Some Parallels Between Currency and Banking Crises

Peter Isard, Andrew Rose, and Assaf Razin
Published Date:
January 2000
  • ShareShare
Show Summary Details


Department of Economics, Dartmouth College, Hanover, NH 03755


There is a sizeable literature on the causes of speculative attacks on fixed exchange rates and a large literature on the determinants of bank runs. Surprisingly, these two literatures rarely overlap, even though both types of crises involve attacks on asset price-fixing schemes. This paper draws a number of parallels between the work on currency crises and the work on banking crises and examines some of the new insights that are coming out of a more integrated approach in the aftermath of the Asian financial crises.

I. Introduction

Robert Flood has made important contributions to our understanding of speculative attacks on fixed exchange rates. Indeed, his many papers on currency crises have advanced a fruitful research agenda focused on the economics of process-switching. The “Floodfest” conference in his honor gives me the opportunity to take another look at the way economists think about crises and their causes.

In the 1990s, financial crises in emerging markets have been characterized by the collapse of both fixed exchange-rate regimes and financial intermediaries such as banks. While some have argued that these crises were essentially currency crises, others have pushed the view that the crises were fundamentally banking crises, where the fixed exchange-rate regime played no precipitating causal role. Still others have suggested that the currency and banking crises were closely intertwined.

The parallels between currency and banking crises are striking. Both involve attacks on asset price-fixing schemes. Both occur when the government can no longer credibly commit its assets in support of a fixed price, be it a fixed price between home and foreign currency or a fixed price between currency and bank deposits. Indeed, the government assets backing the exchange rate and bank deposits can ultimately be the same assets. Once the asset backing is gone, or the government chooses to halt its further depletion, price-fixing schemes collapse. Exchange rates can go into free fall and banks can become insolvent.

There is a large literature on the causes of speculative attacks on fixed exchange rates. (See the survey by Flood and Marion, 1997). There is also a substantial literature on the determinants of bank runs.1 (See the survey by Calomiris and Gorton, 1991). Prior to the Asian crises, however, these two literatures rarely, if ever, overlapped. What we find are currency crisis models that ignore the private banking sector and bank-run models absent open-economy features. The lack of overlap is even more remarkable since both literatures embrace the same two approaches to explaining crises. Both literatures suggest that speculative attacks may be either the anticipated outcome of inconsistent policies or the unanticipated outcome of self-fulfilling changes in market expectations.

Since the 1997-98 financial crises in Asia, like some earlier crises, involved both foreign exchange markets and private financial intermediaries, economists have begun drawing from both currency and banking crisis literatures to enhance their understanding of these events.

In this essay, I first describe the main views about the causes of currency and banking crises and draw some parallels between the two. I then focus on some of the newer research. The Asian crisis has spawned efforts to consider bank-run models in an open-economy setting or suggest ways banking features such as moral hazard in international lending might lead to an eventual collapse of a fixed exchange rate. I take a look at this new crisis literature and stress the benefits of taking a more integrated approach.

II. Crisis Causes

A. Causes of Currency Crises

Twenty years ago, a “first generation” of currency crisis models pointed to inconsistent government policies as the cause of a speculative attack on a fixed exchange rate.2 In most of this literature, the government fixes the price of foreign exchange but also monetizes a large fiscal deficit. Excessive domestic credit creation leads residents to exchange the unwanted domestic currency for foreign currency, reducing the government’s stockpile of international reserves. The erosion of the reserve stockpile is problematic since, to maintain the fixed price of foreign exchange, a government must have sufficient reserves to sell whenever the price of foreign exchange is about to rise.

If speculators wait until reserves are naturally depleted on their own, then at that point the central bank must abandon the fixed exchange rate and the price of foreign exchange will jump up. Speculators foresee this potential opportunity for a capital gain and compete against each other for the profits. In doing so, they advance the date when reserves will be exhausted. At some point before reserves are exhausted on their own, an attack occurs as speculators rush to purchase the government’s remaining stockpile of international reserves. The large transfer of real resources away from the government at the time of attack can be viewed as a penalty paid by the government for having pursued inconsistent policies.

Although this story tracks well with many past crises, it does not appear to characterize the recent one in Asia. During the mid-1990s, Asian governments were in approximate fiscal balance and not pursuing excessive credit creation. Nevertheless, it is important to remember that monetizing fiscal deficits when the exchange rate is fixed is just one example of inconsistent policies, albeit an historically relevant one. The main message of the first-generation model—that crises may be the predictable outcome of inconsistent policies—is still applicable today. Indeed, some of the new “third-generation” crisis models that focus on the role of government guarantees in promoting excessive investment have merely adapted the message of the first-generation model to a new set of policy inconsistencies.

A “second generation” of currency crisis models received considerable attention after the attacks on European currencies and the Mexican peso in the early 1990s. The second-generation models start from the premise that there is no underlying policy inconsistency before the crisis. Instead, these models consider an interaction between private sector behavior and government behavior that gives rise to several possible outcomes.3 In principle, the economy can jump from one outcome to another. A jump from a “no-attack equilibrium” to an “attack equilibrium” can be triggered by a sudden and unpredictable shift in market expectations.

Often the root of the problem is an underlying tension among the government’s multiple objectives. For example, the government may want to promote price stability or signal the markets of its intention to pursue a disciplined monetary policy in the future. It can advance these sorts of objectives with a fixed exchange rate. On the other had, the government may also wish to limit its debt service obligations, lower the rate of unemployment, or inject liquidity into a troubled banking system. These objectives can better be achieved if it abandons the fixed exchange rate to pursue monetary expansion.

As long as the benefits of the fixed exchange-rate policy exceed the costs, the fixed exchange rate is maintained. A shift in market expectations about the viability of the fixed exchange rate may alter the cost-benefit calculus, however. For example, if private agents start to give more weight to the probability of devaluation, interest rates or wage demands can increase, worsening the prospects for lower debt service, a sounder banking system or reduced unemployment. The government may then decide that maintaining the fixed exchange rate is too costly. The government’s decision to devalue validates expectations, making expectations self-fulfilling.

Second-generation currency crisis models often illustrate the trade-offs faced by a government with a social loss function. For example, the government might conduct exchange-rate policy in order to minimize the deviation of prices and employment from their desired levels in the face of output shocks. If output shocks are small, the government will generally find it optimal to keep the exchange rate fixed even though employment deviates somewhat from its desired level.

While an economy may be fortunate enough to experience small shocks most of the time, it may nevertheless face large shocks every now and then. Consequently, it may be optimal for the government to follow a fixed exchange-rate rule most of the time but abandon the rule on occasion when disturbances to the economy are very large. Of course, the government must face a cost each time it invokes the “escape clause” or it will be tempted to break the rule (devalue) too often.4

Because of the way people form their expectations about the future value of the currency, there may be two (or more) values for the threshold disturbance that triggers the escape clause. Suppose the economy is at an equilibrium where only disturbances exceeding the largest threshold value can cause a crisis. If private expectations suddenly become more pessimistic, then the economy can jump to a different equilibrium where smaller shocks bring about a crisis.

