SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

Sharp exchange rate depreciations, or currency crashes, are associated with poor economic outcomes in industrial countries only when they are caused by inflationary macroeconomic policies. Moreover, the poor outcomes are attributable to inflationary policies in general and not the currency crashes in particular. On the other hand, crashes caused by rising unemployment or external deficits have always been followed by solid economic growth, rising asset prices and stable or falling inflation rates.


I. Introduction and Summary

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

Many economic commentators appear to believe that currency crashes invariably have harmful effects (see e.g. Volcker 2005; The Financial Times 2008; The Wall Street Journal 2008). This paper shows that, for industrial countries, the evidence does not support this commonly held view.

Using quarterly data for 20 industrial countries since 1970, this paper identifies 19 episodes of sharp exchange rate depreciations or currency crashes. The episodes are categorized according to the causal factors that appear to have been important in each case. Three broad groupings of causal factors are identified: inflationary macroeconomic policies, large current account deficits or capital outflows and rising unemployment rates. Often, more than one of these factor groupings appeared to be present.

Crashes were followed by poor macroeconomic outcomes (slow GDP growth, rising bond yields and falling equity prices) only when they resulted from inflationary macroeconomic policies. Moreover, currency crashes did not, by themselves, appear to contribute to these poor outcomes. Indeed, the evidence suggests that currency crashes tended to counteract the slowing of GDP growth that often occurs during and after inflationary periods. Ceteris paribus, currency crashes do contribute to inflationary pressure, but there are numerous episodes of crashes in which inflation remained constant or even declined in the quarters immediately after the crash, reflecting the effects of other economic forces.

Before 1986, crashes were almost always caused by inflationary macroeconomic policies. With floating exchange rates, inflationary policies tend to cause an immediate depreciation of the currency. With fixed exchange rates, a currency devaluation may be an integral part of the inflationary policies, or it may occur somewhat later, after strong domestic demand and rising prices cause the current account balance to decline and force the authorities to run down foreign exchange reserves to defend the currency. A common theme among most of the inflationary episodes is the apparent underestimation of the natural rate of unemployment by policy-makers. In the wake of these inflationary crashes, the behaviour of GDP growth depended critically on monetary and fiscal policies. When authorities decided to fight inflation, recession typically ensued. When authorities did not fight inflation, the economy generally continued to grow.1 In most of the episodes driven by inflation, bond yields increased and real equity prices declined, regardless of the stance of macroeconomic policy.

Since 1985, crashes appear to have been caused by negative shocks to economic activity and/or financial market concerns about current account sustainability. Crashes in this period have been less frequent than in the years before 1986 and they have always been followed by periods of strong growth with little or no acceleration of inflation, stable or declining bond yields and rising real equity prices.

II. Previous Studies

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

An extensive literature seeks to explain financial crises or to identify early warning indicators of such crises, particularly in emerging markets. This literature encompasses banking crises, sovereign debt crises and currency crises, where ‘currency crisis’ may be defined to include periods of sharp depreciation as well as periods in which a central bank successfully defends a currency peg from a speculative attack. Good reviews of the early warning literature on currency crises are Edison (2003) and Berg et al. (2004). Three studies that focus on sharp depreciations, or ‘currency crashes’, in emerging markets are Frankel and Rose (1996), Milesi-Ferretti and Razin (1998) and Kumar et al. (2003). These studies find that depletion of foreign exchange reserves is a robust predictor of currency crashes; other, less robust, predictors include rapid domestic credit growth, an overvalued exchange rate, weak GDP growth and high industrial-country interest rates. Somewhat surprisingly, these studies do not find a consistent correlation between the level of foreign debt or the current account balance and subsequent currency crashes. Dooley and Frankel (2003) examine the consequences of currency crises in emerging markets. The papers in their volume highlight the importance of large foreign-currency debts and weak macroeconomic policy institutions in explaining the poor economic outcomes in these emerging-market episodes.

In an interesting study that bridges the gap between emerging markets and industrial countries, Osband and van Rijckeghem (2000) search for ranges of relevant macro and financial variables that have been associated historically with extremely low probabilities of a currency crisis in the following year. They find that high foreign exchange reserves, low foreign debt and a higher (more positive) current account balance imply a very low probability of a currency crisis. After estimating these relationships on developing-country data, they apply them to industrial countries and show that they hold up well. Tudela (2004) applies indicator analysis to industrial countries and finds that import growth, fiscal deficits, non-FDI capital inflows and an appreciated real exchange rate are all positively related to the probability of currency crisis in the following year. Wright and Gagnon (2006) find that current account deficits, high inflation rates and weak GDP growth rates have predictive power for currency crashes in industrial countries.

