CHAPTER 11 International Bailouts, Moral Hazard, and Conditionality

Alessandro Rebucci, and Ashoka Mody
Published Date:
April 2006
  • ShareShare
Show Summary Details
Olivier Jeanne and Jeromin Zettelmeyer1 

The large international bailouts of the 1990s have been criticized for generating moral hazard at the expense of the global taxpayer. We argue that this criticism is misleading because international bailouts create either no or very few costs to the international community. Instead, the problem is that bailouts may be used to facilitate bad domestic policies, thus creating moral hazard at the expense of domestic taxpayers. Ensuring that this does not happen may require a shift toward ex ante conditionality, in the sense that the availability and size of official crisis lending need to be conditional on government policies before the crisis.


The 1990s have seen international official crisis lending of unprecedented size (Table 1). Advocates justified these bailouts as a form of international financial safety net to deal with new risks arising from the international financial integration of emerging economies, most notably the rollover risk on foreign currency debt (Fischer, 1999; and Summers, 1999). At the same time, critics argued that international bailouts were instead a factor contributing to financial crises, because of the moral hazard that they create. There has been a lively debate on whether the practice of large international bailouts should be continued, discontinued, or reformed.2

Table 1IMF-Supported Crisis Packages of the 1990s: Financing Commitments and Disbursements1(in percentage of initial GDP)

Date of ApprovalTotalIMFTotalIMF
MexicoFeb. 199518.
ThailandAug. 199711.
IndonesiaNov. 199719.
Korea, Rep. ofDec. 199712.
BrazilDec. 19985.
Memorandum item

GDP in year of program approval (for Russia, we used 1995).

Russia had several consecutive IMF programs during the 1990s. The first large-scale loan was a stand-by arrangement approved in April of 1995.

Total disbursements in the 1990s.

GDP in year of program approval (for Russia, we used 1995).

Russia had several consecutive IMF programs during the 1990s. The first large-scale loan was a stand-by arrangement approved in April of 1995.

Total disbursements in the 1990s.

That financial safety nets create moral hazard, and possibly excessive moral hazard, seems fairly uncontroversial.3 However, much of the policy debate on how to reduce moral hazard seems to hinge on an erroneous view of how moral hazard is created by international bailouts, namely, the analogy between international bailouts and a standard insurance arrangement (for example, fire insurance or medical insurance). We dismiss this view on empirical grounds and offer an alternative interpretation of how moral hazard might be created by international crisis lending. We then draw the policy implications.4

In a standard insurance contract, the consequences of a bad outcome are mitigated by a state-contingent transfer. For example, if an insured person gets sick, medical insurance will cover part of her treatment costs. Moral hazard arises because the existence of this transfer makes her less anxious to avoid the bad state, at the expense of those who ultimately finance the transfer. One widespread view of international bailouts is based on the transposition of this argument to the international level: a country gets “sick” and receives a transfer, in the form of a subsidized loan from the international community. Moral hazard arises at the expense of those who supposedly pay for that transfer, namely, the global taxpayers. As noted in an April 23, 1998 Wall Street Journal editorial, “What really happens is that the U.S. ends up subsidizing the IMF’s growing practice of making large loans at low interest rates to very risky economies … and U.S. government money comes from taxpayers.”

This view is inconsistent with the available evidence on the repayment record on international official lending. The IMF’s large-scale crisis loans of the 1990s were made at a low interest rate, but in most cases they have been repaid or are in the process of being repaid, as Figure 1 shows. This is typical of the IMF repayment record with countries that have access to international financial markets. Thus, the international safety net does not involve any state-contingent transfers from the global taxpayer to crisis countries. Instead, it involves state-contingent loans, which do not embody any significant subsidy.

Figure 1Use of IMF Credit by Mexico, Thailand, Indonesia, the Republic of Korea, Brazil, and the Russian Federation, 1994–2001

(in billions of U.S. dollars)

The fact that the simple insurance analogy is wrong and moral hazard does not arise at the expense of the international taxpayer does not imply that international bailouts cannot create excessive moral hazard. However, it does imply that one needs to think harder about the mechanisms that might be causing it. If official crisis lending comes at no cost to the rest of the world, and is voluntary on the side of the crisis countries—which are free to reject the offer of assistance, and sometimes do—where is the problem? The problem, in our view, lies in the interaction between international bailouts and domestic policies, as pointed out by Calomiris (1998). The international financial safety net increases the scope for bad policies as well as good policies. International bailouts can help countries implement good policies, but they could also make the consequences of bad domestic policies much worse.

For example, consider the design of domestic financial safety nets in a financially integrated emerging market economy. A good policy, in this context, could be defined as a well-designed system of deposit insurance that would provide a limited guarantee on both domestic currency and foreign currency deposits in the domestic banking system. An example of bad policy could be an implicit government guarantee on the debt of the policymaker’s friends and relatives. The international safety net could be used to back one or the other. The fact that its operation costs nothing to the global taxpayer does not imply that it is going to be used to back good policies. It simply implies that the consequences of domestic policies, good or bad, are going to be borne by the domestic taxpayer. Excessive moral hazard might arise not because international bailouts involve an international transfer, but because they facilitate a domestic transfer from the domestic taxpayer to the borrowers that benefit from implicit guarantees (Jeanne and Zettelmeyer, 2001a).

A similar story can be told about sovereign debt crises. The expectation of full repayment even in an economic crisis could make international investors more lenient in granting new loans or rolling over existing loans to a government (“investor moral hazard”). Official crisis lending can provide governments with liquidity to implement such a bailout. However, if it does not carry a subsidy, and if the government maximizes domestic welfare—with due weight given to the domestic taxpayers who will pay for the bailout in the event of a crisis—this cannot generate excessive moral hazard. For moral hazard to be excessive, there must be a discrepancy between the policymaker’s objective and the domestic taxpayers’ long-term interests (Jeanne and Zettelmeyer, 2001b).

