Crisis Resolution: Next Steps
  • 1 0000000404811396 Monetary Fund
  • | 2 0000000404811396 Monetary Fund
  • | 3 0000000404811396 Monetary Fund

Contributor Notes

At the April 2003 meeting of the International Monetary and Financial Committees, it was decided to further encourage the contractual approach to smoothing the process of sovereign debt restructuring by encouraging the more widespread use of collective action clauses (CACs) in international bonds. This decision was shaped partly by Mexico's successful launch of a bond subject to New York law but featuring CACs, and by subsequent issues with similar provisions from other emerging market countries. This paper reviews the developments leading up to that event, its implications, and prospects for the future. It asks whether we can expect to see additional issuance by emerging markets of bonds featuring CACs, whether such a trend would in fact help to make the world a safer financial place, and what additional steps might be taken to further enhance modalities for crisis resolution.


At the April 2003 meeting of the International Monetary and Financial Committees, it was decided to further encourage the contractual approach to smoothing the process of sovereign debt restructuring by encouraging the more widespread use of collective action clauses (CACs) in international bonds. This decision was shaped partly by Mexico's successful launch of a bond subject to New York law but featuring CACs, and by subsequent issues with similar provisions from other emerging market countries. This paper reviews the developments leading up to that event, its implications, and prospects for the future. It asks whether we can expect to see additional issuance by emerging markets of bonds featuring CACs, whether such a trend would in fact help to make the world a safer financial place, and what additional steps might be taken to further enhance modalities for crisis resolution.

I. Introduction

The debate over how to manage and resolve crises in emerging markets, under way now for the better part of a decade, reached a climax at the Spring Meetings of the International Monetary Fund and the World Bank held in Washington in the spring of 2003. Agreement was reached to push ahead with the contractual approach to smoothing the process of sovereign debt restructuring by promoting the further introduction of collective action clauses into bond contracts while continuing to study and develop the statutory approach, in particular the IMF’s Sovereign Debt Restructuring Mechanism (SDRM).2 These decisions were shaped by Mexico’s successful launch the preceding March of a $1 billion global bond, subject to New York law but featuring collective action clauses, at spreads that were if anything slightly tighter than those on its previously-issued New York law bonds.3 Mexico then followed in April with two additional issues also including collective action clauses, and Brazil, South Africa, and the Republic of Korea all issued bonds in New York with similar provisions. These events put paid to the view that investors would not accept bonds that included collective action clauses and that the governments of emerging markets would be unwilling to issue them for fear of higher borrowing costs. They galvanized the debate by demonstrating the feasibility of contractual innovation.

It is tempting for officials and analysts to congratulate themselves on a job well done and turn to other topics. But the process of improving how we go about sovereign debt restructuring, much less the larger task of making the world a safer financial place, is still incomplete. It remains to be seen how many other emerging markets will follow Mexico, Brazil, South Africa, and the Republic of Korea’s examples. And, while collective action clauses provide mechanisms for coordinating the creditors holding an individual bond issue, they do not coordinate the creditors holding different issues. Recall that Argentina had more than 80 separate sovereign bonds in the market at the time of its December 2001 default. Thus, it cannot be taken for granted that the addition of these provisions to individual loan contracts will significantly facilitate creditor coordination and smooth debtor-creditor negotiations.

Above all, there remains the question of how much can be expected of these improvements in procedures for sovereign debt restructuring. Contractual clauses specifying how restructuring is initiated, how the creditors are represented, when legal action can be initiated, and under what circumstances a change in the financial terms of a bond agreed to by a qualified majority of creditors will be binding on dissidents constitute only limited changes to the status quo. Even those limited changes would have applied only to a subset of recent crises. Then there is the critique that the official community directs too much attention to building better morgues. It should devote more effort in this view to preventing crises and to promoting capital transfer from rich to poor economies than to cleaning up after crises when they occur.

In this paper we reassess the efficacy of this strategy for addressing problems of crisis resolution. We focus on two questions, bringing to bear both theory and evidence. First, are speculative credits likely to follow investment grade countries in adding CACs to their loan instruments? While our analysis of sources of resistance to contractual innovation creates reasons for hoping that Mexico’s path breaking issue may have broken an important logjam, both theory and evidence highlight the moral hazard associated with restructuring-friendly provisions for countries with relatively poor credit. They suggest that CACs may raise the cost of borrowing for countries with poor credit ratings, especially in periods when sentiment toward emerging markets is relatively unfavorable, leaving them slow to embrace these provisions.

In addition, we ask how difficult it may be for countries whose existing bond issues feature unanimous action clauses (UACs) to effect the transition to CACs. The concern is that the holders of bonds that require unanimous consent to changes in financial terms may be able to hold out for favorable restructuring terms at the expense of investors holding bonds with collective action clauses. Again, we find this to be a problem mainly for issuers with relatively poor credit. That the pioneers of New York law issues with CACs are mainly countries with good credit (Mexico, South Africa, and the Republic of Korea) is consistent with this view. On a more optimistic note, we also find that reversion from CACs to UACs may be unlikely—that if bonds with CACs reach a critical mass, issuing new bonds with UACs may become less attractive.

Second, we ask whether CACs are sufficient to solve the problem of cross-issue coordination among creditors, the so-called aggregation problem. A multiplicity of bond issues has both benefits and costs. The benefits accrue through being able to establish yield curves in major currencies and to avoid humps in amortization. However, there also are costs of multiplicity, since collective action clauses in individual bonds do not solve coordination problems across issues. The market appears to be most concerned about aggregation in the case of poor credits with limited market access. However, because investors may not anticipate the relapse of good credits into repayment difficulties, cross issue coordination may become a problem for other issuers as well. We therefore conclude that there is a need to encourage the development of super-collective action clauses, bondholders committees, and a code of creditor conduct.

How much difference will collective action clauses make the efficiency of outcomes and for the stability of international financial markets if they become widespread? Our theoretical analysis suggests that by pricing moral hazard, collective action clauses will encourage market discipline. At the same time, by facilitating creditor coordination, collective action clauses will reduce the deadweight disruptive costs of delay. However, the case for collective action clauses is strongest if they are viewed as one of several interdependent changes in the international financial system, which together promise to make the world a safer financial place, but none of which is feasible in the absence of the others. For example, collective action clauses could reduce the likelihood that the IMF and its principal shareholders will feel compelled to extend financial assistance to countries whose debts are already borderline unsustainable, since the consequent restructuring would not be so disruptive in the presence of these contractual provisions. Absent the expectation of IMF bailouts, borrowers and lenders are likely to exercise more discipline, reducing crisis risk and enhancing systemic stability.

The point is that the international financial architecture is made of up a set of interlocking parts. It is hard to change one without also changing the others. Thus, a concerted effort to alter the provisions of loan agreements may hasten progress on other, complementary changes, which will then work together to make the world a safer financial place.

But in the rush to get everything right, it would be a mistake to think that the job of fine tuning contractual provisions is complete. We have yet to see whether a significant number of speculative credits follow investment-grade countries like Mexico and the Republic of Korea in adopting these provisions. Even if they do, countries with low credit ratings may be tempted to require very high qualified majorities and retain other provisions that stymie collective action and encourage holdout litigation. And, as yet, there is no consensus on the desirability of super-collective-action clauses or the design of informal substitutes such as a standing committee of bondholders and a code of creditor conduct. It is on these issues that the next steps in strengthening crisis resolution should focus.

The paper is organized as follows. After reviewing the development of the debate, we present a theoretical framework for analyzing the creditors’ collective action problem. We use it to analyze the issues that remain to be addressed by the contractual approach: the incentives for adoption (in particular, whether emerging markets with sub-investment-grade ratings will be discouraged from adopting CACs by the prospect of higher borrowing costs), the challenge posed by the inherited stock of bonds (the transition problem), the difficulty of coordinating creditors across bond issues (the aggregation problem), and the risk that the entire initiative may be undermined by asset substitution and market migration. We then present new evidence on many of these questions. Finally, we examine various hypotheses why there has not been faster progress in getting these new contractual provisions into the market. In concluding, we return to the question of how much can be accomplished through these improvements to crisis-resolution processes and to how they fit into the larger architecture debate.

II. The Development of the Debate

It is sometimes said that the debate over crisis resolution was initiated by the Mexican crisis, which highlighted the existence of collective-action problems in decentralized securities markets. In fact, the debate goes back further, to Raffer’s (1990) proposal for an international insolvency procedure designed along the lines of Chapter 9 of the U.S. bankruptcy code, to Oeschsli’s (1981) proposal for empowering the IMF to carry out in the sovereign context many of the responsibilities of the bankruptcy court under Chapter 11 of the U.S. code, and to Ohlin’s (1976) argument for the creation of an institution of “honourable bankruptcy.” These authors were all responding in some sense to the difficulty of restructuring defaulted debts, to the impact this had on creditors and debtors alike, and to the uncomfortable implications for the IMF.

Still, the debate took a new turn with the Brady Plan, the resumption of lending to developing countries through the bond market, and the Mexican crisis, which demonstrated the downside of securitized finance. These events prompted Hurlock’s (1995) proposal for the U.S. and U.K. to close their courts to rogue creditors, Macmillan’s (1995) scheme for the introduction into bond covenants of sharing clauses and thresholds for litigation, and Eichengreen and Portes’ (1995) proposal for promoting the more widespread use of collective action clauses and creating a bondholders council to address the creditors’ collective action problem.

These authors differed in their motivations. Hurlock and Macmillan emphasized the threat of disruptive litigation by rogue creditors.4 Eichengreen and Portes, in contrast, were concerned with problems of creditor coordination more generally and argued the need for majority structuring provisions and a committee of bondholders to facilitate restructuring even in the absence of disruptive litigation.5 They emphasized the need for alternatives to large scale rescue operations, a la Mexico, which were unlikely to be either feasible as a response to future crises, given limited the resources of the international financial institutions, or desirable, given problems of moral hazard.

The first official contribution to this debate was the report of the G-10 Deputies (Group of Ten, 1996), written in response to the call at the Halifax Group of Seven Summit for a review of alternatives for the more orderly resolution of sovereign debt crises. The report concluded that encouragement should be given to the use of standstills by authorizing the IMF to lend to countries that suspend payments (that the Fund should be encouraged to “lend into sovereign arrears” when a crisis country was in desperate need of working capital and making a good-faith effort to negotiate with its creditors).6 It endorsed the more widespread use of contractual clauses providing for the collective representation of debt holders, for qualified majority voting on changes in financial terms, and for provisions requiring that amounts recovered from the debtor be shared among all issue holders.

This G-10 report defined the terms of subsequent discussions but in the short run provoked only an inconclusive debate. Then came the Asian crisis of 1997-98. That crisis pointed up the inadequacy of official resources relative to rapidly expanding international financial markets, again highlighting the need for other mechanisms for resolving crises. The report of the Group of Twenty Two working group on international liquidity crises (G-22, 1998) echoed the case for collective action clauses. The G-7 then placed the issue on its work program for reforming the international financial system with the goal of reaching a consensus by the Cologne Summit in 1999. That consensus (G-7 1999) endorsed the more widespread use of collective action clauses and for the first time urged G-7 governments to consider including them in their own debt instruments. In 1999 U.S. Treasury Secretary Summers endorsed their more widespread utilization, as did the Executive Board and the International Monetary and Financial Committee of the IMF in a series of reports and communiqués.

But translating these words into deeds proved to be difficult. Table 1 (reproduced from IMF 2003, p. 5 8) shows that the share of new issues by emerging markets that included the relevant contractual provisions remained stagnant, even falling. Spokesmen for the creditor community repeatedly warned that the more widespread adoption of collective action clauses would be perceived as an erosion of creditor rights (Institute of International Finance, 1996). Some worried that requiring their more widespread use would limit the demand for emerging market bonds and “generally inhibit market access for those emerging market countries implementing correct reform policies” (Rhodes et al. 1999, p. 2). More apocalyptically they warned of “a prohibitive increase in borrowing costs at a time when trillions of dollars are needed for infrastructure finance…” (Folkerts-Landau, 1999, p. 2). Officials of emerging-market countries similarly regarded the initiative with skepticism, reflecting fears that the new provisions would raise the cost of borrowing.

The resulting lack of progress, against the backdrop of continuing crises (Russia, Ukraine, Brazil, Ecuador, Pakistan, Turkey), led some to consider approaching the problem from the other end. To the extent that creditors and perhaps also debtors might be reluctant to accept the addition of restructuring-friendly provisions because they preferred to receive IMF assistance—which the Fund would feel compelled to extend so long as workouts remained inefficient and costly—the solution was to limit IMF lending and therefore to sharpen the incentive for market participants to pursue alternatives. Thus, a task force of the Council on Foreign Relations (1999) recommended that the Fund should adhere to its normal lending limits of 100 percent of quota in a year and 300 percent of quota over the life of a program, except in instances where the stability of the global financial system was threatened.7 The Bank of Canada and Bank of England (2001) similarly recommended clear presumptive limits on IMF lending. The Meltzer Commission (2000) proposed that the IMF should limit the kind of countries that qualified for assistance.

This approach essentially assumed a solution to the IMF’s time-consistency problem. It assumed that the Fund could credibly commit not to intervene on behalf of a country whose only alternative might be a disruptive, costly, and disorderly default. Others argued that the predominant direction of causality ran from first creating socially acceptable alternatives to IMF bailouts by adopting restructuring-friendly contractual provisions and from there to limiting IMF rescues.

So the debate stood in the summer of 2001, when Argentina’s crisis erupted. Argentina pointed up the dilemma created by the absence of an alternative to IMF assistance. The most revealing turn of events came in August 2001 when the Fund and its shareholder governments agreed to provide the country with an additional $8 billion of assistance. There were doubts at this late date that Argentina’s debt was sustainable, but there were also widespread fears, borne out in the event, that a default would be highly disruptive. The IMF earmarked $3 billion to be brought forward from later disbursements to support a voluntary, market-based operation to improve Argentina’s debt profile—in effect, for a restructuring designed to reduce the country’s immediate debt-servicing obligations. Frustratingly, however, no one could figure out how to use that $3 billion. Investors were reluctant to agree to a restructuring precisely because it was voluntary; they preferred to wait and see whether the multilateral would provide additional assistance. In the end, the official community saw no alternative but to lend, because doing nothing and thereby forcing the country into a messy and difficult restructuring risked endangering Argentina’s neighbors and a fragile international financial system. At the same time, officials feared that this action only put off necessary institutional and political reform. (Fischer, 2002, p. 37) summarized the lesson as follows. “Under present circumstances, when a country’s debt burden is unsustainable, the international community—operating through the IMF—faces the choice of lending to it, or forcing it into a potentially extremely costly restructuring, whose outcome is unknown.”

