Timing of International Bailouts
Author: Mr. Se-Jik Kim

Contributor Notes

Author’s E-Mail Address: skim@imf.org

This paper proposes that international rescue financing should not be provided to a country where a crisis first occurs, but rather to any country that suffers a subsequent crisis. Such a timing-based lending facility can be Pareto-superior to both laissez-faire and existing international crisis lending facilities, when domestic governments have more information on their own economies than does the international lender of last resort. The new facility mitigates moral hazard owing to information asymmetry by not rescuing the first-hit country. At the same time, it limits crisis contagion by rescuing countries in subsequent crises. Even in the presence of common shocks, the timing-based facility can reduce global risks of crisis because it induces countries to undertake greater crisis-prevention efforts so as not to become the first country hit.

Abstract

This paper proposes that international rescue financing should not be provided to a country where a crisis first occurs, but rather to any country that suffers a subsequent crisis. Such a timing-based lending facility can be Pareto-superior to both laissez-faire and existing international crisis lending facilities, when domestic governments have more information on their own economies than does the international lender of last resort. The new facility mitigates moral hazard owing to information asymmetry by not rescuing the first-hit country. At the same time, it limits crisis contagion by rescuing countries in subsequent crises. Even in the presence of common shocks, the timing-based facility can reduce global risks of crisis because it induces countries to undertake greater crisis-prevention efforts so as not to become the first country hit.

I. Introduction

This paper proposes a new international rescue lending facility based on the timing order of crises. It also shows that under information asymmetry, the new timing-based facility can function more effectively than both laissez-faire and existing international crisis-lending facilities in simultaneously resolving the problems of moral hazard and crisis contagion (and even in resolving a coordination problem arising in the case of common shocks and vulnerabilities).

When financial crises swept over Asia in 1997 and 1998, following the Mexican crisis of 1994–95, the international community came to recognize that forestalling new crises was now more important than resolving the ones already taking place. In particular, traditional lending facilities, such as the Stand-By Arrangement of the International Monetary Fund (IMF), were blamed by some for focusing on bailing out countries that were already in the midst of crises rather than preventing crises and, hence, raising moral hazard despite the IMF’s role in abating contagion of crises across countries (see, e.g., Meltzer, 1997; Calomiris, 1998; Fischer, 1998 and 1999; Eichengreen, 2000; Jeanne and Wyplosz, 2001; Krueger, 2002; and IFIAC, 2000). Accordingly, there was a marked growth in demand for a new lending facility aimed at crisis prevention to obviate future threats to the stability of international financial system as compared with such traditional lending facilities for crisis resolution.2

Given these circumstances, the IMF adopted the Contingent Credit Lines (CCL) as a lending facility for crisis prevention in 1999, but the CCL turned out to be ineffective owing to information problems. Under the CCL, only the countries that met certain prequalifications were eligible for financial assistance in the event of a crisis. So countries were expected to pursue sound economic policies and financial systems to meet the prequalifications; that would help to contain the contagion of crises and, at the same time, abate moral hazard, thus minimizing global risks. However, not one member country applied for the CCL, which was terminated in 2003, suggesting that the initial demand for the facility was being eclipsed by its potential problems. Particularly notable was the problem owing to information asymmetry between countries and the international crisis lender.3 Given that only the countries that met certain prequalifications were eligible to use the CCL, a critical condition had to be satisfied for the CCL to effectively prevent both moral hazard and crisis contagion; that is, accurate, near-perfect monitoring of the international crisis lender on the countries.4 In reality, however, substantial information asymmetry exists between residents and foreigners, which may hinder perfect monitoring by international organizations. In such a situation, the CCL may have been effective in containing crisis contagion but ineffective in deterring moral hazard.

Since the CCL turned out to be ineffective, a question arises: is there any other way to forestall new crises and, at the same time, resolve the ones already occurring? Is there any way to resolve both moral hazard and crisis contagion at the same time, particularly in the real world of information asymmetry?

This paper explores an approach that would simultaneously resolve the problems of moral hazard and crisis contagion, and, hence, effectively both prevent new crises and resolve the ones already taking place. More specifically, it proposes a new lending facility that would differentiate rescue financing to different countries by the order in which crises occur. Under this facility, which focuses on the timing of crises, rescue financing would not be provided to the country where the crisis first began but rather to any country faced with a crisis within a short period after the initial crisis occurred.

The key ideas underlying this new facility, particularly for the case of crisis contagion, are twofold. First, the facility differentiates rescue financing to different countries based on whether a country is more likely to be the one that originates the crisis or not; it aims at not rescuing the crisis originator but the victims of contagion. So it can be likened to holding the house that initiated the fire responsible but rescuing the houses that caught fire from the other houses. By doing so, the facility provides solid incentives for countries to prevent crises and stops a crisis, if one breaks out, from spreading further.

Second, the facility distinguishes between the crisis originator and the victims of contagion based on the timing order of crises. Regarding the point in time, the country where the crisis erupted first is very likely to be the crisis originator, and the ones having crises after the first are more likely to be the infected. This facility uses the potential time lag between the crises in the crisis originator and the infected countries. In this regard, it can be likened to identifying the house that originated the fire and the houses that caught fire from neighboring houses by determining the timing order of the fire. The intuition behind this method of identification is straightforward. If one house was on fire and then, half an hour later, a neighboring house started burning, the chance that fire spread from the latter to the former is almost nil, while the chance of the fire having spread the other way around is very high.

The strength of this proposed facility is that the identification depends on the simple and objective rule that is based only on the timing of crises, which is observable and verifiable by anybody. Therefore, the identification under this new facility, unlike existing international crisis lending facilities, does not have to depend on the monitoring capability or the discretional judgment of the international crisis lender.5

To develop the theory more formally, this paper proposes a simple model of crisis contagion and moral hazard (owing to rescue financing) under information asymmetry. In particular, the model assumes crisis contagion, broadly defined as the transmission of crises across countries, following a large literature on crisis contagion (see, for example, Eichengreen, Rose, and Wyplosz, 1996; Glick and Rose, 1998; Kaminsky and Reinhart, 2000; Allen and Gale, 2000; Jeanne, 2000; Calvo and Mendoza, 2000; Goodhart and Huang, 2000; Forbes and Rigobon, 2002; and Kodres and Pritsker, 2002). This assumption captures the fact that the bulk of the empirical literature suggests that crisis contagion is systematically present in the data (Kaminsky, Reinhart, and Végh, 2002). For example, using data for five different currency crises during the 1970s through the 1990s, Glick and Rose (1998) find that currency crises affect countries tied together by international trade. Based on an analysis of Argentina after the Mexican crisis and of Indonesia after the crisis in Thailand, Kaminsky and Reinhart (2000) suggest that crisis contagion transmitted through financial linkages is important.6

In addition, the model assumes that there are two types of crises depending on the origin of the crisis: one is autonomous crisis that occurs internally because of weaknesses in the domestic economy, and the other is imported crisis, which is the outcome of crisis contagion from foreign countries.7 The chances of a country’s having an autonomous crisis can be significantly lowered by its own voluntary crisis-prevention efforts (for example, implementation of structural reforms in financial and corporate sectors and sound macroeconomic policies), though the chance is not fully eliminated.8 The chance of a country’s having an imported crisis after another country has already experienced a crisis can be reduced when the international crisis lender bails out the crisis country. In this situation, the crisis lending by the international crisis lender may contain crisis contagion but raise borrower countries’ moral hazard (by reducing the country’s crisis-prevention efforts).

Finally, the model assumes information asymmetries between the government of a country and the international crisis lender. The international crisis lender cannot accurately observe each country’s crisis prevention efforts. Nor can it accurately tell whether a crisis that hit a country originated there or was imported.

Within the model described above, several important conclusions follow. First, both laissez-faire and existing lending facilities could be ineffective in simultaneously resolving the problems of moral hazard and crisis contagion. The laissez-faire approach, in which the international crisis lender does not provide rescue financing to any country, could be effective in containing moral hazard. Nevertheless, it would be ineffective in abating crisis contagion. Meanwhile, traditional crisis lending facilities of the IMF such as Stand-By Arrangements could effectively prevent crisis contagion. However, the traditional lending facilities tend to induce a moral hazard problem, since countries, which expect to receive rescue financing in the event of crisis, would not exert much effort to prevent crises. Furthermore, given information asymmetries, the CCL may be as ineffective as other traditional IMF crisis financing facilities in dealing with the problem of moral hazard.

Second, the new lending facility can effectively abate crisis contagion without inducing moral hazard under information asymmetry. The new facility based on the timing order of crises allows the international crisis lender to distinguish an autonomous crisis from an imported crisis even under information asymmetry. By not rescuing the country hit first, the timing-based crisis-lending facility can penalize the country in an autonomous crisis (which has a high chance of not having made adequate crisis-prevention efforts) and, hence, deter the moral hazard problem as effectively as the laissez-faire approach. By rescuing the countries hit later, the new facility could assist the countries that are more likely to have been hit by imported crises and deter crisis contagion as effectively as the existing lending facilities of the IMF. In this way, the new facility could overcome the shortcomings of laissez-faire and existing lending facilities.

Third, when both problems of moral hazard and crisis contagion are serious, the new lending facility can be Pareto-superior to both laissez-faire and existing crisis lending facilities. If there is no crisis contagion, laissez-faire could be the best approach. If the moral hazard problem is negligible, the existing lending facilities of the IMF would induce the optimal outcome. However, when both problems of moral hazard and crisis contagion are serious, both laissez-faire and existing lending facilities can be suboptimal: laissez-faire is not effective in deterring crisis contagion and existing lending facilities are not effective in containing moral hazard. Therefore, a new timing-based facility that would overcome the shortcomings of the other approaches can provide the highest world welfare.

The model used in this study can be adapted to accommodate three varied situations: a case in which a string of crises are generated by high common vulnerabilities to a common external shock, rather than by crisis contagion;9 a case in which countries may reverse the timing order between autonomous and imported crises; and a case in which timing-based lending induces greater crisis-prevention efforts than laissez-faire. These variations, however, do not alter the results of this study on the effectiveness of the timing-based lending facility. For example, a variant model shows that even for a common shock, the timing-based lending facility can be Pareto-superior to both laissez-faire and existing lending facilities. The intuition behind the result is straightforward. Suppose that the common vulnerability (or the risk of global or regional crises in the event of a common shock) could be substantially reduced only if all the countries simultaneously made their own efforts to prevent crises. Given the situation, a coordination problem may arise under both laissez-faire and existing facilities: since all the crisis countries are treated the same regardless of their crisis-prevention efforts, individual countries will not make strong efforts to prevent a crisis. In contrast, the timing-based lending facility, by penalizing the first crisis country, can promote competition among borrower countries not to become the first one. The competition might then induce all the countries to voluntarily upgrade their prevention efforts to the optimal level for the world (or the region) as a whole, which would help resolve the coordination problem.

