Toward a Lender of First Resort
  • 1 0000000404811396 Monetary Fund

Contributor Notes

Author(s) E-Mail Address:;

If interest rates (country spreads) rise, debt can rapidly be subject to a snowball effect, which becomes self-fulfilling with regard to the fundamentals themselves. This is a market imperfection, because we cannot be confident that the unaided market will choose the "good" over the "bad" equilibrium. We propose a policy intervention to deal with this structural weakness in the mechanisms of international capital flows. This is based on a simple taxonomy that breaks down the origin of crises into three components: confidence (spreads and currency crisis), fundamentals (real growth rate), and economic policy (primary deficit). Theory then suggests a set of circumstances in which a lender of first resort would be desirable. The policy would seek to short-circuit confidence crises, partly by using IMF support to improve ex ante incentives. Theory also illuminates the potential role of collective action clauses in reducing the risk of self-fulfilling debt crises.


If interest rates (country spreads) rise, debt can rapidly be subject to a snowball effect, which becomes self-fulfilling with regard to the fundamentals themselves. This is a market imperfection, because we cannot be confident that the unaided market will choose the "good" over the "bad" equilibrium. We propose a policy intervention to deal with this structural weakness in the mechanisms of international capital flows. This is based on a simple taxonomy that breaks down the origin of crises into three components: confidence (spreads and currency crisis), fundamentals (real growth rate), and economic policy (primary deficit). Theory then suggests a set of circumstances in which a lender of first resort would be desirable. The policy would seek to short-circuit confidence crises, partly by using IMF support to improve ex ante incentives. Theory also illuminates the potential role of collective action clauses in reducing the risk of self-fulfilling debt crises.

I. Introduction

The discussions on the international financial architecture that followed the Asian crisis of 1997–98 revived the debate over the international financial architecture. Drawing the lessons of the crisis, Stanley Fischer (1999) first proposed that the IMF act as international lender of last resort (ILLR). In November 2001, Anne Krueger, his successor as First Deputy Managing Director of the IMF, advocated a Sovereign Debt Restructuring Mechanism (SDRM) to facilitate a declaration of insolvency for an over-indebted country along the lines of Chapter 11 of the U.S. Bankruptcy Code (Krueger, 2001). One institutional manifestation of the Fischer proposal was the Contingent Credit Lines (CCL) facility, which would have enabled a country affected by a contagion crisis to draw on additional lines of credit. No country made use of this facility, however, and it was eventually shelved in early 2004. The Krueger proposal has not been implemented either. Despite subsequent revisions that reduced the role of the IMF (Krueger, 2002), the proposal was also shelved at the April 2003 meetings, partly because it would have required an amendment to the Articles of Agreement (IMF, 2003).

Both these proposals (ILLR and SDRM) have proved too ambitious in the policy environment. As argued by many commentators (e.g., Jeanne and Wyplosz, 2001) an ILLR must have at its disposal either the resources to inject an indeterminate quantity of fresh liquidity or perfect information regarding solvent and insolvent financial intermediaries. As the latter assumption is virtually ruled out by the very nature of financial crises, the former is tantamount to giving the IMF the means to create liquidity ex nihilo. Such a transfer of monetary sovereignty, which was extremely difficult to implement in the European case, seems totally unrealistic on a world scale. If there is to be a world LLR, it is rather for the large central banks (the U.S. Federal Reserve, the European Central Bank, and the Bank of Japan) to play this role.

Anything along the lines of the Fund’s SDRM proposal has appeared to be infeasible for the same political reason. Setting up an international court with authority over the handling of sovereign debt would entail a substantial transfer of sovereignty, in order to give the court the statutory basis for suspending legal procedures against a country. Nevertheless, there is no doubt that the Krueger initiative dramatically changed the terms of the discussion and gave impetus to the adoption of Collective Action Clauses (CACs) (see our discussion in Cohen and Portes, 2003).

Beyond the political constraints, these policy developments also have been criticized on a more analytical ground. When financial crises erupt, the action taken by the IMF either as an LLR or through an SDRM cannot ignore the underlying causes. It is not appropriate to treat in the same manner a country that is the victim of an unforeseeable loss of market confidence and a country where the macroeconomic indicators have long been unsatisfactory and which therefore is borrowing at abnormally high interest rates. It is for dealing with situations in which a country is suffering from a lack of confidence unjustified by any major deterioration in its fundamentals that the ILLR approach would be useful. It is for dealing with situations in which the debt no longer bears a relation to the fundamentals that the procedures involving bankruptcy or debt reduction might be viewed as essential. Correctly applying such a distinction is very difficult under the pressure of time, when a crisis erupts. This is partly because there will always be doubts over the motives prompting investors to withdraw their confidence.

The doubt regarding the nature of crises explains the risk of moral hazard. Because it is not always possible to distinguish the “good” debtors who have been unlucky from the “bad” who have continued to implement unsustainable policies, intervention by the IMF has continually swung between too much and too little. It was to circumvent these difficulties that the Meltzer Commission (2000) proposed confining the Fund’s scope for action to only those countries that “pre-qualify” based on strict criteria of indebtedness and transparency. But this proposal offers nothing for the countries that fall outside the scope of such pre-qualification, and this is hardly feasible.