In second-generation models of currency crises, the concept of “backing” for the fixed exchange-rate commitment is much broader than the international reserves on the government’s balance sheet. Backing encompasses what the government is willing to give up in order to maintain the fixed exchange-rate policy. Not only must the government be willing to lose international reserves in defense of the fixed exchange rate, it must be willing to sacrifice other things as well—such as some employment or the solvency of some banks. Its willingness to give up these other goals depends on the state of the economy. It is harder to sacrifice further employment for the sake of defending the fixed exchange rate if unemployment is already high. It is harder to allow some banks to fail if it then puts many other banks and firms in financial jeopardy. Second-generation models therefore require that the economy’s fundamentals be weak before a shift in expectations can pull the economy into a crisis.

Because the Asian economies enjoyed high growth, low unemployment and low inflation up until the crisis hit, some have questioned the applicability of the second-generation model to the Asian experience. Yet some Asian governments did face a worsening tradeoff between the goal of maintaining a stable exchange rate and the desire to support fragile banks and indebted firms.

The key message of the second-generation model—that crises can be the unpredictable outcome of a change in market expectations—is an important one. Indeed, some “thirdgeneration” crisis models developed in the aftermath of the Asian experience also rely on a shift in market expectations to trigger a crisis.

B. Causes of Banking Crises

Almost twenty years ago, Flood and Garber adapted their perfect foresight model of a speculative attack on a fixed exchange to the case of a closed-economy bank run (Flood and Garber, 1981). They described a situation where commercial banks transform nominal deposit liabilities into both reserves and long-term bonds. The banks also make a commitment to a fixed nominal price. They agree to pay on demand one unit of high-powered money for each unit of home currency deposited with them. In the absence of deposit insurance, the banks can maintain this price-fixing scheme as long as their assets fully cover their deposits.5

Flood and Garber showed that a bank run can be the predictable outcome of inconsistent policies. In the example they constructed, a central bank policy of deflation undermines the commercial banks’ commitment to redeem deposits at par.6 To maintain the nominal value of their assets in a deflationary environment, banks must purchase new assets to offset the capital losses on their currently held assets. For a time, banks are willing to make these new purchases since the earnings on their asset holdings exceed the cost of managing their portfolio and the size of the asset valuation loss.

Over time, however, the deflation reduces asset earnings. Eventually, it is no longer profitable for banks to maintain the assets required for full backing of their deposit liabilities. At that point, there is a bank run. The attack occurs at the last instant when the banks can fulfill their obligation to convert demand deposits into currency at par.7 Since the bank run occurs in a closed-economy environment without a lender of last resort, the run does not generate a claim on the central bank’s domestic or international reserve assets. Rather, the run forces a transfer of real resources from the commercial banks to private agents.

While this early bank-run example lacks some desirable properties of later bank-run models, it nevertheless shows how a policy inconsistency that erodes the value of banks’ net assets can generate a predictable bank run.8 When banks can no longer maintain sufficient assets to meet their nominal liabilities, depositors, faced with incipient capital losses, run the banking system.

A sudden shift in expectations can also produce a bank run. Moreover, the possibility of a run arises because of the underlying tension between economic objectives—the need for flexibility and the economic pay-off from long-term commitment.9 This tension is transparent in the closed-economy bank-run model of Diamond-Dybvig (1983), where banks transform deposits into high-yielding long-term assets that are costly to liquidate in the short term. Agents prefer the high returns associated with long-term investments but may have to consume at an earlier date due to unexpected shocks.

A bank permits private agents to achieve the optimal allocation of investment and consumption since it provides risk sharing among individuals who need to consume at different random times. The bank offers a positive return on deposits and allows deposit withdrawals on demand. The bank stores a fraction of the deposits and invests the rest in the long-term high-yield technology. By the law of large numbers, deposit withdrawals in the short term will generally equal the expected withdrawals of agents who discover they must consume early, and the bank can cover these withdrawals with its liquid reserves. This is the good outcome.

A second equilibrium outcome is a bank run. In the absence of deposit insurance, depositors in the short term may come to believe that the bank is unsafe and everyone else will be withdrawing their deposits. In that case, they will immediately attempt to withdraw their own funds. The bank is forced to liquidate its long-term investment, but since the liquidated value of bank assets is less than the amount people would like to withdraw, the bank fails. The shift to pessimistic expectations brings about the very event depositors feared.

As noted by Calomiris and Gorton (1991), the rush to withdraw arises because of the first-come-first-served rule of deposit withdrawals. Since those who wait may end up with nothing, depositors compete with each other to be the first to withdraw.10 The run imposes real costs on the economy because it results in the early termination of productive investment.

The event that triggers the change in expectations and moves the economy from the “no-run equilibrium” to the “run equilibrium” is left unspecified. Not surprisingly, the Diamond-Dybvig approach is thus called the “random withdrawal” view of bank runs.

A number of researchers [e.g. Chari and Jagannathan (1988), Gorton (1985), Calomiris and Khan (1991) and Calomiris and Gorton (1991)] have suggested that any piece of news that leads depositors to view bank portfolios as riskier might trigger a bank run. Since banks hold nonmarketable assets that make their portfolios hard to monitor, depositors do not know which specific banks will be most affected by the bad news. Consequently, depositors may decide to withdraw a large volume of deposits from all banks. Banks then choose to suspend convertibility and sort out which of them are insolvent. Banks can accomplish this task because they have better information about each others’ portfolios.

Bank runs may therefore resolve the information asymmetry between banks and depositors and help depositors monitor bank performance. This multiple equilibria story is called the “asymmetric information” view of bank runs. It argues that the economy’s jump from the “no-run equilibrium” to the “run equilibrium” can be rationally triggered by movements in a noisy indicator related to the quality of bank assets. In empirical work, indicators correlated with returns on bank investments include the size of liabilities of failed nonfinancial businesses and the failure of a particularly large non-financial corporation.

III. Parallels between Currency and Banking Crises

Some parallels between currency and banking crises are obvious, but worth emphasizing at the outset. Both crises are attacks on price-fixing policies. In one case, speculators bet that the government cannot maintain the fixed price for foreign currency and rush to purchase the government’s remaining stockpile of international reserves before the fixed price is abandoned. In the other case, depositors believe the banks cannot maintain the fixed price between deposits and currency and rush to withdraw their deposits before the conversion rate is abandoned.

Both types of attacks occur when a finite stock of assets is still held by the institutions supporting the fixed-price scheme. Both types of runs produce a discontinuous drop in the asset holdings of the institution directly under attack. In a currency crisis, the assets of the monetary authority are depleted; in the case of a banking crisis where there is no deposit insurance, the banks’ assets are depleted.

Currency and banking crises spring from the same two causes. They may be the predictable outcome of policy inconsistencies or the unpredictable outcome of sudden shifts in market expectations.

A predictable currency crisis can arise because the government fixes the price of foreign currency yet allows domestic credit growth to erode the stock of international reserves in a predictable way. More broadly, the crisis can result from pursuing any policy that depletes the reserves needed to back the fixed exchange-rate commitment. In the case of a predictable bank run, commercial banks promise to exchange currency for deposits at par on demand and yet the government pursues a policy that predictably erodes the value of the assets backing the commercial banks’ commitment. By causing a predictable deterioration in the assets backing the nominal commitment, the policy inconsistency leads to the inevitable collapse of the fixed-price promise.

An important contribution of the literature on predictable runs is to show that a crisis need not be the result of irrational investors, market manipulators or big shocks to the economy. Rather, a crisis can be the outcome of rational agents who observe the deterioration in the resources backing the nominal commitment and try to profit from dismantling the underlying policy inconsistency.