If the problem is that the international community creates moral hazard by playing the role of accomplice in bad domestic policies, what is the appropriate policy response? Increasing the interest rate on international loans, or collateralizing IMF lending, will clearly not do—such proposals miss the point. Instead, our preferred answer is to make the international community an accomplice in the implementation of good domestic policies. We argue in this paper that this requires shifting the weight of conditionality from traditional conditionality ex post, as an accompanying element of official lending after the crisis, to conditionality ex ante, in the sense that the availability and size of official lending need to be conditional on government policies before the crisis. The international community’s lending policies must create a link between domestic efforts at preventing crises ex ante and the extent of insurance ex post. In our conclusion, we discuss some problems involved in making further progress in this direction, as well as possible solutions.

The IMF Lending Record

In this section, we address a question involving a (hypothetical) emerging economy that must repay on its sovereign or private debt 10 percent of its GDP to foreign creditors—the rough order of magnitude of the large bailouts of the 1990s (see Table 1, or the average fiscal cost of banking crises in Frydl, 1999). The country could default, or, alternatively, it could repay its foreign creditors using a loan from the international community amounting to 10 percent of GDP. Based on the historical record, how much of this 10 percent could the country hope to transfer to the global taxpayer by not repaying the loan? The answer, we are going to argue, is “not very much”—at most a fraction of a percentage point of GDP.

Our answer is based on the historical lending record of the IMF. The justification for concentrating on the IMF is its role as the preeminent official crisis lender. While other multilaterals (in particular, the World Bank, the Asian Development Bank, and the Inter-American Development Bank) also have played a role, their financial involvement in a crisis context has been much more modest than that of the IMF (whose share in multilateral crisis lending has ranged from about 63 percent for Thailand to 100 percent for Mexico). Large-scale bilateral crisis lending has taken place in only three cases: Mexico (1995), Thailand (1997), and Brazil (1998). In two of these—Mexico and Brazil—bilateral lenders have by now been repaid in full. In Thailand’s case, the bi-laterals are being repaid in tandem with the IMF. The two most recent crisis packages, for Turkey and Argentina, were largely financed by the IMF on its own.

Ignoring bilateral lending might be misleading in one important sense. Conceivably, the repayment of IMF loans could be financed by bilateral loans—that is, the IMF could be bailed out by its own shareholders. Bulow, Rogoff, and Bevilaqua (1992) present evidence suggesting that the repayment of nonconcessional World Bank loans to a group of low- and middle-income countries during the 1980s was financed either by bilaterals or by the World Bank’s own concessional lending. Thus, for these countries, the repayment record of international financial institutions (IFIs) could give an overly optimistic picture of the repayment record of the official sector as a whole. However, we think that this is unlikely to be a problem in the class of countries that are candidates for large international bailouts—relatively advanced emerging economies with access to international capital markets. In Appendix I, we present evidence indicating that repayments to the IMF are not financed by new bilateral debt to this group of countries, although the same analysis supports Bulow and others’ (1992) results for the group of heavily indebted poor countries (HIPCs). Given that the focus of this paper is on IMF lending to the first group, we ignore bilateral financing in the remainder of the paper.

How do we measure the potential transfer entailed in IMF lending? We say that IMF loans involve a transfer if their interest rate is not high enough to offset the default risk faced by the IMF. In other words, we define the absence of transfer as a situation where the IMF makes zero profit on average. Conceivably, the IMF could charge a higher interest rate and make a strictly positive profit. Some would argue that the failure to do so—to exploit the IMF’s monopoly rent in crisis lending—constitutes an implicit transfer, or a subsidy. This is not, however, what the IMF critics mean when they claim that its loans are subsidized.

The IMF’s default risk is not necessarily the same as that faced by private lenders. The ability to impose conditionality ex post, which is typically justified by the need to ensure repayment, might imply that the IMF has a more direct influence on measures that improve country solvency than do private institutions. In addition, the IMF may be in a better position to deter default. It can deny further IMF lending and thus effectively bar access to the multilateral financial safety net in the future (this occurs automatically if a country accumulates arrears to the IMF). It can also exert pressure on a country through the actions of its major shareholders, which may link bilateral lending, or other bilateral policies, to a country’s good standing with the IMF.

It follows that the question of whether IMF lending implies a transfer cannot be answered by simply comparing IMF interest rates to market interest rates prevailing during crises, as has been argued in, among other places, the Wall Street Journal’s editorial dated April 23, 1998 (quoted in the previous section) and Calomiris (2001). Instead, the relevant comparison is between the interest rate charged by the IMF and the default risk it faces.

What Interest Rate Does the IMF Charge?

The relevant interest rate for the bulk of IMF lending is the “basic rate of charge on ordinary resources.”5 The rate of charge is set as low as possible while covering the IMF’s financing costs, its administrative budget, and the accumulation of some “pre cautionary balances” (IMF, 1998). The IMF’s financing costs, in turn, depend on the portion of member currency holdings that are remunerated, and the rate of remuneration that is applied. Prior to 1969, the IMF did not compensate its member countries for their currency holdings. The remuneration of member holdings was introduced after 1969 and was increased in various steps during the 1970s and 1980s (Boughton, 2001, Chapter 17, provides the details). Since 1987, the rate of remuneration has been equal to the SDR interest rate (a weighted average of three-month money market interest rates of the five major currencies). Since all but a small portion of member currency holdings are remunerated, this implies that since 1987, the basic rate of charge is effectively set as a markup over the SDR interest rate.

Figure 2 shows that prior to the mid-1980s, the IMF’s basic rate of charge was lower than both the U.S. three-year government bond rate—which is a useful benchmark in view of the average maturity of IMF nonconcessional loans (of two to five years)—and the market-determined SDR interest rate. Thus, in this period, IMF interest rates generally implied a subsidy, in the sense that a borrowing country could have made a small, risk-free profit on IMF loans by maintaining reserves borrowed from the IMF in liquid money market instruments denominated in SDR currencies. Since the mid-1980s, the IMF’s rate of charge has fluctuated around the U.S. three-year government bond rate. Moreover, since the introduction of the Supplemental Reserve Facility (SRF) in December 1997, the IMF has begun to impose surcharges for large-scale lending. Since November 2000, surcharges apply to most nonconcessional IMF lending that exceeds 200 percent of quota—see IMF (2000b) for details.

Figure 2IMF Basic Rate of Charge and Market Interest Rates

(in percentage points)

The IMF Repayment Record: Is the Past a Good Guide for the Future?