Another key reason why distinguishing between confidence crises and crises of fundamentals is difficult is that the former often rapidly turn into the latter: if interest rates rise, debt can rapidly be subject to a snowball effect, so that the initial worries about debt levels then become self-fulfilling. In other words, there are multiple equilibria: low rates represent one equilibrium, high rates another. This is the argument used by Williamson (2002) to characterize the Brazilian situation at the time: the debt was at a level made unsustainable by high interest rates but would rapidly be brought down to a sustainable level (given the government’s primary surpluses) by low interest rates.

It is this dual dimension—ambiguity regarding crisis situations, partially self-fulfilling capacity of negative judgments on a country’s situation—that leads us to propose a policy intervention to deal with this structural weakness in the mechanisms of international capital flows. We argue that IMF members should be able to commit themselves ex ante, should they so wish, to an “indebtedness regime” (similar to the “fixed exchange rate regime” to which they subscribed for many years) that makes it possible for them to carry out preventive action regarding the evolution of debt. The idea is to give them the means to act before the snowball effect comes into play, given that analysis of the debt build-up mechanism shows that it takes time, and therefore provides time, before the situation becomes explosive. This indebtedness regime would be based on the spreads paid. For the sake of simplicity, let us suppose here that a country undertakes never to borrow at spreads greater than 400 basis points. The indebtedness regime signifies that the country will take all necessary steps to hold its indebtedness down to a level compatible with this level of interest rate. If the regime is “credible,” in other words if investors are convinced that rates will never go above this level, multiple equilibria can be ruled out, in that the mechanism “coordinates” expectations on a low level. Moreover, and in our view more importantly, this indebtedness regime has the merit of committing the country to a prudent strategy. It would avoid the widespread temptation to allow problems to accumulate before tackling them and in so doing to become vulnerable to a crisis of confidence, which it is then too late to avoid.

An indebtedness regime of this kind implies active commitment, on the part not only of the country but also of the IMF. Rather than intervening ex post, when the country has lost its access to the financial markets, the mechanism we propose is that the IMF should launch an adjustment program with the country early on, while the country keeps its access to the markets. Arguably, any adjustment program will be milder if applied before the crisis rather than after. The debt will be lower and the country will have “demonstrated” its resolve to act promptly. The preventive measures will avoid questions being posed too late, when the only remedies are extreme shock treatment or default. Furthermore, to the extent that fresh IMF resources may make a difference, they reduce the importance of the snowball effect of high interest on the dynamics of debt. This point, however, is secondary in our view to the merit of taking preventive action.

Obviously, these preventive measures will not solve all the problems. If a country fails to avert the crisis, steps have to be taken. The resolution of the crisis, through a restructuring of the debt in the case where it is unsustainable or through its consolidation in the case of a liquidity crisis, would take place, as usual, under the aegis of the IMF. On the side of the creditors, CACs are the essential instruments making it possible to reach rapid agreement. We propose two simple innovations in this respect. First, if the markets themselves do not implement CACs comprehensively, then the principal financial centers (in particular, New York and London) should adopt a coordinated measure prohibiting debt issues that do not contain CACs. Second, we propose the creation, alongside the Paris Club (dealing with sovereign debt) and the London Club (dealing with bank debt) of a new club to handle bond debt, that might be called the New York Club. A slim-line committee to coordinate these three clubs could also be set up.

The rest of the paper proceeds as follows. We first examine the differences between the debt crises of the 1980s and the debt crises of the 1990s. We show that confidence (or lack thereof) played in the 1990s a substantial role in explaining the debt dynamics of a number of countries. Specifically, we propose a simple taxonomy making it possible to break down the origin of crises into three components: a crisis of confidence (spreads and currency crisis), a crisis of fundamentals (real growth rate), and a crisis of economic policy (primary deficit). As we shall report, there is no such thing as a “pure” situation in which the confidence crisis fully explains the debt crisis, as in the multiple-equilibrium model. On the other hand, there are indeed cases (more in line with intuition) where up to 40 percent of the accrued debt of the most indebted countries stems from interest payments and currency crises.

We then present a theoretical model in which we show that self-fulfilling debt crises (à la Calvo, 1988; or Cole and Kehoe, 1996, 2000) can happen only when debt restructuring is expected to be inefficient, ex post. In our model, a country whose fundamentals are not affected by debt dynamics is indeed immune to confidence crises. In cases where this is not the case, in practice the most likely scenario, we then show the benefit of a commitment device on debt dynamics in order to avoid confidence crises. In the last section of the paper, we draw the policy implications of our analysis: the comprehensive use of CACs to facilitate efficient debt restructuring and the creation of an International Lender of First Resort (ILFR), implemented by the IMF, to provide the commitment device that precludes self-fulfilling crises.

II. The Financial Crises of the 1990s Were Different from Those of the 1980s

In the period leading up to 1982, when Mexico suspended payment on its debt, spreads were very low, rarely exceeding 200–250 basis points, as most bankers at the time thought that countries did not default. Spreads on both Mexican and Brazilian debt did rise in the few months before the debt moratorium, but the syndicated bank lending of the 1970s and early 1980s showed no signs of repeating the 1930s. Although spreads did vary somewhat with the characteristics of the borrower, there was no perceptible market discipline. The bulk of the financial crises involved syndicated loans with very low spreads, and the average real rate of interest on sovereign borrowing in the 1970s was negative. The debt crisis of the 1980s was not anticipated by the lenders. The resolution of the crisis took several painful years, during which Latin American economies stagnated—to the point where income per capita returned to the late 1960s’ level, in what has often been called a lost decade.

