On contingent liabilities and the likelihood of fiscal crises *.

Burnside, Craig

Contingent liabilities are commitments to take on actual liabilities if specific uncertain events occur at some point in the future. There are many types of contingent liabilities: social security payments, loan guarantees, performance guarantees, commitments to provide commodities to the public at fixed prices, deposit insurance, and guarantees to the creditors of public sector enterprises. Increasingly, international organizations emphasize the dangers of contingent liabilities when providing advice to their member governments. Why are these warnings sounded? One answer is obvious--if significant contingent liabilities are realized the government can face substantial fiscal costs that it may be unprepared to meet. Even when governments are willing to prepare themselves, they may face difficulty in doing so: the potential magnitude of a contingent liability, and the probability with which it could be realized are often not that well understood in advance. Furthermore, the timing of a liability's realization can be very difficult to predict.

But there are additional more subtle reasons for worrying about contingent liabilities. By taking them on, governments can make the underlying events that convert them into actual liabilities more dangerous. First, for a given probability of the events occurring, the total economic costs associated with these events can rise. Second, the probabilities of these events occurring can actual increase. Two examples are helpful. First, suppose a government provides free earthquake insurance. This creates a contingent liability--if a damaging earthquake occurs the government will bear a fiscal cost. But standard arguments suggest that the free provision of insurance will encourage greater risk taking by the private sector. More households and firms will locate in earthquake prone areas than otherwise would, and they will also take less precautionary measures against earthquake damage than they otherwise would. This means that when the government provides free insurance it likely raises the overall cost to society of each earthquake. On the other hand, the probability of the earthquake happening is independent of the government's action. However, there are examples where government's action actually raises the probability of a damaging event occurring. Suppose that a government, fixing its exchange rate, implicitly issues a guarantee that in the event that the fixed exchange rate regime is abandoned, it will honor the liabilities of the banking system should this be necessary. (1) This will change the behavior of bank creditors, who are the recipients of the insurance, by making them less cautious in their lending. But it will also change the behavior of banks--they too will alter their behavior in apparently reckless, but profit maximizing, ways. These changes in the behavior of private agents will raise the likelihood of a banking crisis occurring.

In this paper, I focus on this second danger of contingent liabilities. I describe how banks' behavior, in the face of government guarantees, changes in such a way that the banking system becomes more fragile. Banks become more likely to take on exchange rate risk. Therefore, the government becomes more likely to incur the fiscal cost associated with bank failures. Furthermore, the more credible is the government's guarantee, the more likely a crisis becomes.

The first section of the paper presents a simplified version of the model in Burnside, Eichenbaum and Rebelo (2001a), to show how credible government guarantees to bank creditors can make a banking system more fragile. The model is a very simple one in which there is no real source of uncertainty. All uncertainty is associated with the exchange rate--specifically, whether the fixed exchange rate regime will be abandoned. We will see that when governments issue guarantees to the foreign creditors of local banks, these banks will not hedge against exchange rate risk. (2)

In Burnside, Eichenbaum and Rebelo (2001c) the fact that banks do not hedge exchange rate risk raises the specter of self-fulfilling speculative attacks against a currency. If agents come to believe that the exchange rate regime will collapse, they will speculate against local currency--ultimately causing the central bank to float the exchange rate. (3) The central bank's decision to float, in turn, will lead to a depreciation of the currency--in anticipation of the government choosing to print money to finance a bank bailout--which will ultimately lead to the failure of unhedged banks. These bank failures will, in turn, require the government to honor its bailout guarantee. When it does so by printing money, it rationalizes the speculative attack.

For the purposes of this paper, I do not emphasize this latter effect--that banking and currency crises become jointly more likely. Rather I focus on how a banking crisis becomes more likely for a given probability that a currency crisis will occur. In particular, we will see that in equilibrium banking crises will always be associated with currency crises in economies where governments have taken on a contingent liability associated with bank bailouts. This paradox, that government guarantees make for less, not more, stability is consistent with empirical evidence in Demirguc-Kunt and Detragiache (2000). In economies where governments do not issue guarantees to bank creditors, banking crises will not occur even if there is the risk of a currency crisis.