In both types of predictable runs, the time of attack is motivated by profit opportunities. In a perfect-foresight attack on a fixed exchange rate, for example, speculators compete with each other for potential gains in the currency market and end up advancing the time of attack to where there are no realized profit opportunities. When there is uncertainty and the time of the attack can not be perfectly foreseen, successful speculators are awarded capital gains. In the perfect-foresight bank-run example, an attack occurs the moment banks no longer find it profitable to maintain full backing for their deposits.

Perfectly foreseen runs are sudden but orderly. The exchange of assets is consistent with the desires of private agents. Speculators attacking a fixed exchange rate obtain a stock of reserves that just matches the decrease in their demand to hold domestic currency. Depositors running the banks obtain a stock of liquid bank assets that just matches the value of their desired deposit withdrawals. When the exact time of a crisis is not perfectly foreseen, runs are characterized by a panic atmosphere in which agents queue up to acquire international reserves or bank assets according to a “first-come-first-served” criteria.

In both the currency and banking crisis literature, there is also a family of models that relies on an unpredictable shift in market expectations to trigger a crisis. When the markets become more pessimistic about the government’s commitment to the fixed exchange rate, the government may decide to abandon the fixed exchange rate and validate market expectations. When the markets become more pessimistic about the bank’s commitment to its fixed price for currency in terms of deposits, it can trigger a bank run that confirms people’s fears.

In both cases, beliefs are shaped by more than the assets directly backing the fixedprice policy. In the currency crisis case, beliefs are also influenced by the government’s willingness to give up other goals. In the banking crisis case, there is never sufficient asset backing in the short term to support the fixed-price promise, so a shift in expectations that triggers a run is either a random occurrence or a result of some generalized bad news.

Even though both currency and banking crises can be unpredictable, it is not necessarily true that crises “come out of the blue.” In second-generation currency crisis models, a shift in expectations triggers a crisis if fundamentals are already weak. In some bank-run models, a shift in expectations generates a bank run only if it is preceded by bad news about the economy that implies an inability of some banks to support their fixed-price policy. Only the Diamond-Dybvig story of a bank run suggests that a random shift in expectations can precipitate a crisis.11

Moreover, both literatures on unpredictable runs face challenges about whether there really are viable multiple equilibria. If economic incentives are used to distinguish among equilibria in second-generation currency crisis models, then speculators who are awarded capital gains in a crisis may prefer to settle on the equilibrium where attacks are most frequent.12 If uninsured banks never have sufficient backing in the short term to sustain their fixed-price policy, then the run equilibrium might be the only relevant equilibrium.

IV. Banking Crises with Open-Economy Features

Since the start of the Asian crisis, economists have developed open-economy versions of predictable and unpredictable bank runs. We shall first describe key features of these models and then consider the new insights from adding open-economy features.

A. Predictable Bank Runs with Open-Economy Features

Open-economy versions of predictable bank runs show how explicit or implicit government guarantees of resident foreign-currency liabilities can promote excessive investment, often in overly risky projects. Moreover, these guarantees increase the contingent claims on the government’s international reserve assets. Once these claims on reserves increase to a certain threshold, capital inflows can suddenly become capital outflows and a crisis occurs.

The underlying policy inconsistency arises because the government provides an insurance guarantee for the currency-deposit conversion rate yet liberalizes and deregulates the financial sector so that the available backing to support new guarantees declines. The problem is exacerbated if the country has a fixed exchange rate since the international reserves available to support the guarantees must also be available to support the fixed exchange-rate policy.

One example of a predictable bank-run model with open-economy features is the one by Dooley (1997).13 His model has two important features in common with the closedeconomy bank-run model of Flood and Garber (1981). First, the price-fixing scheme to trade deposits for currency at par can only be supported as long as deposit liabilities are fully covered. Second, a policy inconsistency leads to a situation where deposits can no longer be fully covered, triggering a run.

In Dooley’s set up, foreign-currency deposits are backed in part by private bank assets and in part by government insurance that is paid out using international reserves. Because the government insures poorly regulated domestic financial markets, banks increase their international financial liabilities at a more rapid pace than the government increases its reserve assets. When these liabilities are about to exceed their backing, a run is triggered.

The Dooley story has several attractive features. It rationalizes the international capaital that could beital flows to emerging markets before the crisis by showing that a positive shock to the macroeconomic environment can turn government net reserves positive and thus create an insurance incentive for foreign investors. It also relies on the profit motive to explain the timing of the speculative international capital outflow that takes the form of a bank run. When foreign investors no longer earn above-market rates of return, they have an incentive to pull their funds out of the banks. It also shows that the speculative attack is an attack on the government’s international reserve assets whether or not the country has a fixed exchange-rate regime.

The sequence of events can be illustrated with Dooley’s diagram, labeled here as Figure 1. The positive vertical axis in the top panel measures government assets that could be liquidated in order to bail out resident banks if they should default. These are foreign-currency denominated assets, or international reserves, and they include some limited lines of credits from other governments or international organizations. The negative vertical axis measures the government’s liabilities. These liabilities represent the government’s noncontingent foreign liabilities—the foreign exchange the government owes its foreign creditors in the absence of bank runs—and its contingent liabilities—the foreign currency the government owes in case resident banks default. Initially, government assets are not even adequate to cover its noncontingent liabilities. As a result, foreign investors have no desire to deposit their funds in resident banks; government guarantees are not credible because there are no available assets to back them up.

Figure 1.Dooley government guarantees on foreign borrowing.

Now suppose a change in the macroeconomic environment at time t1, such as a drop in international interest rates, reduces the value of noncontingent government liabilities so that net assets become positive. (Net assets equal gross assets minus noncontingent liabilities.) These positive net assets can now support an implicit or explicit government insurance guarantee for bank liabilities.

The middle panel of Figure 1 shows the growth of insured liabilities over time (line D).14 Once the government’s net assets become positive, banks have an incentive to seek deposits from foreign investors. The reason is that they plan to pay back only a fraction (1—α) of these deposits and rely on government insurance to cover the remaining fraction, α (0 < α < 1). (Banks will appropriate αD of the proceeds for themselves.) The promise of a government bailout sets the stage for an ongoing erosion of the government’s financial position. In poorly regulated and supervised financial markets, α may be a large fraction and the flow of new insured liabilities (the slope of line D) may also be large. Hence the government’s required insurance payments in case of a run (line αD) may grow alarmingly over time.

The bottom panel of Figure 1 illustrates the covered interest differential in favor of insured liabilities. Because banks do not plan on repaying the full amount of their insured liabilities, they can afford to offer an above-market yield to foreign investors. Essentially banks compete with each other to obtain foreign deposits by offering a share of their appropriation with foreign investors. As long as foreign investors earn above market yields, they have no incentive to mount an attack on the government’s reserves.

The increase in foreign deposits in domestic banks causes government reserve assets to increase, but not one-for-one, since some reserve assets are spent in support of resident purchases of foreign goods, services or financial instruments. In addition, reserve assets earn the risk-free rate of return, not the higher return offered to foreign investors. The top panel of Figure 1 illustrates that the growth in government liabilities exceeds the growth in its reserve assets.