This paper is organized as follows. Section II proposes a new crisis-prevention facility. Section III presents a simple model as an analytical framework. Sections IV and V analyze the cases of the laissez-faire and existing crisis-lending facilities, respectively. Section VI explores the new timing-based facility. Section VII discusses some extensions of the model. Section VIII concludes.

II. Proposal for a New Crisis-Prevention Facility

In this section, I propose a new lending facility for crisis prevention, entirely different from existing international lending facilities including the Stand-By and the CCL of the IMF and their basic frameworks.

In concrete terms, the paper proposes a two-step crisis-resolving lending facility that differentiates rescue financing by the timing order of crises. More specifically, the new facility operates based on the following rules:

  1. no country where a crisis first occurs (after a considerable period of time, say more than a year, from the previous international financial crisis)10 shall receive rescue financing; and

  2. all countries where crises erupt (within a short period of time, say less than a year) after the initial crisis in (1) shall receive rescue financing.

Such rules use the information derived from the timing of crises; the country hit by crisis first is very likely to have the crisis because of its own problems, while the country facing a crisis shortly after the first crisis has a high chance that contagion causes the crisis. Given the information, the new lending facility aims at punishing the country of the first crisis to abate moral hazard but saving the country of later crisis to prevent contagion.11

To speak figuratively, this timing-based rescue lending facility can be considered as a facility that distinguishes between the house where the fire first started and the houses that caught fire from the first house. What the facility does is to put out the fire on houses that later caught fire in order to stop further spreading, that is, crisis contagion; but not to save the house that first started the fire to deter moral hazard.

An important strength of the timing-based facility is that it can fulfill its role of obviating moral hazard under information asymmetry better than existing crisis lending facilities. It can again be likened to a fire fighter who does not know the story of each and every house: instead of monitoring all those houses one by one to prevent fire (which could be almost impossible), she can induce their awareness of the dangers of fire and voluntary preventive efforts, which are more effective for fire prevention.

In the following sections, I present a simple model to more clearly demonstrate the difference between existing international rescue lending facilities and the new timing-based lending facility, which can deter both moral hazard and contagion effect at the same time.

III. The Basic Model

The desirability of a crisis lending facility may hinge on its ability to resolve the problems of crisis contagion and moral hazard. To see how effectively the newly proposed lending facility can resolve these problems, I propose a simple model of crisis that focuses on crisis contagion and moral hazard.

A. Two Types of Crises

There are two periods, t = 1, 2. The world consists of a continuum of countries defined along the interval [0, 2N]. Each country, indexed by i, is populated with a continuum of consumers whose measure is one. There are two types of countries: borrowing country (i ∈ [0, N]) and lender country (i ∈ (N, 2N]). Each lender country has k amount of capital but does not have production technology, so that it has to lend its capital to borrower countries. Each borrower country has k amount of initial capital and production technology, and borrows ki,f amount of capital from lender countries at the beginning of the first period.12 The loan contract specifies the amount of repayment a borrower country has to make at the end of the final period as rki,f. If the borrower country cannot repay rki,f, it is in default.

Each borrower-producer country can fall into an economic crisis that depresses productivity, and there are two kinds of economic crisis depending on the origin of the crisis. One is autonomous crisis that occurs internally as a result of domestic factors, and the other is imported crisis, which is the outcome of crisis contagion from foreign countries. The possibility of having an autonomous crisis can be lowered by voluntary efforts of crisis prevention on the part of each country. For simplicity, assume that the probability of having an imported crisis can hardly be reduced by such voluntary efforts.13

To discuss the two types of crises more formally, assume that each borrower-producer country produces by the following linear production function14

yi = Aiki, i  [0, N],(1)

where Ai represents the productivity shock that takes A or B(< A), and ki is capital input employed by the i-th country. Total capital employed by a borrower country is the sum of domestic capital (ki,d) and foreign borrowings, that is, ki = ki,d + ki,f. Capital depreciates 100 percent at the end of t = 2. The initial capital endowment of each country (k) is normalized at k = 1/2, which yields ki,d = ki,f = k = 1/2 and ki = 2k = 1.

Completion of production, defined in Eq.(1), takes two periods, as production goes through two stages. The first stage occurs in the first period (t = 1), when an autonomous crisis can occur. The second stage happens in the second period (t = 2), when a bad productivity shock due to contagion can be created. During the second period, crisis contagion could take place in two rounds: the crisis in the country having an autonomous crisis in the first period is spread to some countries (the first round of contagion), and the crisis is further spread from the infected countries to other countries (the second round of crisis contagion).

For simplicity, the basic model assumes that autonomous crises occur only in the first period while imported crises occur only in the second period. This assumption formalizes the idea that an imported crisis is very unlikely to occur without the preceding occurrence of a crisis from which it is infected. To be more general, however, Section 7.2 allows autonomous crises to occur in the second period as well and imported crises to occur in the first period as well, and derives similar results.

Given the above production process, the productivity shock is revealed in each period. Let Ati denote the productivity shock (which takes A or B) in t (= 1, 2). Assume that the productivity that a country has for the whole production process, Ai, is determined as

Ai=min[A1i,A2i].(2)

So if the production process goes without a bad shock in any of the two stages, the productivity is determined to be A. But if a country experiences a bad shock in one of the two production stages, the productivity is given by B. Whether a country has a good productivity shock (A) or a bad productivity shock (B) in each period (t = 1, 2) is observable by anybody (including the international crisis lender).

Let θ1i be the probability of a country’s having a crisis in the first period, and θ2i be the probability that the country does not have a crisis in the first period but in the second period. Let θAi be the probability of a country’s having an autonomous crisis, and θCi be the probability that the country is hit by an imported crisis. Then the crisis probability θi of each country is given by

θi=θ1i+θ2i=θAi+(1θAi)θCi.(3)

Then the probability distribution of productivity, Ai, can be summarized by

Ai={Awith probability (1θi)B(<A)with probability θi,(4)

which tells us that a country has productivity B with probability θi, or productivity A, which is bigger than B, with probability (1 − θi). As a result, the country’s final output, yi, is determined to be Bki with probability θi, and Aki with probability (1 − θi) at the end of the final period (t = 2). Suppose that r is given as B < r < A.

The above assumptions imply that the output of a country in crisis is equal to B, regardless of the origin of the crisis. Therefore, in this model it is impossible to distinguish between autonomous crisis and imported crisis by observing the amount of output. It is possible, however, to distinguish between the two types of crises by observing the timing order of crises, given that an imported crisis is unlikely to occur in the first period.

B. Probability of Autonomous Crisis

A borrower-producer country may have an autonomous crisis in the first period. The probability that a borrower country is hit by an autonomous crisis, θAi, can change according to how much the country strives to prevent crisis by resolving its own macroeconomic or structural problems at the beginning of the first period (but after foreign borrowings are made). That is, the probability of autonomous crisis occurring depends on each country’s efforts for crisis prevention. We denote the efforts of each country for crisis prevention as ei, defined on the interval ∈ [0, 1]. The probability of a country’s having an autonomous crisis, θAi, is given by

θAi=θA(ei),(5)

which suggests that θAi is affected only by ei.

Here we assume the relationship between ei and θAi as

θAiei<0,(6)

which tells us that the more efforts for crisis prevention, the lower the possibility of having an autonomous crisis.

Meanwhile, a crisis does not always break out for want of efforts by a country. That is, θA(0) = 1 will not be the case, and we have

θA(0)=θAmax<1,(7)

where θAmax is the maximum of θA(ei).

In addition, a country may have a crisis even in the case where it makes its maximum effort to prevent crisis. That is,

θA(1)=θAmin>0,(8)

where θAmin is the minimum of θA (ei).

Under the above assumptions, rescue financing provided by the international crisis lender can create a moral hazard problem. Particularly if the international community resolves any country’s financial crisis through bailouts, the countries would not put as much effort into crisis prevention as they would have without bailouts, which in turn creates moral hazard and raises θAi.

C. Probability of Crisis Contagion

Borrower countries may have crisis contagion in the second period. During the second period, countries may be infected by the crisis of a country having autonomous crisis in the first period or by the crises of the countries that were already infected in the second period. For simplicity, here I do not specify the specific contagion mechanism, while it could be through trade links or capital market links, or something else.15

The chance of crisis contagion in the second period (θCi) rises with the number of countries hit by crisis in the first period. In addition, the chance of crisis contagion is reduced by rescue financing in the first and second period, partly because the adverse effect that a financial distress in a crisis country could have on the production possibility of the other countries with close trade or financial links can be reduced by rescue financing (for simplicity we here do not explicitly introduce the mechanism through which rescue financings abate contagion effects).

To be more specific, let nt denote the measure of the countries that have a crisis in period t(= 1, 2),16 and ϕt be the fraction of the crisis-hit countries that receive rescue financing from the international crisis lender among all the crisis countries in period t. Now assume that the probability of crisis (due to contagion) is given by

θCi=γ(aϕ1)(bϕ2)n1, a,b>1,(9)

where γ(> 0) captures the strength of crisis contagion, and a(> 1) and b(> 1) represent the part of crisis contagion that is not affected by the international crisis lender’s rescue financing in the first and second period, respectively.17

This assumption suggests that θCin1>0 and θCiϕ1,θCiϕ2<0. That is, the more countries in the world experience a crisis in the first period or the less crisis-countries are bailed out by the international crisis lender, the more countries have a crisis (due to contagion) in the second period.

The assumption of a, b > 1 implies that θCi=γ(a1)(b1)n1>0 when ϕ1= ϕ2 = 1. That is, even in the case where all crisis countries receive rescue financing, crises due to contagion will occur (though its chance will be reduced given n1).

For notational convenience, define δ1θCi, while the range of δ is given by: δ ∈ [δl, δu] where δl1γabNθAmax and δu1γ(a1)(b1)NθAmin.

D. International Crisis Lender

There is the international crisis lender (or the international lender of last resort) which may provide rescue financing to crisis countries. If it decides to rescue a country in crisis, it provides financing which amounts to ρ = rki,f.18 This amount of rescue financing is just large enough for the country in crisis to make repayments. As rki,f amount of crisis lending is provided, the country in crisis makes the required repayment rki,f to the lender countries (therefore it is no longer in default).