The second section of the paper extends the analysis to the case where the government's guarantee is not perfectly credible. It shows that the less credible the government's guarantee is, the less likely is a banking crisis. There is a lower limit on the credibility of the government guarantee, below which the banking system protects itself from exchange rate risk. The third section explores whether the results are sensitive to assumptions about the structure of banks' assets and liabilities. The fourth section provides some concluding remarks.

1. A Baseline Model of How Guarantees Affect Fragility

Here I consider a simple partial equilibrium model of a small open economy. There is a single good, no trade barriers, and purchasing power parity holds:

(1) P = S[P.sup.*]

Here P and [P.sup.*] denote the domestic and foreign price level respectively, while S denotes the exchange rate defined as units of domestic currency per unit of foreign currency. For simplicity I let [P.sup.*] = 1, so that one unit of the single good and one unit of foreign currency are equivalent.

The economy consists of four types of agents: domestic banks, foreign creditors of domestic banks, borrowers from domestic banks and the government. There is no source of fundamental uncertainty in the model. To highlight the role that government guarantees play in making the banking system more fragile, I allow for only one potential source of uncertainty: possible abandonment of a fixed exchange rate regime.

1.1. The Exchange Rate

Below, I will analyze the behavior of banks facing a potential currency crisis in some discrete time period. I will assume that a fixed exchange rate regime, with S = [S.sup.I], has been in effect for all prior periods. I assume that all agents understand that the probability distribution governing the exchange rate in the upcoming period is as follows:

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Implicit in equation (2) is the notion that there is some probability q with which the fixed exchange rate regime will be abandoned within the period.

1.2. The Banking Sector

In this subsection I present a model of banks similar to the ones in Chaff, Christiano and Eichenbaum (1996), Edwards and Vegh (1997) and Burnside, Eichenbaum and Rebelo (2001a, d). (4) In the model, banks are perfectly competitive and finance themselves by borrowing foreign currency from foreign investors. (5) The banks, in turn, lend in local currency to domestic borrowers at fixed nominal interest rates, and are therefore exposed to exchange rate risk. But the model allows banks to hedge exchange rate risk in frictionless forward markets. Therefore, if banks take on a currency mismatch it is because they are willing to do so. (6)

My main focus is on banks' portfolio decisions under the fixed exchange rate regime. I show that banks will fully hedge exchange rate risk when there are no government guarantees to foreign creditors. On the other hand, banks will not hedge when there are government guarantees. In fact, in the presence of guarantees, banks prefer to declare bankruptcy and minimize their residual value in the event that the currency is floated. (7)

In the model, banks are perfectly competitive and their actions are publicly observable. (8) Let L denote the number of dollars a bank borrows from foreign investors at the beginning of a period. The bank converts these dollars into local currency at the rate [S.sup.I], and lends the resulting L[S.sup.I] units of local currency to domestic borrowers at a fixed gross interest rate [R.sup.a]. (9) To carry out its lending, the bank incurs a real transactions cost of [delta]L. Since banks will only be repaid by domestic borrowers after the realization of the current period's exchange rate, the gross value of the bank's lending portfolio, net of these transactions costs, will be given by

(3) [V.sup.L] = ([R.sup.a] [S.sup.I]/S - [delta]) L

The important aspect of (3) is that it shows that the gross value of the bank's lending portfolio diminishes if the fixed exchange rate regime is abandoned, that is, if S = [S.sup.D].

Banks can hedge exchange rate risk by entering into forward contracts. (10) Let F denote the one-period forward exchange rate defined as units of local currency per dollar. We assume that the forward contracts are priced risk-neutrally in a frictionless market. Under these assumptions, the forward rate, F, is

(4) 1/F = (1 - q) 1/[S.sup.I] + q 1/[S.sup.D] = E(1/S)

This condition implies that the expected real payoff from buying or selling a unit of the local currency forward is zero. (11) Let X denote the number of dollars the bank purchases forward. By definition this means the bank receives X dollars at the end of the period in exchange for delivery of XF units of local currency, which, at that stage will be worth XF/S dollars. So the bank's profits from hedging are

(5) [V.sup.H] = X(1 - F/S).

