of a general reserve issue may spend rather than on average hold them.⁴⁶ Another possibility is to rely more on the lender of last resort or on precautionary lines of credit of the Argentine type. It is not possible without a more detailed analysis to decide which approach is preferable, but I expect that the recent experience of crises will lead to larger holdings of reserves, one way or another.

Third is private sector involvement in the solution of financial crises. A first and constructive possibility is the already-noted approach followed by Argentina and a few other countries, of putting in place precautionary lines of credit from private sector lenders. This is a useful supplement to the holding of reserves, and could possibly be cheaper than increasing reserves. A second and very important approach, suggested in a report by the G-10 deputies after the Mexican crisis, is the proposal that bond contracts should be modified to permit a country to reschedule payments in the event of a crisis. More generally, clauses on collective representation and majority decisions by creditors could be included in bond and other contracts, to facilitate the reaching of agreements with creditors in times of crisis.^{47 48} Another suggestion, most prominently associated with Jeffrey Sachs, is the possibility of a formal imposition of a stay on payments by a country in crisis.⁴⁹ In addition, the Executive Board of the IMF has agreed that the Fund may lend to countries in arrears to private creditors, provided they are pursuing appropriate policies and making good faith efforts to cooperate with the creditors.

Critical as the issue of private sector involvement is, it has to be approached carefully, lest proposed solutions increase the frequency of crises. For instance, the formalization of a requirement that the banks, or any other set of creditors, always be forced to share in the financing of IMF programs, would be destabilizing for the international system. If such a condition were insisted on, the creditors would have a greater incentive to rush for the exits at the mere hint of a crisis. This is a real dilemma, one that suggests not only a role for a lender of last resort, but also the need for a differentiated approach to involving the private sector that depends on the circumstances of each country: sometimes a formal approach may be necessary, as in Korea at Christmas in 1997; at other times less formal discussions could serve better; and on occasion, if a country enters a program sufficiently early, there might not be any need to approach the creditors.

Fourth, work is under way to define a set of international standards and to encourage countries to adopt them. The best known standards are those for banking, defined by the Basle Committee on Banking Supervision, including its Core Principles set out in 1997. The IMF's Special Data Dissemination Standard (SDDS) has just gone into full operation. Codes of fiscal practice and monetary and financial transparency are also being prepared by the IMF in cooperation with other institutions. Among other important international standards already developed or in the process of development are international accounting standards, IOSCO standards for the operation of securities markets, and an international standard for bankruptcy regulations.

If countries adopt these standards, their financial systems should work better, and the likelihood and intensity of crises would be reduced. At this point, the main incentives for a country to adopt the standards are the expectation that the economy would operate more efficiently, and the hope that international investors would treat the economy more favorably. The evidence that most leading emerging market countries have subscribed to the IMF's SDDS suggests that these incentives will have a positive effect. In addition, a major international effort will be undertaken to improve banking standards, in part through international monitoring and IMF surveillance. Nonetheless, further incentives are needed, for instance the risk weights assigned by regulators in creditor countries could reflect the recipient country's observance of the standards.

At the end of 1997, the IMF introduced the Supplemental Reserve Facility (SRF), which can make short-term loans in large amounts at penalty rates to countries in crisis. SRF loans have been made to Korea, Russia, and Brazil, subject to policy conditionality. At present, the Executive Board of the IMF is considering the possibility of introducing a contingency or precautionary facility, to supplement the reserves of countries threatened by a crisis but not yet in one. The recent loan to Brazil included several of the features contemplated for the contingency facility. The lending terms for the contingency facility would be similar to those for the SRF, but some thought is being given to the possibility that countries could prequalify for assistance.

The major reforms of the international system now on the agenda would have to be implemented for these lender of last resort-like facilities to operate effectively. In particular, the improvement of standards, other changes to improve transparency and increase relevant information,⁵⁰ together with improved procedures to bail in the private sector, would reduce the frequency and scale of crises. The design of procedures for private sector bail-ins is the most difficult issue of systemic reform to be confronted in the next few years, and the standards initiative is no less important.