For simplicity, assume that the reserves for rescue financing are raised from contributions made by all the countries. So each country’s contribution to the reserves can be seen as insurance premium, and rescue financing can be thought of as insurance payments. As a result, the countries that receive rescue financing here will not have to repay ρ to the international crisis lender later. Regarding the timing of making contributions, assume that contributions were already made in the past (before t = 1). Therefore, the international crisis lender does not have to raise the reserves in the initial period (t = 1). Of course, we may instead assume that the international crisis lender raises the reserves at the beginning of the initial period, rather than before the initial period. Or we may assume ρ to be just a pure lending that should be repaid later. But these variations do not change the main results (see Appendix B for the analysis on these cases).

Furthermore, there is information asymmetry between the international crisis lender and the countries. The international crisis lender cannot accurately observe nor verify the effort made by each country for crisis prevention, ei, possibly reflecting the situation with huge state verification costs (Townsend, 1979).

E. Government

The government of each borrower-producer country can reduce the chance of crisis by exerting efforts to prevent autonomous crisis at the beginning of the first period. Crisis prevention indubitably calls for expenses, time and efforts. The cost (in terms of the utility of the representative consumer) that a country has to entail for crisis prevention, denoted by υ(ei), is given by

υ(ei)=ψei,(10)

where ψ(> 0) captures the cost per unit of crisis prevention effort.

The government seeks to maximize the expected utility of the representative consumer by choosing ei:

maxei E(u(ci,ei))  = E(u(ci))v(ei)= E(ln ci)Ψei,(11)

which suggests that more consumption or less efforts to prevent crisis induce higher utility, while consumption raises utility following a logarithmic function.19

F. Costs of Debt Repudiation

At the end of the second period when production is completed, borrower countries are supposed to repay rki,f to lender countries. Assume that any borrower country repays to lender countries in two cases: where it is not hit by crisis, or where it is hit by crisis but receives rescue financing from the international crisis lender. For simplicity, however, the borrower country does not repay in the case where it is hit by crisis and the international crisis lender does not provide rescue financing.20

If a borrower country repudiates its foreign debts, lenders would impose sanctions. As a result, the borrower country would not be able to borrow foreign capital nor have access to trade credits which are imperative for the next production cycle (while the current model does not explicitly introduce the next production cycle by assuming that there is only one production cycle that takes two periods).21 So defaulting on foreign debts would be very costly. Reflecting such costs, I assume that if a country does not repay rki,f, it incurs costs which amounts to λ (in terms of the utility of the representative consumer).

A country has the highest welfare (or utility of the representative consumer) when it is not hit by crisis. It has the second highest welfare when it is hit by crisis but receives rescue financing, and the lowest welfare when it is hit by crisis and does not receive rescue financing. More specifically, we have ln(Ar/2) > ln B > ln B − λ.22

IV. Laissez-Faire

Under the environment described in the basic model above, what would be the best way to minimize both moral hazard and contagion? This section examines how the equilibrium levels of moral hazard and crisis contagion are determined under laissez-faire.

In this case of laissez-faire, no rescue financing is provided to the countries in crisis, that is, ϕ1 = ϕ2 = 0. Given Eq.(9), it then follows that θCi=γabn1. So crisis contagion occurs to its maximum given n1.

Then, if a borrower country does not have a bad productivity shock in any of the two periods, the consumption of the representative consumer is given by Akirki,f = (Ar/2) (recall that ki = 1 and ki,f = 1/2). If the country was hit by a bad productivity shock in one of the two periods, it does not repay rki,f. So the consumption is given by Bki = B, while the borrower country incurs the cost due to debt repudiation, λ.

Using these, the expected utility in the absence of international crisis lending is given by

E(u(ci,ei))=E(lnci)ψei   =(1θi) ln(Ar/2)+θiln Bθiλψei.(12)

For notational convenience, define wiE(u(ci, ei)). Then the government of each borrower country (i ∈ [0, N]) will solve the following optimization problem [P1]:

[P1]maxwiei=[1  θA(ei)  (1  θA(ei))(γabn1)] ln (Ar/2)+[θA(ei)+(1  θA(ei))(γabn1)](lnB  λ)  ψei.(13)

In solving [P1], I focus on the case where in the absence of the international crisis lender, each borrower country makes positive efforts for crisis prevention. For this, I assume that the following condition holds.

Condition (i)

For any ei ∈ [0,1] and δ ∈ [δl, δu], it holds that

ψ+θA(ei)eiδλ<θA(ei)eiδ(ln(Ar/2)+lnB)<ψ(14)

which tells us that the cost of crisis prevention (ψ) and the cost of debt repudiation (λ) are not negligible.

Under Condition (i), differentiating wi with respect to ei gives

wiei=[θAiei(1γabn1)+(1θAi)γabn1θAiθAiei] (ln(Ar/2)+lnBλ)ψ.(15)

Note that given the assumption of a continuum of borrower countries, the measure of the countries having an autonomous crisis (n1) is given by n1=0NθAjdj. It then follows that

n1θAi=0,(16)

which suggests that the measure of the countries having an autonomous crisis is not affected by the change in the i-th country’s probability of having an autonomous crisis, θAi.

Using Eq.(16), the derivative in Eq.(15), regardless of the other governments’ effort for crisis prevention, becomes

wiei=θAiei(1γabn1)(ln(Ar/2)+lnBλ)ψ.(17)

which, given Condition (i), is positive for all ei ∈ [0, 1].

So the optimal solution of the i-th government for crisis prevention in this case (denoted by ei,L) is given by

ei,L=1,(18)

regardless of the other countries’ efforts to prevent their own autonomous crises.

This holds for all the borrower-producer countries, and hence the Nash equilibrium is determined as a case where all the countries make maximum efforts to prevent crisis: ei,L = 1, for all i ∈ [0, N]. So in this case, there is no moral hazard problem.

As a result, the probability of a country’s having an autonomous crisis is at its minimum: θA(1)=θAmin>0. Note that though all the borrower countries make efforts for crisis prevention to its maximum, some of them experience autonomous crises (Eq.(8)). The measure of those countries is given by n1=0NθAmindj=NθAmin.23 Given a positive n1, the chance that a country has a crisis due to contagion, θCi, is given by θCi=γabn1=γabNθAmin. This suggests that in equilibrium contagion occurs, particularly at its maximum for a given n1.

In this case, lender countries (i ∈ (N, 2N]) receive rki,f from the borrower countries that did not have a crisis, and nothing from the countries that had a crisis. So each lender country, which invests evenly across borrower countries for risk pooling, will receive rki,f(1θi)=(r/2)(1θAmin(1θAmin)γabNθAmin).

Now define the world’s welfare to be the sum of the utility of the representative consumer over all the borrower and lender countries (that is, 0N[E(lnci)ψei]di+N2NE(lnrki,f)di). Then the world’s welfare in the absence of the international crisis lender, denoted by WL, is given by

WL=N[(1θAmin(1θAmin)γabNθAmin)ln(Ar/2)+(θAmin+(1θAmin)γabNθAmin)(lnBλ)ψ]+N[ln(r/2)+ln(1θAmin(1θAmin)γabNθAmin)].(19)

Note that in the absence of the international crisis lender, moral hazard can be prevented. However, crisis contagion cannot be effectively deterred under the laissez-faire approach. As a result, the welfare of the entire world may not attain its maximum possible value. The reason is that externalities due to crisis contagion cannot be resolved by each individual country under laissez-faire. Hence there may arise the need for an international coordinator or crisis lender.

V. Existing Crisis-Lending Facilities

How can the international crisis lender or the international lender of last resort resolve the problem of externality due to crisis contagion? What would be the effect of a contagion-preventing rescue policy on the problem of moral hazard?

A. Traditional Crisis Lending Facilities

First we examine how traditional crisis lending facilities of the IMF such as the Stand-By Arrangement affect each of contagion and moral hazard problems. Traditional lending facilities are often offered to countries in crisis, regardless of whether the countries are hit by autonomous crisis or imported crisis.

Suppose that under such lending facilities, all the crisis-hit countries receive rescue financing, and hence we have ϕ1 = ϕ2 = 1. Given Eq.(9), it then follows that θCi=γ(a1)(b1)n1. So crisis contagion could be minimized given n1.

Recall that the international crisis lender is assumed to provide rescue financing which amounts to ρ = rki,f to each country in crisis. Furthermore, if a borrower country does not have a bad productivity shock, the consumption of its representative consumer is given by (Ar/2), as under laissez-faire. If the country is hit by a bad productivity shock, the representative consumer consumes B, while the country does not incur the cost due to debt repudiation, λ. Using these, the maximization problem of each borrower country can be expressed as

[P2]maxwiei=[1θA(ei)(1θA(ei))(γ(a1)(b1)n1)] ln (Ar/2)+[θA(ei)+(1θA(ei))(γ(a1)(b1)n1)] ln Bψei.(20)

Using Eq.(16), the derivative of wi with respect to ei is given by (regardless of the other countries’ efforts for crisis prevention)

wiei=θAiei(1γ(a1)(b1)n1)(ln(Ar/2)+lnB)ψ,(21)

which, under Condition (i), is negative for all ei ∈ [0, 1], suggesting that a country’s raising efforts for crisis prevention reduces its welfare.

Therefore the optimal solution of ei for any borrower country in this case, denoted by ei,T, is given by

ei,T=0,(22)

regardless of the other countries’ decisions. Therefore, in Nash equilibrium, we have ei,T = 0 for all the counties. This suggests that traditional crisis lending facilities, which provide countries with an insurance against the happening of a bad productivity shock, create moral hazard.

In the equilibrium, the probability that a country has an autonomous crisis is at its maximum: θA(0)=θAmax. Meanwhile, the chance that the country has contagion, θCi, is given by θCi=γ(a1)(b1)NθAmax.

In this case, lender countries receive rki,f (= r/2) from the borrower countries regardless of the occurrence of crisis as the international crisis lender fully bails out defaulting countries. The world’s welfare in this case (denoted by WT) is then given by

WT=N[(1θAmax(1θAmax)γ(a1)(b1)NθAmax) ln (Ar/2)+(θAmax+(1θAmax)γ(a1)(b1)NθAmax) ln B]+N[ln(r/2)].(23)

In sum, traditional crisis-lending facilities are able to contain the problem of contagion but begot the problem of moral hazard. From the standpoint of the world as a whole, however, the optimal level of efforts to prevent crisis may be positive. In this regard, traditional lending facilities could be sup-optimal.