Does the interest rate charged by the IMF since 1987 reflect the default risk to which it is exposed? If so, the default risk facing the IMF would need to be very low, since the SDR and U.S. interest rates comparable with the IMF basic rate of charge are virtually risk free. Given that the IMF typically lends to crisis countries that appear very shaky—and are often cut off from private financing or face prohibitively high market interest rates precisely for that reason—this seems a tall order. However, it is exactly what the IMF’s repayment record suggests, at least on first appearance. The recent HIPC Initiative is the first case of IMF claims reduction, and arrears cases to the IMF are few and far between. Most of them are settled eventually, and they generally involve countries that go through wars or violent internal conflicts.6

However, the relatively small number and volume of arrears cases are evidence of low default risk only if we assume that the IMF’s repayment record over the last 50 years will continue to apply in the future. Some skeptics have expressed doubts about this. They argue that the IMF’s past repayment record was made possible by the IMF’s practice of refinancing its loans to countries that were unable to repay. If this is true, bad loans will increasingly tend to crowd out good ones in the IMF loan portfolio, and we eventually should observe either an increase in the rate of default on IMF lending, or debt reduction, as in the case of the HIPC initiative.

The fraction of IMF programs that end up in long-term lending relationships is examined in Table 2. We report the number of “complete” and “incomplete” lending cycles over IMF history. A lending cycle is defined as an uninterrupted period of strictly positive “Total Fund Credits and Loans Outstanding,” as reported in International Financial Statistics. A lending cycle is “complete” if it ends prior to 2000. We refer to countries with outstanding debt in 2000 as being in an “incomplete” lending cycle.

Table 2Completed and Incomplete Debt Cycles, 1947–20001
Number of CountriesComplete Debt CyclesIncomplete Debt CyclesAverage Duration of Cycles (years)
All countries186165887.117.9
Industrial countries253104.7n.a.
Developing countries161134887.617.9
Middle East141426.59.5
Western Hemisphere3757147.618.1
HIPC countries24225386.123.5
Non-HIPC developing countries119109508.013.6
EMBIG countries32738157.813.8
Non-EMBIG countries13496737.618.8
Memorandum item (excluding cycles initiated after 1991)
EMBIG countries3273787.920.6

Number of countries with outstanding debt in 2000.

HIPCs (see Appendix II for full list).

Countries whose bond spreads are tracked by J.P. Morgan’s “EMBI Global” Index (see Appendix II).

Number of countries with outstanding debt in 2000.

HIPCs (see Appendix II for full list).

Countries whose bond spreads are tracked by J.P. Morgan’s “EMBI Global” Index (see Appendix II).

Overall, and in most regions, the number of complete lending cycles exceeds that of incomplete ones. There are two main exceptions: Africa and non-OECD (Organization for Economic Cooperation and Development) Europe. The situation in the latter is due to a number of relatively recent lending cycles with transition economies that borrowed from the IMF for the first time in the 1990s. In contrast, the large proportion of incomplete lending cycles in Africa reflects the prevalence of successive borrowing arrangements from the IMF over many years.

Unsurprisingly, the occurrence of long incomplete cycles is especially high in HIPCs. Almost all of these countries (38 out of 42) have ongoing lending relationships with the IMF that started more than twenty years ago, on average, and incomplete lending cycles far outnumber completed cycles. The IMF experience with HIPCs illustrates the possibility that very long lending cycles might turn into “bad loans” that the international community will eventually have to forgive, at least in part. However, this experience concerns very poor countries that have little in common with the group of countries that are the main focus of this paper—the relatively advanced emerging economies that have benefited from the large international bailouts of the 1990s. The question is how widespread the long-term continuous lending relationships with the IMF are in the latter group.

To answer this question, we look at the length of lending cycles in the emerging market–country group tracked by J.P. Morgan’s “EMBI Global” Index (EMBIG). This is a group of 27 relatively advanced emerging market countries—the most internationally financially integrated countries in the nonindustrial country group. Appendix II has a full list of these countries. All major capital account crises of the 1990s involved members of the EMBIG group. Table 2 shows that the number of incomplete lending cycles in the EMBIG group (15) is low relative to the number of completed cycles, and that the incomplete cycles’ average length is substantially shorter than that of the non-EMBIG developing countries (13.8 versus 18.8). However, as the last line of the table shows, even in the EMBIG group, about half of the incomplete lending cycles in 2000 have been ongoing for, on average, more than 20 years. This is driven mainly by four long-term lending relationships: Côte d’Ivoire (since 1974), Panama (since 1974), Peru (since 1976), and the Philippines (since 1968). Thus, among the more advanced developing countries, which are the focus of this paper, long-term lending relationships with the IMF are not the rule, but neither are they nonexistent. On this basis, we proceed under the assumption that the IMF’s repayment record so far need not necessarily reflect the steady state even in this group of countries.

Estimating an Upper Bound for the Transfer Element in IMF Lending

If one accepts that, in principle, the IMF’s excellent repayment record need not continue indefinitely, estimating the subsidy element in IMF lending raises a fundamental difficulty. An estimate must be conditional on an assessment of the future repayment prospects of countries that are currently indebted to the IMF, and such an assessment is necessarily highly subjective. Rather than speculating about default probabilities for individual countries, we address these issues by computing a crude upper bound to the redistributive element implicit in IMF lending.

Our methodology relies on the notion of a hypothetical long-run IMF lending regime. This long-run regime involves two types of IMF lending cycles: those that are eventually repaid, and those that would go on forever unless the country either defaulted or received some multilateral debt relief. We refer to the first type as “finite,” and to the second as “infinite,” cycles. The latter are motivated by the long lending cycles observed in Table 2, and the worry that such cycles could reflect a process by which the IMF refinances its own obligations, leading to increasing country debt levels that need to be eventually recognized as unsustainable—as in the case of HIPCs. We estimate the fraction μ, of “infinite” cycles, using the method presented in Box 1.

Table 3 summarizes our results. It presents point estimates and p-values for μ, for the countries in Table 2, along with δ, the maximum amount that countries can borrow from the IMF based on their quotas (see Box 2). The maximum ex ante transfer for each subsample (assuming a zero recovery rate for infinite cycles) is then given by the product of the two (τ = μδ). Thus, τ is the maximum transfer that a country can expect ex ante, i.e., based on our best estimate of the fraction of infinite cycles and without knowing whether its lending cycle is of the infinite type or not. For purposes of comparison with δ, we also show the actual outstanding average debt to the IMF in 2000 for each subsample, in terms of that year’s GDP.