What about the Bagehot lessons, that in a crisis the lender of last resort should lend freely, at a penalty rate, on good collateral, but that institutions that would be bankrupt in normal times should not be saved? Both the penalty rate and the notion of lending freely have been incorporated in the SRF. Policy conditionality can be interpreted as a further element of the penalty, as seen from the viewpoint of the borrower country's policymakers. It is not obvious how to judge whether loans made under the SRF correspond adequately to "lending freely", but given that loans have to be made to an individual country, some caution about excessive financing is warranted in the light of moral hazard. Possibly "lending freely" in the international context translates to the condition that the international lender of last resort should stand ready to lend early and in sufficient amounts to other countries that might be affected by contagion from the crisis. The Fund's capacity to do that has been enhanced by the agreement on the quota increase.

With regard to bankruptcy, private sector debtors in the crisis country should be covered by national bankruptcy laws. This is one of the reasons that a major effort is under way to strengthen these laws. There is no bankruptcy status for a sovereign, but workout procedures, including those of the Paris and London Clubs, and possibly those to be developed as bail-ins are considered further, play a similar role. The economics of such a quasi-bankruptcy for a sovereign debtor is complex, for the ability to generate repayments is more a matter of political than of economic feasibility.

The Articles of Agreement permit the Fund to ask for collateral, but it has only once done so.⁵¹ The Fund and the World Bank are regarded as preferred creditors, who have a first claim on payments made by countries in debt to them, and their collateral is in a sense the denial of market access to countries that would default. In considering the explicit provision of collateral for Fund programs, it has been argued that there is a tradeoff between the amount of policy conditionality that accompanies a loan, and the amount of collateral -- and that policy conditionality is more important. However, if the contingency facility moves in the direction of lending in anticipation of a crisis, there could be a greater reliance on collateral and somewhat lesser reliance on conditionality. This would also have good incentive effects on the country: if it had to have internationally acceptable assets available to serve as collateral, it would presumably be discouraged from running down its reserves too far before calling on the Fund for assistance.⁵²

How would these changes in the system deal with the moral hazard problem? Recall that three mechanisms were discussed in the domestic context: official regulation; private sector monitoring and self-regulation; and the imposition of costs on those who make mistakes. The adoption of international standards would raise the quality of official regulation. Improvements in transparency and the provision of information by the public sector and improved regulation, together with bail-in procedures that set the right incentives, would encourage better monitoring and self-regulation by the private sector. The charging of a penalty rate would discourage borrower moral hazard and the new procedures to bail in the private sector would greatly reduce investor moral hazard.⁵³

In considering how to limit moral hazard, we should also distinguish the hazards associated with different types of international lending.⁵⁴ The problem is far more serious for interbank lines of credit than, say, for equity investment. The responsibility for dealing with the moral hazard problem for interbank lines of credit lies as much with the government of the lender as with the borrower government, for it is the former which supervises and tends to protect its banks. Lender supervisory authorities will have to recognize the responsibilities of their institutions to participate in bail-ins and workout procedures.

The SRF and the contingent facility now under consideration, together with the changes to the international system now under discussion, would go a long way towards making the international capital markets operate as well as the better domestic capital markets. But there remains the nagging question of how to provide incentives for countries to adopt the necessary international standards.

An important suggestion in this regard has been made by Calomiris (1998) and Calomiris and Metzler (1998). They recommend that the IMF act only as lender of last resort, under Bagehot rules, and only to countries that meet a stiff set of requirements, most importantly on the banking system. Among these conditions is the requirement that foreign banks be allowed to operate in the country, a change that should be adopted in any case. Loans would be made to qualifying countries on the basis of collateral, and without policy conditionality.

Without going into the overall merits of their analysis,⁵⁵ I would like to draw on the suggestion that only countries that meet specified standards be eligible to borrow from the lender of last resort. Recall the concern that countries will not have sufficient incentives to meet the new international standards. Suppose that only countries that adopt certain standards would have access to, for example, the new contingency lending facility. This would provide a powerful additional incentive to adopt these standards.

For such a scheme to work, lender-of-last-resort loans would have to be denied to countries that do not qualify. Too big to fail makes this a very difficult task -- and contagion makes too big to fail a rational strategy. One way of dealing with the problem is to vary the rate of charge depending on the extent to which a country has met the relevant standards. Another is to allow a non-qualifying country to borrow only with tougher policy conditionality, which would make sense given the structural weaknesses implied by the country's failure to meet the standards.