B. Contingent Credit Lines (CCL)

Now we explore how the Contingent Credit Lines (CCL), which was introduced by the IMF in 1999, would function under information asymmetry. As discussed earlier, a key idea of the CCL is prequalification. When countries that satisfy some prequalification are hit by a bad productivity shock (either in the first or second period), the CCL facility provides rescue financing.

The key to efficient operation of the CCL is how accurately the international crisis lender can observe and verify ei. If there is no information asymmetry between the international crisis lender and each country, the CCL would work as follows. The international crisis lender would set a threshold for the country’s efforts to prevent crisis (denoted by eth ∈ (0, 1]), and it would allow only the countries that meet the threshold level to use the CCL facility in the event of economic crisis. More specifically, when the country’s actual effort for crisis prevention, ei, is less than the threshold level, eth, the rescue financing would be set to be zero (ρ = 0), and when it is higher than or equal to the threshold, then the rescue financing would be set positive at ρ = rki,f = r/2.

However, due to informational asymmetries, it is often very difficult for the international crisis lender to accurately observe and verify each country’s efforts to prevent crisis, ei. Recall that the basic model assumes the case of information asymmetry where ei cannot be observed nor verified accurately by the international crisis lender. In this case, its rescue financing (ρ) has to rely on the observed efforts, denoted by e˜i, not the actual efforts ei. More specifically, the rule for rescue financing under the CCL is set as

ρ={0if e˜i<ethrki,fif e˜ieth.(24)

That is, when the observed effort for crisis prevention (e˜i) is lower than the threshold level, no rescue financing is provided, and when it is higher than or equal to the threshold, then the amount of rescue financing that is just enough to satisfy the repayment requirement rki,f is provided.

To highlight the role of information asymmetry, suppose that the international crisis lender’s observation (e˜i) on the actual effort by the government of each country (ei) is inaccurate. Particularly assume that

e˜i={0if ei<α1if eiα,(25)

which suggests that the observed effort (e˜i) takes one if the true effort is not lower than α, and zero otherwise.

Then a country will receive a positive rescue financing in the event of crisis as long as its true effort (ei) is greater than or equal to α. It is because in this case the country’s effort observed by the international crisis lender will be e˜i=1. But the country will receive no rescue financing in the event of crisis if its true effort (ei) is lower than α.

Given these, the government of each borrower country with private information solves the following optimization problem [P3]:

[P3]maxei wi=Φ(ei)[(1θA(ei)(1θA(ei))γ(a1)(b1)n1) ln (Ar/2)+(θA(ei) +(1θA(ei))γ(a1)(b1)n1)lnBψei]+(1Φ(ei))[(1θA(ei) (1θA(ei))γabn1)ln(Ar/2)+(θA(ei)+(1θA(ei))γabn1)(lnBλ)ψei],(26)

where Φ(ei) is an index number which takes one if eiα and zero otherwise.

To solve the optimization problem, consider first the case where ei < α. If a borrower country makes crisis prevention efforts within this range, the international crisis lender observes e˜i=0. Therefore, there is no rescue financing (ρ = 0). In addition, Φ(ei) = 0, and hence the optimization problem [P3] is reduced to [P1], which is for the case of laissez-faire. Recall that given Condition (i), it holds wiei>0. Therefore, it would be better for the government to raise ei as high as possible within the range of ei (< α).

Now consider the case where eiα. For this range of ei, the international crisis lender observes e˜i=1. So there will be full bailouts (ρ = rki,f) regardless of the amount of ei. We also have Φ(ei) = 1, and hence the optimization problem [P3] is reduced to [P2], which is for the case of traditional lending facilities. Under Condition (i), it holds wiei <0 for this range of ei. Hence the government would reduce ei as low as possible within the range of eiα.

For simplicity, assume that the expected utility evaluated at ei = α (that is, E(u(ci, α))), is not greater under laissez-faire than under traditional lending facilities. It then follows that the optimal level of crisis prevention effort in this case (ei,C)is given by

ei,C=α,(27)

which suggests that each borrower country sets ei at the lowest level among the eis which satisfy the prequalification (that is, e˜ieth).

Under the CCL, any country satisfying prequalification receives rescue financing when it faces a bad productivity shock. Therefore in this case, contagion can be contained by rescue financing, as under traditional lending facilities. In addition, the possibility of an autonomous crisis is determined to be θA(α), which is higher than under laissez-faire but lower than under traditional lending facilities. This suggests that the CCL may improve over traditional lending facilities by reducing the moral hazard problem.

However, the extent to which it abates the moral hazard problem depends on the degree of information asymmetry. Particularly when the problem of information asymmetry is substantial, the CCL might not operate effectively.

To illustrate this, consider the case where α is close to zero. Then a country will receive rescue financing in the event of crisis, even when it made only a little effort for crisis prevention. In this case, the CCL would become almost the same as in the Stand-By facility case, and the world’s welfare under the CCL could be approximated by that under traditional rescue financing facilities (that is, Eq.(23)).

The above discussions suggest that under information asymmetry, the CCL may not be effective in reducing moral hazard. When ei cannot be observed and verified accurately, setting the threshold level of ei and determining the prequalification based on that threshold level can lead to such critical problem.

VI. Timing-Based Crisis-Lending Facility

In the previous sections, we have shown that existing international lending facilities may be ineffective in reducing moral hazard, particularly under information asymmetry. This section will examine how the new lending facility proposed in Section 2 works in our model, and show that it can effectively resolve both problems of moral hazard and crisis contagion.

A. Timing-Based Crisis-Lending Facility

Within the model, the new facility can be formulated as follows:

  1. ϕ1 = 0, that is, in the first period, no country receives rescue financing; and

  2. ϕ2 = 1, that is, in the second period, any country with a bad productivity shock receives rescue financing.

So rescue financing (ρ = rki,f) is provided to any crisis country in the second period, but to none in the first period.24 Given Eq.(9), we then have: θCi=γa(b1)n1. In addition, a country, if hit by crisis in the first period, produces Bki = B, and does not repay rki,f, while it incurs the cost due to nonrepayment, λ. So the utility from consumption in the event of crisis in the first period is ln B − λ. Meanwhile, the country, if hit by crisis in the second period, produces B, but receives rescue financing and repays ρ = rki,f. So the utility from consumption is ln B. If the country does not have a bad productivity shock, the utility from consumption is ln(Ar/2).

Given these, the government of each borrower country solves the following optimization problem [P4]:

[P4]maxei wi=[1θA(ei)(1θA(ei))(γa(b1)n1)] ln (Ar/2)+θA(ei)(lnBλ)+(1θA(ei))(γa(b1)n1)lnBψei.(28)

Given n1θAi=0, differentiating wi with respect to ei gives

wiei=θAei[(1γa(b1)n1)(ln(Ar/2)+lnB)λ]ψ,(29)

which, given Condition (i), is positive for any ei.

So the optimal solution for the government of any borrower country in this case, denoted by ei,N, is given by

ei,N=1,(30)

and therefore we have ei,N = 1 for all i ∈ [0, N] in Nash equilibrium.

This suggests that the timing-based crisis prevention facility raises the countries’ effort for crisis prevention to the maximum under information asymmetry, in sharp contrast with existing international lending facilities. By exerting maximum efforts for crisis prevention (ei,N = 1), each country reduces its probability of having an autonomous crisis to θAmin. Note that given ei is not greater than one, the new lending facility and the laissez-faire approach induce the same amount of crisis prevention efforts: ei,N = ei,L = 1. However, if we allow ei to be unbounded from the above, the timing-based lending facility can induce higher crisis prevention efforts than laissez-faire (see Section 7.3).

In addition to raising crisis prevention efforts, this timing-based facility may help contain crisis contagion. Note that it does not provide rescue financing to the countries experiencing autonomous crises in the first period (ϕ1 = 0). Therefore, once autonomous crises occurred, it cannot prevent the first round of crisis contagion. However, it lowers the chance of crisis contagion to γa(b1)θAminN by reducing the number of countries in autonomous crisis to θAminN compared to existing lending facilities. Furthermore, the new facility helps contain the second round of crisis contagion by providing a full rescue financing to the countries having crises owing to contagion during the second period (ϕ2 = 1).

The above discussion suggests that the new lending facility can effectively resolve both problems of moral hazard and contagion under information asymmetry. The new facility based on the timing order of crises allows the international crisis lender to distinguish between autonomous crisis and imported crisis under information asymmetry. By not rescuing the country hit first, the timing-based lending facility can punish the country in autonomous crisis and hence deter the moral hazard problem as effectively as laissez-faire. By rescuing the countries hit later, the new facility can assist the countries hit by imported crises and, therefore, deter crisis contagion as effectively as existing lending facilities of the IMF. In this way, the new facility may overcome the shortcomings of laissez-faire and existing lending facilities.

Now we examine the world’s welfare. In this case, lender countries receive nothing from the countries that had a crisis in the first period, but rki,f from the countries that did not have a bad shock in any period and those that had a bad shock in the second period. So each lender country will receive rki,f(1θAmin)=(r/2)(1θAmin). Then the world’s welfare in this case is given by

WN=N[(1θAmin(1θAmin)γa(b1)NθAmin) ln (Ar/2)+θAmin(lnBλ)+(1θAmin)(γa(b1)NθAmin)lnBψ]+N[ln(r/2)+ln(1θAmin)].(31)

In the following subsections, I compare the world’s welfare under the timing-based lending with those under laissez-faire and existing international lending facilities.

B. Welfare Comparison with Laissez-Faire

I first compare the world’s welfare under the timing-based lending facility (Eq.(31)) with that under laissez-faire (Eq.(19)). The difference in welfare, denoted by ΔWN−L(≡WNWL), is given by

ΔWNL=N[(1θAmin)γaNθAmin(ln(Ar/2)lnB)+(1θAmin)γabNθAminλ]+N[ln(1γabNθAmin)],(32)

which suggests that an introduction of the new lending facility in an laissez-faire economy reduces the chance of contagion without causing a moral hazard problem (and hence without raising the chance of autonomous crisis).

From Eq.(32), it is obvious that

ΔWNL>0,(33)

that is, WN is greater than WL. This result suggests that the timing-based lending facility can bring higher welfare than laissez-faire in the presence of crisis contagion. The intuition behind this result is clear. The timing-based lending facility is as effective as laissez-faire in preventing moral hazard, because it induces countries to put as much voluntary crisis prevention efforts as laissez-faire. Moreover, the new facility generates additional benefits of reducing crisis contagion, which the laissez-faire approach cannot.