As long as the financial liabilities of banks are backed fully by a combination of bank and government assets, there will be no run. The run occurs at time t2, the last moment when the government’s reserve assets can cover its liabilities.15

In the Dooley model, the government stock of international reserve assets is lost because the government chooses to honor its contingent liabilities. Once lost, there are no reserves left to stabilize the nominal exchange rate. It follows that if the country had a fixed exchangerate regime, it will likely collapse with the bank run.16

By considering a predictable bank run in an open-economy setting, the Dooley model and others along the same lines do three things. First, they illustrate the expansion of the government’s nominal commitments. The government now stands ready to support the fixed rate between bank liabilities and currency and the fixed rate between home and foreign currency. If bank liabilities are in domestic currency, then a government bailout of the banks requires an injection of liquidity that might undermine the fixed exchange-rate policy. If bank liabilities are in foreign currency, then the government’s commitments to the banks and to the fixed exchange rate become even more closely intertwined. A bank bailout draws on the same resources needed to support the fixed exchange rate. The government’s commitment to the banks is only as good as its commitment to the fixed exchange rate. As the reserve backing erodes, both commitments are undermined.17

Second, the Dooley story shows that the private sector always has an incentive to transfer to its own balance sheet the government assets backing a nominal commitment. When a positive shock produces an increase in the government’s net worth, agents may take advantage of the government’s nominal commitments to increase their own net worth at the government’s expense. Just as currency speculators try to purchase the remaining government reserves in the hopes of making capital gains, bank owners may steal a fraction of foreign-currency deposits, knowing that the government will draw on its reserves to bail out depositors.

Third, the Dooley model shows that an economy can experience international capital inflows and an increasing stock of international reserves up until the moment of attack. What matters is the government’s net reserve stock. When the government’s contingent foreign-currency liabilities are increasing at a faster pace than its international reserves and lines of credit, its net stock of foreign-currency reserves is deteriorating in a predictable fashion, bringing closer the time of inevitable collapse.

B. Unpredictable Bank Runs with Open-Economy Features

Several papers have extended the Diamond-Dybvig (1983) model of an unpredictable bank run to an open-economy setting.18 In these papers, the value of bank assets accessible in the short term continues to fall short of potential withdrawals, and bank runs are still generated by self-fulfilling shifts in expectations. Goldfajn and Valdes (1997) show that intermediation of foreign funds through the banking system increases the probability of currency crises and bank runs. They also show how intermediation magnifies the size of capital outflows associated with crises. Chang and Velasco (1998) explore the implications of global financial liberalization for banks’ vulnerability to runs as well as the link between financial fragility and the fixed exchange-rate system. We shall focus on the Chang-Velasco open-economy extension of the Diamond-Dybvig model.

The Chang-Velasco model adds a world capital market to the three-period Diamond—Dybvig framework. One unit of a good can be invested in the world capital market at date to to yield one unit in either time t1 or t2. The domestic production technology retains the same characteristic as in Diamond-Dybvig—it is quite productive if the investment is held for two periods, but it is costly to liquidate early. Suppose R > 1 is the return on the investment if it is held for two periods, but L < 1 is the return if it is liquidated after one period. Only domestic residents have access to this technology.

Demand deposit contracts require agents to surrender their endowment, e, and their rights to invest or borrow abroad to the bank at to. Agents have the option to withdraw C1* units of consumption from the bank in period 1 or C2* units of consumption in period 2, where consumption is greater if deposits are held two periods, C2* > C1*. The bank can borrow bo from abroad at time to and b1 at t1. The bank faces an overall international credit ceiling of b¯=b0+b1 The bank invests the endowment and funds initially borrowed from abroad in the long-term illiquid technology (I = e + bo).

Chang and Velasco make two assumptions about foreign debt (that are later relaxed): (1) the bank always repays its foreign debt, and (2) any foreign debt of one-period maturity acquired at time to can be automatically renewed at t1 on the same conditions as before.

The Chang-Velasco story can be illustrated in Figure 2. There are two possible equilibrium outcomes. In one equilibrium, which corresponds to the optimal allocation, only agents who find out they must consume early withdraw deposits at time t1, and the bank can fully cover the withdrawals by borrowing from abroad at t1. The bank does not have to liquidate its long-term asset nor hold some liquid assets between dates to and t1.

Figure 2.Chang-Velasco bank run.

Alternatively, all domestic agents may attempt to withdraw their deposits at time t1 because they believe everyone else will be doing the same. In that case, the uninsured bank will fail since its obligations to domestic depositors exceed the value of available assets.19 As in the Diamond-Dybvig model, this open-economy version is silent on what causes the economy to jump from the no-run equilibrium to the run equilibrium.

In the Chang-Velasco model, a shift to more pessimistic expectations by foreign creditors increases the vulnerability of banks by reducing the amount of liquidity banks have available in the short run. The expectational shift induces foreign creditors to stop lending and makes them unwilling to roll over the short-term debts banks previously incurred. As a result, the value of assets available to the bank at time t1 is reduced, increasing its vulnerability to a run.20

Figure 3 illustrates the result of a shift to pessimism by foreign creditors. Note how banks that previously had the greatest access to the international capital markets now face the largest drop in available assets when the adverse shift in expectations occurs. The reason is that banks with greater access to external funding initially faced a higher credit ceiling and acquired more short-term debt, b0. When this debt cannot be rolled over, the drop in available bank resources, (1LR)b0, is greater. Without these resources, banks are more fragile. Indeed, greater fragility may increase the chance of a run.21

Figure 3.Chang-Velasco foreign creditor panic.

Chang and Velasco illustrate that in an open-economy version of bank runs, the central bank can try to save the banks or preserve the fixed exchange rate regime, but it cannot do both. If there is a bank run and the central bank does not supply the banks with extra liquidity (that is, act as lender of last resort), then the banks will fail. The central bank would be constrained from supplying the extra liquidity if it operated a fixed exchange rate and a currency board, for example. In that case, it could not ensure new liquidity in the absence of an equivalent amount of new foreign-currency assets coming in. Alternatively, if the central bank did not have a currency board, it could issue the domestic liquidity to keep the banks solvent, but then depositors could attempt to trade the home currency withdrawn from the banks for foreign currency (in the amount of C1* if the exchange rate equals one). The fixed exchange rate collapses because the foreign-currency assets available to meet this demand are only bI+L(1b¯R). The central bank can borrow from abroad and it can liquidate the assets it takes over from the private banks, but the total amount of foreign-currency assets it can acquire in this manner still falls short of demand since bI+L(1b¯R)<C1*. Unless it can acquire additional foreign assets from international lending organizations or foreign governments, the fixed exchange-rate regime will collapse.

Several important themes come out of Chang and Velasco’s open-economy version of the Diamond-Dybvig bank-run model. First, newly liberalized domestic banks often borrow foreign currency from abroad and take in foreign-currency deposits but still lend mostly in domestic currency. These banks can become illiquid when their short-term liabilities in foreign currency exceed the amount of foreign currency they can get access to on short notice. This point reinforces the message by Kaminsky and Reinhart (1999), who found that 70 percent of the banking crises they studied were preceded by financial sector liberalization in the previous five years and that financial liberalizations accurately signaled 67 percent of all banking crises studied. It also corresponds to the findings of the predictable bank run models that show that deregulation combined with explicit or implicit guarantees for poorly-supervised banks can generate overlending, excessive risk taking and an eventual crisis.

Second, the open-economy version of the Diamond-Dybvig bank run shows that the accumulation of short-term external liabilities by banks can be risky. If foreign creditors should panic and refuse to roll over bank debts, the banks have less foreign-currency assets to draw on and are more vulnerable to a run. Thus the short maturity of capital inflows can contribute to bank fragility. A related point is that while a crisis can still be triggered by a shift in the expectations of domestic depositors, as in the closed economy story, a shift to pessimism by foreign creditors may also help precipitate a crisis. Indeed, the distinction between a foreign creditor panic and a domestic bank run may be blurred or both may occur at the same time and reinforce each other.