If the bank purchases dollars forward (X > 0) then it will make profits from hedging when S = [S.sup.D], and losses when S = [S.sup.I]. Choosing such a forward position is a natural way for the bank to hedge the risk to its lending portfolio that comes from the possibility that the exchange rate regime will be abandoned. In a sense, it allows the bank to transfer some of the profits its lending portfolio will generate in the good state of the world (S = [S.sup.I]) to cover some of the losses its lending portfolio will generate in the bad state of the world (S = [S.sup.D). Below I discuss limits that counterparties in forward markets will put on the hedging positions of banks.

The combined gross value of the bank at the end of the period is

(6) [V.sup.G](L,X;S) = [V.sup.L] + [V.sup.H],

where I have now made explicit the fact that the bank's value depends on its portfolio decision, (L,X), and on the realization of the exchange rate, S.

All that remains to be specified is the rate at which banks can borrow. Foreign investors are assumed to be risk neutral, so, in equilibrium, the expected return on any investment they make will be R, the world risk free interest rate. This does not, however, mean that local banks can borrow at the rate R.

Since I have abstracted from informational problems in financial markets, the portfolio choices of banks are observable and the rates banks can borrow at will be given by a competitively determined schedule [R.sup.b](L,X). Why? The reason is simple: under limited liability some feasible portfolio choices, (L,X), will make it optimal for a bank to default in some states of the world. When [V.sup.G](L,X;S) [direct sum] [R.sup.b](L,X)L, it is optimal for a bank to repay its creditors because it can then distribute non-negative profits [V.sup.G] (L, X; S) - [R.sup.b] (L, X) L to its shareholders. On the other hand, if [V.sup.G] (L, X; S) < [R.sup.b] (L, X) L it is optimal for the bank to default by declaring bankruptcy. In this case, the bank surrenders its gross assets (say, to a bankruptcy court) and pays nothing to its shareholders. However, as we will see below, limited liability, in and of itself, does not cause the borrowing rate to deviate from R. This requires that default is costly, in the sense that some of the bank's gross value is destroyed when it defaults. Here, I assume that bankruptcy reduces a bank's gross value by the amount [omega]L with 0 < [omega] < R.

Given these assumptions, the schedule [R.sup.b] (L, X) is set in such a way that the expected gross return of banks' foreign creditors will be equal to R. However, the exact form of the schedule depends on the nature of government policy, in particular, it depends on whether the government issues guarantees to banks' creditors.

1.3. The Borrowing Rate Schedule

There are some portfolio choices that allow local banks to borrow at the rate R. Consider Figure 1. In the region described as "fully hedged" banks set L and X consistent with [X.sub.0](L) [less than or equal to] X [less than or equal to] [X.sub.1](L), where [X.sub.0](L) = -[[R.sup.a] [S.sup.I]/[S.sup.D] - (R+d)] L / (1-F/[S.sup.D]) and [X.sub.1](L) = [[R.sup.a] - (R+d)] L / (F / [S.sup.I]-1). Setting X [direct sum] [X.sub.0](L) means that the bank is buying enough dollars forward to have [V.sup.G](L, X; [S.sup.D]) [direct sum] RL. Setting X [direct sum] [X.sub.1](L) means that the bank is buying few enough dollars forward that [V.sup.G](L, X; [S.sup.I]) [direct sum] RL. So, for (L, X) such that [X.sub.0](L) [less than or equal to] X [less than or equal to] [X.sub.1](L), the borrowing rate schedule is [R.sup.b](L, X) = R.

[FIGURE 1 OMITTED]

In the region described as "default when S = [S.sup.I]" the bank is buying too many dollars forward, in the sense that its hedging losses when S = [S.sup.I] make [V.sup.G](L, X; [S.sup.I]) < RL. This means foreign creditors will charge a higher borrowing rate to the bank. The reason is straightforward. Although the hedging losses in state S = [S.sup.I] imply higher hedging profits in state S = [S.sup.D], the foreign creditors suffers an additional loss: part of the bank's gross value is destroyed in state S = [S.sup.I]. So [R.sup.b](L, X) must be set consistent with

(7) RL = (1 - q) [[V.sup.G](L, X; [S.sup.I]) - [omega]L] + q [R.sup.b](L, X) L.

