Ch. 14: Exchange Rate Crises: How Pegs Work and How They Break

Based on Feenstra and Taylor, Chapter 20 (KOM chapters 21 and 22). There are no lecture slides for this topic on the site yet; check against the lecture when it is given.

The previous chapter asked whether a country should peg. This chapter asks how a peg actually works in the central bank's books, and why pegs so often end in a crisis. The plan: (1) facts about exchange rate crises and their costs; (2) the central bank balance sheet and how a peg is maintained; (3) currency boards and the backing ratio; (4) what happens with money demand shocks, bank bailouts and risk premiums; (5) the first-generation crisis model (bad fundamentals, predictable collapse); (6) the second-generation model (self-fulfilling crises, multiple equilibria); (7) contagion, the IMF and policy lessons, with the cases of the ERM 1992, Asia 1997 and Argentina 2002.

Facts about exchange rate crises

An exchange rate crisis is a large and sudden depreciation that ends a fixed (or tightly managed) exchange rate. FT use a rule such as a depreciation of 15% or more after a period of stability. Key facts:

• Crises are common. Since 1970 there have been dozens of them, in rich and poor countries alike. Pegs often last only a few years.
• Crises are costly, especially in emerging markets. Output often falls sharply in the year of the crisis and the following years. Losses of several percent of GDP are normal; in severe cases (Indonesia 1998, Argentina 2002) output fell by around 10% or more.
• Crises come in bundles. Exchange rate crises often occur together with banking crises and sovereign default crises: twin crises (two at once) or triple crises (all three). Each type makes the others more likely, for example through foreign currency debt on bank balance sheets.
• Political costs. Finance ministers and central bank governors often lose their jobs after a devaluation.

Why are crises in developing countries so costly? Foreign currency debt (original sin, previous chapter) means the depreciation causes balance sheet losses; banks fail; credit dries up; capital flows reverse (a sudden stop). In advanced countries with debt in home currency, a devaluation can instead be expansionary (Britain after 1992).

How pegs work: the central bank balance sheet

To understand a peg we need the central bank's simplified balance sheet.

Because the balance sheet balances:

M = B + R

The central bank controls M only by buying or selling assets. Buying domestic bonds raises B and M; buying foreign currency raises R and M.

Fixing, interest rates and the backing ratio

Under a credible peg E = Ē with capital mobility, UIP gives i = i*. With PPP (or simply with prices given in the short run), the money market equilibrium is:

M = P × L(i*) × Y, with P = Ē × P*

All the variables on the right are given from the central bank's point of view, so money supply is determined by money demand. The central bank cannot choose M; it can only choose how M is split between B and R. The backing ratio is R/M: the share of the money supply backed by foreign reserves.

Currency boards

A currency board is a peg with a backing ratio of at least 100%: every unit of home money is backed by foreign reserves (B = 0, R = M), usually written into law. Examples: Hong Kong since 1983 (HK$7.8 per US$), Estonia and Lithuania before they joined the euro, Bulgaria, and Argentina from 1991 to 2001 (1 peso = 1 dollar). With R ≥ M, the central bank can always meet any demand to convert home money into foreign currency, so the peg cannot be broken by running out of reserves. The price: the central bank cannot lend to banks or the government, so it cannot act as a lender of last resort.

Shocks to a pegged economy (balance sheet diagrams)

FT draw the balance sheet as a diagram with reserves R on the horizontal axis and domestic credit B on the vertical axis. The money supply line M = B + R is a downward sloping 45 degree line. Points on the horizontal axis (B = 0) have a backing ratio of 100% (currency board). Reserves cannot fall below zero, so the line ends at the vertical axis: that point (R = 0) is where the peg breaks. We describe the main cases in tables with a running example: initially M = 1,000, B = 500, R = 500, backing ratio 50%.

Case 1: a fall in money demand

Suppose home output falls, so money demand falls by 10% (L(i*)Y down). With P fixed by the peg, M must fall to 900. How? Residents who hold too much home money exchange it at the central bank for foreign currency. B is unchanged; R falls by 100.

Rule: under a peg, ΔR = ΔM when B is unchanged. A fall in money demand is absorbed by reserves. The peg survives as long as the fall in money demand is smaller than R. Under a currency board (B = 0), any fall in M can always be met: the backing ratio stays at 100%.

Case 2: an increase in domestic credit

Suppose the central bank buys 100 of domestic bonds (for example to finance the government). B rises to 600. This would raise M, lower i below i* and put downward pressure on the currency. To keep the peg, investors who now hold excess money convert it into foreign currency, and the central bank sells 100 of reserves. Net result: M unchanged at 1,000, B = 600, R = 400.