The nature of the debt crises changed in the 1990s. The agents were different: corporate borrowers joined sovereign debtors. Lenders were different, too, comprised increasingly of bondholders rather than bank loan syndicates. The 1980s story, according to which high public deficits created high debt and eventually interest rate rises brought major crises, is not the only one at hand. Confidence crises, through exchange rates or through interest rates, created new scenarios. Crises became more complex: the Asian crises, the Mexican crisis, the Russian crisis give a range of cases that are difficult to subsume under one story. Some crises were expected, some were unexpected, and quite often, for good reasons.

As examples of “foretold” crises, take the cases of Argentina and Ecuador; at the other extreme, take Korea or Mexico.

Table 1.

Crisis Episodes: Selected Indicators

(data two years before the crisis, percentage points unless otherwise indicated)

article image
Source: authors’ calculations based on International Financial Statistics, IMF; and Global Development Finance, World Bank.

From the comparison of these two cases, it is fairly clear that Argentina and Ecuador were fundamentally insolvent, at least with respect to one of the two criteria that are commonly used: debt-to-export ratio above 200 percent and/or debt-to GDP ratio above 50 percent (note, however, that it takes both indicators to anticipate a crisis, on which more later). Huge spreads were paid, and at the time when the crisis erupted, no lender could claim that it was taken by surprise. Yet despite this apparent market awareness, many lenders were taken by surprise, and the discipline of higher spreads had little perceptible effect on the policies of Argentina or its creditors. Argentina was able to borrow at excessive spreads, which simply worsened its fiscal position and exacerbated the crisis and its consequences. This is a case where a write-down of the debt was needed in order to return as soon as possible to sustainable growth.

Case 2 is exactly the opposite. No major macroeconomic disequilibria were observable, insofar as the outstanding stocks were concerned; spreads were correspondingly low. In the case of Mexico, however, it was clear that the large current-account deficit was creating liquidity pressures. In contrast, Korea failed by none of these criteria. Indeed, its weakness came from elsewhere, i.e., the short-term nature of its debt. As the current account demonstrates, however, there was no particular need for a major exchange rate adjustment.

In Case 3, the sovereign risk pertains to the nature of the debtor. Despite good macroeconomic performance, creditors could examine the macroeconomics and perceive the risk of default that the shaky government or the shaky banking system could create. The spreads were correspondingly high.

Let us summarize the discussion so far with the following:

Table 2.


article image

Compared to the 1980s, then, it does not appear to be the case that large disequilibria went unnoticed by the markets. As we now discuss, the high-debt/low-spread cell is empty. In this sense, market discipline has improved.

III. Debt Crises of the 1990s: A Taxonomy

We present in the Appendix the list of countries that signed a program with the IMF during the 1990s. We distinguish three groups of countries according to the nature of the program. Group A (“hard crises”) includes all countries which have experienced one EFF (Extended Fund Facility); group B includes all (other) countries which have experienced more than one SBA (Standby Agreement) in a row (intermediate crises); group C includes all countries which have experienced only one SBA (short crises).

Except for a few cases to which we shall return, the three groups behave as one would expect. Debt is high in group A, moderate in group B, low in group C. More specifically, the debt-to-GDP ratio is significantly higher in group A, where it stands at 75 percent, on average. In both groups B and C it is a little over 50 percent, which is the conventional wisdom threshold for a risk of debt crisis (see Cohen, 2003). While the debt-to-GDP ratio is a good predictor for being in A rather than in B or C, the debt-to-export ratio is instead a discriminating factor for being in B or C: it stands at 200 percent for group B (again, 200 percent is the conventional wisdom number); it stands well below on average for group C, at about 150 percent.

There are a few exceptions to this broad pattern. In group A we find Russia, which despite good macroeconomic data had to resort to an EFF in the face of its inability to raise foreign funds (as reflected by the huge spread paid on its debt). A similar story comes from Colombia, a country where domestic policy uncertainties were the critical problem, more than any macroeconomic imbalance. In group B, there are a few exceptions to the rule that debt-to-GDP is high, but these countries, such as Brazil or India, often have a high debt-to-export ratio (well above the 200 percent threshold); these are relatively closed countries for which both indicators are needed to assess the overall solvency of the country. The only exception in group B is Uruguay, where both ratios are relatively low and which appears to be a prima facie case of contagion from risky neighbors. In group C, Nigeria is a mirror image of Brazil or India: high debt to GDP but low debt to exports, which is easily explained by the outward orientation determined by oil exports.

An additional statistic shows the share of public external debt in total external debt for each of the groups. Public debt represents, respectively, 90 percent, 80 percent and 70 percent of total external debt in groups A, B, and C.

The key to our story is the spread paid on the debt. All countries in groups A and B paid high spreads well before (at least two years before) the crisis occurred. At the other extreme, all countries in group C were paying low spreads even one month before the crisis exploded. Their crises were basically unpredictable, or if predicted, not expected to last very long.

We can summarize these findings so far as follows: Major crises (types A and B) are old-style crises: high debt (in the sense of either debt/GDP above 50 percent or debt/exports above 200 percent) and mostly public. They are predictable at least two years ahead of time.