The development of the standards, and of the international mechanisms to monitor them, will take time. It will take more time for countries to adopt and implement the standards, even though a few emerging market countries may meet some of them already. We should thus expect that it will take at least five years until a new system, based on prequalification for access to a new facility, could begin operating. In the meantime a transition process could be designed, in which countries can qualify for the new facility based on their making good faith efforts to come closer to meeting the standards, possibly with an initial emphasis on the banking standard -- just as a transition process was put in place for countries to meet the Special Data Dissemination Standard. We should not underestimate the complexity of the task and the resources that would be required to improve standards in this way, but any cost-benefit analysis for the world economy of a successful effort to reduce the frequency and scale of crises would justify making the attempt.

The crises of the last five years have revealed major weaknesses in the structure of the international economy. Much good work has been done to analyze the sources of weaknesses, and present potential solutions. With economic policymakers' attention still focused on these problems, it is urgent to start developing and implementing the solutions. Among them is the further definition and development of the role of the international lender of last resort.

These changes should not result in an increase in IMF lending. For if the reforms succeed, there will be fewer international crises, and fewer occasions for crisis lending. And that, surely, is a goal we all share.

Bagehot, Walter (1866). "One Banking Reserve or Many?", The Economist, Vol. 24 (September 1), 1025-6.

     , (1873). Lombard Street: A Description of the Money Market. London: William Clowes and Sons.

Baring, Sir Francis (1797). Observations on the Establishment of the Bank of England and on the Paper Circulation of the Country. To which is added Further Observations. New York: A. M. Kelley (reprint, 1967).

Benston, George, Robert Eisenbeis, Paul Horvitz, Edward Kane, and George Kaufman (1986). Perspectives on Safe and Sound Banking: Past, Present, and Future. Cambridge, Massachusetts: MIT Press.

Bordo, Michael (1989). "The Lender of Last Resort: Some Historical Insights", National Bureau of Economic Research Working Paper Series, 3011 (June).

Boughton, James (1998). "From Suez to Tequila: The IMF as Crisis Manager". Unpublished; Washington: International Monetary Fund.

Brown, Gordon (1998). "Rediscovering Public Purpose in the Global Economy". Speech delivered at the Kennedy School, December 15.

Calomiris, Charles (1998). "Blueprints for a New Global Financial Architecture". Unpublished; New York: Columbia Business School.

     , and Allan H. Meltzer (1998). "Reforming the IMF". Unpublished; New York: Columbia Business School.

Calvo, Guillermo (1995). "Varieties of Capital Market Crises". University of Maryland Center for International Economics, Working Paper 15.

Capie, Forrest (1998). "Can there be an International Lender of Last Resort". Unpublished; London: City University Business School.

Claassen, Emil-Maria (1985). "The Lender-of-Last-Resort Function in the Context of National and International Financial Crises", Weltwirtschaftliches Archiv, 121, 2, 217-237.

Clapham, Sir John (1944). The Bank of England. Cambridge: Cambridge University Press.

Chang, Roberto and Andres Velasco (1998). "The Asian Liquidity Crisis". NBER Working Paper 6796 (November).

Diamond, Douglas and Philip Dybvig (1983). "Bank Runs, Deposit Insurance, and Liquidity", Journal of Political Economy (June), 91, 3, 401-419.

Fischer, Stanley (1998). "Reforming the International Monetary System". Unpublished; Washington: International Monetary Fund.

Garcia, Gillian and Elizabeth Plautz (1988). The Federal Reserve: Lender of Last Resort. Cambridge, Massachusetts: Ballinger.

Giannini, Curzio (1998) "'Enemy of None but a Common Friend to All'? An International Perspective on the Lender-of-Last-Resort Function". Unpublished; Washington: International Monetary Fund.

Goodfriend, Marvin and Robert King (1988). "Financial Deregulation, Monetary Policy, and Central Banking" in W. S. Haraf and R. M. Kushmeider (eds), Restructuring Banking and Financial Services in America. Washington: American Enterprise Institute for Public Policy Research.

Goodhart, Charles A. E. (1995). The Central Bank and the Financial System. Cambridge, Massachusetts: MIT Press.

     , and Haizhou Huang (1998). "A Model of the Lender of Last Resort". Unpublished; Washington: International Monetary Fund.

     , and Dirk Schoenmaker (1995). "Should the Functions of Monetary Policy and Bank Supervision Be Separated?", Oxford Economic Papers, 47, 539-560.