These were the events that brought forth Krueger’s proposal for a Sovereign Debt Restructuring Mechanism in November 2001. The SDRM initiative aligned the IMF with those calling for alternatives for dealing with sovereign debt crises. Faced with the possibility of a more radical solution, market participants, until recently unrelenting critics of collective action clauses, embraced them as, from their perspective, the lesser of evils. In April 2002, a special committee of the Institute of International Finance endorsed their broad-based use (IIF, 2002). It is hard to imagine that this organization, which for years had opposed the more widespread use of collective action clauses, would have done such a dramatic about face in the absence of Krueger’s initiative. In May, six creditor organizations (“the gang of six”) then issued a report acknowledging that CACs had utility for sovereign debt restructuring (Emerging Markets Creditors Association and others, 2002) and suggesting the form that model clauses might take.8

Canada, the members of the European Union, and Switzerland agreed to lead by example, but the most dramatic development on this front was Mexico’s issuance of a $1 billion global bond in New York, underwritten by J.P. Morgan and Goldman Sachs, that included a majority restructuring provision permitting financial terms to be altered with the approval of bondholders holding 75 percent of the principal. Mexico then issued two more bonds in April with collective action clauses, together amounting to $2.5 billion, and Brazil followed later that month with a $1 billion issue in New York that also included collective action clauses (though requiring an 85 percent qualified majority to change payment terms rather than the 75 percent that applies to the Mexican bonds and is typically used under U.K. law). South Africa and the Republic of Korea followed closely on their heels. We expect to see additional emerging-market bonds including collective action clauses issued in New York by the time this paper is published. Thus, we may eventually have a real time test of whether the more widespread use of collective action clauses will help to make the world a safer financial place.

III. Theory

We use a simple model of sovereign debt to analyze whether collective action clauses can raise welfare and to suggest observable differences in the pricing of loan contracts. Debt repayments are bounded by the willingness of the debtor government to pay to avoid the costs of default (following Eaton and Gersovitz, 1981). Our model is based on the infinite-horizon model of debt renegotiation by Kletzer and Wright, 2000 in which the debtor’s willingness to pay takes into account future credit market access and debt renegotiation.9 Willingness to pay can be expressed as the maximal equilibrium present value (in expectation) of current and future net repayments that are time consistent for the debtor, since this is the best that creditors can assure themselves (in present value) by restructuring debt repayments.

If collective action clauses reduce the power of holdout creditors and ease sovereign debt restructuring, then the probability of default can rise because the incentives for debtor discipline diminish. We incorporate this effect by allowing the debtor government to have information about its willingness to pay that is unknown to its creditors. We assume that the government is better informed than foreigners about the political feasibility of raising revenues from the private sector or of implementing contractionary macroeconomic policies to facilitate debt repayment. The government knows its capacity to repay and chooses how much to pay given the consequences of default and the extent of its informational advantage over creditors. In this characterization of moral hazard, foreign creditors only know a range for the government’s willingness to pay, given mutually observable fundamentals such as current output, the level of indebtedness, and the terms of trade. The wider is this range, the greater is the moral hazard. Foreign creditors also face risk regarding future observable fundamentals.

A. The Basic Model

We begin with the collective action problem for debt renegotiation with a single bond issue, following Kletzer, 2003. The outstanding debt, Dt, exceeds the government’s willingness to pay, Πt. Renegotiation in this framework is a simple game between creditors, given the maximum amount they can receive in the aggregate in present value, Πt. Because of the use of proportionate sharing and acceleration clauses, we impose the assumption that bondholders receive a pro rata proportion of any settlement, Πt when they agree to restructure an outstanding debt. However, bondholders who refuse to participate may be able to force full repayment of the bonds they hold before the remaining bonds are renegotiated. An essential element of a constrained optimal contracting model of sovereign debt (such as Kletzer and Wright, 2000) is that the debtor’s willingness to pay does not increase during delay. As a result, delays in agreement and repayment are costly for creditors as a group.

If a proportion of the bonds, x, is held out of the renegotiation of a bond issued with UACs, then the holders of the remaining bonds can offer full repayment, xDt, to the holdouts and restructure the rest of the debt, or they can delay the agreement in anticipation of the possibility of a better outcome. By paying off the holdouts, the holders of (1 -x)Dt of the debt receive, Πt, – xDt, in present value. If they refuse to pay the holdouts, then the most that they can receive is e–rΔt(1 – x) Πt, where Δt is the delay in negotiations if the bondholders do not cooperate and et is the discount factor for the creditors applied to settlements of postponed negotiations.10 By comparing the payoffs for each of these choices, the holdouts will be successful in demanding full repayment if


which implies that x satisfies


The maximal proportion of debt, St, that can be held out successfully is less than one if willingness to pay is less than the outstanding debt.

The return from holding out for any creditor under unanimous action clauses is at least as great as that from agreeing to renegotiate. A coordination problem arises because only the holders of no more of than a share St of the debt can demand full repayment and succeed in equilibrium. This game is similar to the simple war of attrition. When all bondholders are identical, it has an equilibrium in which each creditor plays a mixed strategy in each round of negotiations, choosing to renegotiate with a positive probability less than one that is the same for all creditors and dates.11 The probability of accepting a restructuring will vary across bondholders if they are heterogeneous. The probability that a restructuring is successful equals the probability that the share of bonds held out at any time is less than or equal to St. In equilibrium, this probability, P, is positive, though less than one. Allowing negotiations to take place continuously, P is the hazard rate that negotiations conclude. The aggregate expected return to all creditors equals


In this equilibrium, creditors compete to be members of the successful coalition of holdouts. The model implies that bonds issued with UACs are subject to renegotiation delays that are costly for creditors.

B. Collective Action Clauses

Bonds that include collective action clauses allow delays in renegotiation to be avoided if the size of the requisite qualified majority is sufficiently small so that holdouts cannot be successful. If a qualified majority of the holders of a bond with CACs vote to restructure, then all bondholders receive their pro rata share of the repayment, Πt. The settlement is imposed on the dissenting minority. If the required share of outstanding bonds that must be held by a qualified majority, m, is less than 1 – St then a group of bondholders can only hold up restructuring if their share of the debt exceeds St. In this case, they cannot gain by holding out because the remaining bondholders will not agree to pay them off, and any individual bondholder who completes the qualified majority realizes a higher return by joining than by holding out. Inequality (1) implies that the required qualified majority, m, needs to satisfy


or, equivalently,


to rule out delay. The middle term of the expression (4) is the largest qualified majority requirement that eliminates the ability of holdouts to be successful. This increases with the secondary market discount on the debt (which equals 1 – Dtt) and falls with the interest rate, r.

In the absence of moral hazard, the actual and reported willingness to pay of the debtor government are identical. In this case, the probability of default is identical under different governing laws, but the return to bonds issued with CACs is higher than to bonds with UACs because renegotiation delays for bonds issued with UACs are costly for creditors in the event of default.

In the presence of asymmetric information, however, bonds with collective action clauses are also subject to greater moral hazard and, hence, are subject to a higher probability of default.12

Moral hazard is introduced to the model by allowing the debtor government to have private information about its true willingness to pay. In Appendix I, we formalize this extension of the basic model and let true willingness to pay be increasing in the mutually observable fundamentals, yt, and in a debtor characteristic known only to the government. The debtor’s type summarizes the government’s private information about its capacity to repay, and the dispersion of this private information is a measure of the importance of moral hazard. The greater is uncertainty about the government’s true willingness to pay given observed fundamentals, the greater is the degree of moral hazard.

In Appendix I, we discuss the renegotiation of a bond issue in this extended model with and without CACs. In the event of default, the government reports that it is unable to repay its outstanding debt, Dt, and makes an equilibrium payment, Πt, that is less than its true willingness to pay. For bonds issued with CACs, the probability of default is higher than in the absence of asymmetric information. This is because the government can default when it knows that its true willingness to pay exceeds Dt but creditors are uncertain whether its true willingness to pay is less than or greater than Dt, With asymmetric information, the debtor government has an incentive to default and negotiate a restructuring of its debt even though it is actually willing to pay the debt as contracted. Further, the government generally pays less when it defaults than it would in the absence of moral hazard.

The incentives for the government to misrepresent its willingness to pay its debt obligation can be dampened by the presence of unanimous action clauses because the delays associated with debt restructuring impose costs, in the form of foregone output and investment, on the debtor country above and beyond eventual debt repayment. Drawing from the experience of debt restructurings in emerging market crises, we assume that the deadweight losses incurred during delayed debt restructurings exceed any gains from postponing eventual settlements. Renegotiation delays raise the cost of default for the debtor. In Appendix I, we represent the costs of delay by a constant amount, q, per period, so that the expected cost of delay rises with the length of the delay. If the costs of delay are sufficiently high, then the government will only default when its true willingness to repay is less than its debt obligation so that the probability of default is the same as in the absence of moral hazard. The probability of default will be lower if the debt is issued without CACs. Therefore, unanimous action clauses act as commitment devices.

The introduction of moral hazard has an ambiguous effect on the relative returns to creditors for bonds issued with or without CACs. In states in which default is necessary because the government’s true willingness to pay is less than its debt obligation, the returns to bondholders are reduced by restructuring delays under UACs. In these states, creditor returns are higher for bonds issued with CACs. However, for states in which the government will default on bonds with CACs but not on bonds with UACs, the returns to bondholders are lower for bonds with CACs than for bonds with UACs. As the importance of moral hazard rises, the expected return to bonds with CACs relative to bonds with UACs decreases and can become negative. This model implies that the total amount lent ex ante can be either smaller or larger under CACs than under UACs. It predicts that countries for which moral hazard is important may receive lower capital inflows under CACs while countries for which moral hazard is relatively unimportant may receive larger inflows.

C. Implications for Spreads and Swaps

With asymmetric information, debtor characteristics and circumstances will affect the relative returns to bonds issued with or without CACs in our model, yielding predictions for our empirical analysis below. Differences in observable fundamentals and in the importance of moral hazard will, therefore, affect the interest rate spread on bonds with CACs relative to bonds with UACs.

A decline in observable fundamentals raises the probability of default on both types of bonds. In Appendix I, we explain how deteriorating fundamentals can raise the interest rate differential for bonds with CACs over bonds without CACs under fairly general assumptions. Information regarding fundamentals (for example, the expected growth of future output) that affect repayments is treated as a noisy signal, subject to uncertainty. This implies that the interest rate spread for bonds with CACs rises as fundamentals decline until the probability of default on either type of bond is sufficiently high. When the probability of default on bonds with CACs approaches unity, this effect reverses and the interest spread decreases. It must become negative, so that bonds with UACs pay a premium over bonds with CACs, as the probability of default on bonds with UACs nears one. When debt restructuring is unlikely, bonds with CACs should yield a premium over bonds with UACs, but when debt restructuring is very likely, the opposite may hold.

The importance of moral hazard also affects the interest rate differential between bonds issued with CACs and with UACs. A decline in moral hazard follows from a reduction in the importance of the debtor government’s private information about its true willingness to pay given mutually observable information. In our model, this reduces the probability of default for bonds with CACs and increases the amount repaid in the event of a debt restructuring. As moral hazard decreases, the interest rate spread for bonds with CACs over bonds with UACs decreases. Because bonds with UACs have lower returns than bonds with CACs under symmetric information, the spread must become negative as asymmetric information about the debtor government’s willingness to pay vanishes.

Under asymmetric information, poor fundamentals generate a premium spread on bonds with CACs relative to bonds with UACs and more so as information asymmetries increase. While we cannot observe the extent of information asymmetries, it seems reasonable that an increase in information asymmetry and, hence, the degree of moral hazard, should contribute to a lower credit rating for the country, as would poor fundamentals. Thus, for low-rated countries, our model predicts that the higher moral hazard and probability of default when bonds include CACs could more than offset the benefits of restructuring, resulting in wider spreads than for bonds with unanimous action provisions. As credit quality improves, this premium will gradually disappear, and bonds including CACs will pay lower spreads, reflecting the shorter period of time spent agreeing to the restructuring. In fact, this is what we find in the empirical analysis below.

These differences in the performance characteristics of the two types of bonds can allow for welfare-improving debt swaps when the probability of default becomes large (holding the degree of moral hazard fixed), as illustrated by the current Uruguay situation. Consider a country with a modest degree of information asymmetry and fundamentals that place it midway in the credit rating spectrum. Bonds with UACs may command a lower spread because country commitment to avoid default has value. Therefore, bonds with UACs might be chosen initially in equilibrium. But if fundamentals deteriorate to create a high probability of default, the spread on bonds with UACs relative to bonds with CACs can become positive. In this case, a debt swap into bonds with CACs can increase the present value of the debt to the bondholders without any additional cash and increase debtor welfare even though it increases the probability of ex post renegotiation. This conclusion assumes correct pricing of the bonds with CACs, so that bondholders have rational expectations about the prospects of subsequent restructuring.

D. The Aggregation Problem: Are Collective Action Clauses Sufficient?

Collective action clauses are structured to facilitate coordination by the holders of a specific bond issue by making a qualified majority vote to change the financial terms of that issue binding on all holders, whether they vote positively or not. This limits the danger that holdouts will refuse to accept the change in terms and will have to be bought out at a higher price, thereby reducing the resources available to service the debts of the majority and causing the agreement to unravel. But such provisions are bond specific: they do not provide mechanisms for coordinating the holders of separate sovereign issues. It is possible that same coordination problem that can cause agreements to restructure debt by the holders of an individual bond to unravel might arise when creditors holding one or more of a government’s multiplicity of separate sovereign issues similarly refuse to agree.