Table 3Maximum Average Ex Ante Subsidy(in percentage of debtor country GDP in 2000)
Total Number of Countriesμ1p-Value (H0: μ=0)Debt Limit2Actual Debt in 20003Maximum Subsidy τ = μδ
All countries, average1860.250.0011.24.42.81
Industrial countries, average250.001.004.1n.a.0.00
Developing countries, average1610.280.0012.34.43.46
Middle East140.000.897.24.60.00
Western Hemisphere370.090.1610.52.50.99
HIPC countries4420.590.0018.45.610.84
Non-HIPC developing countries1190.
EMBIG countries5270.050.667.01.70.37
Non-EMBIG countries1340.290.0012.52.63.58

Maximum-likelihood estimate of fraction of “infinite” lending cycles for various subsamples.

300 percent of quota in 2000, as a percentage of 2000 GDP.

In percentage of 2000 GDP.

Heavily indebted poor countries (see Appendix II for full list).

Countries whose bond spreads are tracked by J.P. Morgan’s “EMBI Global” Index (see Appendix II).

Maximum-likelihood estimate of fraction of “infinite” lending cycles for various subsamples.

300 percent of quota in 2000, as a percentage of 2000 GDP.

In percentage of 2000 GDP.

Heavily indebted poor countries (see Appendix II for full list).

Countries whose bond spreads are tracked by J.P. Morgan’s “EMBI Global” Index (see Appendix II).

Box 1.Estimating the Fraction of “Infinite” Cycles

Suppose that a fraction μ of lending cycles is of the infinite type, and denote the average net present value of debt owed to the IMF at the time the lending cycles are recognized as unsustainable as δ. Then, given that the IMF lends at approximately the riskless interest rate, the average ex ante transfer is equal to the frequency of infinite lending cycles, times the country’s indebtedness to the IMF, times the loss rate on infinite cycles:

τ = μδ(1−ρ)

where ρ is the average IMF “recovery rate” for infinite lending cycles. The strategy is to estimate an upper bound for τ by assuming an upper bound for δ and a lower bound for ρ, while attempting to estimate μ from the data. The most pessimistic assumption one can make about the recovery rate is obviously that ρ = 0, i.e., that debt owed on infinite cycles is forgiven in full. To set δ, we make the radical assumption that all countries on infinite lending cycles will be allowed to borrow up to the full amount of their cumulative debt ceilings prior to defaulting or receiving debt relief, namely, 300 percent of quota. This is about three times as high as the average actual indebtedness to the IMF in 2000. While the 300 percent cumulative debt ceiling can be waived in exceptional circumstances, this has occurred only in rare cases—which in the 1990s have mostly coincided with the large crisis packages we briefly discussed in the introduction.

We estimate ¼ using a simple parametric duration model (Greene, 1993). To do so, we need to assume a distribution for the duration of the finite lending cycles. We use a standard generalization of the constant hazard rate model, the Weibull model, which allows for increasing or decreasing the hazard rate as a function of duration.

For the whole sample, the estimated fraction of “infinite” cycles, μ is about 25 percent. However, this masks enormous disparities across regions and income groups. In the HIPC sample, the estimated fraction is almost 60 percent—which is perhaps not surprising given that we know that countries in these groups are eligible for multilateral debt relief. For the 119 non-HIPC developing countries, the estimated proportion of cycles that look like they could go on forever is only 11 percent. For the EMBIG countries, the fraction falls to just 5 percent, and is insignificantly different from zero. The more restrictive the definition of the subsample in terms of income levels and access to capital markets, the lower the estimated fraction of “infinite cycles.”

Box 2.Quotas and IMF Credit Limits

A “quota” is a country’s capital contribution to the IMF. It determines voting power within the institution as well as access to IMF financing. Quotas are set as a function of a country’s GDP, openness, the volatility of exports (or current receipts), and the level of reserves (see IMF, 1998, Box 3). Large countries with diversified exports tend to have much smaller quotas, as a share of GDP, than do small countries whose exports are concentrated in a few sectors. For example, the median quota in 2000 was about 2.7 percent of GDP, but the United States’ quota was only about 0.5 percent, and Zambia’s was about 18 percent. Quotas are reviewed periodically (about once every five years).

Until 1963, a country’s total borrowing limit was 100 percent of quota. This was gradually relaxed during the 1960s and 1970s, in step with the creation of new lending facilities, reaching about 300 percent by 1978 (see Boughton, 2001). Following the second oil shock, access limits were briefly raised much higher, and remained at approximately 450 percent until 1992, when the Ninth Review of Quotas increased quotas by 50 percent. Since then, the limit on total cumulative borrowing from the IMF has been 300 percent of quota, with the understanding that this could be exceeded in “exceptional circumstances.” Recently, two new facilities have been created: the Supplemental Reserve Facility (SRF), for cases of “exceptional balance of payments difficulties,” and the Contingent Credit Lines (CCL), as a defense against contagion. These facilities are exempted from the formal access limit, but are also subject to much shorter repayment periods and much higher rates of charge.

Since 1992, the 300 percent cumulative ceiling has so far been exceeded in six cases: Mexico (peak: 607 percent of quota in 1995), Thailand (400 percent in 1998), the Republic of Korea (1,500 percent in 1998), Indonesia (431 percent in 1998), Turkey (445 percent in early 2001), and briefly Russia, whose outstanding debt reached 318 percent of quota in 1998. The FY 1998/99 Brazil package stayed within the 300 percent cumulative limit, in spite of the use of the SRF.

At present (in mid-2001) two countries still have liabilities to the IMF that exceed the standard cumulative debt ceiling, namely Turkey and Indonesia. For most countries, actual levels of debt outstanding in 2000 are one-third to one-fourth, on average, of the theoretical debt ceilings (see Table 3).