Rule: under a peg, ΔR = -ΔB. Domestic credit expansion does not change the money supply; it only changes its composition (sterilization in reverse). The movement is along the M line, up and to the left. If B keeps rising, R eventually reaches zero and the peg fails. This is the core of the first-generation model.

Case 3: bank bailouts and the lender of last resort

In a banking panic depositors withdraw funds. The central bank may lend to the banks to stop the panic (lender of last resort), raising B. Under a peg, this is Case 2: reserves fall one for one. So a banking crisis puts the peg at risk, and the government faces a choice: save the banks (and risk the peg) or save the peg (and risk the banks). A currency board cannot bail out banks without breaking its own rule. Argentina in 2001 faced exactly this dilemma.

Case 4: defending the peg with interest rates when there is a risk premium

If investors fear a devaluation or default, they demand a premium: UIP becomes

i = i* + (ΔE^e/E) + risk premium

The currency premium compensates for the expected depreciation; the default risk premium compensates for possible default. A higher i lowers money demand L(i)Y, so M falls and, with B fixed, reserves fall. Expectations of a crisis therefore drain reserves even before anything happens. The central bank faces a dilemma: if it keeps B fixed and lets i rise, reserves fall and output suffers; if it tries to keep i low by raising B (sterilization), reserves fall even more. Either way pressure grows.

First-generation crisis model: inconsistent fundamentals

The first-generation model (Krugman 1979, Flood and Garber 1984) explains crises caused by bad fundamentals, usually a government that finances persistent fiscal deficits by central bank credit.

Setup

• The government runs a deficit financed by the central bank, so domestic credit B grows by a constant amount each period.
• Under the peg, M = B + R is fixed by money demand, so reserves fall by the same amount each period: ΔR = -ΔB.
• The policy is inconsistent: reserves will eventually run out, and then the country must float. After the float, money grows with B, so prices and the exchange rate rise: inflation π and depreciation each period equal the money growth rate μ.

The key step: money demand drops when the peg ends

Before the collapse, i = i*. After the collapse the currency depreciates at rate μ, so by UIP i = i* + μ. A higher interest rate means lower money demand. So at the moment of the switch from peg to float, M must fall by a discrete amount. Under the peg the only way M can fall is by a fall in reserves.

Why the attack comes early

Speculators know that once the float starts, the currency will depreciate. Anyone holding home money when the peg ends would make a capital loss. There can be no predictable jump in the exchange rate (that would be an arbitrage opportunity). So investors attack the peg at the exact moment when the remaining reserves equal the drop in money demand: they buy all remaining reserves at the peg rate, M falls by that amount, and the economy moves smoothly from peg to float with no jump in E. The shadow exchange rate (the floating rate that would prevail if the peg were abandoned at that moment) equals the peg exactly at the time of the attack. Before that point the shadow rate is below the peg (floating would mean appreciation), so nobody attacks; after it, it would be above the peg.

Lessons: the crisis is predictable and unavoidable given the policy; it comes before reserves would have run out naturally; it looks sudden but is caused by the fundamentals; the only cure is to fix the fundamentals (stop monetizing the deficit). The model fits many Latin American crises of the 1970s and 1980s (for example Mexico in 1976 and 1982).

Second-generation crisis model: self-fulfilling crises

Some crises hit countries whose fundamentals did not look unsustainable: Britain and Italy in 1992 were not monetizing deficits. The second-generation model (Obstfeld 1994) treats the government as making a choice: it keeps the peg as long as the benefits exceed the costs, and the costs depend on what markets expect.

Benefits and costs

• Benefits of keeping the peg: credibility, low inflation, integration gains, avoiding the balance sheet losses of depreciation, political reputation.
• Costs: keeping i = i* + expected depreciation. If the market expects a devaluation, i must be high to defend the peg, causing output losses (unemployment, higher mortgage costs, bank losses).

Multiple equilibria

Consider two possible market beliefs: no devaluation expected (premium zero, low cost of pegging) or devaluation expected (high premium, high cost of pegging). FT show three regions depending on the strength of the fundamentals (for example the state of the economy, the level of unemployment):

In the intermediate region, the outcome depends on beliefs. If markets believe the peg will hold, interest rates stay low and the government keeps the peg: beliefs are confirmed. If markets believe it will fail, interest rates jump, the cost exceeds the benefit and the government devalues: beliefs are again confirmed. A shift in sentiment (a "sunspot", a speech, a large speculator) can move the economy from the good to the bad equilibrium. The peg is ripe for attack without being doomed.