Similarly, in the region described as "default when S = [S.sup.D]" the bank is buying too few dollars forward, in the sense that its hedging profits when S = [S.sup.D] are insufficient to make [V.sup.G](L, X; [S.sup.D]) [direct sum] RL. When there are no government guarantees, this again means that foreign creditors will charge a higher borrowing rate to the bank. In particular, [R.sup.b](L, X) must be set consistent with

(8) RL = (1 - q) [R.sup.b](L,X) L + q [[V.sup.G](L, X; [S.sup.D]) - [omega]L].

Since a fixed exchange rate regime is often considered an implicit guarantee against a devaluation, it is interesting to consider the case where the government guarantees that foreign creditors are repaid at least RL just when S = [S.sup.D]. In this case, (8) is replaced by [R.sup.b](L, X) = R.

1.4. Optimal Hedging Strategies

Given limited liability, banks will choose L and X to maximize their expected value to their shareholders, which is given by

(9) V(L,X) = E[max{[V.sup.G](L, X; S) - [R.sup.b](L,X) L, 0}].

The max operator in this expression reflects the fact that limited liability protects shareholders from any losses made by the banks. (12) Notice that (9) can be rewritten by using the fact that max{[V.sup.G](L, X; S) - [R.sup.b](L, X)L, 0}) = [V.sup.G](L, X; S) - min{[R.sup.b](L, X)L, [V.sup.G](L, X; S)}. So,

(10) V(L, X) = E[[V.sup.G](L, X; S)] - E[min{[R.sup.b](L, X) L, [V.sup.G](L, X; S)}].

Evaluating E[[V.sup.G](L, X; S)] using (3)-(6) we have

(11) V(L, X) = ([R.sup.a] [S.sup.I]/F - [delta])L - C(L, X),

where

(12) C(L, X) ... E[min{[R.sup.b](L, X)L, [V.sup.G](L, X; S)}]

is the bank's expected cost of borrowing. In states of the world where the bank is solvent--where the "min" in the expression for C is [R.sup.b](L, X)L--the bank will repay its creditors [R.sup.b](L, X)L out of its gross value [V.sup.G]. In states of the world where the bank is bankrupt--where the "min" in the expression for C is [V.sup.G](L, X; S)--the bank will simply gives up its gross assets [V.sup.G] to the bankruptcy judge. Given (11), it is clear that the bank's optimal hedging position is determined by minimizing C with respect to X. In the rest of this section, I show how the optimal choice of X depends on whether the government issues guarantees to banks' creditors.

Given L, suppose a bank chooses X so that (L, X) lies in the region of Figure 1 described as "fully hedged." Recall that in that region, the bank's borrowing rate is R, and that RL < [V.sup.G](L, X; S) for all S. Hence, (12) implies that C(L, X) = RL.

Suppose the bank chooses X so that (L, X) lies in the region of Figure 1 described as "default when S = [S.sup.I]". In this case, (12) can be rewritten as C(L, X) = (1 - q)[V.sup.G](L, X; [S.sup.I]) + q[R.sup.b](L, X)L. But, since [R.sup.b](L, X) is set according to (7), notice that this means C(L, X) = RL + (1 - q)[omega]L. Since this exceeds the expected cost of borrowing for a fully hedged bank, a bank will never choose X in such a way that it defaults when S = [S.sup.I].

Suppose the bank chooses X so that (L, X) lies in the region of Figure 1 described as "default when S = [S.sup.D]". In this case, (12) can be rewritten as C(L, X) = (1 - q)[R.sup.b](L, X)L + q[V.sup.G](L, X; [S.sup.D]). In the absence of government guarantees, [R.sup.b](L, X) is set according to (8), implying that C(L, X) = RL + q[omega]L. Since this exceeds the expected cost of borrowing for a fully hedged bank, in the absence of government guarantees, a bank will never choose X in such a way that it defaults when S = [S.sup.D].