In the last column, we multiply cumulative debt limits with the estimated fraction of “infinite cycles.” As one would expect, this results in a further magnification of the cross-regional variations that were noted earlier. For the HIPC group, the upper bound for the ex ante transfer is more than 10 percent of debtor country GDP. Of course, the actual transfer that these countries can expect from the IMF as a result of the HIPC initiative is far lower, both because the actual debt levels are much below the assumed debt ceilings, and because the HIPC Initiative does not envisage a complete debt write down, as assumed in Table 3; see IMF (1998). For the group of emerging economies that are our main focus here, i.e., the EMBIG countries, the estimated upper bound is only 0.37 percent of GDP. This is driven by the fact that the estimated fraction of “in finite” cycles is very small in this group.

We can now answer the question that was asked at the beginning of this section. Based on the historical record, an EMBIG economy that borrows 10 percent of its GDP would fail to repay with a probability of at most 5 percent—the probability of an “infinite cycle” for this class of countries. The implicit transfer, which results because the interest rate charged by the IMF fails to reflect this default risk, is thus less than 0.5 percent of the country’s GDP. If the country represents 1 percent of the world population and GDP (this corresponds to a large emerging economy, between Argentina and Brazil in size), the per capita cost of the bailout for the global taxpayer would amount to less than 0.0005 times that borne by the domestic taxpayer. It bears emphasizing that even these small numbers are based on an extreme assumption underpinning our hypothetical worst-case scenario, namely, that none of the outstanding debt on “infinite” lending cycles will be recovered. Thus, a reasonable estimate of the ex ante subsidy implicit in IMF lending is likely to be much smaller.

Policy Discussion

In light of the evidence presented in the previous section, we now return to the main question motivating this paper—namely, how the international community can best limit moral hazard. In addition to the papers and editorials cited in the introduction, a number of reports have been written on the reform of international financial institutions in the aftermath of the Asian crisis (see Williamson, 2000 for a review). One report that has been especially influential in the policy debate was prepared by the International Financial Institution Advisory Commission established by the U.S. Congress and chaired by Allan Meltzer (IFIAC, 2000, hereinafter referred to as the Meltzer report). We present a brief review of the different policy approaches in these reports, before presenting the case for ex ante conditionality and concluding with a discussion of some objections against it.

Policy Approaches

Proposals to reform IMF lending policies for the purposes of reducing moral hazard have varied along three main dimensions:

First, increasing the price (making it more costly for countries to borrow from the IMF). The Meltzer report argues that the IMF should charge a penalty rate (above the borrower’s recent market rate) and secure its loan by a clear priority claim on the borrower’s assets. As documented in the previous section, the IMF increased some of its lending rates in November 2000.

Second, decreasing the quantity (limiting the size of bailouts). The Council on Foreign Relations Task Force (1999) proposes that the IMF return to nor mal lending limits (100 percent to 300 percent of quota) for “country crises,” that is, for crises that do not threaten the performance of the world economy. Larger loans could be made under a systemic facility, which would require a supermajority of creditors to be activated. Arguments that there should be more “private sector involvement” in financing crises (Haldane, 1999; and Roubini, 2000) are in the same vein. Eichengreen (1999, 2000) advocates improving crisis resolution mechanisms through changes in the law governing private debt contracts, or through officially sanctioned standstills as a way to resolve investor panics. These can be interpreted as proposals that seek to reduce the need to resort to large official bailouts as the main way of mitigating the consequences of an international financial crisis.

Third, introducing ex ante conditionality (sometimes also referred to as “selectivity” or “prequalification”). The idea is to make the generosity of international bailouts in a crisis conditional on the quality of domestic policies, particularly in the financial sector, before a crisis erupts. The Council on Foreign Relations Task Force suggests that the IMF should distinguish among three categories of countries on the basis of their compliance with a set of standards and good practices, and publish regular reports assessing countries’ progress in meeting these standards. (The standards and practices include, among others, the IMF’s Special Data Dissemination Standard and the Basel Committee’s Core Principles of Effective Banking Supervision.) A country’s class would then determine the availability of official assistance and the interest rate at which it would be charged in the event of a crisis. The Meltzer report takes a similar approach, but in a more extreme form. It recommends that IMF lending be restricted to a group of countries selected for the soundness of their banking policies, with other countries being altogether ineligible for official crisis lending. The idea of ex ante conditionality is also embodied in a new IMF facility, the Contingent Credit Lines (CCL), which provided for exceptional access to IMF resources for countries that qualified ex ante on the basis of sound policies and progress toward meeting internationally accepted standards. (For a precise statement of the conditions that were required to qualify for the CCL, see IMF, 2000a.)7

The Case for Ex Ante Conditionality

We do not think that there is much to be gained from measures that increase the price of official crisis lending—at least in terms of reducing moral hazard. It amounts to fixing a problem that does not exist. The welfare losses stemming from excessive moral hazard, if any, are in the countries receiving international financial assistance, not in the contributing countries. Nor will the proposals in this category work as an incentive mechanism: in countries with severe policy failures, governments are unlikely to be deterred by higher interest rates, since they ultimately do not bear the costs of such interest rates.

Institutional and legal changes that reduce the “demand” for big bailouts by making debt restructurings less costly are clearly desirable from the point of view of reducing moral hazard. Moreover, even if such changes are not successful in eliminating the need for an international financial safety net, systematically smaller bailouts would be optimal if one viewed domestic policy failures as pervasive. In other words, it would be optimal to systematically restrict the policy options of countries if one thought that they would systematically use them to implement bad policies. This is a possible view of the world, but we see it as overly pessimistic. There might be “good countries”—that is, those that put the international financial safety net to good use—and their welfare would be decreased by a one-size-fits-all reduction in bailouts.

Ideally, one would like to reserve the international financial safety net for countries with good policies and deny it to countries with bad ones. This, in essence, is what ex ante conditionality at tempts to achieve. There are two benefits. First, this targets financial assistance to countries that are likely to make the best use of it. Second, it gives all countries an incentive to improve their policies (Jeanne and Zettelmeyer, 2001a).

The conclusion that crisis lending should be linked to policies before the crisis does not rely on a specific model of how policy failures arise. Rather, it follows from the general trade-off between incentives and insurance. The better the institutions and policies that prevail before the crisis, the more “insurance” can be provided, in the form of an international financial safety net, without destroying private sector incentives. Conversely, the worse institutions are, the greater the degree of market discipline required to offset their adverse incentive effects.