Let us now shed some light on the nature of the debt dynamics. The self-fulfilling story is one in which a high spread causes high debt rather than the other way around. Although this phenomenon is theoretically plausible, it is not easy to show empirically that it is indeed convincing. In order to shed some light on this debate, we have decomposed the debt dynamics into the following identity:

Increase of the Debt-to-GDP ratio =

real interest rate * Debt-to-GDP ratio

- Growth rate of the economy * Debt-to-GDP ratio

- Primary Surplus/GDP

The real interest rate is the nominal rate (risk-free rate + spread) adjusted for the deviation of the exchange rate from PPP. The dynamics are computed up to the year of the debt crisis itself. We present this decomposition in Table 3 by dividing each of the three terms of the right-hand side by the sum of their absolute values (the sum of absolute value then adds to one).

Table 3.

Decomposition of Debt Dynamics 1/

article image

Each item expressed as a fraction of the sum of the Absolute value.

Source: Authors’ calculations based on International Financial Statistics, IMF; Global Development Finance, World Bank.

The first term can be interpreted as a confidence premium, the second term as a measure of the underlying fundamentals, and the third term as a measure of the policy choices. We see that on average, the growth component (second column) is the critical factor behind the dynamics of debt. The confidence premium factor (first column) is the second important item, while the deficit itself appears to play the least important role. Some countries are heavily burdened by the confidence crisis term: Brazil, Ecuador, Turkey, Russia are all instances where it almost entirely cancels the growth factor.

This decomposition suggests two policy implications. First, given the fact that bad “fundamentals” are a major part of the story, we conclude that debt write-off may be needed. Second, the role of the confidence term suggests that efficient measures (taken ex ante and ex post) could alleviate the importance of that term.

IV. Confidence Crisis and Crisis of the Fundamentals: A Theoretical Benchmark

Self-fulfilling debt crises are a phenomenon whose theoretical rationale has been pointed out by Calvo (1988), Cole and Kehoe (1996, 2000) and others. The intuition is quite simple: perception of high risk increases the spread, which in turn raises the debt service burden, which in turn provokes the crisis. Beliefs are self-fulfilling because the fundamentals are themselves partly endogenous to the beliefs. If default reduces the amount that a country pays to its creditors below what it would normally pay (say down to nothing in the case of outright default) then lenders’ perceptions do change how much a country will eventually pay. This is less likely in the case of corporate debt if default simply amounts, say, to changing the management of the firm. As we now demonstrate in a simple model, by avoiding outright default and, say, making debt repayment contingent on a country’s income, orderly debt workouts have the potential to avoid self-fulfilling debt crises. We view this as a major argument in favour of orderly workouts ex post.

The model also touches upon another aspect of self-fulfilling crises. Cole and Kehoe (2000) show that debtors that are in the midst of a confidence crisis may want to take drastic corrective action in order to escape from it. We draw upon this result to demonstrate how an ex ante commitment device can reduce the scope of confidence crises. If a country is willing to commit itself not to borrow at high spreads, then under certain circumstances that we explore this can reduce, indeed sometimes eliminate, the risk of confidence crises.

A. A Model of Sovereign Debt

Consider a two-period model with time t ∈ {1,2}. At the beginning of period 1, the country inherits a stock of financial obligation D1 From the beginning to the end of the period, the debt grows from D1 to V1 due to interest rate dynamics. At the end of period 1, the country is endowed with Q1 units of output and then decides on making a repayment R1 to its creditor. At the beginning of time t=2 the country then inherits a new stock of debt, D2, which is itself turned into future debt obligation, V2, at the end of the period. At the end of period 2, a new endowment of resources and the amount of repayment is made. In period 2, we assume that output, Q˜, is stochastic and can take two values:

Q˜=Q+ with probability p.

Q˜=Q with probability (1-p).

and we call R2+ and R2 the repayments that will be made in period 2 in each of the two states of nature.

Let us finally write debt dynamics as follows:


in which x is the interest rate at which the debt is rolled over. Finally, call r the riskless rate of interest. Assuming here competitive risk-neutral lenders, we have:


In order to determine how R2+ and R2 are determined, let us now introduce the risk of default.

B. Default Risk

Assume that the country can always decide to renege on its debt and then pay a penalty which amounts to a fraction, λ, of its income. The analysis will now depend on what happens to the creditors in case of default. We shall examine two cases. In the first instance, we assume that a default involves an outright loss of λ Q that no one can capture, so that default creates a panic. In the second instance, we shall assume that default is efficiently managed by a combination of contractual and institutional measures, so that the debt due is scaled down appropriately in an orderly workout.

Case 1: Outright Default

Default will happen in the bad state if
Assume also, for the time being, that
(otherwise the country would default in all cases and would then find no lenders at time 1). In that case, we can write (1) as:
so that default (in the bad state of period 2) occurs if:

Case 2: Orderly Workout

Assume that default is not just a deadweight loss, but instead the trigger for renegotiation. In the context of our model, let us then assume that large debt can be written down in case of a bad shock. Again call x the interest due in period 2. Restructuring the debt means that the repayments will be scaled down to what the country can afford, namely

R2=Min[λQ˜,D2(1+x)]. In that case, the competitive equilibrium for the banks imposes the following no-profit condition:

C. Equilibrium Strategies

For simplicity, we assume that the debtor attempts to maximize: u(C1)+βE1C˜2, in which u( ) is a concave function and is C˜2 consumption in period 2, a stochastic variable that depends on the state of nature (the simplicity of choosing a linear function in period 2 is aimed at mimicking long-run dynamics, see, e.g., Fischer 1980).