However, if there are government guarantees, [R.sup.b](L, X) = R. In this case, the bank's expected cost of borrowing is C(L, X) = (1 - q)RL + q[V.sup.G](L, X; [S.sup.D]). Notice that, by definition, [V.sup.G](L, X; [S.sup.D]) < RL for any (L, X) in the region labeled "default when S = [S.sup.D]." Hence, under government guarantees, C(L, X) < RL implying that banks will choose X < [X.sub.0], and they will default when S = [S.sup.D]. What is the optimal value of X < [X.sub.0]? Given the expression for C(L, X), it is clear that banks will choose X to minimize [V.sup.G](L, X; [S.sup.D]). This choice needs to be constrained given the description of the model. Notice that if a bank defaults, the gross assets managed by the bankruptcy judge are [V.sup.G]. But by definition, these assets are net of the banks' net profits from hedging. Implicit in our model's setup is the notion that counterparties in forward contracts always get paid first, and that there are sufficient funds to pay them, i.e. [V.sup.G](L, X; S) [direct sum] [omega]L, for all S. If this last assumption were not imposed, then presumably there would be substantial defaults associated with forward contracts, and forward prices would vary significantly across individual contracts. This does not appear to be the case in the real world. Furthermore, the assumption is consistent with the notion that forward contracts are collateralized by the value of the bank's loan portfolio net of bankruptcy costs; in the real world forward contracts are heavily collateralized. (13) Given our restriction on [V.sup.G], it is optimal, under government guarantees, for the bank to set X, so that [V.sup.G](L, X; [S.sup.D]) = [omega]L. In this case, the bank's expected cost of borrowing is just C(L,X) = [(1 - q)R + q[omega]]L.

The basic intuition at work in these results is as follows. A bank's expost cost of borrowing is always what it gives up in the end of the period: if it is solvent it repays its foreign debt, if it is bankrupt it gives up its gross assets. In the absence of government guarantees, what a foreign creditor receives is exactly what the bank gives up, minus any bankruptcy costs. It follows immediately that a bank's expected cost of borrowing can never be lower than the foreign creditors expected return, plus expected bankruptcy costs. As long as bankruptcy is costly, this means the optimal strategy of a cost minimizing bank is to fully hedge any exchange rate risk. In the presence of government guarantees, however, the government drives a wedge between what the bank gives up and what the foreign creditor receives in the state S = [S.sup.D]. Now, as long as bankruptcy is not incredibly costly, it is optimal for the bank to leave as little on the table as it can in the state where S = [S.sup.D]. It does this, using forward contracts, by transferring as many profits as it can to the other state, S = [S.sup.I]. Essentially, what the government has done is give the bank a free option contract that it can only profit from by bankrupting itself in the event that the exchange rate is floated.

Our results allow us to draw our main conclusion: under government guarantees, a banking crisis in which banks default, and the government must honor its guarantee to foreign creditors, will always coincide with the abandonment of the fixed exchange rate. In the absence of government guarantees, a banking crisis will never coincide with the abandonment of the fixed exchange rate. Hence, guarantees make banking crises more likely.

1.5. Optimal Lending

Given our results on optimal hedging behavior, we can now study banks' lending decisions. Notice that in the absence of government guarantees, banks set fully hedge and C(L, X) = RL. Hence, the bank chooses L to maximize V(L, X) = ([R.sup.a][S.sup.I]/F - [delta])L-RL. Clearly, the bank has an infinitely elastic supply curve at the interest rate

(13) [R.sup.a] = (R + [delta]) F/[S.sup.I].

On the other hand, under government guarantees we have C(L,X) = [(1 - q)R + q[omega]]L so that V(L,X) = ([R.sup.a][S.sup.I]/F - [delta])L - [(1 - q)R + q[omega]]L. In this case the bank will have an infinitely elastic supply curve at the interest rate

(14) [R.sup.a] = [(1 - q)R + q[omega] + [delta]] F/[S.sup.I],

which is lower than the interest rate in the absence of guarantees.

If we assume that there is a downward sloping demand curve for loans L = D([R.sup.a]), this means that (13) and (14) represent equilibrium interest rates, and that the amount of lending will be greater under government guarantees. (14) Not only do guarantees reduce the incentive to hedge, but they also act like a subsidy on lending. Our results imply that under either scenario V = 0, in equilibrium. Furthermore, it is easy to show that in either scenario the shareholders of the banks receive 0 from the banks in both states of the world.