An additional benefit of linking crisis lending to ex ante policies is that they could reduce the extent and intrusiveness of ex post conditionality. After the Asian crisis, IMF structural conditionality was criticized for being excessive (Goldstein, 2000a). One reason why more conditions may have been imposed than were perhaps necessary to ensure repayment (the standard justification for IMF conditionality), is that the IMF sought to take advantage of the time window in which it had some leverage over the governments—that is, while the countries needed IMF support. Imposing some conditions for crisis lending ex ante rather than ex post would give the IMF more leverage in good times, allowing it to allocate conditionality over time in a more efficient and balanced way.

Problems with Ex Ante Conditionality

Greater reliance on ex ante conditionality raises a number of questions related to design and implementation. One issue is precisely what ex ante conditions should be imposed. This is beyond the scope of our paper, but the basic principle is clear: the maximum level of crisis assistance ex post (and possibly the interest rate at which it is provided) should depend on the sustainability of fiscal policies, public debt management, and the extent to which a country implements financial sector policies that mitigate excessive risk taking. The last factor includes policies for effective banking supervision along the lines described in the Basel Committee’s “Core Principles” (Basel Committee on Banking Supervision, 1997), accounting standards, and public disclosure requirements.

One important problem with ex ante conditionality is that it may not be time-consistent (De Gregorio and others, 2000). The temptation to bail out all investors is typically greater ex post than ex ante. This is a real problem, but it is not specific to international bailouts; it arises, to some extent, with all financial safety nets (Rochet, 2000). The institutional design of these safety nets is aimed precisely at solving this problem. The solution, as is the case with time-consistency problems in monetary policy, involves rules, delegation, and reputation. Nor is the time-consistency problem specific to ex ante conditionality. Conventional IMF conditionally faces a similar problem, since it is often optimal ex post to release the next disbursement even if preceding performance criteria have been violated. In practice, this is dealt with through rules that determine in which circumstances program requirements can be waived.

A related issue concerns the room that should be left for “constructive ambiguity.” The need to preserve a measure of constructive ambiguity is the most serious objection against ex ante conditionality. On the one hand, it is important that the link between the quality of domestic policies and the extent of crisis lending be governed by explicit rules—both to deal with the time-consistency problem and to ensure universality of treatment ex ante. On the other hand, there will need to be exceptions to these rules. For example, a crisis that is clearly triggered by a financial panic may call for a rescue package even when the policy record prior to the crisis was mixed, or the risk of international contagion may require intervention even when the country would not, in isolation, have qualified for a loan. Conversely, limited fund availability may prevent the IMF from fulfilling its commitments to prequalified countries if a large number of them are simultaneously hit by a crisis. These circumstances cannot always be incorporated into the rules because they are difficult to describe ex ante or to verify ex post.

However, the need for constructive ambiguity does not in itself invalidate the case in favor of some ex ante conditionality. Domestic attempts at reconciling the two are instructive in this regard. For example, the reform of the U.S. deposit insurance system that followed the saving and loan crisis attempted to mitigate the moral hazard resulting from constructive ambiguity by establishing a system of checks and balances that places stringent accountability conditions on economic officials when they decide to rescue a bank because it is “too large to fail” (Council on Foreign Relations Task Force, 1999; Tirole, 2001).8 This approach is consistent with an important lesson from the theory of incomplete contracts: when contracts cannot fully specify the actions of the contracting parties, they should instead focus on the rules by which these actions will be decided—the allocation of “decision rights.” Thus, behind the move toward ex ante conditionality in international bailouts looms the broader question of the governance structure—the rules and procedures by which the international community determines its response to international financial crises.

Appendix I. IMF Lending and Bilateral Lending

This appendix deals with Bulow, Rogoff, and Bevilaqua’s (1992) argument that “any ability IFIs [international financial institutions] have to extract repayments ahead of private creditors may come almost entirely at the expense of bilateral government creditors.” We undertake a simple empirical exercise showing that the relationship between repayments to the IMF and bilateral lending by OECD members (both concessional and nonconcessional) looks quite different depending on the level of income and capital market integration of the creditor countries.

Consider a regression of changes in IMF debt stocks on contemporaneous and lagged changes in the stock of bilateral developing country debt held by OECD countries (these data, going back to 1975, are available from the OECD). In the extreme case, in which repayments to the IMF are entirely financed by new bilateral debt, the contemporaneous and recent lagged changes in OECD debt should have coefficients that add up to −1. In the less extreme case, in which new OECD debt contributes to, but does not fully finance, IMF repayments, the coefficients should be negative, and add up to less than unity in absolute value. Table A1 presents results from such a regression based on a panel of 161 developing countries, restricting attention to episodes of repayments to the IMF (i.e., conditioning on negative changes in the IMF debt stock).

Table A1Correlations Between IMF and Bilateral Debt Stock Changes, 1975–99
Regression Sample
Independent variableAll developingHIPC1Non-HIPCNon-EMBIGEMBIG2
OECD debt change-0.0010.001-0.0010.000-0.098
OECD debt change (-1)-0.026-0.016-0.032-0.0980.027
OECD debt change (-2)0.128-0.0190.1390.0080.207
Regression constant-153.6-14.3-235.8-37.3-577.6
Number of countries11740779324
Number of observations954370584771183
Notes: Dependent variable is the change in total IMF credits and loans outstanding; panel regressions are estimated using fixed effects, with t-values in parentheses. OECD denotes the Organization for Economic Cooperation and Development.

Group of heavily indebted poor countries.

Group of emerging market countries whose sovereign bond spreads are tracked in J.P. Morgan’s “Emerging Market Bond Index Global.” (See Appendix II.)

Notes: Dependent variable is the change in total IMF credits and loans outstanding; panel regressions are estimated using fixed effects, with t-values in parentheses. OECD denotes the Organization for Economic Cooperation and Development.

Group of heavily indebted poor countries.

Group of emerging market countries whose sovereign bond spreads are tracked in J.P. Morgan’s “Emerging Market Bond Index Global.” (See Appendix II.)