Whichever repayment profile will be implemented, we then see that the first period problem, here, will always appear to be equivalent to maximizing:


in which E1 Y2 = pQ+ + (1 − p)(1 − k)Q, in which k = λ in the case of outright default in the bad state of nature in period 2, and k=0 otherwise. This is formally equivalent to:

Maximize u(Q1R1)+β(1+r)R1(2)

We then see, given the linearity of period 2 utility, that the problem such as it is stated here always has a unique solution:


The country seeks to guarantee a certain level of consumption in period 1 and bears the consequences, whatever they may be, in period 2. Since we know that Q1 = C1 + R1, this sets the level R1 that the country is willing to repay (perhaps borrow if R1 is negative) and a corresponding level D2* of debt to be borrowed at the end of period 1. For a given value of V1, this translates one for one into higher values of D2*. In what follows we assume D2*>0.

D. The Potential for Multiple Equilibria

Case 1: outright default

If the amount of debt D2* borrowed by the country is such that:

there is a risk of multiple equilibria, in the case where renegotiations take place in period 2. At the riskless rate, the country is safe in both states of nature; at the risk-adjusted rate, it will default in the bad case, justifying the risk premium. All then depends on how lenders expect the country to behave in case of a bad shock.

Case 2: orderly workouts

We can now readily see that self-fulfilling crises would be impossible in the second case in which debt renegotiations are possible ex post. As is straightforwardly apparent from equation (1’), one sees that:

In other words, if the country plans to default (here not to pay the debt in full) at the risk-adjusted rate, it would also default at the riskless rate. Contrary to the previous case, here a country cannot be drawn into a self-fulfilling debt crisis by worried creditors. This can be summarized as follows:

Proposition: An ex post efficient debt resolution mechanism destroys the risk of self-fulfilling debt crisis.

The intuition behind this proposition and its (lack of) generality is straightforward. A self-fulfilling debt crisis originates from the fact that the fundamentals out of which the debt is repaid may be endogenously shifted in case of outright default. By raising the payment falling due, the debt contract may shut the access to a stream of income that does occur in the bad state of nature. When an efficient debt workout is implemented instead, the fundamentals {λQ+, λQ}are unaffected by the debt contract. One sees that any first period action that would have a bearing on second period output would potentially turn around our results. With debt overhang, even ex post rescheduling could deter investment, although as shown in Cohen (1993) this could also be turned around by an appropriate rescheduling strategy (for an analogue in the area of venture capital, see Chamon, 2002).

E. Ex Ante Devices to Avoid Self-Fulfilling Debt Crises

Assume however that there is no way to ensure ex ante that there will exist ex post an efficient mechanism. In the zone of multiple equilibria, the country may want to restrain its debt below the level that is suggested by the myopic optimization of (2’) In order to see why, call R¯1 the level of repayment for which:


which is the level of repayment above which the country can avoid the self-fulfilling crisis. If it happens that:


then the country will decide to repay more than it would be rational to do otherwise in order to avoid the risk of multiple equilibrium. Rewriting the condition, this will happen if:


This is the more likely if (1-p) (the risk of a bad shock) is large. For small risk, the country will never depart from Program (1). Otherwise the country will accept to “over-repay” its debt in order to avoid the risk of a big interest hike.

Lender of first resort

Note that the country would want to implement R¯1 only in order to get the riskless rate; if it is denied the riskless rate, then the self-fulfilling trap remains. If lenders set the price first before the country decides on how much debt it takes, the incentive to behave prudently disappears. This is where the IMF can step in. By giving the country enough liquidity to settle the terms of new borrowing, giving the market enough time to gather the relevant information on the country’s policy (on which we assume the IMF has superior knowledge), the Fund makes it possible for the ex ante trap to disappear. In the more general setting of Cole and Kehoe, it takes T periods for the country to implement a tough policy that allows the country to escape the bad equilibrium—this would be the time frame where the commitment is needed.

We refer to this situation as one where a lender of first resort is needed: the country still has access to the financial market, but at a price that does not incorporate its willingness to take appropriate action to restore confidence.

Of course, this should not keep the lenders inactive. They have to see through the program to make sure that it is indeed sufficient to bring the country into the safe zone.

Lender of last resort

It may however be the case that the country could actually default on its debt in period 1. This will happen if D2 (1 + x) ≥ λQ+. At the risk-adjusted interest rate, no new lender will appear in period 1. In the self-fulfilling debt crisis regime, this may happen D2(1 + r) ≥ λQ, that is, despite the fact that the country might be entirely solvent in the safe equilibrium. Mathematically, this could happen if QpQ+. In that case, failure to coordinate on R¯1 may have devastating implications. The country may want to implement R¯1, but if it is not trusted, it has to default immediately.

We can refer to this situation as the more traditional one where a lender of last resort is needed. The country has lost its access to the financial market and must rely on a lender that trusts it in order to return to solvency.

We see, however, that there are actually two LLR cases. The first case is one in which the country wants to take action to restore confidence and needs to buy time. The LLR that is needed here is a contingent LLR, which intervenes under the condition that the country implements the program. But this is valid only for certain values of the debt stock. Beyond that level, we find a case that is more conventional. This is the large bailout case, where no action from the country can restore confidence. What is needed to restore confidence is a big window at the riskless rate. This is the pure case of a confidence crisis with no exit unless a confident lender steps in.