A previous paper of mine (Gagnon 2009b) is the only extant study of the effects of currency crashes in industrial countries. Using a different data set and crash definition than I do in this paper, I found that currency crashes were followed by rising bond yields only when inflation was high and/or rising before the crash, and that no currency crash since 1985 was followed (or preceded) by rising bond yields. Crashes since 1985 also tended to be associated with rising real equity values, but the paper did not explore the effects on economic growth. Croke et al. (2006), Algieri and Bracke (2007), Debelle and Galati (2007) and Freund and Warnock (2007) examine reversals of major current account deficits in industrial countries. There is some overlap between episodes of currency crashes and episodes of current account adjustment, with crashes tending to occur before adjustments. But many currency crashes are not followed by current account adjustments and many current account adjustments are not preceded by currency crashes. Three of these studies (Debelle and Galati is the exception) find that episodes with the largest currency depreciations are generally those with the most benign outcomes for GDP growth.

III. Data Sources

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

This paper uses quarterly average effective (that is trade-weighted) nominal and real exchange rates available from the Bank for International Settlements (BIS) based on a group of 26 relatively advanced economies. The effective exchange rates are available from 1964, but the analysis starts in 1970 because most of the other economic series are available only from 1970. The 20 countries studied are Australia (AL), Belgium (BE), Canada (CA), Denmark (DK), Finland (FI), France (FR), Germany (GE), Greece (GC), Ireland (IR), Italy (IT), Japan (JA), the Netherlands (NL), New Zealand (NZ), Norway (NO), Portugal (PT), Spain (SP), Sweden (SD), Switzerland (SZ), the United Kingdom (UK) and the United States (US).2

Most of the other data used in this paper are from the Organisation for Economic Co-operation and Development (OECD) Economic Outlook database. The primary exceptions are equity prices (from the OECD Main Economic Indicators database) and foreign exchange reserves and dollar exchange rates (from the International Monetary Fund (IMF) International Financial Statistics database). Where possible, gaps in the data were filled in using the IMF International Financial Statistics database.3 Government balances for all countries are at an annual frequency.4 Inflation rates are based on GDP deflators.

IV. Identification of Crashes

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

A necessary condition for a crash in this paper is an exchange rate depreciation that exceeds 15% over four quarters.5 In some cases, the exchange rate continued to decline at a rate in excess of 15% for several quarters, but I group such closely connected periods of depreciation as single episodes.

  • To qualify as a separate episode, the four-quarter exchange rate depreciation must exceed 15% and it must not have exceeded 15% in any of the previous four quarters. Applying this criterion yields 27 episodes.

In some of these 27 episodes, the sharp depreciation reflected a quick reversal of a previous sharp appreciation. In some other cases, the sharp depreciation was itself reversed soon afterwards. In order to focus on episodes that represent a sustained change in the economic environment, I apply an additional criterion to filter out transitory exchange rate fluctuations.

  • The exchange rate must have depreciated at least 15% over the 12-quarter period centred on the four quarters used to identify the start of a crash.
  • This criterion filters out the following episodes: IT1995, JA1979, JA1990, JA1996, NZ1986, NZ1998, SD1977 and UK1986.

I believe that excluding most of these episodes is reasonable. In particular, contemporaneous economic reporting generally did not view these episodes as periods of currency crisis. Two potential exceptions are NZ1998 and SD1977, which were viewed as periods of currency crisis by some contemporaneous observers. Including these in our analysis would not have affected the results of this paper; the Swedish experience of 1977 fits in well with the inflationary crashes and the New Zealand experience in 1998 accords well with the crashes driven by rising unemployment.

A third criterion that some previous studies of developing countries have imposed (Frankel and Rose) is that the depreciation is not simply a gradual acceleration of a pre-existing downtrend. That criterion does not appear to be relevant for industrial countries. In all cases, the four-quarter exchange rate depreciation at the start of a crash was at least 7.5 percentage points greater than that over the previous four quarters.