The results indicate that the Bulow, Rogoff, and Bevilaqua (1992) hypothesis is not supported on the entire sample of developing countries, where we see small and insignificant coefficients on the contemporaneous OECD debt change and its first lag, and a significant positive coefficient on the second lag. But the picture is strikingly different if one considers only heavily indebted poor countries. Both the first and second lags are now negative and highly significant, and add up to about -0.25. If attention is restricted to all “non-emerging” developing countries, the negative effect is smaller (about -0.1), but still significant. In contrast, the relatively richer and “emerging” developing countries show a significant positive relationship between repayments to the IMF and to OECD members.

These findings may not be entirely surprising. As Bulow, Rogoff, and Bevilaqua point out, one would expect bilateral transfers to facilitate multilateral debt repayment if bilaterals cared about multilaterals being repaid, and the amounts involved were small. But crisis lending to middle-income countries is often large, and may be politically much more costly than lending through specialized multilateral institutions. This is suggested by the shrinking role of bilateral crisis lending after the Mexico crisis and its complete absence in the recent lending packages to Argentina and Turkey.

Appendix II. Countries in HIPC and EMBIG Groups: Definitions Used in Previous Section

Heavily Indebted Poor Countries (HIPCs): Angola, Benin, Bolivia, Burkina Faso, Burundi, Cameroon, the Central African Republic, Chad, Congo, Côte d’Ivoire, the Democratic Republic of the Congo, Equatorial Guinea, Ethiopia, The Gambia, Ghana, Guinea, Guinea-Bissau, Guyana, Honduras, Kenya, the Lao People’s Democratic Republic, Liberia, Madagascar, Malawi, Mali, Mauritania, Mozambique, Myanmar, Nicaragua, Niger, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Somalia, Sudan, Tanzania, Togo, Uganda, Vietnam, the Republic of Yemen, and Zambia.

EMBI Global Index (EMBIG) countries: Algeria, Argentina, Brazil, Bulgaria, Chile, China, Colombia, Côte d’Ivoire, Croatia, Ecuador, Hungary, the Republic of Korea, Lebanon, Malaysia, Mexico, Morocco, Nigeria, Panama, Peru, the Philippines, Poland, Russia, South Africa, Thailand, Turkey, Ukraine, and República Bolivariana de Venezuela.


    Andrews, David, Anthony R.Boote, SitadRizavi, and SukhwinderSingh, 1999, Debt Relief for Low-Income Countries. The Enhanced HIPC Initiative, IMF Pamphlet No. 51 (Washington: International Monetary Fund). Also available on the Web at

    • Search Google Scholar
    • Export Citation

    Aylward, Lynn, and RupertThorne, 1998, “An Econometric Analysis of Countries’ Repayment Performance to the International Monetary Fund,”IMF Working Paper 98/32 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Basel Committee on Banking Supervision, 1997, “Core Principles for Effective Banking Supervision,” documentavailable on the Web at

    • Search Google Scholar
    • Export Citation

    Boughton, James M., 2001, Silent Revolution: The International Monetary Fund, 1979–1989 (Washington: International Monetary Fund).

    Bulow, Jeremy, KennethRogoff, and Alfonso S.Bevilaqua, 1992, “Official Creditor Seniority and Burden-Sharing in the Former Soviet Bloc,” Brookings Papers on Economic Activity: 1,Brookings Institution.

    • Search Google Scholar
    • Export Citation

    Calomiris, Charles W., 1998, “The IMF’s Imprudent Role as Lender of Last Resort,”Cato Journal,Vol. 17, No. 3.

    Calomiris, Charles W., 2001, “How New Is the ‘New IMF?’” Testimony Before the Joint Economic Committee of the United States Congress (March8).

    • Search Google Scholar
    • Export Citation

    Council on Foreign Relations Task Force, 1999, Safeguarding Prosperity in a Global Financial System: The Future International Financial Architecture (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation

    De Gregorio, José, BarryEichengreen, TakatoshiIto, and CharlesWyplosz, 2000, An Independent and Accountable IMF, Geneva Reports on the World Economy Report, Vol. 1 (September).

    • Search Google Scholar
    • Export Citation

    Dell’Ariccia, Giovanni, IsabelGödde, and JerominZettelmeyer, 2000, “Moral Hazard and International Crisis Lending: A Test” (unpublished; Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Edwards, Sebastian, 1998, “Abolish the IMF,” Financial Times, November13.

    Eichengreen, Barry, 1999, Toward a New International Financial Architecture. A Practical Post-Asia Agenda (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation

    Eichengreen, Barry, 2000, Can the Moral Hazard Caused by IMF Bailouts be Reduced? Geneva Reports on the World Economy, Special Report No. 1.

    • Search Google Scholar
    • Export Citation

    Feldstein, Martin, 1998, “Refocusing the IMF,”Foreign Affairs,Vol. 77.

    Fischer, Stanley, 1999, “On the Need for an International Lender of Last Resort,”Journal of Economic Perspectives,Vol. 13 (Fall).

    Frydl, Edward, 1999, “The Length and Cost of Banking Crises,”IMF Working Paper 99/30 (Washington: International Monetary Fund).

    Goldstein, Morris, 2000a, “IMF Structural Programs” (unpublished; Washington: Institute for International Economics).

    Goldstein, Morris, 2000b, “Strengthening the International Financial Architecture: Where Do We Stand?”Institute for International Economics Working Paper No. 00–8 (Washington).

    • Search Google Scholar
    • Export Citation

    Greene, William H., 1993, Econometric Analysis, 2nd ed. (New York: Macmillan).

    Haldane, Andy, 1999, “Private Sector Involvement in Financial Crisis: Analytics and Public Policy Approaches,” Bank of England Financial Stability Review (November).

    • Search Google Scholar
    • Export Citation

    International Financial Institutions Advisory Commission (IFIAC), 2000, “Report of the International Financial Institution Advisory Commission,” Allan H. Meltzer, Chairman (Washington) (also referred to as the Meltzer report).

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF), 1998, Financial Organization and Operations of the IMF, IMF Pamphlet No. 45 (Washington: International Monetary Fund). Also available on the Web at

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF), 2000a, “Summing Up by the Acting Chairman of the IMF Executive Board: Contingent Credit Lines,” IMF Executive Board Meeting, November 17 (Washington: International Monetary Fund). Also available on the Web at

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF), 2000b, “Review of Fund Facilities: Proposed Decisions and Implementation Guidelines,” prepared by the Legal, Policy Development and Review, and Treasurer’s Departments, November2. Available on the Web at

    • Search Google Scholar
    • Export Citation

    Jeanne, Olivier, and JerominZettelmeyer, 2001a, “International Safety Nets, Private Debt Crises and Moral Hazard” (unpublished; Washington: International Monetary Fund), forthcomingas an IMF Working Paper.