F. Conclusion of the Model

We then see that the debt of a country can take four critical values:

One value, call it Da, above which the country has to pay a risk premium which is justified by an intrinsic risk of default in period 2.

One value, call it Db < Da, which is such that within the range [Db, Da], multiple equilibria can occur.

One value Dd such that Db < Dd < Da, which is such that within [Db, Dd] the country will want to shift to the safe zone, through tougher repayment, but would not default if denied the riskless rate. This is a zone for a lender of first resort.

One value Dc such that Dd < Dc < Da, which is such that within [Db, Dc] the country will want to shift to the safe zone, through tougher repayment, but would default if denied the riskless rate. This is a zone for a contingent lender of last resort.

Finally the range [Dc, Da] is the large bailout case, when only fresh money at the riskless rate can avoid default.

V. Our Policy Implications: Collective Action Clauses and a Lender of First Resort

A. Collective Action Clauses

Our theoretical model shows how important ex post efficiency can be, not only from an ex post perspective, but also ex ante, if one wants to avoid a self-fulfilling crisis. Even in the absence of an SDRM, ex post efficient outcomes remain important. We therefore support a new institutional framework that would operate in the case of a “solvency crisis,” when debt restructuring would be necessary.

There would first be new contractual arrangements: CACs. Debt contracts are incomplete, and, as we have seen, the consequent problems are more severe for international lending. The institutional and legal structure for such lending must respond to this fundamental problem. The broad phrase “collective action clauses” has been extended to cover a wide range of proposals. As specified in Cornelli and others (1995) and supplemented recently by Taylor (2002), these would bring into bond contracts (and indeed to bank lending instruments) a range of clauses that would, we believe, promote orderly workouts of international debt, rather than the chaotic sequel to default that we observe now (in the Argentine case, for example). These would include initiation and engagement clauses detailing how negotiations would proceed; a clause permitting changes by a qualified majority of the terms of the debt, including amounts and dates payable; a sharing clause that would require pro rata distribution to all bondholders of any payment made to any one of them; and a nonacceleration clause to avoid having one missed payment trigger an immediate full repayment obligation. An additional contractual innovation that would facilitate restructurings would be to utilize the trust deed form for bonds (common under U.K. law but not in New York—see Buchheit and Gulati, 2002). Here the trustee acts for all holders of a given security and centralizes enforcement of any decisions (in particular, the trustee shares among the bondholders the proceeds of any settlement).

In addition to the existing Paris Club and London Club mechanisms, which deal with debt to governments and to banks, respectively, there would be a permanent (but “light”) bondholders committee—the “New York Club,” say. It would oversee bondholders’ negotiations with the debtor. There might also be a new mediation agency—again, an administratively “light” structure that would coordinate the Paris Club, London Club, and New York Club, primarily by ensuring the timely exchange of information and comparison of assumptions. It would verify claims and oversee bondholder voting. It might take on other roles, such as endorsing (or not) a proposed standstill. The proposal of the Institute for International Finance to bring all creditors into a single negotiating committee overriding the Paris and London clubs seems unnecessary, as the existing structures work efficiently.

The markets now seem to be accepting CACs more or less universally, since the Mexican issue of February 2003. But if the process reverses, there is a relatively simple, feasible way of implementing these proposals. The mandates of the American Securities and Exchange Commission (SEC) and the British Financial Services Authority (FSA) include duties to protect investors and to maintain orderly markets. That is sufficient justification and authority, without new legislation, for them to intervene. It is clear from the case of Argentina that those markets were and are disorderly and that investors have not been adequately protected against the eventuality of default by having adequate post-default procedures in place.

Thus we propose that the American, British, and other major financial center regulatory authorities stipulate that bonds issued or traded in their markets must include CACs and other workout-friendly clauses. The IMF could organize and indeed help to fund a voluntary exchange program (with enhancements) for outstanding stocks of securities without such clauses. And the Fund should make access to the SFR (indeed, any Fund program) open only to countries that use CACs.

B. Lender of First Resort

Let us now follow the insights of the theoretical model that we have presented to highlight its policy implications for ex ante policies. As we argued in the theoretical model, unless one can be made 100 percent sure that ex post resolution of debt crises will be efficient, there remains room for improving the nature of ex ante equilibria.

In the model of self-fulfilling debt crises of the kind analyzed above, a debtor that is the victim of a confidence shock may want to get out of the danger zone by taking stringent actions. A country that could gain credibility by reacting to such fears needs to buy time to get out of the danger zone. Let us proceed here to see how this could be achieved in practice.

A commitment device for the indebted countries

Assume that a country initially borrows at low spreads: think of a well-performing emerging market country, and assume that a new shock suddenly lowers the market’s assessment of its creditworthiness. The country still has market access, but at terms that may not take account of its willingness to take appropriate action to deal with the shock. If the country accepts higher spreads, it “gambles for resurrection” by taking the chance that things will eventually settle down, or simply buys time in order to make internal adjustments. The problem with this option is that if spreads do not return to sustainable levels, the debt will meanwhile spiral upwards, making it more difficult ex post to get the country to act decisively. For a country that is committed to (say) a 400 basis-point spread, the IMF should work with the country on an analysis of the cause of the problem and of the remedies that could resolve it. A programme would then be designed, which, if agreed upon by the country, could grant access to IMF money if needed.