The severity of this problem is unclear. Most emerging market sovereigns have only a handful of issues in the market. Ukraine had five and Ecuador had six at the time of their respective defaults. On the other hand, Argentina had more than 80 separate sovereign issues outstanding.

If a country has many different bonds with CACs in the market, any single bond might be a small enough share of the total debt so that its qualified majority can hold out for a larger than pro rata share in the overall debt restructuring. This could apply to a large number, if not all, of the bond issues. The share of the debt owed in any single bond issue needs to be less than St defined in equation (1). In this case, the qualified majority of each of the bond issues plays the war of attrition game. If the majorities of a share of the debt greater than or equal to 1 – St agrees to restructure, then the qualified majority of any one of the remaining issues maximize their return by holding up negotiations for a larger settlement. The return to holding out exceeds the return to joining debt restructuring negotiations for any qualified majority leading to an inefficient equilibrium. The symmetric equilibrium of the game is the same as before: the holders of each bond adopt a mixed strategy in which a qualified majority votes to join the renegotiation with positive probability, p, in each round. This leads to a probability, P, that only a share of the debt less than or equal to St is held out allowing restructuring to occur. In equilibrium for the coordination game, the holders of a bond representing a fraction x of the total debt receive


where E(sDt) is the expected payout to holdouts.13 This is less than xΠt which they would realize if the qualified majorities of all the bonds could agree to restructure without delay. These payoffs are similar to those of the prisoners’ dilemma.

The coordination game between holders of various bond issues implies that aggregation can be a problem even with collective action clauses and that additional contractual innovation might be necessary to support cross-issue coordination. Possible mechanisms for coordinating investors include information sharing, the adoption of codes of conduct, the formation of bondholders committees, and super-collective action clauses. Super-collective action clauses could allow a qualified majority of all bondholders to vote on the terms of restructuring in the event of default. These could add a provision to each bond issue that allows agreement by the qualified majorities of a sufficient percentage of all outstanding bond issues, or by a qualified majority of all bondholders regardless of what issues they hold, to be binding on all bondholders. The universal adoption of such covenants, or others to the same effect, would allow a qualified majority of the holders of all bond issues to avoid the coordination problem for renegotiation of bonds issued with CACs (the required majority needs to satisfy the condition of equation (4)).14

Though they differ in important respects, as discussed below, the two-step approach to debt restructuring recently proposed by J.P. Morgan (2002) and Uruguay’s recent restructuring both seek to implement super-collective action clauses through a debt exchange. When the probability of default is high, a swap into collective action clause bonds that carry a common majority action clause can be welfare improving for both bondholders and the debtor, as discussed above. The outstanding bonds can be swapped for bonds that differ with respect to maturity and other terms, while the common majority action clause resolves the aggregation problem in the event of renegotiation. Our model implies that the swap may require little or no additional funds from official sources in order to be successful, even though it can increase the probability of renegotiation, and even though this increase is anticipated by creditors. This result is most likely to obtain when the debtor has experienced an adverse event that makes the restructuring of outstanding bonds issued with UACs very likely.

E. Implications for the Transition

We consider how the interest spread on a bond with CACs is affected by the proportion of debt that currently lacks them. This case is important for asking whether the existence of bonds with UACs in the market creates inertia that discourages the issuance of new bonds with collective action clauses. If the majority of the debt is held in the form of bonds issued with UACs, then the holders of a bond issued with CACs may be willing to renegotiate that bond outside default to avoid costly debt restructuring delays that would follow the exercise of cross default clauses by holders of bonds with UACs. The holders of bonds issued with CACs accounting for a share of the outstanding debt equal to x should accept such terms of renegotiation if


This implies that the interest spread for a bond issued with CACs relative to the spread for bonds with UACs rises with the share of the outstanding debt that features UACs. The probability that the debtor country will be able to restructure the bond with CACs without renegotiating the debt with UACs decreases if moral hazard is absent. The interest spread differential for the minority share of bonds with CACs relative to bonds with UACs should be smaller the higher is a debtor country’s credit rating.15 This could explain why the transition from bonds with UACs to bonds with CACs may exhibit inertia, especially in the case of sub-investment-grade credits. This inertia should be less for investment-grade credits. We find empirical support for these predictions below.

Do borrowers have an incentive to begin issuing bonds with UACs when a majority of its debt is already issued as bonds with CACs? This case addresses the possibility that progress in introducing collective action clauses into the market could be reversed subsequently. If the costs of delays to agreement borne by the debtor country are independent of the proportion of debt which remains in default, then there should be no interest rate differential between a minority bond issued with UACs and bonds issued with CACs. This is because the probability of default is independent of the percentage of debt issued as bonds with CACs.

An alternative assumption is that the costs of debt restructuring delays for the country are small if the share of debt that remains in default is small. It may be reasonable to presume that the disruption of domestic finance, economic activity and cross-border capital flows diminish significantly if a sizable majority of the debt can be restructured successfully. In this case, the debtor government and the qualified majorities of the holders of bonds with CACs (which comprise a majority of the debt) can agree on a debt restructuring without the participation of the holders of the minority bonds issued with UACs, who cannot agree on restructuring. If the costs to the country are small, these can be shared with the holders of the bonds with CACs to achieve a restructuring of the majority of the debt. If the holders of the bonds with UACs cannot agree to renegotiate their bonds with the majority of bonds issued with CACs, then they suffer delay costs. The costs borne by the country reduce the probability of default, but this increases the returns to bonds with CACs as well. The bonds with UACs are disadvantaged rather than advantaged when a larger share of the debt is held in bonds with CACs. Put another way, this argument implies that the interest spread on bonds with UACs relative to bonds with CACs should rise with the share of outstanding debt that includes CACs. This should reassure those worried that progress in introducing CACs may be reversed subsequently. We find support for this prediction in the empirical analysis below.

IV. Evidence

We now present empirical evidence on several of the issues highlighted by the preceding theoretical analysis.

A. Borrowing Costs

The most prominent worry—or at least the one that has received the most scholarly attention—is that creditors might regard collective action clauses as weakening their rights, rendering it more costly for emerging markets to borrow. Qualitative evidence is not very helpful for settling this debate. Those who are skeptical that collective action clauses will significantly affect borrowing costs observe that these provisions are not often referred to by market participants. Those who suspect the existence of an effect, on the other hand, can cite instances where market participants have commented on their presence (see UBS Warburg, 2003, for an example). They can observe that the existence of an effect depends on awareness only on the part of the marginal investor.

Quantitatively, the issue has been studied by Eichengreen and Mody, 2000a,b, and Becker, Richards, and Thaicharoen, 2000, using data on primary market (launch) spreads.16 Neither study supports warnings that collective action clauses would increase borrowing costs across the board. But while Becker, Richards, and Thaicharoen find no significant impact of the presence or absence of collective action clauses in their overall sample of new issues, Eichengreen and Mody distinguish bonds by the credit rating of the issuer and find that the presence of collective action clauses reduces spreads for issuers with investment grade ratings but widens spreads for sub-investment-grade credits. Their interpretation, as in the model of Section C, emphasizes the tradeoff between the efficiency advantages of more orderly restructuring, which dominates for high quality borrowers who are unlikely to default opportunistically, and the associated moral hazard, which dominates for low-quality borrowers whose motives and response are suspect. In addition, they find that the magnitude of the penalty for sub-investment grade borrowers using collective action clauses depends on market sentiment: when the Emerging Market Bond Index (EMBI) is relatively stable and the EMI spread is low, this penalty is limited. In contrast, when the EMBI is volatile—when investors are particular uncertain about the prospects for emerging bond markets—the penalty is greater, suggesting that investor fear that a broader range of speculative credits may use this uncertainty as cover for opportunistic behavior (see Mody, 2003).

These differences are important for understanding the obstacles to the more widespread use of collective action clauses. Becker, Richards, and Thaicharoen suggest that emerging markets have no reason to wait. Eichengreen and Mody, in contrast, suggest that countries with lower rates may be deterred by the specter of higher borrowing costs.

The recent issues by Mexico and Brazil, subject to New York law but including majority restructuring provisions, are relevant to this debate but cannot resolve it. Mexico’s first issue featuring CACs, scheduled to mature in 2015, was priced to yield 6.92 percent, a spread of 313 basis points over 10-year U.S. treasuries. While exact benchmarks are not easy to construct, market analysis suggests that bond was priced at a premium of about 8-10 basis points. It is not unreasonable to assume that the first issue to break new ground was charged a premium for doing so. A similar bond placed in April 2033 was, however, thought to have been issued at a small discount. The Eichengreen-Mody results suggest that a country which has just succeeded in obtaining an investment grade rating (Mexican debt was rated BBB-, its lowest investment grade rating, by Standard & Poor’s, and Baa2, one step above the lowest investment grade, by Moody’s) should have enjoyed a discount on bonds with collective action clauses of about 25 basis points relative to the yield curve.

Brazil’s $1 billion issue with CACs placed in New York in late April is potentially informative because Brazil was the first speculative grade issuer to take this step.17 However, Brazil’s bond included an 85 percent majority restructuring provision, which places the threshold about halfway between the standard unanimous and majority action provisions, damping the impact on spreads.18 The model of Section C suggests that as we move from a 75 percent to an 85 qualified majority, the majority must be roughly twice as patient to prevent holdouts from expecting that they have a reasonable shot at receiving more than a pro rata share of the settlement and by thus creating incentives for holdouts to generate an impasse in debt restructuring.19

Not surprisingly, opinions regarding the pricing of this issue differ.20 These complications remind us not to make too much of a couple of data points. Progress on this question is only likely to take place through the analysis of substantial new data sets. Gugiatti and Richards, 2003, take a step in this direction by analyzing the pricing of a substantial set of bonds on the secondary market at a single point in time. Using Merrill Lynch Global Index data for 31 January 2003, they regress the log of the ratio of the yield on the emerging market bond relative to the yield on a corresponding mature market bond on the country credit rating, the duration of the bond, issue size, a dummy variable for whether the issue is dollar denominated, a dummy variable for the inclusion of collective action clauses, and various interaction terms.21 Credit ratings are coded on a scale from 1 to 18, 1 being Al, the highest rating in the sample, and 18 being CC3, the lowest.22 The key interaction term is between the rating and the presence of collective action clauses; this speaks to the hypothesis that the effects of CACs are different for high-and low-rated issuers. If the dummy variable for the presence of CACs is negative, while that on the interaction term is positive and sufficiently large, issuers with sufficiently poor credit ratings pay wider spreads when including CACs, even if investment-grade issuers do not.

The authors omit issuers rated below B–/B3 and issues not in U.S. dollars or major European currencies.23 They include dummy variables for bonds issued by Mexico, Brazil, and Turkey. Why only these countries deserve dummy variables is unclear, although it is true that they had multiple bonds in the market and their bonds were three of the more important components of the EMBI index. While Gugiatti and Richards find that collective action clauses are associated with smaller spreads for high-rated issuers and higher spreads for low-rated issuers (as in the earlier Eichengreen-Mody study), many of the coefficients they estimate, and specifically that on the interaction term between credit rating and the presence of CACs, are statistically insignificant.24

Gathering the same data independently, we were able to replicate their results. But it is unclear whether the insignificance of the coefficients reflects the relatively small size of the sample of secondary market spreads, the omission of very low (C-rated) bonds from the analysis, the state of the secondary market on the particular date they analyze, and the nature of the country sample, or whether there really is no effect of CACs on secondary-market spreads that varies with credit quality. The secret of successful empirical work, according to Frankel, is to define the question so that failure to reject the null of no effect can be counted as success. Authors whose null is that there is no impact on spreads that varies with credit quality will see no need to pursue the questions raised in this paragraph. We, on the other hand, do.

Standard advice for the econometrician faced with poorly determined coefficients is to gather more data. We therefore gather the same data for four additional points in time and pool the five cross sections (for sources and further details, see Appendix II). Our earlier analysis of launch spreads (viz. Mody, 2003) suggested that the point on the credit-quality scale at which the spread discount for bonds with collective action clauses becomes a premium (where spreads are wider for bonds with CACs than bonds without them) depends on the state of the markets: when the markets are less skeptical or uncertain about emerging market debt, they apply a penalty for collective action clauses to a narrower range of issuers. We therefore gathered data for the four most recent dates at which the EMBI spread was at a local peak or trough: April 12, 2002, when the EMBI spread was at a local trough; September 30,2002, when it was at a local peak; September 6, 2000, the prior local trough; and November 2, 2001, the prior local peak. The dependent variable is Merrill Lynch’s option-adjusted spread, which is based not just on the difference between the bond in question and a benchmark bond in the same currency but also takes into account the implications of the yield curve in discounting future cash flows.25

Table 2 reports the results. The first column reports random effects estimates, which capture both the time series and cross section variation in the data. These suggest that bonds of countries with high credit ratings that feature CACs enjoy tighter spreads, other things equal; this is evident from the negative coefficient on the CACs dummy. The variable in question is statistically different from zero at the 90 percent confidence level. In addition, however, as credit quality deteriorates (the rating variable rises according to the current metric), this spread differential first narrows and then reverses sign, as we found in our earlier work with launch spreads. The fixed-effects estimates in column 2, which include fixed effects for both countries and periods, are essentially the same. The point where the spread discount on bonds with collective action clauses turns to a premium is approximately where credit quality, as presently scaled, falls to 10 (equivalent to a rating of Bl, that of the Romania in early 2003).

By including the entire vector of country dummies, we have essentially generalized the specification of Gugiatti and Richards, who included dummy variables for Mexico, Brazil and Turkey alone. But to probe further whether the results are being driven by the observations for particular countries, we dropped the observations for a series of countries, one at a time. Columns 3, 4, and 5 show the results when we drop the observations for Brazil, then for Mexico, and then for Turkey. The results are again unchanged.