2. Imperfect Credibility

In this section, I consider a situation in which the government's guarantee is viewed as only partially credible. In particular, I assume that foreign creditors assume that there is some probability [alpha] with which they will be bailed out by the government in the event that banks default in the state S = [S.sup.D].

To evaluate banks' optimal decisions in these circumstances we need only consider their expected cost of borrowing, C(L, X), if they choose X [less than or equal to] [X.sub.0], and compare it, once more, to RL. For a bank with X [less than or equal to] [X.sub.0], notice that the condition implying that foreign creditors expect a return of R on their loans, (8), is now given by:

(15) RL = (1 - q)[R.sup.b](L, X)L + q(1 - [alpha])[[V.sup.G](L, X; [S.sup.D]) - [omega]L] + q[alpha]RL.

Recall that a bank that defaults when S = [S.sup.D] has C(L, X) = (1 - q)[R.sup.b](L, X)L + q[V.sup.G](L, X; [S.sup.D]). Comparing this expression to (15), we see that

(16) C(L, X) = [alpha]q[V.sup.G](L, X; [S.sup.D]) + q(1 - [alpha])[omega]L + (1 - q[alpha])RL.

Minimizing this expression with respect to X, we get [V.sup.G](L, X; [S.sup.D]) = [omega]L, and

(17) C(L, X) = q[omega]L + (1 - q[alpha])RL.

This is lower than RL (the expected cost of borrowing if the bank fully hedges), as long as [alpha] > [omega]/R. In other words, it remains optimal for banks to choose X [less than or equal to] [X.sub.0] and default when S = [S.sup.D] as long as [alpha] > [omega]/R. If the government's guarantee is so lacking in credibility that this condition is violated, banks will fully hedge.

3. Would Banks Prefer to Lend in Dollars, or Borrow in Local Currency?

One question that arises is whether the contractual structure in the banking system described here is optimal. One might ask, for example, whether firms would prefer to make their loans to firms in dollars, since their borrowing is done in dollars. Ignoring the possibility that firms might have difficulty repaying loans in dollars when S = [S.sup.D] we can answer this question in a straightforward way. If banks lend in dollars, then the expression for [V.sup.G] becomes

(18) [V.sup.G] = ([R.sup.a] - [delta])L + X(1 - F/S).

Notice that this will not change bank's incentive to default when S = [S.sup.D]. The existence of hedging contracts allows banks to reduce their value in that state of the world to the point that they cannot repay their creditors. And this allows them to exercise the valuable option of transferring higher hedging profits to the other state of the world, S = [S.sup.I].

Another possibility is that banks would borrow local currency from abroad, or that the interest rates on their borrowing abroad would be indexed to the state of the world. Again, this would not change the outcome of the model. The independent existence of forward contracts as a means to default in the state of the world where S = [S.sup.D] means that banks would alter their hedging positions in order to default in that state. The message of the earlier sections depends only on the government issuing a guarantee, and the banks having an investment vehicle with which to take advantage of it.

4. Conclusions

In this paper I have shown how government's can affect the behavior of private agents when they take on contingent liabilities. In particular, I have used a model of banks in a small open economy to show how government guarantees to banks' creditors can make the banking system more fragile. The model in this paper is one in which there is some exogenous probability with which the government will abandon a fixed exchange rate regime. If the government issues a guarantee to the creditors of local banks, that it will bail them out in the event of a devaluation, this causes banks to make portfolio decisions consistent with default in precisely that state of the world. If the government doesn't issue guarantees, the banks protect themselves by hedging exchange rate risk in the forward market. In this way, guarantees make the banking system more fragile, because the abandonment of the fixed exchange rate regime necessarily brings on a banking crisis. In the absence of guarantees, these banking crises will not occur. (15)

This is not to say that all banking crises in the real world are related to currency crises. In the real world there are many sources of uncertainty that could potentially lead to a banking crisis. The point of this paper is to highlight a potential danger of a particular government policy that can raise the probability of banking crises. Governments may well have good reasons to issue guarantees, but they should be warned of their potential cost.