C. Welfare Comparison with Existing Facilities

Next, I compare the world’s welfare between the timing-based facility and existing international rescue lending facilities. The difference in welfare, denoted by ΔWN−T(≡WNWT), is given by

ΔWNT=N[((θAmaxθAmin)+((a1)θAmax(1θAmax)aθAmin(1θAmin))γN(b1))(ln(Ar/2)lnB)]+N[θAminλψ +ln(1θAmin)].(34)

The terms inside the first bracket of Eq.(34) represent the welfare gain of the timing-based facility compared to existing lending facilities, and the terms inside the second bracket capture its welfare costs. The welfare gain depends on how effectively the new facility can reduce the chances of crises by abating the moral hazard problem that arises under traditional lending facilities. As Eq.(34) indicates, the welfare gain due to the crisis prevention effect critically depends on θAmin for a given θAmax. The smaller θAmin is, the larger the welfare gain is.

Assume that when θAmin goes to zero (so the crisis prevention effect is at its maximum), the welfare gain due to the crisis prevention effect is greater than the welfare cost. More specifically, assume that

θAmax[1+(1θAmax)(a1)γN(b1)](ln(Ar/2)lnB)>ψ.(35)

Given (35), it can be shown that there exists a threshold level of θAmin, denoted by θAth, below which we have

ΔWNT>0.(36)

This tells us that as long as θAmin is lower than θAth for a given θAmax (that is, the crisis prevention effect is greater than a certain threshold level), we have WN > WT (see Appendix A for the proof).

This result suggests that the new lending facility can induce higher world’s welfare than existing lending facilities in the presence of the moral hazard problem. The timing-based facility, by inducing ei,N = 1, can prevent moral hazard more effectively than existing lending facilities. It can also help reduce the chance of crisis contagion by reducing the number of the countries hit by autonomous crisis, let alone by providing rescue financing to the countries hit by crisis in the second period. In contrast, existing lending facilities contain crisis contagion only by providing rescue financing to the countries that have already had a crisis.

Note that the new facility can help resolve a coordination problem that may arise under existing lending facilities. When all the countries make efforts to reduce their own chances of autonomous crisis, there will be a substantial decline in the number of the countries hit by autonomous crises in the first period. Given crisis contagion and consequent externality, such efforts will also reduce the number of the countries hit by imported crises in the second period. As a result, the world’s welfare gain from the countries’ crisis prevention efforts can far exceed an individual country’s welfare gain. Under existing facilities, however, individual countries, which seek to maximize their own welfare, do not make crisis-prevention efforts up to the optimal level for the world as a whole. In contrast, the new facility induces countries to undertake crisis prevention efforts to the optimal level.

In sum, the above results suggest that the timing-based facility can be Pareto-superior to both laissez-faire and existing lending facilities. In the absence of crisis contagion, the laissez-faire approach could be the best. Without the moral hazard problem, existing crisis lending facilities could be the optimal. In the presence of both problems of moral hazard and crisis contagion, however, the new lending facility can provide the highest welfare to the world by effectively resolving both problems.

VII. Extensions and Discussions

The basic model allows for some extensions to explore various situations that may potentially affect the effectiveness of the timing-based lending facility. In this section, I address three such extensions: the case of high common vulnerabilities to a common shock; the case where the reversion of the timing order of crises is allowed; and the case where crisis prevention efforts could be without bound. I will show that those extensions do not alter the main results.

A. Common Shocks and Vulnerabilities

Some strings of crises may occur due to crisis contagion while others due to high common vulnerabilities to a common shock. The basic model focuses on the former case given that a vast majority of existing studies find evidence of crisis contagion (Kaminsky, Reinhart and Végh, 2002). However, we can modify the basic model to show that even in the case of common shocks and vulnerabilities (Mody and Taylor, 2003), the timing-based facility would bring an optimal outcome.

Consider a variant model where a series of crises occur due to high common vulnerabilities combined with a common shock. Let ϵ denote a common shock, and χ ∈ [χmin, χmax] denote common vulnerabilities. Under common vulnerabilities, borrower countries face the same probability of having a crisis. The probability of a country’s experiencing a crisis depends on the common shock and common vulnerabilities as

θi=χϵ.(37)

So the higher χ or ϵ is, the more likely the country is to suffer a crisis.

The productivity shock in any of the two periods here is the result of a common shock. It is not an autonomous nor an imported shock. For simplicity, assume that a country may have a bad productivity shock in any of the two periods, but not in both periods. Let θ˜1i be the chance of having a crisis due to the common shock in the first period, and θ˜2i be the chance of having a crisis due to the common shock in the second period. Then we have: θi=θ˜1i+θ˜2i=χϵ.

The possibility of a country’s having a crisis in any of the two periods, θi, cannot be lowered only by its own voluntary efforts to reduce it, given θi = χϵ. However, if a country strives to prevent crisis by solving its own structural or macro problems (at the beginning of the first period), it may delay the occurrence of crisis.25 More specifically, assume that

θ˜1iei=θ˜2iei<ψ/λ.(38)

which tells us that a country’s crisis prevention efforts cannot lower its chance of crisis throughout the two periods (θi), but its chance of having crisis earlier (θ1i). The more efforts for crisis prevention, the lower the possibility of having a crisis in the first period. Eq.(38) also indicates that the reduction in the chance of crisis in the first period, induced by a country’s crisis prevention efforts, will be offset by the same increase in the chance of crisis in the second period.

The magnitude of common vulnerabilities (χ) depends on how many countries make efforts to prevent crisis. It rises with the number of the countries that make crisis prevention efforts. To be more specific, assume that

χ=χ(0Neidi),χ(0Neidi)<0.(39)

Given the assumption of a continuum of borrower countries, we have χei = 0, suggesting that a single country’s crisis prevention efforts hardly affect the common vulnerability. However, if all the countries put efforts to prevent crisis at the same time, it is lowered.

Under the variant of the model, now examine how the equilibrium levels of global financial risks are determined under laissez-faire, existing crisis lending facilities, and the timing-based lending facility. First, consider the laissez-faire case. In this case, no crisis-hit country receives rescue financing. As a result, the government of each borrower country solves the following optimization problem (at the very beginning of the first period):

maxei wi=[1θ˜1iθ˜2i]ln(Ar/2)+[θ˜1i+θ˜2i](lnBλ)ψei.(40)

Under Eq.(38), together with the assumption of a continuum of borrower countries, we have wi ei=ψ<0, regardless of the size of the common vulnerability, χ. So the optimal solution of a borrower country is given by

ei,L=0,(41)

regardless of the other countries’ effort to prevent crisis. As this holds for all the borrower countries, the Nash equilibrium is determined as a case where all the borrower countries make no efforts to prevent crisis. As a result, the common vulnerability is at its maximum: χ(0N0di)=χ(0)=χmax.

The intuition behind the result is straightforward. Under laissez-faire, there is no difference in rescue financing between the first-hit country and the others. So borrower countries have no incentive to avoid being the first-hit country. Therefore, no country will make efforts to prevent crisis. So in equilibrium, global risks will be at its maximum.

Second, consider traditional crisis lending facilities, under which any crisis country receives rescue financing. In this case, the optimization problem of each borrower country is given by

maxei wi=[1θ˜1iθ˜2i]ln(Ar/2)+[θ˜1i+θ˜2i]lnBψei,(42)

which yields wi ei=ψ<0 and ei,T = 0, regardless of the other countries’ decisions. Therefore, in Nash equilibrium, we have ei,T = 0 for all the borrower countries, and χ(0) = χmax, the same as under laissez-faire. The reason for the common vulnerability’s being at its maximum here is similar to the laissez-faire case. Under traditional lending facilities, rescue financing is provided to any crisis country regardless of the timing of crises. So countries would have no incentive to strive not to be the first crisis country.

Finally, consider the timing-based lending facility. In this case, rescue financing is provided to any crisis country in the second period, but to no country in the first period. The maximization problem of each borrower country is then given by:

maxei wi=[1θ˜1iθ˜2i]ln(Ar/2)+θ˜1i(lnBλ)+θ˜2ilnBψei.(43)

Differentiating wi with respect to ei gives

wiei=θ˜1iei(lnBλ)+θ˜2ieilnBψ=θ˜1iei(λ)ψ,(44)

which, given Eq.(38), is positive. So the optimal solution for any borrower country in this case is given by

ei,N=1,(45)

regardless of the other countries’ efforts for crisis prevention. As a result, all the borrower countries make maximum efforts to prevent crisis in Nash equilibrium. In addition, the common vulnerability is at its minimum: χ=χ(0N1di)=χ(N)=χmin.

This suggests that in the case of common shocks and vulnerabilities, the timing-based lending facility can raise the countries’ efforts for crisis prevention to its maximum and, by doing so, reduce the risk of a large-scale global crisis. This is in sharp contrast with both laissez-faire and existing crisis lending facilities, under which the countries do not make any crisis prevention effort. It can also be easily shown (in an analogous way to Section 6) that in the case of common shocks, the timing-based facility can be Pareto-superior to both laissez-faire and existing lending facilities.

Note that the timing-based lending facility can resolve a coordination problem that arises under both laissez-faire and existing lending facilities. In the presence of common vulnerabilities, the risk of global crisis could be substantially lowered only when all the countries put crisis prevention efforts at the same time. From the standpoint of an individual country, however, its crisis prevention efforts can little affect the chance of its having a crisis, but only delay the timing of crisis. Under both laissez-faire and existing facilities, however, borrower countries do not have any incentive to make crisis prevention efforts, because all the crisis countries are treated the same regardless of the timing of crises. In contrast, the timing-based lending facility, by differentiating the first crisis country from the others, promotes competition among borrower countries not to become the first country. The competition could then induce all the countries to undertake their voluntary crisis prevention efforts, which helps resolve the coordination problem.

The above discussion, together with Section 6, suggests that the timing-based lending facility can be Pareto-superior to both laissez-faire and existing facilities regardless of whether crises are caused by contagion or common shocks. Therefore, even under uncertainty over whether crises are due to crisis contagion or common shocks, the timing-based facility can bring about a better outcome than laissez-faire and existing lending facilities.26

B. Reverse Timing Order and Racing for Delays

The basic model assumes that autonomous crises occur earlier than imported crises, given that in reality an imported crisis is less likely to precede the occurrence of a crisis from which it is infected. However, we cannot totally rule out the possibility that an autonomous crisis takes place after an imported crisis, particularly when countries try not to become the first crisis country. A variant of the basic model can be easily applied to the case of potential reversions of the timing order between autonomous and imported crises.