Guesnerie, Roger (1987). "Hidden actions, moral hazard and contract theory", in John Eatwell, Murray Milgate and Peter Newman (eds), The New Palgrave: A Dictionary of Economics, Volume II, 646-651. London: The Macmillan Press.

Guttentag, Jack and Richard Herring (1983). "The Lender-of-Last-Resort Function in an International Context" Princeton University Essay in International Finance No. 151 (May).

Hirsch, Fred (1977). "The Bagehot Problem", The Manchester School, 45, 3 (September), pp.241-257.

Holmstrom, Bengt and Jean Tirole (1998). "Private and Public Supply of Liquidity", Journal of Political Economy, 106,1 (February), pp. 1-40.

Humphrey, Thomas (1975). "The Classical Concept of the Lender of Last Resort", Federal Reserve Bank of Richmond Economic Review, 61 (February), pp. 2-9.

     , and Robert Keleher (1984). "Lender of Last Resort: A Historical Perspective", Cato Journal, 4, 1, pp. 275-321.

Jeanne, Olivier (1998). "The International Liquidity Mismatch and the `New Architecture'". Unpublished. Washington: International Monetary Fund.

Kaufman, George (1988). "The Truth about Bank Runs" in C. England and T. Huertas (eds), The Financial Services Revolution: Policy Directions for the Future. Boston: Kluwer Academic Publishers.

      (1991). "Lender of Last Resort: A Contemporary Perspective", Journal of Financial Services Research, 5, pp. 95-110.

Keynes, John Maynard (1943). "A Comparative Analysis of the British Project for a Clearing Union (C.U.) and the American Project for a Stabilization Fund (S.F.)" in D. Moggridge (ed), The Collected Writings of John Maynard Keynes, Vol. XXV (1980). London: MacMillan.

Kindleberger, Charles (1978). Manias, Panics, and Crashes: A History of Financial Crisis, New York: Basic Books.

      (1986). The World in Depression, 1929-1939. Revised and enlarged edition. Berkeley: University of California Press.

Kreps, David M. (1990). A Course in Microeconomic Theory. Princeton, New Jersey: Princeton University Press.

MacKay, Charles (1841). Extraordinary Popular Delusions and the Madness of Crowds. New York: Farrar, Straus and Giroux (reprint, 1932).

Meltzer, Allan (1986). "Financial Failures and Financial Policies" in G. G. Kaufman and R. C. Kormendi (eds), Deregulating Financial Services: Public Policy in Flux, Cambridge, Massachusetts: Ballinger.

      (1988). "What's Wrong with the IMF? What Would be Better?" paper prepared forFederal Reserve Bank of Chicago Conference: Asia: An Analysis of Financial Crisis, October 8-10.

Mints, Lloyd W. (1945). A History of Banking Theory. Chicago: University of Chicago Press.

Mundell, Robert (1983). "International Monetary Options", Cato Journal, 3, 1, pp. 189-210.

Radelet, Steven, and Jeffrey D. Sachs (1998). "The East Asian Financial Crisis: Diagnosis, Remedies, Prospects", Brookings Papers on Economic Activity, 1, 1-74.

Rockoff, Hugh (1986). "Walter Bagehot and the Theory of Central Banking" in F. Capie and G. E. Wood (eds), Financial Crises and the World Banking System. New York: St. Martin's Press.

Schwartz, Anna (1988). "Financial Stability and the Federal Safety Net" in W. S. Haraf and G. E. Kushmeider (eds), Restructuring Banking and Financial Services in America. Washington: American Enterprise Institute for Public Policy Research.

Smith, Vera (1990). The Rationale of Central Banking and the Free Banking Alternative. Indianapolis: Liberty Press.

Solow, Robert (1982). "On the Lender of Last Resort" in C. P. Kindleberger and J. P. Laffargue (eds), Financial Crisis: Theory, History, and Policy. Cambridge: Cambridge University Press.

Tirole, Jean (1988). The Theory of Industrial Organization. Cambridge, Massachusettes: MIT Press.

Thornton, Henry (1802). An Enquiry into the Nature and Effects of the Paper Credit of Great Britain. Edited by F. A. Hayek. Fairfield: Augustus M. Kelley Publishers (reprint, 1978).

Zettelmeyer, Jeromin (1998). "International Financial Crises and Last Resort Lending: Some Issues" Unpublished. Washington: International Monetary Fund.