Nothing should be automatic in this process. Countries signal ex ante their willingness to avoid the snowball effect of rising spreads and rising debts and seek to avoid it at an early stage. Furthermore, IMF resources, to the extent that they are cheaper, may halt the snowball effect. IMF support remains conditional on taking appropriate measures, so that it is not a free lunch. But the Fund programme and its implementation will give the country credibility that could soon restore market access at or below the “trigger” rate. This is, then, an appropriate catalytic role for IMF lending. Note that the absence of automaticity avoids the danger of overlending as the country’s market rate approaches the trigger—the markets cannot simply assume that this is a ceiling at which an automatic bailout would kick in.

Although there is no automaticity in the mechanism, there is also no requirement that the country or the IMF need distinguish ex ante between a potential crisis of fundamentals and a potential self-fulfilling confidence crisis. The country acts and the Fund responds when spreads reach the trigger level, regardless of the fundamentals. Only subsequently, in the formulation of the IMF programme, does the Fund assess the fundamentals.

Note also that there is no question of the country having to “tie its hands” ex ante and then confronting a time-consistency problem. The point is that in the circumstances appropriate for the ILFR mechanism, the country will want to utilize it and to take on the appropriate Fund programme. This is incentive-compatible.

One may fear that the informational content of spreads will be reduced as they become a policy variable. It is true that lenders, being aware that countries will take actions against rising spreads, will change their pricing policy. If, as a result, spreads become lower, this is in itself a good thing as it reduces the snowball effect. But it is very unlikely that they could fail to detect a country that becomes insolvent. Indeed, actions to correct imbalances are voluntary, not automatic. Lenders must then keep track of a debtor’s solvency. But the policy may achieve the objective of making self-fulfilling spread crises if not impossible, at least less likely.

The trigger rate is similar to a guaranteed price or rate of interest on the country’s debt. Since debt instruments are priced as present values of the expected future stream of payments (allowing for possible default), our proposed facility would result in some distortion of debt prices at any date, regardless of whether on that date the risk premium hits the trigger level or is below it, as long as agents expect that there are possible paths leading to states in which the country risk rate would exceed the trigger level. That is, the “contract” with the IMF would affect the rate of discount on future debt payments and hence affect the conditional expected present values as of any date (we owe this point to Enrique Mendoza). But this distortion seems a small price to pay for the advantages of the scheme.

It might also be feared that the amount of lending necessary to reestablish confidence could be very large. We do not think this likely, even in the case of a large country, simply because the mechanism would kick in before the debt reached unsustainable levels—indeed, while the markets were still offering access, although at a potentially unsustainable, destabilizing spread. There could be a problem, however, if an exogenous shock (e.g., a rise in U.S. interest rates) hit all debtors simultaneously. Empirically, spreads are an increasing function of the overall level of rates, and so we might see many countries hitting the trigger spread simultaneously. The problem could be avoided by fixing the trigger as a margin over the average Emerging Market Bond Index (EMBI) spread. This would not satisfactorily deal with all self-fulfilling crises—it might well be the case that if the average EMBI spread were to rise sufficiently, many countries could still face self-fulfilling confidence crises at the same time. As a practical matter, however, we believe that there is a merit to pragmatically boiling down our proposal to cases where countries are hit by an idiosyncratic shock. We then suggest to act preventively in cases where spreads rise over the EMBI index. In this case, a number like 150 basis points over the EMBI average could be the trigger point.

In our view this mechanism could replace the now defunct Contingent Credit Line (CCL) facility, although it is quite different both in intent and design. The CCL was created to help “first-class policy” countries to deal with contagion crises. As Stanley Fischer emphasized, such countries do not need to be “perfect.” They need to obey international standards such as the Basel Committee’s Core Principles for Banking Supervision, the code on Transparency in Monetary and Financial Policies, etc. The country must have enjoyed “constructive relations with its private creditors and be taking appropriate measures to limit its external vulnerability,” the latter including exchange rate viability and the absence of arrears on sovereign debt. The idea was to create a first-class policy group to discriminate between the implementation of good and bad policies and eradicate the moral hazard risk. The reason why no country decided to use the CCL was the fear of sending a wrong signal to the market, despite the quasi-pre-qualification clauses that were attached to it. Our mechanism instead is one that only relies on market signals (spreads), so that it would not run into such a risk. The reason why we attach so much importance to spreads is that they both reveal a problem and contribute to creating it.


Table A1.

Taxonomy of Debt Crises A

(Crisis A = one Extended Fund Facility (EFF))

article image
Source: Cohen and Portes (2003).

As share of long-term debt at three-year average.

Table A2.

Taxonomy of Debt Crises B

Crisis B = two to three consecutive Standby Arragements (SBA)

article image
Source: Cohen and Portes (2003).

As share of long-term debt at three-year average.

Brazil, Turkey, and Uruguay also benefited from SRF combined with SBA (two for Brazil, one for Turkey and Uruguay).

Three SBA in the 1900s. Data in (t) are averages for the period 1990-2000.

Table A3.

Taxonomy of Debt Crises C

(Crisis C= one Standby Arragement (SBA))

article image
Source: Cohen and Portes (2003).

As share of long-term debt at three-year average.

Table A4.

Pre-Crisis Spreads

article image
Source: Cohen and Portes (2003).