In columns 6, 7, and 8 we drop the observations for Lebanon, Egypt and Qatar. These observations warrant special attention because these three countries have in fact issued bonds subject to New York law that include majority restructuring provisions.26 (The relevant bonds are coded as including CACs throughout our analysis of secondary market spreads.) The provisions of these issues are not well known, however, because they are private placements and because they are not therefore included in the EMBI. Moreover, the fact that they are private placements means that they have a relatively illiquid secondary market, raising questions about whether they should be included in an analysis of secondary market spreads in the first place.27 Reassuringly, the results are again the same when we drop the observations for these three countries, one country at a time.28

Recall that Gugiatti and Richards drop Japanese law bonds from the sample and that we have followed their convention. There may be some reason to think that such bonds enjoy a captive market of Japanese institutional investors, and that they are therefore priced differently on the secondary market. The same argument can presumably be made about German-law bonds.29 As a further form of sensitivity analysis, we therefore dropped German law bonds from the sample. The consequences can be seen in the first two columns of Table 3. Again, the key results continue to hold: tighter spreads for bonds with CACs when the issuer is high quality, wider spreads when it is of low quality.30

In columns 3 and 4 we allow the premium (discount) for bonds with CACs to vary with market sentiment, since our earlier analysis suggested that this could be the case (and, specifically, that the point on the credit rating scale where poorer credits began paying a premium for CACs kicked in at higher credit rates when investor sentiment was less favorable toward emerging markets). We add the triple interaction of the credit rating, the EMBI spread, and the dummy variable for CACs to test whether the range of sub-investment grade credits that incur a penalty for using CACs varies with market sentiment. A positive coefficient would suggest that the point on the credit quality scale at which the addition of CACs leads to a spread premium rather than a spread discount occurs lower on current scale (at a better credit rating) when the EMBI spread is high and investor sentiment toward emerging markets is relatively poor. For safety, we also include the two-way interaction of the credit rating and the EMBI spread to ensure that the coefficient on the triple interaction term, which is the one of particular interest, is not being contaminated by the impact of changes in market sentiment on how investors value claims on issuers of different credit quality.

The fixed-effects estimates in column 4 support this hypothesis. The random effects estimates in column 3 are less supportive, although the signs of the coefficients are the same. When the EMBI spread is high (investor sentiment toward emerging markets is relatively unfavorable), the range of issuers with relatively poor credit ratings that pay a spread penalty for including CACs is larger than when the EMBI spread is low (investor sentiment is relatively favorable). Moreover, when the EMBI spread is 700 (indicative of relatively favorable sentiment), the bonds of issuers whose credit quality is relatively poor do not trade at a spread premium as a result of the inclusion of CACs. But when the EMBI spread increases further from there, a spread premium appears for issuers at the relatively poor end of the credit quality spectrum.

These findings go some way toward reconciling previous studies of the primary market. Recall that some of those studies found that speculative borrowers face higher funding costs when using collective action clauses (Eichengreen and Mody, 2000a,b) but another (Becker, Richards, and Thaicharoen 2001) did not. Here we see that the point where this penalty kicks in varies with market sentiment, and that when sentiment is particularly favorable this turning point can be located very low on the credit quality scale, or disappear entirely. This is reassuring insofar as it explains how previous investigators could have reached seemingly contradictory conclusions. The finding is also reassuring since, to the extent that any penalty depends on market conditions, it suggests that other measures that work to limit market volatility may make the use of collective action clauses attractive even for sub-investment grade countries.31

To dismiss these results as spurious it is not enough to say that the inclusion or exclusion of CACs is a matter of market convention. That would explain a set of zero coefficients, but not the pattern we obtain for both primary and second markets. In addition, the skeptic would need to come up with an unobserved characteristic of some high-rated countries that further enhanced their credit worthiness (reducing their borrowing costs) and also encouraged them to borrow in London and Luxembourg, and another unobserved characteristic of some low-rated borrowers that similarly enhanced their credit worthiness and simultaneously encouraged them to borrow in New York. The skeptic would also have to explain what omitted shift variable leads us consistently to find, using two different data sets, that investors penalize a wider range of speculative issuers for using collective action clauses in periods when investor sentiment is less favorable toward emerging market bonds.

Having obtained the same results on two entirely different data sets, we continue to believe that the use of collective action clauses will if anything reduce funding costs for investment-grade issuers, for whom investors welcome the existence of mechanisms to facilitate orderly restructuring. There is still the possibility that they will raise funding costs for speculative credits, although the extent of this effect is likely to depend on market conditions.

B. The Transition Problem

The transition problem is that more than two-thirds of emerging market debt presently lacks collective action provisions, and that even if these clauses were included in all new issues going forward, the better part of a decade might have to pass before existing bonds with unanimous action provisions matured and were retired (IMF 2002).

Roubini and Setser (2003) suggest ten years is not an unreasonable period of time to complete a process that has already been debated for a decade. Russia, Ukraine and Lithuania already use English law in their dollar-denominated bonds. Argentina could introduce clauses into 10 to 20 percent of the stock of emerging market bonds if it included these provisions in its restructuring bonds (as the IMF will no doubt encourage it to do). And, as we have seen, other countries like Uruguay might do similarly even in the absence of default. The prospect of major Brady swaps by countries like Brazil, which are already on the markets’ radar screen, would be another vehicle for introducing clauses into many of the remaining bonds. It would also be possible to expedite the process with a broad-based debt exchange, as analyzed by Group of Thirty (2002).

As explained in Section C, in the absence of a broad-based exchange, investors asked to take up new issues with collective action provisions might worry that their instruments were less senior than the existing stock of claims with unanimous action provisions. Because bonds with CACs are easier to restructure, they may be restructured more frequently or on less favorable terms. Thus, there may be some reluctance to accept new instruments with collective action clauses when there is an existing stock of instruments that omits them.

Mexico’s global issue featuring collective action clauses, in the presence of a large inherited stock of debt that does not include them, suggests that this problem is unlikely to be serious. We can provide further evidence by considering pricing in the primary market generally. The primary market is the relevant one in the present context, since we are now talking about the incentive to take up new issues.

We extend the Eichengreen and Mody, 2000a,b, analysis of launch spreads by constructing an independent variable that measures the share of the existing stock of bonds that includes the relevant contractual provision and interacting it with the governing law on the new issue. We use data from Capital Bondware on bonds placed internationally by the governments of emerging market economies between 1991 and 2000 (see Appendix II for details on data used).32 Our dependent variable is the launch spread, defined as the yield to maturity at time of issue minus the yield on a low-risk bond of comparable maturity 33. As controls we use the standard explanatory variables utilized in previous studies of emerging-market bonds: these include characteristics of the issue (its amount, its maturity, whether it bears a fixed or floating rate), characteristics of the issuer (the continent on which it is located, its credit rating, its recent growth rate, the volatility of its exports, its debt/GNP ratio, its reserves to short-term debt ratio, and its ratio of domestic credit to GDP), and characteristics of the global financial environment (the ten-year U.S. Treasury rate, the U.S. high-yield spread, and the volatility of the Emerging Market Bond Index during the quarter the bond was issued).

Table 4 shows the results. Overall, the spread on a new bond with CACs is not higher when the existing stock of debt is dominated by bonds with unamimous action clauses (see column 1)—consistent with the Mexican example cited above. However, when we distinguish issuers by credit quality we do we see an effect. For consistency with our own previous work on launch spreads, we measure credit quality using Institutional Investor ratings, which vary from 0 (worst credit) to 100 (best credit).34 We partition issues into three groups: those with ratings between 0 and 35, those with rates from 36 to 50, and those with ratings of 51 and above.35 Issuers with the lowest credit quality, who are presumably most likely to restructure, do incur higher costs from issuing bonds with CACs when the existing stock is dominated by bonds with unanimous action clauses. This result is what we would predict on the basis of Section C above. In contrast, there is no evidence of this effect for better credits, which explains why we did not obtain it in the first column.

We obtain a similar result when we analyze the impact of a large stock of bonds with CACs on the spread on a newly-launched UAC issue. The coefficient on the share of debt with CACs in the existing stock is positive for sub-investment grade issuers but zero for investment-grade credits. It could be that holders of UAC bonds fear that they will be left hanging when a government restructures the majority of its debt using majority action provisions. Again, this is the prediction of the theoretical analysis in Section C above.

C. Asset Substitution and Market Migration

Then there is the possibility that investors not enamored of collective action clauses might substitute bank loans or other credit instruments for bonds if renegotiation-friendly provisions are added to the latter. However, our evidence from the primary and secondary markets does not support the view that investors will find these provisions repulsive and substitute away from them. The danger that bond flotations might migrate from markets where collective action clauses are required by regulation or statute to markets where they are not seems exaggerated, for the same reason. In addition, most issuers now prefer to issue global bonds that meet registration requirements in all major markets in order to maximize the size of the potential customer base. From this point of view, it seems unlikely that the use of clauses in the New York market would cause the market to migrate into unregistered securities or illiquid locales.36

D. Aggregation

As explained in our theoretical discussion, collective action clauses are structured to facilitate coordination by the holders of a specific bond issue by making a qualified majority vote to change the financial terms of that issue binding on all holders, but they do not provide mechanisms for coordinating the holders of separate issues. How much to worry about this is unclear. On the one hand, the special difficulties of restructuring the debts of countries with many separate debt issues may be a serious concern. On the other hand, there may exist other mechanisms—information sharing, a code of conduct, bondholders committees, or super-collective action clauses—through which investors can be coordinated.

If aggregation is costly, then investors will presumably demand a premium in order to hold claims on an issuer with multiple instruments in the market, especially when there is a significant likelihood that its obligations may have to be restructured. It should therefore be possible to test for the presence of an aggregation problem using evidence from the primary market. Again, launch spreads rather than secondary market spreads are the relevant dependent variable, since we are concerned with the incentives of how to structure new issues. The key explanatory variable is the number of separate issues that the sovereign already has in the market at the time a new bond is launched.37 The controls used for this purpose, as listed in Table 5, are the same as those used for the transition problem discussed in subsection (b) and listed in Table 4. In particular, we include the country’s debt/GNP ratio to be sure that our measure of the number of separate sovereign issues is not simply picking up the level of indebtedness of the country.

The coefficient on the number of separate sovereign issues (the “multiplicity premium”) is reported in the first column of Table 5.38 We do see evidence of an aggregation problem. The coefficient on number of bonds is positive and statistically significant at standard confidence levels. The point estimate suggests that distributing the same amount of debt among an additional ten bonds would raise spreads on the tenth bond by about 2 percent, or about 8 basis points. These findings are not driven by the observations for Argentina, a country with an exceptionally large number of bonds in the market; we get essentially the same results after dropping the Argentine observations.

This effect is not large, but its impact is quite a bit larger for countries with low credit ratings. This is intuitive: if our variable is really picking up costs of aggregation that come into play during restructuring negotiations, then it should have the largest effect on the obligations of countries whose perceived probability of having to restructure is high. It should have the largest effect, in other words, on countries with poor credit ratings.

Again measuring credit quality using Institutional Investor country ratings, which range from 0 (worst credit) to 100 (best credit), we now allow the effect of the number of bond issues to differ by rating, again distinguishing three credit-rating groups on the Institutional Investor scale (0–35, 36–50, and 51–100). The estimated effects, in the second column of Table 5, confirm that the largest multiplicity premium is demanded of countries with the lowest credit ratings (0–35). For countries with intermediate ratings (36–50), in contrast, the effect is of the same order of magnitude as the full-sample estimates reported before. For countries with relatively high credit ratings (above 50), the coefficient for the number of separate bond issues turns negative. Arithmetically, the relatively small positive coefficient on number of issues for the sample as a whole is thus an average of a large positive effect for the lowest rated countries, a small positive effect for countries with intermediate ratings, and a negative effect for the highest rated countries. This presumably reflects the interaction of two offsetting economic forces. Having an additional debt instrument in the market complicates future restructuring negotiations; this is the dominant factor for low-rated issuers, for whom the likelihood of future restructuring is high and for whom this factor consequently carries considerable weight. At the same time, continuing interaction with the market builds reputation and can be taken as a sign of a country’s commitment to maintain its good credit; this effect dominates for high-rated issuers.39

The case of Argentina provides a useful perspective on these results. Recall that the government had upwards of 80 bonds in the market prior to its default. Most of the outstanding bonds were contracted when it was in the intermediate rating category (an Institutional Investor rating between 36 and 50), where, our results suggest, investors demanded only a small additional premium to compensate them for potential costs of aggregation, reflecting the fact that the perceived probability of default, while not negligible, was still limited, and where the government’s continued interaction with the market was taken as a reassuring indication of its commitment to the maintenance of its credit. When Argentina’s creditworthiness deteriorated subsequently, however, the fact that the government had many issues in the market, raising the specter of complex restructuring negotiations, became a significant concern. This concern increased the difficulty for the Argentine authorities of attempting to meet their immediate financial needs by floating another issue, independently of standard debt-sustainability considerations.

We can also ask whether aggregation costs vary depending on whether or not an issue includes collective action clauses. The answer, shown in Table 6, is no. There we estimate the same spread regressions as in Table 5 (with the spread as the dependent variable, the number of separate issues as the key explanatory variable, and the same list of controls), but separately for bonds that are subject to U.K. and U.S. law (and which therefore do and do not include collective action clauses, respectively). We also distinguish bonds subject to other miscellaneous laws, some of which include collective action clauses and other which do not. The results for bonds with collective action clauses are in the first column, while those for bonds without them are in the second. Comparing the two columns, there is no indication that the presence or absence of collective action clauses significantly conditions the perceived costs of aggregation. A test for the significance of differences in coefficients across the columns of this table leave us unable to reject the null of equality of the coefficients for bonds with and without CACs. That we do not find significant differences in the multiplicity premium as a function of whether or not a country’s bonds have collective action clauses suggests that the latter may not be enough, by themselves, to solve problems of cross-issue aggregation.

One can imagine responding to these findings in different ways. One response is that a multiplicity premium of 8 basis points for the sample as a whole is not enough to lose sleep over. Aggregation difficulties are minor compared to other factors that inflate borrowing costs for emerging markets. At the other extreme, one might worry about the significantly larger multiplicity premium affecting countries with low credit ratings and advocate the use of “super-collective-action clauses”—provisions in each bond issue that provide for a binding supermajority vote of not just holders of that issue but other issues also. The challenge then would be to get these provisions into the market and gain investor acceptance. Adding a super-collective-action clause could not proceed on a bond-by-bond basis. Rather, all issues would have to be converted simultaneously, which is likely to be possible only for actual or potentially distressed debtors, like Argentina and Uruguay.