    • Search Google Scholar
    • Export Citation

    Jeanne, Olivier, and JerominZettelmeyer, 2001b, “International Bail-Outs, Domestic Policy Failures and Moral Hazard” (unpublished; Washington: International Monetary Fund), forthcoming as an IMF Working Paper.

    • Search Google Scholar
    • Export Citation

    Lane, Timothy, and StevenPhillips, 2000, “Moral Hazard in IMF Financing,”IMF Working Paper 00/168 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    McBrady, Matthew, and Mark S.Seasholes, 2000, “Bailing-In” (unpublished; University of California, Berkeley).

    Meltzer report—see IFIAC.

    Mussa, Michael, 1999, “Reforming the International Financial Architecture: Limiting Moral Hazard and Containing Real Hazard” (unpublished; Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Nunnenkamp, Peter, 1999, “The Moral of IMF Lending: Making a Fuss About a Minor Problem?” Kiel Discussion Paper No. 332 (Kiel, Germany: Kiel Institute of World Economics).

    • Search Google Scholar
    • Export Citation

    Rochet, Jean-Charles, 2000, “The Future of Banking Supervision: Market Discipline vs. Centralized Supervision” (unpublished; Toulouse, France: Institut d’Economie Industrielle, Université des Sciences Sociales).

    • Search Google Scholar
    • Export Citation

    Roubini, Nouriel, 2000, “Bail-In, Burden-Sharing, Private Sector Involvement (PSI) in Crisis Resolution and Constructive Engagement of the Private Sector” (unpublished; Department of Economics, New York University).

    • Search Google Scholar
    • Export Citation

    Schwartz, Anna J., 1998, “Time to Terminate the ESF and the IMF,” Cato Institute Policy Briefing No. 48 (Washington: Cato Institute).

    • Search Google Scholar
    • Export Citation

    Shultz, George, William E.Simon, and Walter B.Wriston, 1998, “Who Needs the IMF?” Wall Street Journal,February3.

    Summers, Lawrence H., 1999, “The Right Kind of IMF for a Stable Global Financial System,” remarks to the London School of Business, December14.

    • Search Google Scholar
    • Export Citation

    Tirole, Jean, 2001, “Financial Crises, Liquidity Provision and the International Monetary System,” manuscript prepared for the Paolo Baffi Lecture on Money and Finance.

    • Search Google Scholar
    • Export Citation

    Williamson, John, 2000, “The Role of the IMF: A Guide to the Reports,” IIE Policy Brief No. 00–5 (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation

    Zhang, Xiaoming Alan, 1999, “Testing for ‘Moral Hazard’ in Emerging Markets Lending,” IIF Research Paper No. 99–1 (Washington: Institute of International Finance).

    • Search Google Scholar
    • Export Citation

This paper was first published in Economic Policy in 2001 and has not subsequently been revised or updated. It appears here with the permission of Blackwell Publishing, the journal’s publisher. An earlier version was presented at the 33rd Economic Policy Panel Meeting, which was held in Stockholm April 6–7, 2001. We are grateful to our two discussants at the Panel, Philippe Bacchetta and Andrew Scott, as well as the editor, Charles Wyplosz, for very useful comments and suggestions. We also thank Mark Allen, Reza Baqir, Eduardo Borensztein, Barry Eichengreen, Gaston Gelos, Ken Kletzer, Tim Lane, Paul Masson, Carlos Medeiros, Christian Mulder, Mike Mussa, Sanjaya Panth, Steve Phillips, Andrea Richter, Ken Rogoff, Nouriel Roubini, Alexander Swoboda, Tessa van der Willigen, Alexander Wolfson, and especially Daniel Cohen for discussions and/or comments on earlier drafts. Our thanks also go to several colleagues at the IMF—including Jim Boughton, Christina Daseking, Leyla Ecevit, Doris Ross, and Barry Yuen—for data and information on IMF financial operations. Any remaining errors, however, are our sole responsibility.


The distinction between moral hazard and excessive moral hazard is emphasized by Mussa (1999). The relevant question from a policy perspective is not whether international bailouts generate moral hazard per se, but whether this moral hazard is excessive in view of the benefits of the international financial safety net.


Note that we are not concerned with quantifying the extent to which international bailouts create moral hazard. This is the subject of a recent set of papers by Nunnenkamp (1999); Zhang (1999); Lane and Phillips (2000); Dell’Ariccia, Gödde, and Zettelmeyer (2000); and McBrady and Seasholes (2000).


This is the interest rate that applies to the standard (nonconcessional) IMF lending facilities. The interest rate charged by the IMF on its concessional facilities—the Enhanced Structural Adjustment Facility (ESAF) and its successor, the Poverty Reduction and Growth Facility (PRGF)—is much lower (currently 0.5 percent). Concessional lending targets very poor countries and has not been used to deal with the international financial crises that we focus on in this paper.


Total arrears to the IMF in 2000 amounted to US$3 billion, about 4.5 percent of its total loans and credits outstanding. The following countries were in arrears: Afghanistan, the Democratic Republic of the Congo, Iraq, Liberia, Somalia, Sudan, and the Federal Republic of Yugoslavia. For a comprehensive documentation of IMF arrears cases, see Aylward and Thorne, 1998.


No country has applied for the CCL, in part because it was viewed as offering few advantages when compared with the Supplemental Reserve Facility (SRF), which has been used for large-scale lending without formal prequalification. In response, in November 2000, the IMF lowered the rate of charge on the CCL relative to the SRF, and made several other changes designed to enhance the CCL’s attractiveness.


The Federal Deposit Insurance Corporation Improvement Act (FDICIA) of 1991 makes it harder for regulators to bail out uninsured creditors. An extension of the guarantee to all bank creditors is possible in exceptional circumstances, but this requires the explicit consent of the secretary of the treasury in consultation with the president of the United States, two-thirds of the governors of the Federal Reserve System, and two-thirds of the directors of the FDIC.

    Other Resources Citing This Publication