Finally, as Chang and Velasco emphasize, “the combination of an illiquid financial system and fixed exchange rates can be lethal” (1998, p. 41). If the central bank does not act as a lender of last resort (because it operates a currency board, for example), then bank runs can occur. If the monetary authority does act as a lender of last resort in domestic currency, then bank runs can be eliminated but only at the cost of undermining the fixed exchange rate, since private agents will try to convert the newly issued domestic currency into foreign exchange. If the monetary authority tries to act as a lender of last resort in foreign currency, then it is drawing on the same reserve assets needed to support the fixed exchange-rate commitment. Consequently when there is a fixed exchange rate and domestic banks are internationally illiquid, there will be either a banking crisis or a currency crisis should expectations turn pessimistic.

V. Conclusion

In the 1990s, emerging markets have often been confronted by both currency crises and banking crises. As noted by Kaminksy and Reinhart (1999), these joint crises did not emerge in the post-war period until the 1980s, when some developing countries such as Chile liberalized financial sectors where banks played a dominant role and simultaneously dismantled controls on capital-account transactions. The linkage has become a striking aspect of the financial crises in the 1990s.

When the government fixes the price of foreign currency and stands ready to bail out banks that promise to pay depositors currency on demand, it takes on two nominal commitments. Currency and banking crises are attacks on these asset price-fixing schemes. They are attempts by the private sector to transfer the government resources backing the nominal commitments to private portfolios. They can be the predictable outcome of inconsistent policies that erode the resource backing for the fixed prices, or they can be the unpredictable outcome of a change in expectations about the credibility of the fixed-price pledges.

The literature designed to explain the causes of currency crises and the literature on the determinants of bank runs in closed economies have developed along separate but parallel lines. In response to the Asian crises, there have been attempts to develop bank-run models with open-economy features, but as yet no efforts to specify currency-crisis models with a private banking sector trading in international assets. Even though much work still needs to be done to untangle the complexity of the Asian financial crises, the new work highlights the old dilemma of having too few tools to achieve too many targets. For a number of emerging economies that have liberalized and globalized, the attempt to sustain two fixed-price schemes with the same stock of international reserves has proved to be an impossible task.

The term “bank run” is often used synonymously with “bank panic” and is meant to describe an event involving a significant number of banks. It also refers to a situation where depositors suddenly demand conversion of their deposits into currency.

Robert Flood played an important role in developing these models. The seminal contributions were made by Steve Salant and Dale Henderson (1978), Paul Krugman (1979), and Bob Flood and Peter Garber (1984a).

For examples, see Flood and Garber (1984b), and Obstfeld (1986, 1994, 1997).

The escape-clause model was developed by Flood and Isard (1989) and Persson and Tabellini (1990) and applied to fixed exchange-rate regimes by Obstfeld (1997). Flood and Marion (1997) illustrate that the escape clause finds its advantage over a permanent fixed exchange-rate rule only if the economy faces very large shocks on occasion.

Irving Fischer (1911) was among the first economists to suggest that a bank run can occur when the bank’s assets no longer cover its nominally fixed liabilities, namely its demand deposits. Note that the run is predicated on there being no lender of last resort.

In a later example meant to capture features of the U.S. savings and loan crisis, Garber (1981) modeled a bank run where the policy inconsistency involves thrift institutions who hold long-term, fixed-interest assets while the central bank produces accelerating inflation and offers only limited deposit guarantees.

Bank assets fully cover bank deposits as long as R + PBB = D, where R is the book value of bank reserves, PBB is the market value of long-term bonds whose price is PB and whose supply is B, and D is the book value of deposit liabilities. (For simplicity, bank capital is zero, no interest is paid on deposits, and there are no deposit guarantees by the monetary authority.) Banks suffer capital losses on their bond holdings due to deflation. They are willing to acquire additional bonds to offset these capital losses as long as the earnings from their bonds net of capital losses exceed the costs of managing their portfolios. The condition required to maintain full asset backing is:

where PB(t)Y¯ is nominal GDP, θ is capital’s share of GDP, B(t)/B¯ is the fraction of total bonds, B¯, held by banks. P˙B(t)B(t) is the capital loss on bonds held by banks, and T (t) is the cost per bond of managing the banks’ portfolio. Bond holders such as banks receive the income earnings of physical capital once deflation causes capitalists to default on their bond payments. Equation (la) says that as long as these earnings net of capital losses exceed bank costs, banks have the incentive to maintain the market value of their assets.

Dividing both sides of Equation (la) by PB(t)B (t), we can restate condition (la) as:

The left-hand side of the inequality in (lb) is the nominal interest rate earned by banks. The right-hand side of (lb) is the lowest nominal interest rate that gives banks the incentive to maintain full backing. (This lowest rate, or floor rate, is assumed to be a constant, k.) A run on the bank can occur as soon as the interest rate falls to the floor.

For example, this early model does not consider the implications of government guarantees of bank deposits. Neither does it rationalize the special role played by private banks in transforming liquid liabilities into illiquid assets. In addition, the banks operate in a closed-economy environment.

Krugman (1998) has stressed this point in his description of the Asian crisis. Recall that second-generation currency crisis models are also based on an underlying conflict between economic objectives.

Jacklin (1987) notes that markets must also be incomplete so that agents cannot trade claims on physical assets to cover unexpected short-term consumption needs. Bryant (1980) and Waldo (1985) have developed similar models of a bank run.

Note that in the Diamond-Dybvig model, investment returns are certain. Thus bank runs are purely speculative in origin and, according to Diamond and Dybvig, can be triggered “by almost anything.” (Diamond-Dybvig, 1983, p. 404). If we introduce the possibility that bank investments may be risky, then any information that leads individuals to alter their beliefs about the quality of bank loan portfolios might trigger a run. Gorton (1985) and others have developed models in which bank runs are motivated by shifts in beliefs about bank portfolio “fundamentals.” These types of bank runs may have a predictable component.

Other examples of a predictable collapse of financial intermediaries in an open-economy setting are Corsetti, Pesenti and Roubini (1998) and McKinnon and Pill (1996). An important early paper that contains some of these same ideas is Diaz-Alejandro (1985). Krugman (1998) and Schneider and Tornell (1998) also develop models with explicit or implicit government guarantees to explain the Asian financial crisis but these models do not rely on open-economy features. Moreover, Krugman notes that a crisis can also be triggered by a sudden shift in expectations if the markets come to believe that government guarantees will not be forthcoming.

Dooley’s model actually applies to any domestic resident institution or individual, not just banks. Moreover, resident liabilities need not be just bank deposits but could extend to equities, corporate bonds, government securities and other instruments. In my discussion of Dooley’s model, I shall focus only on banks and their financial liabilities to foreigners.

It is assumed it takes time to attract these deposits, so liabilities do not jump up at time t1, but rather grow smoothly over time.

With uncertainty, doubts about the government’s commitment to exhaust its reserves or confusion about the size of government assets and liabilities can also generate speculative attacks.

Dooley suggests it is possible for the country to continue a fixed exchange-rate regime after the bank run if a tiny positive shock turns net reserve assets positive.