Uruguay has undertaken an experiment along these lines. To address its serious debt problem, the government offered an innovative bond exchange in April and May of 2003. The new bonds include super-collective-action clauses allowing changes in financial terms if 75 percent of an issue agrees or if 85 percent of all series and 66.66 of each affected series agree. Exit consents, which deface the old bonds, and warnings by the government that if the exchange failed it would have no alternative but to default, were used to encourage participation. Our theoretical analysis above suggests that under Uruguay’s circumstances (when a country has experienced a large negative credit event due to circumstances largely not of its own making) a high level of participation is to be expected, since it will be welfare improving and, in particular, has the potential to raise the value of debt for creditors.40 Consistent with this conjecture, Uruguay’s exchange offer did in fact elicit a high rate of participation.41 Favorable market conditions may have also helped.

Yet another solution would be to rely on information sharing and procedural conventions to encourage holders of different issues to coordinate on the cooperative equilibrium, as discussed in Section C above. Communication and information sharing reduce the scope for strategic behavior by creditors that may result in their selecting the noncooperative equilibrium. (Think, as discussed above, of the classic prisoners’ dilemma, in which the noncooperative equilibrium depends on the inability of the prisoners to communicate.) A code of conduct leading to common procedures and a bondholders committee where information can be pooled may then limit opportunistic behavior.

This is how the official community and the markets have approached the issue of cross-issue coordination. EMCA, 2002, and Taylor, 2002a,b, have emphasized the desirability of engagement and initiation clauses which would specify the actions the sovereign and investors would take in the event of a credit default. In Taylor’s, 2002a, p. 2, words, engagement and initiation clauses would “provide for early dialogue, coordination and communication among creditors and a sovereign and limit disruptive legal action.” The Bank of France has similarly suggested a code of conduct for sovereign debt restructuring, which can be understood (in part) as a set of procedures and conventions to encourage information sharing and standardized procedures.42 The Bank’s code includes nine main principles: (1) early engagement with creditors; (2) fair information sharing among all interested parties; (3) fair representation of all creditors; (4) an expeditious and cooperative process; (5) comparable treatment among creditors; (6) fair burden sharing between debtor and creditor; (7) good-faith negotiation; (8) preservation of the debtor’s financial situation; and (9) rapid restoration of financial stability. While the code would not be legally enforceable, it still would provide some useful structure and guidance for negotiations.

The Eichengreen-Portes, 1995, idea of a New York Club, popularized by Hubbard, 2002, and Kroszner, 2003, is another mechanism whereby creditor coordination might be encouraged through information sharing and repeated interaction. Miller, 2003, cites spokesmen for the creditor community as arguing that what creditors want are collective action clauses, a Code of Good Practice, and a forum for negotiations as a tripartite approach to facilitating orderly restructuring without creating moral hazard or unpredictability.43 Others are skeptical that a code of conduct could help investors coordinate on the good equilibrium. Roubini and Setzer, 2003, write, “No matter what the code aims to do, particular attention needs to be given to the set of incentives that will lead all parties to have an interest in abiding by a non-binding code. In theory, adherence to the code during the restructuring could be a condition for creditors’ final agreement on restructuring terms. However, this raises obvious problems of time consistency.” This, of course, is simply the distinction between a situation in which one ends up in the noncooperative equilibrium with probability one, in which case a code is useless, and a situation with multiple equilibria, where a code can help investors coordinate on the better equilibrium. And, as shown in Section C, the conditions under which multiple equilibria exist are quite general.

V. Why the Reluctance to Adopt?

We reserve for later the question of whether the more widespread use of collective action clauses will significantly reduce the frequency and severity of crises. But the official community, for its part, evinces little doubt that this innovation would at least be a step in the right direction. This makes the failure of official rhetoric to promote progress a source of frustration. What explains the failure of investors and issues to embrace bonds with these provisions more rapidly? And what does their reluctance to do so in the past imply for the future? 44

The absence of collective action clauses from bonds issued in the United States is a long-standing phenomenon. The need for bondholder coordination first attracted attention in the nineteenth century, when large corporations relying on external finance first appeared on the economic scene.45 The railroads were the largest such corporations; they relied most heavily on debt finance, and they had to overcome many of the same challenges as infrastructure projects and infrastructure finance in modern-day emerging markets (see Eichengreen, 1996). The combination of widely disbursed bond holdings and high costs of liquidation made it inefficient to allow a single creditor or a small minority of creditors to force a railroad to liquidate (since track and related investments typically had greater economic value in place than as salvage). The same was true, then as now, of a variety of industrial corporations. In England, a market solution was found in the introduction of majority action clauses into bonds starting in the 1870s. These clauses, like those in English-law bonds today, allowed a super-majority of the bondholders to agree to reduce the amount due under a bond and rendered their decision, when ratified by a vote of the specified majority, binding on all bondholders, including any who had not endorsed the change.

This contrasts with the situation in the United States, where collective action clauses were never widely utilized and investors instead relied on the courts to avoid inefficient liquidation. The explanation may lie in the exceptionally convoluted capital structure of U.S. corporations, especially railroads, which made it difficult to implement the English-style market-based approach (Skeel, 2002). Another possibility is that bonds including collective action clauses may not have been regarded as unconditional promises to pay under the terms of the U.S. Negotiable Instruments Act. Whatever the reason, before World War I, most U.S. corporate insolvencies were reorganized through a court-led procedure known as “equity receivership.” Once the U.S. Congress amended the Bankruptcy Act in the 1930s to facilitate supervision of corporate reorganizations by a bankruptcy judge, they proceeded under the familiar Chapter 11 (and other chapters) of the U.S. Bankruptcy Code.

Thus, even at its height, in the 1930s, the practice of including English-style collective action clauses in bond contracts extended to only perhaps 10 percent of U.S. corporate bonds. The Trust Indenture Act of 1939 gave the U.S. approach official sanction. William O. Douglas, influential member and chair of the Securities and Exchange Commission, saw collective action clauses as allowing corporate and Wall Street insiders to take advantage of small bondholders in corporate reorganizations, which were not infrequent in the 1930s.46 The result was the Trust Indenture Act, which included a Section 316(b) that prohibited reductions in amounts due under a publicly-issued corporate bond without the consent of each and every bondholder. This restriction was feasible, in the sense that it did not lead to a spate of inefficient liquidations, because of U.S. bankruptcy law allowed the courts to substitute for the missing provisions.

This history helps to explain why majority action clauses are not included in corporate bonds issued in the United States. But it cannot explain why such provisions are excluded from sovereign bonds, to which the Trust Indenture Act does not apply. Indeed, there would be no rationale for applying it, given the absence of a bankruptcy court to substitute for the missing collective action provisions—which is of course the problem that the reforms under discussion here are designed to address.

To explain the transfer of this “genetic code” from corporate bonds in the 1930s to sovereign bonds in the 1990s, one must tell a story like the following. Virtually no bonds of foreign sovereigns were issued in New York between the 1930s and the 1980s. The bond market only started up again following the advent of the Brady Plan in 1989. At that point there were no practicing attorneys in New York experienced in drafting sovereign bond covenants. Falling prey to the block-copy command, they simply transferred the template used for corporate bonds.

This history may explain the origin of current practices. But to say that a phenomenon is historically rooted is not the same as suggesting that it is historically determined. That a practice has historical roots does not mean that it cannot change, even rapidly under some circumstances. Thus, for the absence of collective action clauses from the U.S. market to be a path-dependent (historically-determined) equilibrium requires not only the initial conditions given by this ancient history but also a lock-in mechanism that significantly slows the process of change. Why then, if collective action clauses have attractive features, has change not been faster? Financial markets are hardly slow to innovate; they are criticized for many things but only rarely for their reluctance to develop new financial instruments.

Allen and Gale (1994) suggest five reasons why socially desirable financial innovations sometimes fail to emerge.

  • Product uncertainty. Investors may be uncertain about the performance characteristics of the new instrument—for example, about whether greater ease of restructuring will make restructuring more frequent—causing them to demand a premium in order to hold it. That premium may discourage borrowers from utilizing it. Even if countries can educate investors, convincing them that they are not likely to act opportunistically, doing so may have costs that deter use of the new financial product.

  • Competitive structure of the financial industry. Some of the costs of designing the new clauses and educating investors about them will be incurred by the financial firm underwriting the issue. There could be a higher financial cost associated with drafting and marketing new provisions. “[0]ff-the-shelf language costs less,” as the IMF (2002, p. 10) puts it. Insofar as other firms may be able to quickly enter the market for these instruments, the returns on the initial investment will be competed away. Pioneering the innovation will therefore be unattractive.

  • Coordination problems. It may be necessary for a number of borrowers to issue these instruments simultaneously for the development of a deep and liquid secondary market on which investors can effectively spread risk. This creates a first-mover problem: individual borrowers have no incentive to internalize these risk-sharing benefits and liquidity effects insofar as these also impact other countries. In addition, the idea that creditors holding bonds with collective action clauses may believe that their instruments are effectively less senior and secure if other bonds of the same issuer lack such provisions may require all creditors to accept the new instruments simultaneously.

  • Implications for systemic stability. The new instrument may have positive externalities for the stability of the international system. That the costly and disruptive nature of debt restructuring under present arrangements places pressure on the IMF to extend financial assistance, which encourages excessive lending and borrowing in expectation of an official bailout and thereby heightens crisis risk, is an example of such an externality. But individual countries, with only weak incentives to internalize this externality, may display a reluctance to adopt CACs that is excessive from a social point of view.

  • Political distortions. To these market failures one may add government failure. Politicians with uncertain reelection prospects may have higher discount rates and shorter time horizons than society as a whole. Consequently, they may prefer inflexible contractual provisions that reduce costs of borrowing now, tying the debtor government to the mast by creating costs of restructuring that are inefficiently high from a social point of view. Alternatively, debtors and creditors may prefer a regime where they are bailed out to one in which debt is restructured, and they may be able to resist the adoption of restructuring friendly rules and regulations that limit the pressure for official assistance.

Once upon a time, product uncertainty may have mattered. Issuers repeatedly invoked uncertainty about how investors would receive bonds subject to New York law but including collective action clauses as a reason for their reluctance to include them. But even if one insists that there once was uncertainty about how investors in New York would price bonds with these provisions, the fact that Mexico, Brazil, South Africa, and the Republic of Korea have now issued bonds with collective action clauses in that jurisdiction renders the argument of only historical interest.47

The fixed costs of innovation and competitive structure of the financial industry similarly strike us as weak explanations for the failure of collective action clauses to work their way into the New York market more quickly. Off-the-shelf language may cost less, but even if the only language on the shelf in the United States requires unanimous consent, the language of collective action clauses can be easily imported from abroad. Insofar as there remain costs of adapting that language to U.S. legal circumstances, the fixed costs can be shared by encouraging governments and market participants to jointly contribute to their design. In fact, this is what the G-7 and gang of six associations of market participants have done in cooperating on the design of model clauses. Their coordinated intervention may have broken down any residual effects of this barrier to innovation.

As for the need for several consequential issuers to move simultaneously to create a liquid market in bonds with collective action clauses in New York, such a market now exists, courtesy of Mexico, Brazil, South Africa, and the Republic of Korea’s recent issues.48 From all appearances, the bonds in question are trading smoothly on the secondary market.

The remaining explanations may have more sway. The high discount rates of short-lived governments can clearly lead myopic politicians to undervalue costly steps that offer benefits only down the road. Creditors and debtors, especially sub-investment grade debtors if we are right that speculative credits will have to pay a premium when issuing bonds with collective action clauses, may prefer a regime where they are bailed out to one in which debt is restructured, and they may be aware of the political pressure on the IMF to lend when mechanisms for dealing with problems of creditor coordination are absent. These political distortions may explain sub-optimal rates of adoption.

Similarly, economic distortions may result in the suboptimal use of collective action clauses, specifically if the smoothing of procedures for sovereign debt restructuring has positive externalities for systemic stability. In other words, use will be suboptimal if the benefits of the decision to adopt do not accrue exclusively to the adopting country. If one country’s adoption leads to a generalized reduction in the moral hazard associated with IMF rescues, then the system as a whole may grow more stable. Investors will apply more rigorous market discipline, and governments will more carefully limit their demand for foreign capital. This will mean fewer crises and less of a tendency for crises to spill across borders. Of course, whether these reforms will significantly enhance systemic stability continues to be debated.49 But, leaving aside that debate for now, the notion that the benefits are systemic, and not merely country specific, is a consistent explanation for why there is a gap between what is socially optimal and privately practiced.

If the obstacles created by product uncertainty, fixed costs, and coordination problems have been removed by Mexico, Brazil, South Africa, and the Republic of Korea’s recent issues, then we should see more bond issues in New York with collective action clauses in coming months and years. If Becker, Richards and Thaicharoen (2001) are right, a long list of speculative credits will soon join this parade. But if we are right, the movement may be limited to countries with relatively good credit ratings. And, to the extent that the political distortions described above are significant, adoption will in any case remain suboptimal from a social point of view.

This suggests offering pecuniary incentives or taking regulatory action to offset the distortion, as suggested by Roubini and Setser (2003). These authors suggest that U.S. government should start by arm-twisting the major investment banks. If this doesn’t work, the Securities and Exchange Commission could require the use of clauses in SEC registered bonds. Although G-7 governments have embraced the argument for collective action clauses, they remain reluctant to alter securities registration requirements and exemption rules to require their use. The role for regulators is traditionally seen as protecting investors from fraud and assuring the integrity of markets, not as reforming the international financial architecture, rendering officials reluctant to go down this road. Treasury would presumably have to convince the SEC that majority action provisions provided a crucial protection for bondholders. If this campaign failed, the Trust Indenture Act could be amended to make the inclusion of collective action clauses a matter of statute rather than regulation.