  • Bank for International Settlements, 1996, The Resolution of Sovereign Liquidity Crises (Basel).

  • Buchheit, Lee, and Gaurang Gulati, 2002, “Sovereign Bonds and the Collective Will,” Georgetown University Law Center Working Paper 34 (Washington: Georgetown University Law Center). Available via the Internet:

    • Search Google Scholar
    • Export Citation
  • Calvo, G., 1988, “Servicing the Public Debt: The Role of Expectations,” American Economic Review, Vol. 78, pp. 64761.

  • Chamon, Marcos, 2004, “Are Debt Crises Self-Fulfilling?IMF Working Paper 04/99 (Washington: International Monetary Fund).

  • Cohen, Daniel, 1991, Private Lending to Sovereign States (Cambridge, Massachusetts: MIT Press).

  • Cohen, Daniel, 1993, “Low Investment and Large LDC Debt in the 1980s,” American Economic Review, Vol. 83, No. 3, pp. 43749.

  • Cohen, Daniel, 2001, “The HIPC Initiative: True and False Promises,” International Finance, Vol. 4, No. 3, pp. 36380.

  • Cohen, Daniel, 2003, “How to Avoid Self-Fulfilling Debt Crises?” (unpublished; Paris: École Normale Supérieure).

  • Cohen, Daniel, and Richard Portes, 2003, Crise Souveraine: Entre Prévention et Résolution, Conseil d’Analyse Economique, Rapport No. 43 (Paris: La Documentation Française).

    • Search Google Scholar
    • Export Citation
  • Cole, Harold, and Timothy Kehoe, 1996, “A Self-Fulfilling Model of Mexico’s 1994–1995 Debt Crisis,” Journal of International Economics, Vol. 41, pp. 30930.

    • Search Google Scholar
    • Export Citation
  • Cole, Harold, and Timothy Kehoe, 2000, “Self-fulfilling Debt Crises,” Review of Economic Studies, Vol. 67, pp. 91116.

  • Cornelli, Francesca, Barry Eichengreen, Leonardo Felli, Julian R Franks, Christopher Greenwood, Hugh Mercer, Richard Portes, and Giovanni Vitale, 1995, Crisis? What Crisis? Orderly Workouts for Sovereign Debtors (London: Centre for Economic Policy Research). Available via the Internet:

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, and Richard Portes, 1986, “Debt and Default in the 1930s: Causes and Consequences,” European Economic Review, Vol. 30, pp. 599640.

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, and Richard Portes, 1999, “Reforming the International Financial System,” Economic Journal, Vol. 109, No. 549, pp. 55776.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2003, “Report of the Managing Director to the International Monetary and Financial Committee on a Statutory Debt Restructuring Mechanism” (Washington). Available via the Internet:

    • Search Google Scholar
    • Export Citation
  • Jeanne, O., and C. Wyplosz, 2001, “The International Lender of Last Resort: How Large is Large Enough?NBER Working Paper No. 8381 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Krueger, Anne, 2001, “International Financial Architecture for 2002: A New Approach to Sovereign Debt Restructuring,” address given at the National Economists’ Club Annual Members’ Dinner, Washington, November 26.

    • Search Google Scholar
    • Export Citation
  • Krueger, Anne, 2002, “New Approaches to Sovereign Debt Restructuring,” address given during conference on “Sovereign Debt Workouts: Hopes and Hazards,” Washington, April 1.

    • Search Google Scholar
    • Export Citation
  • Lerrick, Adam, and Allan Meltzer, 2001, “Blueprint for an International Lender of Last Resort,” paper presented at Carnegie-Rochester Conference Series on Public Policy, Rochester, April. Available via the Internet:

    • Search Google Scholar
    • Export Citation
  • Meltzer Commission, 2000. “Report of the International Financial Institution Advisory Commission” (Washington). Available via the Internet:

    • Search Google Scholar
    • Export Citation
  • Sachs, Jeffrey, 1995, “Do We Need an International Lender of Last Resort?Graham Lecture, Princeton University, Princeton, New Jersey.

    • Search Google Scholar
    • Export Citation
  • Taylor, John, 2002, “Sovereign Debt Restructuring: A U.S. Perspective,” address given at the Institute for International Economics Conference, Washington, April 2. Available via the Internet:

    • Search Google Scholar
    • Export Citation
  • Williamson, John, 2001, “The Role of the IMF: A Guide to the Reports,” International Economics Policy Brief 00-5 (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation
  • Williamson, John, 2002, “Is Brazil Next?International Economics Policy Brief 02-7 (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation

Cohen is professor of economics, École Normale Supérieure (PSE), director of CEPREMAP and Research Fellow, CEPR; Richard Portes is professor of economics at London Business School and École des Hautes Études en Sciences Sociales and President of CEPR. This paper was written while Cohen was a visiting scholar at the IMF and Portes was visiting Columbia Business School. It draws on a report prepared for the Conseil d’Analyse Economique (Cohen and Portes, 2003). The authors are grateful for comments by Laurence Bloch, Eduardo Borensztein, Jean Pisani-Ferry, and participants at a CAE meeting, the IMF-Bank of Spain conference (Madrid, June 2004), the Reinventing Bretton Woods—Austrian National Bank conference (Vienna, June 2004), and the Reinventing Bretton Woods—World Economic Forum conference (Rome, July 2004).