In Corsetti, Pesenti and Roubini (1998), the government bailout of the private sector’s liabilities is covered only in part by international reserves. The rest is covered by explicit taxation. Thus the contingent liabilities of the government represent a fiscal problem that is not apparent until the crisis unfolds. The expectation that tax revenue will fall short of what is required can also generate expectations of future money growth and exchange-rate depreciation that will trigger a currency crisis. The Corsetti, Pesenti, Roubini story therefore has a parallel to the first-generation currency crisis model that emphasizes fiscal deficits and their monetization as an underlying cause of crisis.

Some examples are Garber and Grilli (1989), Goldfajn and Valdes (1997) and Chang and Velasco (1998). It is interesting that open-economy bank run models developed to analyze the Asian crises are based on the Diamond-Dybvig approach since empirical evidence from earlier crises favors the “asymmetric information” view of bank runs over the Diamond-Dybvig “random withdrawal” story.

If at t1 all agents try to withdraw deposits, total withdrawals will be C1*. The value of bank assets at t1 is b1+L(Ib¯R) since the bank borrows b1 from abroad and liquidates (Ib¯R) of the long-term investment at a return of L. Liquidation is only partial because Chang-Velasco assume the bank honors its commitment to repay all its foreign debts at t2. As a result, the bank keeps a fraction (γ) of the investment illiquid to pay off the foreign debt at t2. Repayment of the foreign debt at t2 requires RγI=b¯. Since only (I—γI) of the long-term investment is liquidated early and γI=b¯R, then L(IγI)=L(Ib¯R). To show that at time t1 potential bank liabilities [C1*] exceed the liquidated value of bank assets [b1+L(Ib¯R)], we define π 1 as the probability of consuming early and π 2 as the probability of consuming late and use the solution for the optimal allocation (I = e + bo1C1* = b1, π2C2*=RIb¯, b¯=b0+b1) to write Rπ1C1*+π2C2*=Rb1+(RIb¯)=Rb1+R(e+b0)(b0+b1)=Re+(R1)b¯. With a little bit of algebra, it can then be shown that bank assets at time t1 are b1+L(Ib¯R)=LC1*L[(b¯b0)(1+1L)]<C1*.

If foreign creditors refuse to lend more at time t1, b1 = 0 and the value of assets available to the bank at t1 falls to L(Ib0R). Note that if foreign creditors also refuse to roll over the initial debt, bo must be paid back at t1 and bank assets fall further, to LI—bo.

Chang and Velasco also discuss the effect of financial liberalization, achieved through lower reserve requirements or reduced monopoly power of banks, on the availability of bank assets at time t1. In addition, they analyze the effects on bank assets of asset price booms and busts, an unexpected increase in world interest rates at t1, and a government subsidy for the long-term investment project, They show that the coefficient of risk aversion for the consumer has to be greater than one to ensure the possibility of the bank-run equilibrium. The Diamond-Dybvig model assumes a risk aversion parameter greater than one.



Let me begin by saying that I really enjoyed reading this well written and very interesting paper. What Nancy does in this paper is to present a cohesive survey of three strands of literature: the theoretical literature on currency crises, banking crises, and the more recent papers that have attempted to link the two. It is the “twin crisis” literature, which focuses on banking crises in open economy models, that the paper devotes the most attention to. Like the literature on currency crises, very different interpretations of the causes of the twin crises have been offered and this paper lays out some of the key differences between first generation models, which stress the role of economic fundamentals, and second generation models, which highlight the role of self-fulfilling crises and multiplicity of equilibria.

I am going to divide my remarks into three parts. First, I am going to make some brief comments about the theoretical models discussed in the paper, complementing Nancy’s discussion of these. Second, I am going to turn my attention to the empirical evidence on the links of currency and banking crises. Lastly, I will make some remarks about the scope and direction of future research in these areas.

Banking Crisis Models with Open Economy Features

A predictable open-economy bank-run model, which is discussed at some length in this paper, is that by Dooley (1997). In the Dooley setting, the policy inconsistency arises because the government provides an insurance guarantee for the currency-deposit conversion rate yet allows the backing to support new guarantees to decline over time. This moral hazard story of financial crises has some appeal in explaining capital flows to emerging markets in the wake of the Mexican peso crisis of 1994 and the (then) unprecedented size of that bail out. It also may help explain why domestic-foreign interest rate spreads fail to systematically rise ahead of crises (see Kaminsky and Reinhart, 1996). However, I do have two issues as regards this story that I would like to add to Nancy’s comments on this paper. The first issue has to do with the role of uncertainty. In this framework, moral hazard arises because the guarantee offered by central bank reserves is fully credible. But, of course, in reality there may be both uncertainty and information asymmetries. Surely, as was the case in Thailand, derivative positions (the central bank had borrowed dollars in the forward market) can hide what the “true” level of reserves is. Hence, an investor may not have full information on the extent to which the liabilities being issued are fully backed or not. In a similar vein, as the recent Asian crises have shown, the extent of liabilities or implicit guarantees that are outstanding is often not known until after the crisis. In either case, an investor would face the risk, with a nonzero probability, that the central bank does not actually have enough reserves (either its own or those it can borrow from international organizations) to bail out everyone. In this case, uncertainty about who gets paid and who doesn’t would mitigate investors’ appetite for these bank deposits.

Second, it is important to remember that these grandiose bailouts are a relatively new feature of international capital markets while booms and busts in the capital flow cycle and financial crises have been around for a long time. It must be remembered that foreign investors, harking back to the previous century and Latin American railroad bonds, have lost a lot of money during financial crises. Hence, one can only take the moral hazard argument to explain the ebb and flow of cross border capital movements so far.

Turning to second generation explanations of the twin crises, as Nancy notes, there have been two recent papers that have extended the Diamond and Dybvig (1983) framework to an open economy setting—Goldfajn and Valdes (1996) and Chang and Velasco (1998). The brief remarks that I will make here apply to both of these models. As in the original paper, the crises in these models are owing to a liquidity problem on the part of banks. The banks may be faced with runs and may not be able to borrow from abroad to satisfy deposit withdrawals. The key assumption in the models that gives rise to illiquidity is that banks borrow short (from abroad) and lend long (to domestic projects). It is worth noting, however, that this illiquidity scenario presumably rules out the existence of foreign banks, which would have recourse to liquidity in times of unexpectedly large withdrawals via the parent bank abroad. Secondly, it also rules out banks holding any liquid asset, such as an internationally traded bond, that can be liquidated if the need arises. The introduction of either of these plausible considerations into the models would considerably dampen their explosive behavior.

Empirical Evidence on the Links of Currency and Banking Crises

Given that the theme of this paper has been the parallels and links between currency and banking crises, I feel I should discuss briefly what the empirical evidence tells us about the chronology of these events (Table 1). This part of my discussion is based on Kaminsky and Reinhart (1996), which examined the issue in some detail.

Table 1.The sequence of the twin crises.
Financial liberalization
Banking sector problems begin
Currency crashes
Banking sector problems peak
Protracted output collapse

From the analysis of nearly thirty banking crises, it would appear necessary to start the discussion of these crises talking a little bit about financial liberalization. Most of the banking crises we examined in that paper shared the common feature that the financial sector had been liberalized shortly before the crisis took place. It would appear that the removal of interest rate ceilings and reductions in reserve requirements that are part of the liberalization process in an environment of lax regulation and even more lax supervision is a recipe for an indiscriminate lending boom and an eventual banking crisis. In turn, as the bad bank loans pile up and the financial sector begins to depend on central bank credit and low interest rates, the seeds are sown for a policy inconsistency between the central bank’s exchange rate commitment and its endeavors to act as a lender of last resort to the banks. More often than not, this policy incompatibility ends up as a currency crisis.