Other approaches are less promising. Taylor (2002a) has suggested that the adoption of collective action clauses could be encouraged by making this a condition of access to IMF facilities. In the strong version of this proposal, only countries that already incorporated collective action clauses into their international loan agreements would be eligible for IMF loans. However, this is not an effective incentive for the growing class of investment-grade countries that do not contemplate having an IMF program. At a more fundamental level, this approach comes dangerously close to assuming a solution to the IMF’s time-consistency problem. The IMF’s principal shareholders can aver their reluctance to lend to countries that have not embraced the relevant contractual reforms, but when a crisis looms they will feel pressure to back down and lend to countries whose inflexible loan contracts create a risk that an involuntary restructuring will be difficult, messy and uncertain. Knowing that the IMF has an incentive to disburse anyway, countries will have little incentive to alter their habits.

In the weak version of this proposal, the IMF would lend at preferential interest rates to countries that added CACs to their loan contracts. But when a country is in the throes of a crisis, the interest charge on IMF money is not the first thing on its mind. This approach would also have to surmount legal obstacles. Article V.8(d) of the IMF’s Articles of Agreement guarantees comparability of treatment; this obliges the institution to offer all member countries access to individual facilities on comparable terms. It means, among other things, not discriminating in terms of interest charges.

VI. Will These Reforms Make the World a Significantly Safer Financial Place?

The fundamental question for participants in this debate is whether new procedures for resolving sovereign debt crises will significantly enhance the efficiency and stability of international financial markets and the growth and stability of the developing countries that depend on those markets. Our view is that while these provisions will make a difference, they are only one among many needed improvements. The case for them is strongest if their addition to loan agreements is viewed as one of a number of interdependent changes in the international financial architecture, none of which is feasible in the absence of the others but which together promise to make the world a significantly safer financial place.

It is possible to point to Mexico and question whether a provision that results in such small changes in borrowing costs could really produce significant changes in countries’ ex ante behavior.50 Note, however, that the relevant comparison is not with Mexico’s borrowing cost but with its spread, compared with which the estimated differentials achieved by the new bonds are not negligible. Borrowers traditionally bargain hard for every basis point of investment banking fees, which are of roughly the same magnitude (25 to 50 basis points). Be that as it may, for other countries with a greater perceived probability of having to restructure, the spread differential would surely be larger. So, presumably, would be the incentives to adjust borrowing and lending behavior and hence the risk of future crises. To the extent that collective action clauses raise borrowing costs for such countries, they are, in effect, pricing the moral hazard made possible by the existence of asymmetric information. This fact raises the pressure on countries to increase transparency and strengthen domestic policies in ways that ultimately reduce country and systemic risk.51

Other skeptics (e.g., Mussa, 2002) question whether collective action clauses will make a difference ex post—that is, after crises erupt. They observe that they would have had little impact on most of the major debt crises of the 1990s. In Mexico in 1994—5, the problem was the difficulty of rolling over the tesobonos (the country’s domestic-law debt) and international credit lines to Mexican banks (private debt). In Thailand, Korea, and Indonesia, the problem was again with credits and loans extended to banks and corporations (private debt). In 1998, the GKOs on which the Russian government defaulted were domestic-law bonds. In contrast, Argentina is precisely the kind of case that might have played out differently in the presence of collective action clauses.

But the important question is not how many past debt crises would have developed differently had lenders and borrowers made greater use of collective action clauses—it is how many future debt crises will develop differently in their presence. On the one hand, borrowers and their regulators have learned from past crises about the special risks of short-term debt. It is unlikely that we will see more countries incurring large amounts of short-term, foreign-currency-indexed or denominated debt by issuing 90-day dollar-linked notes (a la Mexico in 1994) or allowing their banks to borrow 90 day money offshore in dollars (as in Thailand and the Republic of Korea). Borrowers and regulators better appreciate the special risks of short-term funding and the advantages of medium-and long-term bonds. And, if borrowing will, in fact, increasingly take the form of bonds, then collective action clauses are likely to be more relevant in the future than the past.52

Working in the other direction is the growing importance of private borrowing. When borrowing is done not by the sovereign but by private enterprises, national bankruptcy courts are available to reorganize unsustainable debts. Strategic behavior by rogue creditors can be restrained by a court-imposed standstill and a court-administered composition plan, complete if necessary with the power to cram down restructuring terms on holdouts. This does not eliminate the role for collective-action clauses. As in the case of nineteenth-century British railways, their presence widens the scope for the consenting adults to agree on the terms of the debt restructuring among themselves, rather than relying on the intervention of the courts. In emerging markets, where independent judiciaries are weak, this is especially desirable. In addition, when corporations have assets abroad as well as foreign debt issues, they face the danger that rogue creditors may resort to foreign courts to attach those assets, making agreement on restructuring terms correspondingly less attractive to other creditors. These arguments suggest that there is a case for private enterprises borrowing internationally to also use collective action clauses (as private entities borrowing in London already do), although the case for these provisions is less urgent than in the case of sovereigns, for whom the option of court-led reorganization is not available.

One way of viewing the resulting dynamics is that collective action clauses will first become more important for future crises, as governments fund themselves at longer tenors, after which their importance will recede, once sovereign borrowing gives way to private borrowing. There has already been a move away from short-term funding in response to the Mexican and Asian crises. The move away from sovereign borrowing will presumably take longer; it presupposes further progress in privatization, improvements in corporate governance, and measures to strengthen domestic bankruptcy and insolvency procedures.

But the case for collective action clauses is strongest if they are viewed as one of several interdependent changes in the international financial system, which together promise to make the world a safer financial place but none of which is feasible in the absence of the others. Collective action clauses could reduce the likelihood that the IMF and its principal shareholders will feel compelled to extend financial assistance to countries whose debts are already borderline unsustainable, since the consequent restructuring would not be so disruptive in the presence of these contractual provisions. Absent the expectation of IMF bailouts, borrowers and lenders will exercise more discipline, reducing crisis risk and enhancing systemic stability. The IMF has introduced greater procedural clarity in its policy on exceptional access, but the official community also recognizes that such limits will be credible and time consistent only if there exist other ways of dealing with impending defaults. Similarly, enhancing the independence and forthrightness of IMF surveillance, as suggested by Balls (2003), may raise the risk that a country will have to restructure by calling attention to its weaknesses; again, it can be argued that more forthright surveillance is feasible only if mechanisms are in place to smooth the consequent restructurings.

The point is that the international financial architecture is made of up a set of interlocking parts. It is hard to change one without also changing the others. Thus, a concerted effort to change the provisions of loan agreements may hasten progress on other, complementary changes, which will then work together to make the world a safer financial place.

It is, of course, on these other changes that reform should focus. Crisis prevention should be at least as high a priority as crisis resolution. The debate over new procedures for crisis resolution should not be allowed to crowd out the international financial institutions’ fundamental work on transparency standards (standards for fiscal, monetary, and financial policy transparency), financial sector standards (banking supervision, securities, insurance, and payment systems), and corporate sector standards (corporate governance, accounting, auditing, insolvency, and creditor rights). It should not be allowed to drain energy from the effort to develop domestic financial markets and thereby attenuate the “double-mismatch” problem.

But it would be a mistake to think that the job of fine-tuning contractual provisions and supplementing them with institutional supports is complete. We have yet to see whether a significant number of speculative credits follow investment-grade countries like Mexico and the Republic of Korea in adopting these provisions. Even if they do, countries with low credit ratings may be tempted to require very high qualified majorities and retain other provisions that stymie collective action and encourage holdout litigation. There may yet be a need for regulatory changes to encourage more countries to adopt workable majority restructuring provisions. Nor is there a consensus on either the need for super-collective-action clauses and the feasibility of getting them into the market or the adequacy of informal substitutes like a standing committee of bondholders and a code of creditor conduct. Progress on these issues should be the next step in the effort to strengthen mechanisms for crisis resolution.


Theoretical Details

This appendix provides some of the analysis relied upon in Section C. The characteristics of debtor willingness to pay are based on the infinite-horizon stochastic model of sovereign debt with self-enforcement constraints imposed on both debtors and creditors by Kletzer and Wright (2000). This model implies that the gains from future smoothing of the debtor’s consumption provide incentives for current repayment. We can infer from the analysis in Kletzer and Wright that these gains will vary with the debtor’s taste and technology, although these comparative dynamics are not considered in that model.

Consistent with the Kletzer and Wright model, we can write the debtor’s objective in reduced form as a function of the present value of payments, Πt, and the debtor’s surplus in the continuation of the international borrowing relationship that follows repayment, wt. The debtor government’s objective is given by u(–Πt,wt), where θ is the debtor’s type and u(–Πt,wt) is increasing, strictly concave and twice continuously differentiable in –Πt, and wt Sovereign immunity is expressed here by requiring that u(–Πt,wt) has a lower bound (the reservation utility in agency models) which yields the true willingness to pay for the debt restructuring model. The true willingness to pay is denoted Vt= V(yt, θ).

Creditors do not know θ but do know that its support is the interval, [θmin, θmax]. For simplicity, we assume that to the knowledge of creditors θ is distributed uniformly over this interval, and we parameterized θ so that θmin = 0. Relying on the general results of the agency literature (for example, see Mas-Colell, Whinston, and Green (1995), chapter 14), creditors can offer an implicit contract to debtor in which wt increases with Πt, until Πt = Dt. The debtor’s surplus, Vt – Πt, is zero for θ = 0 and increases with θ. Because the support and distribution for θ are continuous, we also assume that u1(Πt,wt;θ)/u2(Πt,wt;θ) is increasing in θ These assumptions allow us to assert that Vt(yt, θ)–Πt(yt) and Πt (yt,θ) are increasing with θ until θ is large enough so that Πt, = Dt which defines θ^. For θθ^, the debtor pays Dt. We also define θ¯ by the relationship, V(yt,θ¯)=Dt;θ¯θ^.

Under collective action clauses, the returns to bondholders are given by


For bonds with unanimous action clauses, repayment is delayed and the debtor suffers deadweight costs of delay equal to q per period. The debtor’s gain from default is given by


Because we assume that delays are socially costly, we let this be negative for all θθ¯ without loss of generality. For θ<θ¯, the debtor always defaults because V(yt)<Dt. The return to bondholders under UACs is given by


The difference between the returns to bonds with CACs and bonds with UACs is


The first term in equation (Al) is the cost of restructuring delay for creditors for bonds with UACs, while the second term is the cost of higher default probabilities for bonds with CACs in the presence of moral hazard. This difference is ambiguous as noted in the text. Without further assumptions, deteriorating fundamentals can either increase or decrease the return difference between bonds issued with and without CACs, depending upon the distribution of private information and expected cost of delay under UACs. If creditor ignorance about the debtor’s true willingness to pay is represented by assuming that Vt is uniformly distributed over its support for each yt then the effect of deteriorating fundamentals on the return difference shown in equation (Al) can be signed. This implies that θ is distributed uniformly and V(yt, θ) is linear in θ. Reducing yt increases the difference between θ^ and θ¯, and the probability of default for bonds with CACs rises faster than the probability of default for bonds with UACs if θ¯<θmax.

The results summarized in the text for the effects of deteriorating fundamentals can be demonstrated for a simple case. We assume that V(yt,θ) - Π(yt) is a linear function of θ from θmim = 0 to θ^ for each yt.53 V(yt θ), the true willingness to pay, can be written as


and the reported willingness to pay as


so that the debtor’s surplus due to information asymmetry is


where A > B. The two critical values for θ are given by

θ^=DtV(yt,0)B  andθ¯=DtV(yt,0)A.

The difference in the expected returns to creditors for bonds with UACs and bonds with CACs, equation (Al), becomes


if both critical values, θ^ and θ¯, are less than θmax. This is negative for all 0PP+r1 unless A> 2B in which case it can be positive for small values of PP+r. Since V(yt) is increasing in yt, the expected return differential decreases as yt increases or Dt decreases. This implies that the interest rate spread for bonds with CACs over bonds with UACs rises as fundamentals deteriorate (yt decreases or Dt increases).

For θ^>θmax, equation (7) becomes


In this expression, the expected return to bonds with UACs falls relative to bonds with CACs as θ¯ rises to θmax (the probability of default in the absence on asymmetric information goes to one) for PP+r<1. It is negative when θ¯ equals θmax for PP+r<1 (and zero when delay is costless,PP+r=1). This implies that, given yt, the interest rate spread for bonds with CACs over bonds with UACs decreases when the probability of default on bonds with UACs is sufficiently high. It also implies that this can be true when yt is also uncertain (creditors and debtors are equally informed of yt,).

The effects of changing the support of the distribution of θ can be shown using equation (Al). A shrink in the dispersion of creditor uncertainty about underlying debtor characteristics can be represented as a mean-preserving shrink in the interval [θmin,θmax] given a uniform distribution (allow θmin to differ from zero). For such a shrink, the probabilities of default on bonds with CACs and bonds with UACs become closer in value and the spread on bonds with CACs over UACs shrinks. As θ becomes known with certainty by creditors, the interest differential for bonds with CACs over bonds with UACs becomes negative because delay is costly for bondholders. Therefore, a reduction in the importance of moral hazard reduces the interest spread for bonds issued with CACs and can turn it negative.


Data Sources and Construction of Variables

A. Data on Secondary Market Spreads

Data on secondary market spreads were obtained from the Merrill Lynch Global Index System (MLGIS) provided by Bloomberg. The data correspond to bonds issued by sovereign entities that are included in the Merrill Lynch market indices GOLQ and IPOO. The emerging market countries considered are: Argentina, Bahrain, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Croatia, Dominican Republic, Egypt, El Salvador, Guatemala, Hong Kong, Hungary, Indonesia, Israel, Jamaica, Kazakhstan, Korea, Lebanon, Lithuania, Malaysia, Mexico, Panama, Peru, Philippines, Poland, Qatar, Romania, Russia, Slovakia, Slovenia, South Africa, Thailand, Tunisia, Turkey, Ukraine, Uruguay, and Venezuela. This empirical study considers four different dates corresponding to the two most recent peaks and the two most recent troughs in JP Morgan’s Emerging Market Bond Index Plus (EMBI+) as well as a last fifth date used by Richards and Gugiatti in their similar analysis. Starting with the first trough in EMBI+, the dates are as follows: September 6, 2000, November 2,2001, April 12, 2002, September 30, 2002, and January 31, 2003. The Bloomberg data source also provides the information for each bond on its current ratings, effective modified duration, currency, and par amount. Bloomberg does not appear to provide the governing law; as such, each bond was individually identified and matched to obtain its corresponding governing law from Bondware.