Notes

(1.) Mishkin (1996) and Obstfeld (1998) argue that a fixed exchange rate arrangement effectively is an implicit government guarantee of this kind. The role of the fixed exchange rate as a guarantee is also emphasized by Burnside, Eichenbaum and Rebelo (2001a,b,c) and Corsetti, Pesenti and Roubini (1999). The role of government guarantees to the financial sector--with less specific reference to the exchange rate regime--is also discussed by Diaz-Alejandro (1985), Velasco (1987) and Dooley (2000).

(2.) The lack of hedging by banks plays a crucial role in several papers motivated by the Asian crisis, for example, Aghion, Bacchetta and Banerjee (2000), Chang and Velasco (1999, 2000), and Krugman (1999).

(3.) The view that the behavior of speculators is important in currency crises has been emphasized by Obstfeld (1986a, 1996), Cole and Kehoe (1996), Sachs, Tornell and Velasco (1996), Radelet and Sachs (1998) and Chang and Velasco (1999).

(4.) Other models of the role of banks in currency crises include Akerlof and Romer (1993), Caballero and Krishnamurthy (1998), and Chang and Velasco (1999).

(5.) Burnside, Eichenbaum and Rebelo (2001d) consider examples in which banks have domestic and foreign creditors. This does not substantially change the conclusions drawn here.

(6.) The model could be set up differently so that banks borrow and lend in currencies and with contract arrangements of their choice. The results would be the same. In equilibrium banks would choose to expose themselves to exchange rate risk in the face of government guarantees.

(7.) These results mirror those in Kareken and Wallace (1978) concerning the impact of deposit insurance on bank portfolios.

(8.) In other models of currency crises which involve banks, information plays a more important role. See, for example, McKinnon and Pill (1996, 1998) and Chinn and Kletzer (2000).

(9.) In some crises banks did not literally have a currency mismatch in their portfolios because they made loans to local firms in dollars. See, for example Gavin and Hausmann (1996) who write about the Chilean banking crisis of 1982. See Burnside, Eichenbaum and Rebelo (2001a) for an extension of the model to the case where banks make loans in dollars but, as a result, face credit risk that is correlated with exchange rate risk.

(10.) See Burnside, Eichenbaum and Rebelo (2001a) for evidence on the availability of hedge instruments in countries affected by recent crises.

(11.) This does not violate Siegel's (1972) paradox, which pertains to expected nominal returns to forward contracts.

(12.) Burnside, Eichenbaum and Rebelo (2001a) allow for the possibility that shareholders lose capital invested in the bank. This does not substantially change the results.

(13.) See Sercu and Uppal (1995), Chapter 4.

(14.) Burnside, Eichenbaum and Rebelo (2001a, d) present models in which the demand curve for loans is determined within the framework of a general equilibrium model.

(15.) Burnside, Eichenbaum and Rebelo (2001d) explore how guarantees can make banking and currency crises simultaneously and mutually more likely.

References

Aghion, Philippe, Philippe Bacchetta, and Abhijit Banerjee. "A Simple Model of Monetary Policy and Currency Crises." European Economic Review, 2000, 44, pp. 728-38.

Akerlof, George and Paul Romer. "Looting: The Economic Underworld of Bankruptcy for Profit." Brookings Papers in Economic Activity, 1993, 2, pp. 1-60.

Burnside, Craig, Martin Eichenbaum and Sergio Rebelo. "Hedging and Financial Fragility in Fixed Exchange Rate Regimes." European Economic Review, 2001a, 45, pp. 1151-93.

Burnside, Craig, Martin Eichenbaum and Sergio Rebelo. "On the Fiscal Implications of Twin Crises," NBER Working Paper 8277, 2001b.

Burnside, Craig, Martin Eichenbaum and Sergio Rebelo. "Prospective Deficits and the Asian Currency Crisis." Forthcoming, Journal of Political Economy, 2001c.

Burnside, Craig, Martin Eichenbaum and Sergio Rebelo. "Government Guarantees and Self-Fulfilling Speculative Attacks." Manuscript, Northwestern University, 2001d.

Chang, Roberto and Andres Velasco. "Illiquidity and Crises in Emerging Markets: Theory and Policy," Working Paper, New York University, 1999.