However, the story does not end with the currency crisis, as the devaluation itself appears to have pernicious feedback effects on the banking sector. There are clearly balance sheet effects, among other transmission channels, that merit close scrutiny. Indeed, most often the peak of the banking crisis (if not its beginning) occurs shortly after the currency crisis. Nor does the story end there. As the economy contracts, often severely, the domestic financial sector remains mired in serious difficulty for an extended period of time.

In that analysis, we also show that when currency crises occur alongside banking crises the crises are far more severe than when the currency crisis occurs without banking sector problems (see Table 2 and Kaminsky and Reinhart, 1998a). Also, the recessions are deeper and more protracted and the crash in asset prices far greater. Indonesia’s decline in GDP of nearly 14 percent in 1998 starkly reminds us of the severity of these capital-market crises.

Table 2.The severity of the crises.
Banking CrisesBalance-of-Payment
Severity MeasureTwinSingleTwinSingle
Cost of Bailout13.35.1 *N.A.N.A.
(percent of GDP)
Loss of ReservesN.A.N.A.25.48.3 *
Real DepreciationN.A.N.A.25.726.6
Composite IndexN.A.N.A.25.617.5
Source: Kaminsky and Reinhart (1996).Notes: Loss of reserves is the percentage change in the level of reserves in the six months preceding the crises. Real depreciation is the percentage change in the real exchange rate (with respect to the dollar for the countries that peg to the dol–lar and with respect to the mark for the countries that peg to mark) in the six months following the crises. The composite index is the weighted average of the loss of reserves and real depreciation. Episodes in which the beginning of a banking crisis is followed by a balance-of-payments crisis within 48 months are classified as twin crises.* Denotes that the measure of severity of single crisis episodes is statistically different from the twin crises severity at the 5% level. An N.A. denotes not applicable.

Where Do We Go Next?

While the models that Nancy has reviewed in the paper capture several features of banking and currency crises, these models as well as a much broader family of first and second generation models of currency crises, are silent on several crucial dimensions. Hence, high on the list of topics for future research in the area of financial crises, I would stress three broad themes, that have received comparatively little scrutiny.

First, we should develop models that can explain the self reinforcing vicious circle of banking crises and currency woes. Specifically, the oft-observed pattern of banking crises leading to currency crises and the latter making the financial sector problems even worse. Pinning down the balance sheets of firms and banks and modeling the balance-sheet effects of these crises would clearly be a welcome addition to this literature.

Second, more research needs to be done on what Calvo (1998) has called the sudden stop problem, referring to the sudden stop or drastic reversal of capital inflows and its highly disruptive effects on economic activity. We need to gain a better understanding of the determinants of the output collapses we observe. This would be a departure from existing models. In both first generation and second generation models devaluations are expansionary. In an example of a first generation model, Gerlach and Smets (1995) explain “contagion” following a devaluation in one of two countries that are engaged in bilateral trade with one another by the recessionary effects in the second country of the real appreciation after its trading partner devalued. In that linear model the devaluation in the second country occurs because the decline in output leads to a decline in money demand and a loss of international reserves. In Obstfeld (1994), for instance, the policymaker’s loss function weighs the loss of credibility from devaluing from the economic loss of not doing so. In either case devaluations produce the textbook improvement in economic activity. Clearly, the aftermath of devaluations in emerging markets paint a very different picture, which we have yet fully to understand and formalize via a model. In this regard, understanding the behavior of banks following the crises is of some importance as, for a variety of reasons, bank lending dwindles and banks often hold high levels of excess reserves.

Lastly, the role of foreign banks in propagating disturbances—or, more broadly, the role of common lenders—as a vehicle of contagion is another area where models have been relatively scarce. Yet, some of the recent evidence on the channels of contagion, such as Frankel and Schmukler (1996) and Kaminsky and Reinhart (1998b), point to the importance of a variety of financial sector links.



Nancy Marion has provided us with a very informative and readable paper. Moreover, I am in agreement with Nancy’s basic conclusion that there is a fundamental conflict—embedded in both the speculative attack and banking crisis models—arising from the authorities’ attempt to maintain two prices (a fixed exchange rate and the price of deposits relative to currency) with too few instruments. However, when analyzing some of the policy changes that have occurred in the emerging markets in the period since July 1997,1 think that an equally useful characterization of the dilemma confronting the authorities would be a distinction between maintaining a fixed exchange rate and preserving the stability of the financial system. In most emerging markets, the latter objective is a much broader concept than just maintaining a fixed price between the currency and deposits. Indeed, in many emerging markets, the goal of stabilizing the financial system extends beyond stabilizing just the banking system and typically includes the stability of nonbank financial institutions and other asset markets. The importance of this broader objective can be illustrated by considering the policy actions that have been taken to find additional instruments to solve the Marion policy dilemma.

One policy option has been the Argentine solution—fix the exchange rate with a currency board; address the lender-of-last-resort problem by obtaining international loans (or lines of credit) that can be drawn on during crisis periods; and limit the ability of the banking system to issue insured liabilities and take excessive risks by imposing high risk-adjusted capital ratios, required reserve and liquidity ratios, and improved prudential supervision. This particular solution is not without cost, however, since it reduces the scale of financial intermediation. In Hong Kong, the authorities have sought to stabilize asset prices both by limiting the supply of land available for new housing and by intervention in the stock and stock index futures markets. Another solution, which has been utilized by Malaysia, is to use capital controls to delay and tax the exit of foreign investors. Finally, another alternative which has been utilized by Russia is to default on previously insured liabilities as well as to impose capital controls on outflows. What is not yet fully understood, however, is what the long-run effects of these policy measures are likely to be.

One aspect of the recent experience in Asia that the speculative attack and banking crisis models do not capture very well is what happened during and after the crises to the economies’ institutional structures. In the speculative attack models that Nancy has surveyed, for example, the government uses up its foreign exchange reserves to defend both price commitments, and then ends its commitment to maintain one of the prices (the fixed exchange rate). The authorities typically maintain a fixed conversion price between currency and deposits because they can always print domestic high powered money whereas they cannot print foreign exchange reserves. However, even in the countries that have maintained a fixed price between currency and deposits, the speculative attacks and banking crises in the period since July 1997 have been characterized by extensive institutional changes and failures, both during and after the crises, which greatly amplified the asset price adjustments that occurred. While the collapses of the banking systems of many the Asian economies have been widely recognized, these collapses were also accompanied by sharp changes in the structures of the foreign exchange and domestic money markets, as well as the equity markets. Indeed, a key characteristic of these crises has been that the institutional structure has become endogenous during the crisis. In part, this reflects the important role that banks played in both the foreign exchange and payment systems in emerging markets. Before the crisis, the typical foreign exchange market was an interbank market, with banks willing to take on intraday foreign exchange exposures in order to provide market liquidity and to help match order flows throughout the day. Even where required by regulation to limit overnight foreign exchange exposure, these intraday exposures could be quite large. This type of interbank market totally collapsed during the crisis as banks refused to take intraday open positions (because of the fear that counterparties would not deliver). As a result, many of these foreign exchange markets became broker markets, where brokers took no intraday positions but merely quoted foreign exchange prices with large bid-ask spreads and searched for a match between buyers and sellers of foreign exchange. As a result, quoted exchange rates were often not transactions prices. Moreover, determined attempts to acquire foreign exchange by any market participant could produce large daily movements in the exchange rate due to low market liquidity. Often the central bank was the only seller of foreign exchange.