The secondary market spread used is an option-adjusted spread (OAS) as provided by MLGIS. This OAS is the spread relative to the off-the-run U.S. Treasury curve. This spread equates the theoretical present value of the bonds’ cash flows to their current market price. The value of this OAS is that it utilizes a whole yield curve as a benchmark instead of a specific risk-free asset and it allows investors to directly compare fixed income instruments with similar characteristics, but trade at significantly different yields because of embedded options. OAS can be thought of as the compensation an investor receives for assuming risks (e.g., default risk, liquidity premium, etc…), net of the cost of any embedded options.

B. Primary Bond Spreads

The data on primary spreads was obtained from Bondware and covers the period 1991 to 2000. Bondware provides: (a) launch spreads over risk free rates charged for bonds of comparable maturity and issued in the same currency (spreads are measured in basis points, where one basis point is one-hundredth of a percentage point) (b) the amount of the issue (millions of US$); (c) the maturity in years; (d) whether the borrower was a sovereign, other public sector entity, or private debtor; (e) the governing law under which the bond contract was written; (f) currency of issue; (g) borrower’s industrial sector: manufacturing, financial services, utility or infrastructure, other services, or government (where government, in this case, refers to subsovereign entities and central banks, which could not be classified in the other four industrial sectors); (h) the country and regional identity of the borrower; (i) the nationality of the book runner; (j) the market in which the bond was issued.

In addition, the regressions using primary spreads also used as right-hand side regressor a number of country characteristics and global variables.

Country Characteristics

article image
article image
Sources:International Monetary Fund: World Economic Outlook (WEO) and International Financial Statistics (IFS).World Bank: World Debt Tables (WDT) and Global Development Finance (GDF).Bank of International Settlements: The Maturity, Sectoral and Nationality Distribution of International Bank Lending.Credit ratings were obtained from Institutional Investor’s Country Credit Ratings. Political Risk Index was obtained from the International Country Risk Guide.Missing data for some countries was completed using the U.S. State Department’s Annual Country Reports on Economic Policy and Trade Practices (which are available on the internet from“Global” Variables:The website for the US interest rates and industrial production data is: rates; emerging market spreads; Japanese, German, and U.K. interest rates were obtained from Bloomberg.
Table 1.

Emerging Markets Sovereign Bond Issuances, by Jurisdiction1

article image
Source: IMF (2003).

. Number of issuance is in number. Volume of issuance is in billions of U.S. dollars.

English and Japanese laws.

. German and New York laws. However, the Egyptian issuance of US$1,500 million out of New York in June 2001 contains CACs and has thus been reclassified.

Data for 2003-Q1 are as of February 20,2003.

Table 2.

Panel Regressions of Secondary-Market Spreads

article image
Source: Authors’ calculations based on sources listed in Appendix II.Note: z statistic in parentheses; based on robust standard errors for Random Effects models. * significant at 5 percent; **significant at 1 percent.
Table 3.

Influence of Market Conditions on Bonds with CACs

article image
Source: Authors’ calculations based on sources listed in Appendix II.Note: z statistics in parentheses; based on robust standard errors for Random Effects models. * significant at 5 percent; ** significant at 1 percent.
Table 4.

Implications of Existing Composition of Bonds for Changeover

article image
Source: Authors’ calculations based on sources listed in Appendix II.

In addition, dummy variables were included for different currencies, fixed rate bonds, guarantees, put and call options, and offshore issuance.

Table 5.

The Aggregation Effect: All Bonds and Differentiated by Credit Quality

article image
Source: Authors’ calculations based on sources listed in Appendix II.

In addition, dummy variables were included for different currencies, fixed rate bonds, guarantees, put and call options, and offshore issuance.

Table 6.

The Aggregation Effect: Differentiated by Governing Laws

article image
Source: Authors’ calculations based on sources listed in Appendix II.

In addition, dummy variables were included for different currencies, fixed rate bonds, guarantees, put and call options, and offshore issuance.


  • Allen, Franklin and Douglas Gale, 1994, “Financial Innovation and Risk Sharing” (Cambridge: Cambridge University Press).

  • Argentina, Government of, 2003, “Description of the Process of Sovereign Debt Restructuring,” (unpublished: Buenos Aires, (February).

    • Search Google Scholar
    • Export Citation
  • Balls, Edward, 2003, “Preventing Financial Crises: The Case for Independent IMF Surveillance” (unpublished manuscript: Institute of International Finance,

    • Search Google Scholar
    • Export Citation
  • Bank of Canada, and Bank of England, 2001, “Resolution of International Financial Crises” (unpublished: Bank of Canada and Bank of England).

    • Search Google Scholar
    • Export Citation
  • Bank of France, 2003, “Towards a Code of Good Conduct on Sovereign Debt Re-Negotiation” (unpublished: Banque de France).

  • Becker, Torbjorn, Anthony Richards, and Yungong Thaicharoen, 2001, “Bond Restructuring and Moral Hazard: Are Collective Action Clauses Costly?IMF Working Paper, 01/92 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Bratton, William W., and G. Mitu Gulati, 2002, “Sovereign Debt Restructuring and the Best Interest of the Creditors,” Sloan Program Working Paper (forthcoming in January at, Georgetown University Law Center).

    • Search Google Scholar
    • Export Citation
  • Bulow, Jeremy, and Kenneth Rogoff, 1989, “A Constant Recontracting Model of Sovereign Debt,” Journal of Political Economy, Vol. 97, pp. 155178.

    • Search Google Scholar
    • Export Citation
  • Council on Foreign Relations, 1999, “Safeguarding Prosperity in a Global Financial System: The Future International Financial Architecture” (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation
  • Douglas, William O. 1940, “Democracy and Finance” (New Haven: Yale University Press).

  • Eaton, Jonathan, and Mark Gersovitz, 1981, “Debt with Potential Repudiation,” Review of Economic Studies, Vol. 48, pp. 289309.

  • Eichengreen, Barry, 1996, “Financing Infrastructure Investment in Developing Countries: Lessons from the Railway Age,” in Infrastructure Delivery: Private Initiative and the Public Good, ed. by Ashoka Mody (Washington: The World Bank).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, 2002, “Financial Crises and What to Do About Them” (Oxford: Oxford University Press).

  • Eichengreen, and Ashoka Mody, 2000a, “Would Collective Action Clauses Raise Borrowing Costs?NBER Working Paper No. 7458. (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, and Ashoka Mody, 2000b, “Would Collective Action Clauses Raise Borrowing Costs? An Update and Extension,” World Bank Research Paper No. 2363 (Washington: the World Bank).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, and Ashoka Mody, 2003, “Is Aggregation a Problem for Sovereign Debt Restructuring?American Economic Review, Papers and Proceedings (forthcoming).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, and Richard Portes, 1995, “Crisis? What Crisis? Orderly Workouts for Sovereign Debtors” (London: Centre for Economic Policy Research).

    • Search Google Scholar
    • Export Citation
  • Emerging Market Creditors Association, 2002, “Model Covenants for New Sovereign Debt Issues” (unpublished manuscript, EMCA, May 3),

    • Search Google Scholar
    • Export Citation
  • Fischer, Stanley, 2002, “Financial Crises and Reform of the International Financial System,” NBER Working Paper Noo. 9297 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Folkerts-Landau, David, 1999, “Testimony Before a Hearing of the Committee on Banking and Financial Services,” 20 May,

    • Search Google Scholar
    • Export Citation
  • Gavin, Michael, 2003, “On the Sovereign Debt Restructuring Debate,” UBS Warburg Emerging Markets Daily (March), pp. 57.

  • Goldstein, Morris, 2003, “Debt Sustainability, Brazil, and the IMF,” Working Paper No. 03-1 (Washington: Institute of International Economics).

    • Search Google Scholar
    • Export Citation
  • Group of Seven, 1999, “Strengthening the International Financial Architecture,” Report of the G-7 Finance Ministers to Heads of State and Government, Cologne, Germany.

    • Search Google Scholar
    • Export Citation
  • Group of Ten, 2002, “Report of the G-10 Working Group on Contractual Clauses” (September),

  • Group of Thirty, 2002, “Key Issues in Sovereign Debt Restructurmg” (Washington: Group of Thirty).

  • Gugiatti, Mark, and Anthony Richards, 2003, “Do Collective Action Clauses Influence Bond Yields? New Evidence from Emerging Markets,” Research Discussion Paper 2003-02 (Sydney: Reserve Bank of Australia).

    • Search Google Scholar
    • Export Citation
  • Hubbard, R. Glenn, 2002, “Enhancing Sovereign Debt Restructuring,” Remarks of R. Glenn Hubbard, Chairman of the Council of Economic Advisors, to the Conference on the IIF’s Sovereign Debt Proposal (Washington: American Enterprise Institute).

    • Search Google Scholar
    • Export Citation
  • Hurlock, James, The Way Ahead for Sovereign Debt,” Euromoney (August), pp. 7879.

  • Institute of International Finance, 1996, “Resolving Sovereign Financial Crises” (Washington: Institute of International Finance).

  • Institute of International Finance, 2002, “Crisis Prevention and Resolution in Emerging Markets,” Report of the Special Committee on Crisis Prevention and Resolution (Washington: Institute of International Finance).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 1999, “IMF Policy on Lending into Arrears to Private Creditors” (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2002, “Collective Action Clauses in Sovereign Bond Contracts-Encouraging Greater Use” (Washington: International Monetary Fund at

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2003, “Collective Action Clauses: Recent Developments and Issues” (Washington: International Monetary Fund at

    • Search Google Scholar
    • Export Citation
  • J.P. Morgan, 2002, “Two Step Sovereign Debt Restructuring” (unpublished; J.P. Morgan).

  • Kenen, Peter, 2001, The International Financial Architecture: What’s New? What’s Missing? (Washington, D.C: Institute for International Economics).

    • Search Google Scholar
    • Export Citation
  • Kletzer, Kenneth, 2003, “Sovereign Bond Restructuring: Collective Action Clauses and Official Crisis Intervention.” Paper for Bank of England conference, “The Role of the Official and Private Sectors in Resolving International Financial Crises,” July 23 and 24, 2002, revised.

    • Search Google Scholar
    • Export Citation
  • Kletzer, and B. D. Wright, 2000, “Sovereign Debt as Intertemporal Barter,” American Economic Review, Vol. 90, pp. 621639.

  • Kroszer, Randall, 2003, American Economic Review, Papers and Proceedings (May).

  • Krueger, Anne, 2002, “A New Approach to Sovereign Debt Restructuring” (Washington: International Monetary Fund).

  • Macmillan, Rory, 1995, “Towards a Sovereign Debt Work-Out System,” Northwestern Journal of International Law and Business, Vol. 16, pp. 5775.

    • Search Google Scholar
    • Export Citation
  • Mas-Colell, Andreu, Michael Whinston, and Jerry R. Green, 1995, “Microeconomic Theory” (New York: Oxford University Press).

  • Meltzer Commission [International Financial Institution Advisory Commission], 2000, Report, (Washington: GPO).

  • Miller, Marcus, 2003, “Sovereign Debt Restructuring: Where Things Stand Now” (unpublished manuscript, Institute for International Economics).

    • Search Google Scholar
    • Export Citation
  • Mody, Ashoka, 2003, “Crisis Resolution: Lessons From the Bond Market,” XVIII Meeting of the Latin American Network of Central Banks and Finance Ministries, Inter-American Development Bank, Washington DC, 9 April 2003

    • Search Google Scholar
    • Export Citation
  • Mussa, Michael, 2002, “Latin American Economic Crisis,” Testimony to the Senate Committee on Banking, Housing and Urban Affairs, 14 October,

    • Search Google Scholar
    • Export Citation
  • Oechsli, Christopher, 1981, “Procedural Guidelines for Renegotiation LDC Debts: An Analogy to Chapter 11 of the U.S. Bankruptcy Reform Act,” Virginia Journal of International Law, Vol. 21, pp. 305341.

    • Search Google Scholar
    • Export Citation
  • Ohlin, Goran, 1976, “Debts, Development and Default,” in A World Divided: The Less Developed Countries in the International Economy, ed. by Gerald K. Helleiner (Cambridge: Cambridge University Press).

    • Search Google Scholar
    • Export Citation
  • Raffer, Kunibert, 1990, “Applying Chapter 9 Insolvency to International Debts: An Economically Efficient Solution with a Human Face,” World Development, Vol. 18, pp. 301311.

    • Search Google Scholar
    • Export Citation
  • Rhodes, William, and others, 1999, “Global Private Finance Leaders Stress the Importance of Voluntary Approaches to Crisis Resolution in Emerging Markets,” press release, (Washington: Institute of International Finance).

    • Search Google Scholar
    • Export Citation
  • Rogoff, Kenneth, and Jeromin Zettelmeyer, 2002, “Bankruptcy Procedures for Sovereigns: A History of Ideas, 1976-2001,” IMF Staff Papers, Vol. 49, pp. 470507.

    • Search Google Scholar
    • Export Citation
  • Roubini, Nouriel, and Brad Setser, 2003, “Improving the Sovereign Debt Restructuring Process: Problems in Restructuring, Proposed Solutions, and a Roadmap for Reform” (unpublished manuscript: New York University and Council on Foreign Relations).

    • Search Google Scholar
    • Export Citation
  • Skeel, David A., Jr. (2002), “Can Majority Voting Provisions Do It All?” (unpublished manuscript, University of Pennsylvania).

  • Sturzenegger, Federico (2002), “Default Episodes of the ‘90s: Factbook and Preliminary Lessons” (unpublished manuscript, Universidad Torcuato Di Telia, Buenos Aires (June)).