Case: the ERM crisis of 1992

After German reunification in 1990 the Bundesbank raised interest rates to fight inflation. Countries in the ERM, pegged to the mark, had to match the high rates. Britain and Italy were in recession with rising unemployment: their fundamentals were intermediate. In September 1992 markets (George Soros's fund among them) bet heavily against the pound. On 16 September the Bank of England raised its rate from 10% to 12% and announced 15%, which made the cost of the defense very visible; that evening Britain left the ERM. Italy left too. France, with stronger credibility and German support, survived. In 1993 the ERM bands were widened to plus or minus 15%. After leaving, Britain cut rates and recovered: the devaluation was expansionary, consistent with debt mostly in pounds.

Lesson: the second-generation view explains why countries with reasonable fundamentals can suffer crises, and why the high interest rates needed to defend a peg can themselves trigger the decision to quit.

Contagion

Contagion is the spread of a crisis from one country to others. Channels:

• Trade links: a devaluation in one country makes its competitors' exports less competitive and hurts their partners' exports, worsening their fundamentals.
• Financial links: investors who lose money in one country sell assets elsewhere to meet margin calls or to reduce risk; common creditors (such as Japanese banks in Asia) withdraw from the whole region.
• Wake-up call: a crisis in one country makes investors look more closely at others with similar weaknesses.
• Self-fulfilling shift in beliefs: in the second-generation model, a crisis next door can move a vulnerable country to the bad equilibrium.

Case: the Asian crisis of 1997

In July 1997 Thailand floated the baht after a speculative attack. Within months Indonesia, Malaysia, the Philippines and South Korea suffered large depreciations. These countries had sound budgets and fast growth, but weak banks, large short-term foreign currency debt and pegs to the dollar. The depreciations caused large balance sheet losses and deep recessions (Indonesia's output fell by about 13% in 1998). The crisis spread further to Russia (1998) and Brazil (1999). Many observers called it a "third-generation" crisis, combining banking problems, foreign currency debt and self-fulfilling expectations.

The IMF and the international lender of last resort

A country under attack needs reserves. The International Monetary Fund lends reserves to countries in crisis, giving them time to adjust. Loans come with conditionality: the borrower must adopt policies such as fiscal tightening, monetary tightening, bank restructuring or structural reforms.

• For: an international lender of last resort can stop a self-fulfilling run (second-generation crises), just as a domestic lender of last resort stops bank runs.
• Against: moral hazard, because governments and creditors take more risk if they expect a bailout; conditionality may be too harsh (critics argued the IMF's tight policies deepened the Asian recessions); the IMF's resources are limited relative to private capital flows; and it cannot help much with first-generation crises, where the fundamental problem must be fixed anyway.

Since the Asian crisis many emerging markets have built up large foreign reserves as self-insurance, partly to avoid needing the IMF.

Case: Argentina 2001 to 2002

Argentina adopted a currency board (the Convertibility Plan) in 1991 at 1 peso = 1 US dollar, to end hyperinflation. It worked: inflation fell to near zero in a few years. Problems built up after 1998:

• A recession from 1998, made worse when Brazil, a major trading partner, devalued in 1999 and the dollar was strong, so the peso became overvalued.
• Persistent fiscal deficits (including by the provinces) financed by foreign borrowing, with public debt rising; much private and public debt was in dollars.
• Rising risk premiums made defending the board increasingly costly, lowering money demand and draining deposits.

In December 2001 the government froze bank deposits (the "corralito"), riots followed, the government defaulted on its debt, and in January 2002 the currency board was abandoned. The peso fell to more than 3 pesos per dollar within months. Output fell by more than 10% in 2002. Argentina is a triple crisis: currency, banking and sovereign default. It shows that even a currency board is not a guarantee: a hard peg is only as hard as the political will to accept its costs. It mixes first-generation elements (fiscal problems) with second-generation elements (rising costs of defending the peg).

Preventing crises: policy options

• Sound fundamentals: no monetization of deficits, strong bank supervision, limits on foreign currency debt.
• Floating: a floating rate cannot be attacked in the same way; but fear of floating and foreign currency debt limit this option.
• Harder pegs: currency boards or full dollarization raise credibility, but remove the lender of last resort and monetary policy.
• Capital controls: limit speculative flows (Malaysia 1998, the Chilean tax on inflows), at the cost of less financial integration.
• Reserves and international lending: large reserves or IMF credit lines raise the backing ratio and deter self-fulfilling attacks.