In this variant, assume that it may take time for a country that had a bad productivity shock to have a full-blown crisis. As a result, a bad autonomous shock may generate a crisis (and so observable by anybody) in the second rather than the first period. More specifically, if a country has a bad autonomous productivity shock, it may have a crisis in the second period with probability πAi>0 or in the first period with probability 1πAi. In addition, if it has a bad contagion shock, the country may have a crisis in the first period with probability πCi>0 or in the second period with probability 1πCi. So πAi and πCi represent the chances of the reversion of the timing order of crises (note that the basic model assumes πAi=πCi=0). In the following, we examine both cases where the probabilities of timing reversions are exogenously given and where they are endogenously determined.

First, consider the case where the probabilities of reversions are exogenously given by πAi=πA and πCi=πC. In this case, the probability of a country’s having a crisis in the first period is

θ1i=(1πA)θAi+πC(1θAi)θCi,(46)

and the probability of having a crisis in the second period is

θ2i=πAθAi+(1πC)(1θAi)θCi.(47)

In this case, the crisis probability of each country (θi) is the same as in the basic model (that is, Eq.(3) or θi=θ1i+θ2i=θAi+(1θAi)θCi). Under both laissez-faire and existing lending facilities, the optimization problem of a borrower country in this variant model then becomes the same as that of the basic model. As a result, the optimal level of crisis prevention efforts is determined to be ei,L = 1 under laissez-faire, and ei,T = 0 under existing lending facilities.

Under the timing-based lending facility, the optimization problem in this variant is slightly altered. The government of each borrower country solves the following optimization problem [P4]’ instead of [P4]:

[P4]maxei wi=[1θA(ei)(1θA(ei))θCi] ln (Ar/2)+((1πA)θA(ei)+πC(1θA(ei))θCi)(lnBλ)+(πAθA(ei)+(1πC)(1θA(ei))θCi)lnBψei.(48)

We can easily show that the derivative of wi with respect to ei in Eq.(48) is positive as long as πA is lower than a certain threshold level, πth(< 1). As a result, the optimal solution under the timing-based facility in this variant of the model is the same as that of the basic model: ei,N = 1.

This suggests that even when autonomous crises are allowed to occur later than imported crises, the main results of the basic model would not change as long as the chances of timing reversions are not too high.

Second, consider the case where each country can affect the probabilities of timing reversions. More specifically assume that

πAi=f(si/μs)+πA,πCi=f(si/μs)+πC, f>0, f(1)=0,(49)

where si (∈ [0, smax]) represents a country’s efforts to have a crisis in the second rather than the first period, and µs is the average of si (i.e., μs=0Nsidi/N). The positive first derivative (f′ > 0) tells us that the greater efforts a country makes (compared to those of the other countries), the more likely the country is to have a crisis in the second period. In addition, f(1) = 0 indicates that if all the countries make the same amount of efforts, the chances of the timing reversion do not change. For simplicity, assume that si does not entail any costs. Assume also that each country seeks to maximize the utility of the representative consumer by choosing ei and si.

Under both laissez-faire and existing lending facilities, borrower countries in this variant of the model do not have any incentive to delay the crisis. Therefore, the optimal solution for si could be zero for all the borrower countries.

Under the timing-based facility, each government would try hard to raise its own πAi by choosing the maximum si. Note that such an effort would substantially reduce the chance of having a crisis in the first period only if the other countries did not raise si. However, every country has an incentive to raise πAi by raising si to its maximum. As a result, in equilibrium, all the borrower countries put their maximum efforts: si = smax. Given f(1) = 0, it also follows that no country would succeed in raising πAi.

The above discussion suggests that the optimization problem of this variant of the model becomes the same as that of the basic model. As a result, the introduction of a potential racing for crisis delays would not alter the main results of the paper.27

C. Crisis-Prevention Efforts Without Bound

In the basic model, we focus on the case where the timing-based lending facility and laissez-faire induce the same amount of crisis prevention efforts. A critical assumption needed to make such a case was that crisis prevention efforts, ei, cannot be greater than one. However, we can modify the basic model to allow crisis prevention efforts not to be bounded from the above. In this more general case, the equilibrium level of crisis prevention efforts could differ between the timing-based lending facility and laissez-faire.

Assume that crisis prevention efforts could take any nonnegative value, that is, ei ≥ 0, and suppose that

2θAi(ei)2>0.(50)

We can then have an interior solution for the optimal crisis prevention efforts under both laissez-faire and the timing-based lending facility. Under laissez-faire, we can derive the optimal (and equilibrium) crisis prevention efforts by equating wiei in Eq.(17) to zero, that is,

wiei=θAiei[(1γabn1)(ln(Ar/2)+ln Bλ)]ψ=0.(51)

Under the timing-based lending facility, the optimal crisis prevention effort is derived from

wiei=θAei[(1γa(b1)n1)(ln(Ar/2)+ln B)λ]ψ=0,(52)

which equates wiei in Eq.(29) to be zero.

From a comparison between Eqs. (51) and (52), we have

ei,L<ei,N.(53)

This suggests that the timing-based lending facility can be more effective even in preventing moral hazard (or inducing crisis prevention efforts) than laissez-faire. The reason is as follows. Recall that the utility in the event of crisis in the second period is given by lnB for the case of the timing-based facility, and lnB − λ for laissez-faire. So a country has higher utility in the event of crisis in the second period under the timing-based lending facility than under laissez-faire. The crisis-time utility gain from the timing-based facility is magnified if the chance of crisis in the second period (θ2i) increases, while the chance is raised by crisis prevention efforts (recall that θ2i=(1θAi)θCi ). Therefore, under the timing-based facility, crisis prevention efforts would raise the welfare of a borrower country more than under laissez-faire, which induces the country to exert more crisis prevention efforts.28

VIII. Conclusion

This paper has proposed a new lending facility that differentiates rescue financing provided to different countries based on the timing of crises. This new facility basically would take significant account of the timing of crises: it would refuse loans to the country that started a crisis but unconditionally provide rescue financing to countries where crises occurred within a short period after the initial crisis. Given that the key to developing a desirable rescue-lending facility lies in how to reduce moral hazard and crisis contagion at the same time, the paper has introduced a simple model and used it to explore how laissez-faire, existing crisis-lending facilities, and the proposed timing-based facility each affect the problems of moral hazard and crisis contagion. It suggests that laissez-faire may be effective in abating the moral hazard problem but ineffective in containing crisis contagion. Existing international crisis-lending facilities may be effective in resolving the contagion effect but ineffective in abating the moral hazard problem. In contrast, the newly proposed lending facility could effectively resolve both moral-hazard and contagion problems at the same time, particularly under information asymmetry. Hence, the timing-based lending facility merits attention as an alternative that would overcome the shortcomings of existing lending facilities. Using a variant of the model, this study has shown that even where crises are caused by common shocks instead of crisis contagion, the timing-based lending facility would be more effective in reducing risks of global crisis than the laissez-faire approach or use of existing facilities.

Note that the logic of the timing-based lending facility proposed here is general and could be easily applied to the crisis-managing authorities or lender of last resort in the domestic context, by replacing “countries” by “firms” and “crisis” by “default.” The domestic lender of last resort may use this type of lending facility, based on the timing order of crises, to effectively mitigate both the problems of moral hazard and default contagion.

APPENDIX I.

Proof of (36)

To formally prove this, express ΔWN−T as a function of θAmin, that is, ΔWNT=Γ(θAmin). With θAmin<1/2, we then have

Γ(θAmin)θAmin=N[[1+γa(b1)N(12θAmin)](ln(Ar/2)lnB)λ11θAmin]<0.(54)

Given (35), we also have

limθAmin0Γ(θAmin)=N[θAmax[1+(1θAmax)(a1)γN(b1)](ln(Ar/2)lnB)ψ]>0.(55)

Finally, limθAminθAmaxΓ(θAmin) is given by

limθAminθAmaxΓ(θAmin)=N[(θAmax)(1θAmax)(γN(b1))(ln(Ar/2)lnB)θAmaxλψ+ln(1θAmax)]<0.(56)

Therefore, there exists a threshold level, θAth(<θAmax), which satisfies Γ(θAth)=0. It then follows that for any θAmin<θAth, we have Γ(θAmin)>0, that is WN > WT.

APPENDIX II

Financing of Reserves for Bailouts

The basic model assumes that the international crisis lender already raised the reserves for rescue financing so that countries now do not have to make contributions for that. This appendix considers a different case, where no reserves were already raised and therefore the international crisis lender has to raise it from the countries at the beginning of the initial period. Let R denote the amount of the reserves needed, and w be the fraction of the initial capital endowment that each country has to contribute.

Under laissez-faire, where no rescue financing is provided to the countries in crisis, the amount of the reserves needed for rescue financing (R) is zero, and so is w. Therefore, this variant of the model has the same optimization problem and solution (that is, ei,L = 1) as the basic model.

Under existing lending facilities, the international crisis lender provides rescue financing regardless of the timing of crises. The rescue financing to a crisis country amounts to ρ = rki,f = rk(1 − ω) = (r/2)(1 − ω). If a borrower country is not hit by crisis, its representative consumer consumes (Ar/2)(1 − ω). If the country is hit by crisis, the representative consumer consumes B(1 − ω), while the country does not incur the cost due to nonrepayment, λ. Using these, the maximization problem of a borrower country is given by

[P2]maxei wi=[1θA(ei)(1θA(ei))γ(a1)(b1)n1] ln((Ar/2)(1ω))+[θA(ei)+(1θA(ei))γ(a1)(b1)n1] ln(B(1ω))ψei(57)

which, under Condition (i), yields ei,T = 0, as in the basic model.

Under the timing-based lending facility, rescue financing is provided to any crisis country in the second period, but to no country in the first period. As a result, the optimization problem of each borrower country is given by:

[P4]maxei wi=[1θA(ei)(1θA(ei))γa(b1)n1] ln((Ar/2)(1ω))+θA(ei)(ln(B(1ω))λ)+(1θA(ei))(γa(b1)n1) ln(B(1ω))ψei,(58)

which, under Condition (i), yields ei,N = 1.

The above discussions suggest that the assumption that the international crisis lender needs to raise the reserves for rescue financing does not change the equilibrium level of crisis prevention efforts of the basic model.