    • Search Google Scholar
    • Export Citation
  • Taylor, John, 2002a, “Sovereign Debt Restructuring: A U.S. Perspective,” Treasury News, (2 April).

    • Search Google Scholar
    • Export Citation
  • Taylor, John, 2002b, “Using Clauses to Reform the Process for Sovereign Debt Workouts: Progress and Next Steps,” Office of Public Affairs, U.S. Treasury, PO-3672 (5 December).

    • Search Google Scholar
    • Export Citation
  • UBS Warburg, 2003, “Uruguay Swap: I See the Stick, Where is My Carrot?Global Economic and Strategy Research, Latin America Fixed Income Strategy Note (23 April).

    • Search Google Scholar
    • Export Citation

The authors’ affiliations are the University of California at Berkeley; the University of California at Santa Cruz; and the IMF, respectively. This paper was presented, in somewhat different form, at the Brookings Institution’s Trade Forum conference held in Washington, DC on May 15-16, 2003 and will be published by Brookings Press in the Brookings Trade Forum 2003. The authors are grateful to Adrian de la Garza for valuable research assistance and to Mark Allen, Axel Bertuch-Samuels, Robin Brooks, Eduardo Borensztein, Ricardo Caballero, Susan Collins, Monica de Bolle, Janet Kong, Jens Nystedt, Effie Psalida, Carmen Reinhart, and Tony Richards for comments.


Collective action clauses specify procedures for selecting a bondholder’s representative and enumerate his responsibilities, include majority enforcement clauses in which the litigation decision must be made by a requisite fraction of the bondholders (say, 25 per cent), and require that all funds thereby recovered be distributed in proportion to the principal amount. They specify the share of the bondholders whose vote suffices to amend payment terms like the timing and amount of principal and interest. Changes endorsed by the specified majority are then binding on all bondholders. Traditionally, bonds subject to English law include collective action clauses, while bonds subject to New York law do not. (Bonds governed by New York law normally contain majority enforcement provisions but not majority restructuring provisions.) Thus, the main challenge for the contractual approach is to introduce majority restructuring clauses into bonds issued under New York law. The SDRM would involve an international treaty obligation empowering a qualified majority of all creditors to agree on the binding terms of a structuring offer, to assign seniority to new money, and to create a dispute resolution forum to allocate voting rights and tabulate the results (see Krueger, 2002).


There were a few instances of governments issuing bonds in New York with collective action clauses prior to Mexico, such as Lebanon and Qatar in 2000 and Egypt in 2001. But these issues attracted little notice because they were small. In addition, they were not included in the Emerging Market Bond Index (EMBI), since as private placements (issued under Rule 144A, which provides a safe harbor from registration under the U.S. Securities Act of 1933) they did not have a liquid secondary market. (Unregistered securities placed privately can be resold only to “qualified institutional buyers.”) This last point is important, since these bonds have been used by other authors to draw conclusions about how CACs are priced in secondary markets. We return to this below.


Which remains the preoccupation of recent studies such as Roubini and Setser (2003).


The difference probably reflects that the first set of studies was already underway before the Mexican crisis and that they emerged from the legal community.


The lending-into-arrears policy was conceived in the late 1980s as an element of resolving the debt crisis that erupted in 1982 (IMF, 1999). It has since undergone a number of modifications, and in this context was broadened to encompass arrears on international sovereign bonds and other nonbank private credits. The important modification in the present context was that in 1999, which specified that lending into sovereign arrears to external private creditors can be granted in circumstances in which prompt Fund support is essential for successful implementation of the country’s adjustment program, and the country is pursuing appropriate policies and is making a good-faith effort to reach a collaborative agreement with its creditors.


This idea that the adoption of firm lending limits will encourage a more friendly reception for proposals for restructuring-friendly contractual provisions remains a theme in the subsequent writings of Morris Goldstein (2003), who held the pen for the Council’s report.


In September the G-10 (2000) also issued a report describing a set of model clauses. Roubini and Setser (2003, p. 9) summarize how the G-10 and EMCA et al draft clauses differ from one another.


The results are also consistent with the Bulow and Rogoff (1989) model.


Their return is e-rΔt(1– x) Πt when refusing to pay the holdouts is an equilibrium action.


This equilibrium is a perfect Nash equilibrium. There are other, asymmetric, equilibria. Notably, any given set of holders of exactly St share of the bonds can always hold out while the remaining bondholders always agree to renegotiate.


The presence of cross-default clauses in many sovereign bonds suggests that the perceived difference in default probabilities may be slight. But it is sometimes possible for the issuer to obtain a waiver of default before the event occurs. And, especially with sovereigns, just because one bond is declared as in default, it need not follow that the holders of all other issues also accelerate.


This is less than Πt, and can be calculated using the binomial formula.


Consistent with this intuition, in the U.K., the majority needed to restructure corporate debt in the event of bankruptcy is lower for an adjourned meeting of bondholder representatives than for the first meeting.


Again, this prediction is borne out by the empirical analysis below.


The second set of authors also provides a limited analysis of the secondary market. In discussing prior results, we concentrate on the extension of their analysis of the secondary market by Richards and Gugiatti (2003) below.


The country had a B2/B+ rating.


This follows the EMCA model clauses.


Hence, an 85 percent majority provision, while helpful for collective action, may not go far enough, especially during the period when investors are left holding a highly uncertain asset. Also, provisions for so-called deceleration clauses maintain incentives for holdouts to litigate. In addition, sentiment favored emerging bond markets at the time of Brazil’s issue; capital flows were relatively abundant. As noted, in previous work (and new results presented below), we find that the effect of collective action clauses on spreads is limited under such conditions. Market commentary suggests that the Brazilian authorities were quite adept at timing their bond issue to capitalize on this fact (see J.P. Morgan, Emerging Markets Today, April 11, 2003).


Some observers detect no difference from comparable Brazilian bonds without CACs. Others (e.g., Credit Suisse/First Boston Emerging Markets Sovereign Strategy Daily, 30 April, p. 2) detect a spread penalty of 10-15 basis points, which is consistent with our econometric results.


Presumably this date was selected on the grounds that it provided the most recently available data at the time of writing. We argue below—and provide evidence—that dates are important, since market sentiment, and with it investors’ appetites for bonds with collection action provisions, varies over time.


Thus, a higher number implies a worse rating, which is the opposite of the Institutional Investor scale utilized in our analysis of launch spreads (below). The rating used here is a Merrill Lynch composite based on both Moody’s and Standard & Poor’s. When the rating is available from both agencies for that particular bond, a simple arithmetic average is taken. When there is information from only one agency, the “composite” is the rating of that one agency. When no information is available on the bond itself, the country’s foreign-currency long-term sovereign rating is used. Data on credit ratings is from Standard & Poor’s and Moody’s, while information on governing laws is mainly from Bondware.


This effectively leads to the elimination of Japanese-law bonds.


The same signs and insignificant coefficients were also evident in the more limited analysis of secondary market spreads in Becker, Richards and Thaicharoen (2000).


This is a slightly different measure of the spread than used by Gugiatti and Richards. Conversations with market participants convince us that this is the measure of secondary market spreads relied on by the markets and thus the measure most appropriate for the current analysis. Fortunately, this difference in the definition of the dependent variable does not appear to produce any significant differences from the results obtained by Gugiatti and Richards. We also use another measure of spread used by market participants (the spread based off the swap curve) and again obtain virtually identical results.


As noted in footnote 3 above.


Again, see footnote 3 above. In addition, Lebanon has had a large captive market for its bonds, the bulk of which are bought by domestic commercial banks and the central bank. Hence, while Lebanon receives a relative low credit rating from Moody’s and Standard & Poor’s, its bonds trade at spreads comparable to those of higher-rated issuers.


The results are even stronger when we drop the observations for these three issuers simultaneously. We then find statistically significant patterns (with negative coefficients for the CACs dummy and positive coefficients for the interaction term) for most of the individual cross sections. We suspect that the special nature of the market for their issues creates pricing idiosyncrasies that weaken the statistical results when Egypt, Lebanon and Qatar are included. But rather than add or remove countries or add or remove dummy variables for issuers of a particular nationality, we prefer to hang our hats on the full-sample results.


There is some evidence that German law bonds are priced differently (see Eichengreen and Mody 2000a,b).


Because in some cases we have data for the same bonds at successive points in time, as a form of further sensitivity analysis we added dummy variables for each of these bonds. This allows us to control for individual bond effects (for idiosyncracies of bonds such as their liquidity and special provisions that Gugiatti and Richards do not include among their explanatory variables). Random effects estimates of this specification leave the key results unchanged. Fixed effects estimation is not appropriate, for this would amount to eliminating the effects of other contractual provisions of the bonds in question (including whether or not they include collective action clauses).


Even if they do not, the theoretical analysis suggests that an attempt to lower costs through the use of unanimous action clauses may add subsequent costs that redound unfavorably on the issuer. Thus, why such borrowers remain reluctant to make use of collective action clauses is an interesting issue; we address it below.


This empirical analysis is drawn from Eichengreen and Mody (2003). In principal, this is the universe of new sovereign issues in the period since the developing-country bond market started up again in the wake of the Brady Plan, although in practice the number of observations is slightly smaller than that universe, reflecting problems of missing data.


The definition of the latter depends on the currency in which the emerging-market bond is issued; it is a U.S. treasury bond for U.S. dollar-denominated bonds, a U.K. government bond for sterling-denominated issues, a Japanese government bond for yen-denominated issues, and so forth.


Note that this metric is the opposite of that used by Gugiatti and Richards (and adopted in our preceding analysis of secondary spreads to enhance comparability with their analysis).


The precise cutoffs used to partition the data set by credit quality are imported them from our previous work (Eichengreen and Mody 2000a, 2000b). Our initial idea was to partition the data into credit quality quartiles: 0-25, 26-50, 51-75, and 76-100. The fact that there were no emerging markets with ratings above 75 led us to collapse the last two quartiles into one. And the fact that few countries with very low ratings, in the 0-25 category, were actually able to issue bonds led us to shift first cutoff from 26 to 31 or 35, as here, bringing the number of observations in the first two categories closer to equality. Experimentation with alternative cutoffs does not suggest that the current results are particularly sensitive to this partition.


We calculate this by cumulating new issues and removing earlier issues as they are retired.


These estimates correct for sample selectivity, reflecting the fact that not all potential borrowers are in the market at all times, by estimating a two-equation system of the decision to borrow and the spread, using maximum likelihood. Reassuringly, equations for the spread estimated by ordinary least squares are essentially identical for present purposes.


In addition, it can be argued that having a large number of bond issues is something to strive for since it permits emerging markets to develop benchmark yield curves in all three major funding currencies. Again, however, this argument is likely to apply only (or mainly) to high rated issuers.


As explained in Section C, when there is uncertain about future fundamentals and moral hazard, the value of the debt can rise with a swap of bonds with UACs for bonds with super- CACs, if the probability of default is high (but default is not necessarily certain). In these circumstances, the interest rate on bonds with super-CACs should be lower than that on bonds with UACs. The welfare benefits would include the reduction in the probability of market disruptions caused by prolonged but inconclusive restructuring negotiations.


Uruguay’s exchange offer was special in a number of respects and differs from a proposal advanced by J.P. Morgan (2002) in important ways. The country did not suspend payments prior to initiating the first step of the exchange, as anticipated by the authors of the J.P. Morgan formula. And Uruguay’s super-collective action clauses specified high overall and bond-specific thresholds, both of which must be satisfied in order for the super-collective action clauses to apply.


In contrast, the Institute of International Finance has proposed a code of conduct that mainly enumerates requirements for the debtor and says little about the need for credible commitments on the part of the creditors.


Debtors evidently appreciate this need, although they place the emphasis on the uses of case-specific committees, as opposed to a standing forum. Thus, Argentina, in an early 2003 communication with the bondholders, suggested that as it proceeds with meetings and consultations, it would work with its creditors to put together a number of “coordination groups” to design the ultimate restructuring proposal. Membership in the group would depend on ability of a member to speak for a group of investors, willingness to abide by confidentiality, and “the contribution that such a potential member may offer to a constructive dialogue.” (Government of Argentina 2003).


The simplest explanation would of course be that neither debtors nor creditors view such innovation as desirable. Creditors fear that it would erode their rights, while debtors fear that it would raise their borrowing costs. But as we saw in the previous section, there is little support for this in the data.


A case can be made that the East Asian trading companies of earlier centuries anticipated this financial form, but these equity partnerships were typically wound up following completion of the voyage for which they had been formed. Consequently, inefficient liquidation was not an issue.


Douglas advocated this view in a series of articles and books; see for example Douglas (1940).


Even as a matter of historical interest, its relevance can be questioned, given that 30 per cent of the bonds already in the market (those subject to UK law) already include the relevant provisions. Debtors and creditors can reference these loans (as do researchers) when they want to price similar instruments. Similarly, the large amounts of domestic debt issued by the Russian government but held and traded by individual and institutional investors in the United States include collective action clauses, providing another reference point for market participants wishing to resolve product uncertainty.


In any case, such a market did not have to be created from scratch, since markets already exist in UK-law bonds and domestic-law Russian bonds.


We turn to this question in the next section.


To be clear, ex ante here means prior to any default and subsequent restructuring.


Thus, we are not allied with either side in the debate between “theories of too much” and “theories of too little” capital flows. To be clear, we are not arguing that capital flows are either too large or too small, and that the more widespread use of CACs would offset the associated distortion. Our argument rather is that credits of different types are mispriced relative to one another once one takes the systemic externality into account, and that changes in institutional arrangements and regulations that raise the relative cost of borrowing for risky credits, thereby going some way toward internalizing this externality, are a step in the direction of greater efficiency.


That bonds are involved in the cases we are now seeing—Argentina, for example—is consistent with this view.


Our assumptions also allow a piecewise linear solution for Vt –Πt (assuming risk neutral creditors with a constant common discount rate, as in Kletzer and Wright (2000)).


Cross-border bank claims in all currencies and local claims in non-local currencies of maturity up to and including one year.


Total consolidated cross-border claims in all currencies and local claims in non-local currencies.