In domestic money markets, there was also a high degree of segmentation. As concerns about the solvency of domestic banks increased, many foreign banks would make domestic currency loans in the local interbank market only to other foreign banks. Moreover, some of the stronger domestic banks would only deal with the local foreign banks.

The structure of equity markets also was transformed when broker dealers that acted as market makers could no longer serve that function because of their inability to obtain bank credit. The resulting decline in market liquidity contributed to a sharp fall in equity prices. If we are going to try to limit the scope and duration of financial crises, one of the things that we need to have a better understanding of is how institutional structure changes during such crises and what factors drive these changes.

    CalvoGuillermo. (1998). “Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops.”Journal of Applied Economics13554.

    ChangRoberto and AndresVelasco. (1998). “Financial Crises in Emerging Markets: A Canonical Model.”NBER Working Paper No. 6606.

    DiamondDouglas and PhillipDybvig. (1983). “Bank Runs, Liquidity, and Deposit Insurance.”Journal of Political Economy9140149.

    DooleyMichael P. ().“A Model of Crises in Emerging Markets.”NBER Working Paper No. 6300.

    FrankelJeffrey A. and Sergio LSchmukler. (1996). “Crisis Contagion and Country Funds: Effects on East Asia and Latin America.” InReuvenGlick (ed.) Managing Capital Flows and Exchange Rates: Perspectives from the Pacific Basin. Cambridge: Cambridge University Press pp. 232266.

    GerlachStefan and FrankSmets. (1995). “Contagious Speculative Attacks.”European Journal of Political Economy114563.

    GoldfajnIlan and RodrigoValdes. (1997). “Capital Flows and the Twin Crises: The Role of Liquidity.”International Monetary Fund Working Paper WP/97/87.

    KaminskyGraciela L. and Carmen M.Reinhart. (1996). “The Twin Crises: The Causes of Banking and Balanceof- Payments Problems.”International Finance Discussion Paper No. 544. Washington: Board of Governors of the Federal Reserve. Forthcoming in American Economic Review 1999.

    KaminskyGraciela and Carmen M.Reinhart. (1998a). “Financial Crises in Asia and Latin America: Then and Now.”American Economic Review88 (2) 444448.

    KaminskyGraciela and Carmen M.Reinhart. (1998b). “On Crises Contagion and Confusion.” Unpublished paper.

    ObstfeldMaurice. (1994). “The Logic of Currency Crises.”Cahiers Economiques et Monetaires43189213.

General Discussion

Peter Garber agreed with Mathieson’s characterization of the illiquidities that arise at the time of a crisis. Although models of speculative attacks recognize that exchange rates and other asset prices exhibit discontinuities, model builders have paid little if any attention to the illiquidities that appear at the moments of discontinuity. By contrast, considerable attention has been paid to these illiquidities in the IMF’s International Capital Markets reports, most notably in the April 1993 report following the currency turmoil that shook the European Monetary System during the summer of 1992. Bob Flood had drafted material on speculative attacks for that report and had emphasized that our ability to explain the dynamics of macroeconomic behavior requires an understanding of the illiquidities that appear at times of crisis.

Michael Mussa noted that models of the speculative attack process were applicable to a set of phenomena that was broader than currency crises and banking crises. The same type of process occurs in many cases of ordinary businesses that are about to go bankrupt. When a company’s bank creditors or suppliers of trade credit begin to suspect that the company is in trouble, they start to reduce their credit exposures, which results in fundamentally the same type of phenomenon as a bank run, though usually in slower motion. Mussa also observed that institutions that have the implicit guarantee of the government can operate fairly deeply under water for a considerable period of time before their creditors decide they have to pull the plug. As the scope of implicit guarantees has grown around the world, we’ve come to see very large losses associated with bankruptcies, usually because the problem banks had become deeply submerged by the time they had to be put out of their misery.

Dale Henderson called attention to a paper on banking crises by Douglas Waldo (Waldo, 1985). Waldo’s contribution, which had been written much earlier than it was published, was overshadowed by Diamond and Dybvig (1983) but, in many ways, was more carefully developed than the latter paper, with more attention to spelling out the essentials of banking institutions.

As a comment on Reinhart’s discussion, Henderson felt that in some cases it was misleading to place “financial liberalization” at the top of the chain. The savings and loan crisis in the United States provided an example in which financial institutions had essentially become moribund prior to financial liberalization and confronted perverse incentives once the liberalization came. Henderson thought that in analyzing cases in which financial liberalization had been followed by financial crisis, it was important to address the situation at the time of the liberalization and to try to sort out the extent to which the crisis had resulted from perverse incentives associated with the initial conditions.

In response to Henderson, Reinhart noted that part of the difficulties that seemed to follow financial liberalization could be attributed to the fact that in many countries, banks had little prior history of credit relationships with the private sector. Liberalization tended to induce banks to switch from lending to government to lending to private borrowers, particularly to the extent that it coincided with corrective fiscal measures that reduced the government’s need for credit. In a number of countries, banks that began to lend to the private sector in significant volume following liberalization had suffered from their inexperience.

Michael Dooley commented on another point that Reinhart had emphasized in her discussion, namely, the failure of speculative attack models to pay much attention to the output losses that followed financial crises. In Dooley’s view, this phenomenon was an endogenous and central part of the process. In particular, the prospect that large output losses would be suffered by countries that did not repay their loans, and the incentives that borrowers had to avoid such prospective losses, played a critical role in inducing lenders to extend international credits and in influencing the forms in which such credits were extended. In the absence of a clear understanding of this point, efforts to strengthen the architecture of the international monetary system could become fundamentally misguided.

In responding to the question of whether output collapses implied that IMF programs should recommend more expansionary fiscal policies, Reinhart observed that the output collapses were associated with a drying up of bank credit. Even banks that had liquidity didn’t lend, but rather became very risk averse and held excess reserves. Regardless of fiscal positions, an understanding of this phenomenon in an open-economy setting was central to understanding the output collapses.

Mathieson agreed with Reinhart but emphasized that the collapse of the credit process involved nonbank credits as well as bank credit. In emerging market economies, as well as many mature market economies, credit played a key role in the production process insofar as the provision of raw materials and intermediate goods was heavily dependent on trade credits and supplier credits. In many emerging market economies, credit flows took place in an environment that lacked an effective legal system for enforcing contracts. The basis for credit flows to finance inputs to the production process in such environments—or the basis for the expectation that lenders would be repaid—was the threat that the flow of goods would stop if the credit lines dried up. In this context, Mathieson felt that the recent crisis in emerging market countries had seen the credit process between firms affected as least as dramatically as the credit process between banks and firms. In his view, the collapse of interfirm credits had had a larger and more direct impact on output than the drying up of credits from banks to the larger corporations.

In responding to the discussants, Marion broadly agreed with the comments that had been made. She felt that more attention should be paid to the interactions between currency crises and banking crises, to pinning down the balance sheets of firms and banks, and to modeling the balance-sheet effects of crises. Existing models of speculative attacks provided useful insights and should not be rejected too quickly, but new ways of thinking about crises were also needed.

    Other Resources Citing This Publication