This variant of the model slightly changes the magnitude of the world’s welfare under existing lending facilities and the timing-based facility. Under existing lending facilities, total reserves needed for rescue financing (denoted by RT) and an individual country’s contribution (denoted by ωT)are given by

RT=θAmaxN(1+(1θAmax)γ(a1)(b1)N)(r/2)(1ω),(59)
ωT=θAmax(1+(1θAmax)γ(a1)(b1)N)(r/2)1+θAmax(1+(1θAmax)γ(a1)(b1)N)(r/2).(60)

Using these, the world’s welfare under existing rescue financing facilities (denoted by WT) is

WT=N[(1θAmax(1θAmax)γ(a1)(b1)NθAmax) ln (Ar/2)+(θAmax+(1θAmax)γ(a1)(b1)NθAmax)lnB+ln(1ωT)]+N[ln((r/2)(1ωT))].(61)

In the case of the new lending facility, total reserves needed for rescue financing and an individual country’s contribution are given by

RN=((1θAmin)γa(b1)θAminN2)(r/2)(1ω),(62)
ωN=(1θAmin)γa(b1)θAminN(r/2)1+(1θAmin)γa(b1)θAminN(r/2).(63)

which yields the world’s welfare to be

WN=N[(1θAmin(1θAmin)γa(b1)NθAmin)ln(Ar/2)+θAmin(lnBλ)+(1θAmin)(γa(b1)NθAmin)ln B+ln(1ωN)ψ]+N[ln(r/2)+ln(1ωN)+ln(1θAmin)].(64)

Suppose that the welfare gain from providing rescue financing is larger than the welfare costs due to having reserves, more specifically

γaNθAmin(1θAmin)(ln(Ar/2)ln B)>2 ln(1ωN).(65)

Given (35) and (65), it then follows that there exists a threshold level of θAmin below which we have

ΔWNL>0andΔWNT>0.(66)

which suggests that the main results of the paper are not affected by assuming that the international crisis lender needs to raise reserves for bailouts.

Furthermore, we may introduce another variant of the model to examine the case where the rescue financing provided to crisis countries (ρ) is just a pure lending that needs to be which yields the world’s welfare to be repaid later (see Krueger, 2002; Jeanne and Zettelmeyer, 2001; Joshi and Zettelmeyer, 2003). In this variant, we introduce an additional period, t = 3. Assume that each country has additional income in t = 3, denoted by Vi, which is different from output at the end of the second period, yi (= Aiki). Suppose that countries can repay ω to the international crisis lender in t =3 out of Vi. We can then show, in an analogous way to the above, that this variation of the model does not alter the main results of this study.

References

  • Allen, Franklin, and Douglas Gale, 2000, “Financial Contagion,” Journal of Political Economy 108, pp. 1-33.

  • Berg, Andrew, 1999, “The Asia Crisis: Causes, Policy Responses, and Outcomes,” IMF Working Paper No 99/138.

  • Bulow, Jeremy, and Kenneth Rogoff, 1989, “A Constant Recontracting Model Of Sovereign Debt,” Journal of Political Economy 97, pp. 155-178.

    • Search Google Scholar
    • Export Citation
  • Calomiris, Charles, 1998, “The IMF’s Imprudent Role as Lender of Last Resort,” Cato Journal 17, pp. 275-295.

  • Calvo, Guillermo, and Enrique Mendoza, 2000, “Rational Contagion and the Globalization of Securities Markets,” Journal of International Economics 51, pp. 79-113.

    • Search Google Scholar
    • Export Citation
  • Chopra, Ajai, Kenneth Kang, Meral Karasulu, Hong Liang, Henry Ma, and Anthony Richards, 2002, “From Crisis to Recovery in Korea: Strategy, Achievements, and Lessons,” In: Coe, David, and Se-Jik Kim (Eds.), Korean Crisis and Recovery, IMF and KIEP, pp. 13-104.

    • Search Google Scholar
    • Export Citation
  • Corsetti, Giancarlo, Bernardo Guimarães, and Nouriel Roubini, 2003, “The Tradeoff between an International Lender of Last Resort to Deal with Liquidity Crisis and Moral Hazard Distortions: A Model of the IMF’s Catalytic Finance Approach,” Mimeo, Yale University and New York University.

    • Search Google Scholar
    • Export Citation
  • De Gregorio, Jose, Barry Eichengreen, Takatoshi Ito, and Charles Wyplosz, 1999, An Independent and Accountable IMF, Geneva, International Center for Monetary and Banking Studies and Centre for Economic Policy Research.

    • 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 1, International Center for Monetary and Banking Studies and Centre for Economic Policy Research.

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, and Ashoka Mody, 2000a, “Would Collective Action Clauses Raise Borrowing Costs?,” NBER Working Paper No. 7458.

  • Eichengreen, Barry, and Ashoka Mody, 2000b, “Would Collective Action Clauses Raise Borrowing Costs? An Update and Extension,” World Bank Working Paper No. 2363.

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, Andrew Rose, and Charles Wyplosz, 1996, “Contagious Currency Crisis: First Tests,” Scandinavian Journal of Economics 98, pp. 463-484.

    • Search Google Scholar
    • Export Citation
  • Feldstein, Martin, 1998, “Refocusing the IMF,” Foreign Affairs, March/April, pp. 20-33.

  • Fischer, Stanley, 1998, “In Defence of the IMF: Specialized Tools for a Specialized Task,” Foreign Affairs, July/August, pp. 103-106.

    • Search Google Scholar
    • Export Citation
  • Fischer, Stanley, 1999, “On the Need for an International Lender of Last Resort,” Journal of Economic Perspectives 13, pp. 85-104.

  • Flood, Robert, and Peter Garber, 1984, “Collapsing Exchange-Rate Regimes: Some Linear Examples,” Journal of International Economics 17, pp. 1-13.

    • Search Google Scholar
    • Export Citation
  • Flood, Robert, and Nancy Marion, 2001, “A Model of the Joint Distribution of Banking and Exchange-Rate Crises,” IMF Working Paper 01/213.

    • Search Google Scholar
    • Export Citation
  • Forbes, Kristin, and Roberto Rigobon, 2002, “No Contagion, Only Interdependence: Measuring Stock Market Comovements,” Journal of Finance 57, pp. 2223-2261.

    • Search Google Scholar
    • Export Citation
  • Gai, Prasanna, Simon Hayes, and Hyun Song Shin, 2001, “Crisis Costs and Debtor Discipline: the Efficacy of Public Policy in Sovereign Debt Crises,” Bank of England Working Paper No. 136.

    • Search Google Scholar
    • Export Citation
  • Glick, Reuven, and Andrew Rose, 1998, “Contagion and Trade: Why Are Currency Crises Regional?,” Journal of International Money and Finance 18(4), pp. 603-617.

    • Search Google Scholar
    • Export Citation
  • Goodhart, Charles, and Haizhou Huang, 2000, “A Model of the Lender of Last Resort,” IMF Working Paper.

  • Haldane, Andrew, Adrian Penalver, Victoria Saporta, and Hyun Song Shin, 2003, “Analytics of Sovereign Debt Restructuring,” mimeo, Bank of England and London School of Economics.

    • Search Google Scholar
    • Export Citation
  • International Financial Institution Advisory Commission (IFIAC), 2000, Report of the International Financial Institution Advisory Commission, Department of Treasury, Washington, DC.

    • Search Google Scholar
    • Export Citation
  • Jeanne, Olivier, 2000, Currency Crises: A Perspective on Recent Theoretical Developments, Special Papers in International Economics No. 20, Princeton University.

    • Search Google Scholar
    • Export Citation
  • Jeanne, Olivier, and Charles Wyplosz, 2001, “The International Lender of Last resort: How Large is Large Enough?,” NBER Working Paper No. 8381.

    • Search Google Scholar
    • Export Citation
  • Jeanne, Olivier, and Jeromin Zettelmeyer, 2001, “International Bailouts, Moral Hazard, and Conditionality,” Economic Policy 33, pp. 409-432.

    • Search Google Scholar
    • Export Citation
  • Joshi, Priyadarshani, and Jeromin Zettelmeyer, 2003, “Implicit Transfers in IMF Lending, 1973-2003,” mimeo, IMF.

  • Kaminsky, Graciela, and Carmen Reinhart, 2000, “On Crises, Contagion, and Confusion,” Journal of International Economics 51, pp. 145-168.

    • Search Google Scholar
    • Export Citation
  • Kaminsky, Graciela, and Carlos Végh, 2002, “Two Hundred Years of Contagion,” mimeo.

  • Kaminsky, Graciela, and Sergio Schmukler, 1999, “What Triggers Market Jitters? A Chronicle of the Asian Crisis,” Journal of International Money and Finance 18, pp. 537-560.

    • Search Google Scholar
    • Export Citation
  • Kodres, Laura, and Matthew Pritsker, 2002, “A Rational Expectations Model of Financial Contagion,” Journal of Finance 57, pp. 769-799.

    • Search Google Scholar
    • Export Citation
  • Krueger, Anne, 2002, “Preventing and Resolving Financial Crises: The Role of Sovereign Debt Restructuring,” a speech at Latin American Meeting of the Econometric Society, IMF.

    • Search Google Scholar
    • Export Citation
  • Krueger, Anne, and Jungho Yoo, 2002, “Falling Profitability, Higher Borrowing Costs, and Chaebol Finances During the Korean Crisis,” In: Coe, David, and Se-Jik Kim (Eds.), Korean Crisis and Recovery, IMF and KIEP, pp. 157-196.

    • Search Google Scholar
    • Export Citation
  • Krugman, Paul, 1979, “A Model of Balance-of-Payments Crises,” Journal of Money, Credit, and Banking 11, pp. 311-325.

  • Lerrick, Adam, and Allan Meltzer, 2003, “Blueprint for an International Lender of Last Resort,” Journal of Monetary Economics 50, pp. 289-303.

    • Search Google Scholar
    • Export Citation
  • Meltzer, Allan, 1997, “Moral Hazard Goes Global,” Financial Times, October 27.

  • Mody, Ashoka, and Mark Taylor, 2003, “Common Vulnerabilities,” CEPR Discussion Paper No. 3759.

  • Rose, Andrew, 2002, “One Reason Countries Pay Their Debts: Renegotiation and International Trade,” NBER Working Paper No. 8853.

  • Stokey, Nancy, 2002, “Rules Versus Discretion after Twenty-Five Years,” NBER Macroeconomics Annual, forthcoming.