Chang, Roberto and Andres Velasco. "Financial Fragility and the Exchange Rate Regime," Journal of Economic Theory, 2000, 92, pp. 1-34.

Chari, V.V., Lawrence Christiano, and Martin Eichenbaum. "Inside Money, Outside Money and Short Term Interest Rates." Journal of Money, Credit and Banking, 1995, 27 Pt. 2, pp. 1354-86.

Chinn, Menzie and Kenneth Kletzer. "International Capital Inflows, Domestic Financial Intermediation and Financial Crises under Imperfect Information." Mimeo, University of California, Santa Cruz, 2000.

Cole, Harold and Timothy Kehoe. "A Self-Fulfilling Model of Mexico's 1994-95 Debt Crisis." Journal of International Economics, 1996, 41, pp. 309-30.

Corsetti, Giancarlo, Paolo Pesenti and Nouriel Roubini. "Paper Tigers? A Model of the Asian Crisis." European Economic Review, 1999, 43, pp. 1211-36.

Demirguc-Kunt, Asli and Enrica Detragiache. "Does Deposit Insurance Increase Banking System Stability?" Working Paper WP/00/3, International Monetary Fund, 2000.

Dooley, Michael. "A Model of Crises in Emerging Markets." Economic Journal, 2000, 110, pp. 256-72.

Edwards, Sebastian and Carlos Vegh. "Banks and Macroeconomic Disturbances under Predetermined Exchange Rates." Journal of Monetary Economics, 1997, 40, pp. 239-78.

Gavin, Michael and Ricardo Hausmann. "The Roots of Banking Crises: The Macroeconomic Context," in Ricardo Hausmann and Liliana Rojas-Suarez eds. Banking Crises in Latin America. Washington, DC: Inter-American Development Bank, 1996.

Kareken, John and Neil Wallace. "Deposit Insurance and Bank Regulation: A Partial Equilibrium Exposition." Journal of Business, 1978, 51, pp. 413-38.

Krugman, Paul. "Analytical Afterthoughts on the Asian Crises." Manuscript, MIT, 1999.

McKinnon, Ronald and Huw Pill. "Credible Liberalizations and International Capital Flows: The Overborrowing Syndrome," in T. Ito and A. Krueger, eds. Financial Deregulation and Integration in East Asia, 7-42. Chicago: University of Chicago Press, 1996.

McKinnon, Ronald and Huw Pill. "International Overborrowing: A Decomposition of Credit and Currency Risks." World Development, 1998, 26, pp. 1267-82.

Mishkin, Frederic. "Understanding Financial Crises: A Developing Country Perspective," in Michael Bruno and Boris Pleskovic, eds. Annual WorldBank Conference on Development Economics 1996. Washington, DC: World Bank, 1996.

Obstfeld, Maurice. "Rational and Self-Fulfilling Balance-of-Payments Crises." American Economic Review, 1989, 76, pp. 72-81.

Obstfeld, Maurice. "Models of Currency Crises with Self-Fulfilling Features." European Economic Review, 1996, 40, pp. 1037-48.

Obstfeld, Maurice. "The Global Capital Market: Benefactor or Menace?" Journal of Economic Perspectives, 1998, 12, pp. 9-30.

Radelet, Steven, and Jeffrey Sachs. "The Onset of the East Asian Financial Crisis," NBER Working Paper 6680.

Sachs, Jeffrey, Aaron Tornell and Andres Velasco (1996) "Financial Crises in Emerging Markets: The Lessons from 1995." Brookings Papers on Economic Activity, 1998, 1, pp. 147-215.

Sercu, Piet and Raman Uppal. International Financial Markets and the Firm. Cincinnati, Ohio: South-Western Publishing, 1995.

Siegel, Jeremy. "Risk, Interest Rates and the Forward Exchange." Quarterly Journal of Economics, 1972, 86, pp. 303-09.

* This paper was presented at the 7th Dubrovnik Economics Conference organized by the Croatian National Bank in June 2001. Other papers from the conference will be included in the June 2002 issue of this journal which will be devoted to the conference proceedings.

Craig Burnside
The World Bank
1818 H Street NW
Washington DC 20433, USA