  • Townsend, Robert, 1979, “Optimal Contracts and Competitive Markets with Costly State Verification,” Journal of Economic Theory 21, pp. 265-293.

    • Search Google Scholar
    • Export Citation
1

I am very grateful to Olivier Jeanne, Simon Johnson, Allan Meltzer, Ashoka Mody, Jaewoo Lee, and Jeromin Zettelmeyer for valuable comments.

2

New ways of debt restructuring have also been in high demand. Collective-action clauses and a sovereign debt restructuring mechanism have been proposed as solutions to the restructuring problem of collective representation (see Eichengreen and Mody, 2000a and 2000b; Eichengreen, 2000; Krueger, 2002; and Haldane, Penalver, Saporta, and Shin, 2003). Lerrick and Meltzer (2003) also propose that the IMF provide a cash-support bid to insolvent sovereign borrowers at a discount from the government’s minimum offer on restructured value of debt (while providing liquidity support, tied to preconditions, at a penalty rate to solvent but illiquid borrowers).

3

There were some other problems in implementing the CCL. For example, applying for the CCL could have been interpreted as a signal of weakness of the economy, and so no country would have wanted to apply (see, for example, IFIAC, 2000; and De Gregorio, Eichengreen, Ito, and Wyplosz, 1999).

4

Throughout this paper, I use the terms “international lender of last resort” and “international crisis lender” interchangeably. Jeanne and Wyplosz (2001) use the term “lender of last resort” only for a situation of a self-fulfilling crisis, while many other authors (for example, Fischer, 1999; Calomiris, 1998; and Lerrick and Meltzer, 2003) use the term more broadly and suggest that the IMF could play the role of an international lender of last resort even without providing unlimited liquidity.

5

Gai, Hayes, and Shin (2001) show how critically the efficacy of crisis management by the IMF depends on the quality of its monitoring of debtor countries. Under imperfect monitoring, “rules” offer an advantage over “discretion” by protecting against policymakers’ arbitrary decision making and resulting policy blunders (see, for example, Stokey, 2002).

6

Most empirical studies suggest evidence of excess comovement in asset returns but indicate more limited consensus on whether this comovement increases after a country is hit by crisis. Forbes and Rigobon (2002) find no significant increase in cross-country linkages after a shock to one country, while Kaminsky and Schmukler (1999) find asymmetries between the effects on asset prices of good and bad news from elsewhere.

7

To simplify the model, I distinguish crises into one of the two types, autonomous and imported. However, we may explicitly introduce a model where the actual crisis is a composition of the two types so that both domestic and foreign factors affect the chance of crisis. In this variant model, we could assume that the country hit by crisis first is affected more strongly by domestic than foreign influences, while the countries hit later are affected more strongly by foreign influences. In this case, the main qualitative results of the paper would not change.

8

The early studies on currency crisis (for example, Flood and Garber, 1984; and Krugman, 1979) emphasize the importance of unsound fiscal policy, combined with exchange-rate fixing, as a cause of currency crises. Influenced by the 1997 crisis in East Asia, however, recent studies also focus on structural weaknesses in the financial and corporate sectors as another key cause of financial crises (see, for example, Krueger and Yoo, 2002; Chopra and others, 2002; and Flood and Marion, 2001).

9

Mody and Taylor (2003) suggest that country-specific movements of the exchange market pressure index (EMPI) can be well explained by common vulnerability indicators.

10

In implementing this new lending facility, the international crisis lender may set a time lag that is needed to make an operational definition of the “first” crisis, based on empirical studies on timing patterns of previous crises. For example, based on the time lags among the ERM crisis in 1992–93, the Mexican crisis in 1994–95, and the East Asian crisis in 1997–98, one may suggest that one year or so would be a reasonable criterion to identify the “first” crisis.

11

Here, for simplicity, the new facility assumes that the country of the first crisis shall not get any rescue financing. But we may instead assume that the country of the first crisis shall receive a part of what the country of later crisis would receive as rescue financing. This does not alter the main results of the paper.

12

The current model focuses on borrower-side moral hazard problems. But we may explicitly introduce the moral hazard problem of lender countries (or international investors) as well. In this variant model, however, the main result would not change. More specifically, we may assume that lender countries can at least partially observe the crisis prevention efforts of borrower countries. Then if rescue financing is not provided by the international crisis lender, lender countries would spend resources to distinguish between the borrower countries that make crisis prevention efforts and those that do not, and make loans only to the former. In the case where any crisis country is bailed out, however, lender countries would not make a costly distinction between the two types of borrower countries.

13

We may instead assume that the chance of having an imported crisis is affected by crisis-prevention efforts. However, this modification would not change the results of the paper, as long as crisis-prevention efforts reduce the chance of imported crisis less effectively than that of autonomous crisis.

14

International crisis lending can be justified when there is externality, notably due to crisis contagion, or in the presence of self-fulfilling equilibrium. The current model does not explicitly introduce the element of self-fulfilling crises. But we may introduce such crises, for example, by explicitly introducing complementarity between productivity shock (At) and capital (kt) in the production function, together with the possibility of outflow of foreign capital in the second period.

15

See, e.g., Eichengreen, Rose, and Wyplosz (1996) and Glick and Rose (1998) for the evidence of trade links, and Kaminsky and Reinhart (2000) and Mody and Taylor (2003) for capital market links. For theoretical models of propagation channel, see, e.g., Allen and Gale (2000), Calvo and Mendoza (2000), and Kodres and Pritsker (2002).

16

In the current setup with a continuum of (accordingly, an infinite number of) countries, n1 and N represents the “measure” of producer countries in autonomous crisis and total producer countries, respectively. But we will use the terms “number” and “measure” interchangeably throughput the paper.

17
To explicitly distinguish between the first and the second round of contagion, we may introduce a three period version of the model. Then, the probabilities of crisis contagion in the second and third period (denoted by θC2i and θC3i) can be given by
θC2i=γ2(aϕ1)n1,θC3i=γ3(aϕ2)n2,a>1,

where γ2(> 0) captures the strength of the first round of crisis contagion (the effect of the crisis in the first period on the chance of the crisis in the second period), γ3(> 0) represents that of the second round of contagion, and a represents the part of crisis contagion that is not affected by the international crisis lender’s rescue financing. By adding θC2i and θC2i, we can then derive Eq.(9), which is for the two period version.

18

Without losing generality, we may consider the cases of partial financing (i.e., ρ < rki,f) or different ρs (for example, ρ = rki,f for some countries but ρ < rki,f for other countries).

19

The logarithmic utility function substantially simplifies the calculations in later sections, in addition to assuring risk aversion, while assuming more general utility functions, for example CRRA utility function, would not alter the main results of the paper.

20
This assumption can be derived as a result of the optimal choice of borrower countries. To illustrate this, assume that the cost due to nonrepayment, denoted by λ, satisfies
ln Aln (Ar/2)<λ<ln B ln(Br/2).

Now suppose that there is no international crisis lender, and consider a borrower country which decides whether or not to repay rki,f after the productivity shock is revealed. If the borrower country does not repay rki,f, the consumption of the representative consumer is given by Aki = A(2k)= A in the case of no crisis, and Bki = B(2k)= B in the case of crisis (recall that k = 1/2). So, given the cost due to nonrepayment, the utility from consumption (u(ci)) is given by (ln A − λ) in the case of no crisis, and (ln B − λ) in the case of crisis. Meanwhile, if the borrower country repays rki,f, the consumption of its representative consumer is given by Akirki,f = (Ar/2) in the case of no crisis, and Bkirki,f = (Br/2) in the case of crisis. So the utility from consumption is given by ln(Ar/2) and ln(Br/2). Then given the above assumption on λ, we have ln(Br/2) < ln B − λ so that a defaulting country would not repay. We also have ln(Ar/2) > ln A − λ so that the country which did not have a crisis would make the repayment.

21

See Bulow and Rogoff (1989) for a model which explicitly introduces sovereign default penalties, and Rose (2002) for empirical support for the penalties.

22

Note that without losing generality, we may instead assume that borrower countries in crisis repay to lenders even when the international crisis lender does not provide rescue financing, and that ln(Ar/2) > ln B > ln(Br/2).

23

Given the assumption of a continuum of countries, there are many “first” crisis countries (or countries that have autonomous crisis in the first period). We may interpret this result as representing that there are many regions each of which has its own first crisis country. Or we may explicitly introduce a variant model in which there are a continuum of different regions of borrower countries with measure one, each of which has N countries, and crisis contagion occurs only within each region. Of course, the modification does not alter the results of the paper.

24

I implicitly assume that the policy on rescue financing adopted by the international crisis lender (ϕt) is credible. Of course, we may explicitly introduce the cost due to the loss of reputation large enough to induce the international crisis lender to commit to its policy on ϕt.

25

Without losing generality, we may assume that countries may have autonomous crisis as well and that the voluntary crisis prevention efforts can also reduce the chance of autonomous crisis. For example, we may assume that θi=χє+θAi=θ˜1i+θ˜2i+θ˜Ai and θAiei<0. The main results, however, will not alter as long as the crisis delay effect (i.e., θ˜1iei ) is large compared to the crisis prevention effect (i.e., θAiei ).

26

To illustrate this, for example, consider the case where crises can occur as a result of crisis contagion with probability β ∈ (0, 1) and common shocks with probability 1 − β. In addition, define the world’s welfare to be a weighted sum of the utility in the cases of crisis contagion and common shocks with each having a weight of β and 1 − β. In this case, we can easily show that the new lending facility is Pareto-superior to both laissez-faire and existing facilities. The result is intuitively obvious given that the timing-based facility induces higher welfare in both cases of crisis contagion and common shocks.

27

One may imagine a hypothetical case where under the timing-based lending facility, countries may precipitate rather than postpone crisis: if a country was already hit by crisis, some other countries with high chances of crisis would try to strategically default to use the window of opportunity to receive rescue financing. In reality, however, such a case is unlikely to happen. Given huge welfare cost of crisis (even when being bailed out), the countries would not try to precipitate crisis unless the chance of having a crisis in the near future is extremely high. Moreover, when no country has been yet hit by crisis, each country would exert much crisis prevention efforts so as not to become the first crisis country. As a result, no country would have a chance of crisis high enough to induce them to precipitate crisis after another country is hit by crisis.

28

Corsetti, Guimarães, and Roubini (2003) suggest that contingent liquidity assistance provided by the international lender of last resort can promote countries to undertake efficiency-enhancing reform efforts.