Altogether, then, this paper examines 19 episodes of sharp depreciations, or currency crashes: AL1985, FI1992, GC1980, GC1983, GC1985, IT1973, IT1976, IT1993, NZ1975, NZ1984, PT1977, PT1982, SD1983, SD1993, SP1977, SP1983, SP1993, UK1976 and US1986.6 Most of these episodes were viewed by some contemporaneous observers as periods of currency crisis. Figure 1 plots the four-quarter change in the nominal effective exchange rate for each country in our data set, with the start of each crash episode marked by a vertical line.7 I define the start quarter as the quarter in which the four-quarter depreciation first reaches 15%.

image

Figure 1.  Nominal effective exchange rates (four-quarter percent change)

Download figure to PowerPoint

Following are a few observations on Figure 1. First, in Italy in the 1970s and Greece in the 1980s, there are currency crashes that followed each other within two or three years. These are identified as separate crashes by the above criteria, but none of the empirical results of this paper would be affected if they were treated as single episodes. Second, there is some tendency for crashes to occur in more than one country nearly at the same time. For example, Italy, New Zealand, Portugal, Spain and the United Kingdom experienced crashes in 1975–77. This pattern appears to reflect more than simple financial market contagion. In each case, there are common elements in the underlying fundamentals of the countries experiencing near-simultaneous crashes, and these are discussed below. Another cluster of crashes occurred in 1992–93 in Finland, Italy, Spain and Sweden, as these countries devalued or dropped out of the European Exchange Rate Mechanism (ERM). Portugal and the United Kingdom also devalued in 1993, but their effective depreciations were not large enough to meet the above criteria. The empirical results would not be affected if the latter two episodes were treated as crashes. Other countries that faced speculative attacks in 1992 and 1993, such as Belgium, Denmark and France, suffered only temporary depreciations and never came close to meeting the crash criteria.

Finally, I note that applying the analogous criteria to sharp exchange rate appreciations would identify only nine episodes of sharp and sustained appreciations: AL2003, GE1973, JA1977, JA1986, JA1993, JA1999, SZ1975, SZ1978 and UK1997. The relative rarity of sharp appreciations implies at least a modest degree of asymmetry between appreciations and depreciations.

V. Causes of Crash Episodes

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

The variables identified by the empirical literature as related to currency crashes can be grouped into three broad categories relating to underlying factors.

  • High current or expected inflation. Excessively expansionary macroeconomic policies.
  • Large current account deficit. Decreasing stock of foreign exchange reserves. Large or growing foreign debt.
  • Rising unemployment. Weak aggregate demand. Slow GDP growth.

Moreover, there are interconnections between these factors. In particular, expansionary macroeconomic policies can boost demand for imports and thus lower the current account balance. Also, a decline in foreign demand for a country's exports can create a current account deficit and rising unemployment. As discussed below, the channels for these interconnections and the dynamics of crashes are crucially dependent on whether a country has a fixed or a floating exchange rate.

A. High Inflation

Inflationary policies ultimately put upward pressure on all prices, including the price of foreign currency. A rise in the price of foreign currency is a depreciation of the exchange rate. Whether the exchange rate depreciation occurs early or late in the inflationary process depends critically on a myriad of factors, such as wage and price controls, trade barriers, controls on international capital flows and the exchange rate regime. Perhaps the most important factor is the exchange rate regime.

  • In a fixed exchange rate regime, inflationary macroeconomic policies push up the prices of goods and services and thus the real exchange rate, leading to a decline in the current account balance. Investors may anticipate that a depreciation ultimately will be necessary, leading to outflows of private capital. The downturns in both the current and the capital accounts deplete the central bank's foreign exchange reserves, ultimately forcing it to devalue the currency.
  • In some cases, governments devalue the currency as a deliberate action to stimulate the economy, so that inflation occurs after the crash rather than before.
  • Under floating exchange rates, inflationary macroeconomic policies often depreciate the currency before broad price indices begin to accelerate, reflecting the greater flexibility of asset markets relative to goods markets.

B. Pressures on External Accounts

As noted above, when the exchange rate is not freely floating, a widening current account deficit and diminishing foreign exchange reserves can be part of the process by which inflationary policies lead to currency crashes. However, other sources of change in the demand and supply of a country's imports and exports can also contribute to a widening current account deficit and diminishing reserves. These include, for example, trade policies, global commodity prices, business cycles in key trading partners and investor perceptions of the investment climate. A currency crash occurs when investors suddenly decide to pull their capital out or to stop financing the current account imbalance.

C. Rising Unemployment

A third factor contributing to currency crashes is a negative shock to aggregate demand that slows economic activity and increases the rate of unemployment. In the face of a contractionary shock, central banks tend to ease monetary policy. In the case of fixed exchange rate regimes, the ease takes the form of a devaluation. In the case of floating exchange rate regimes, the ease takes the form of lower interest rates, which in turn push down the exchange value of the currency. If the source of the shock is reduced demand for a country's exports, the current account deficit may also contribute to the currency crash.

D. Historical Experience

Table 1 displays the measures of these three factors around the onsets of currency crashes. The first column shows the difference between inflation in the country experiencing a currency crash and the average over all 20 countries, which is termed ‘excess inflation’. Excess inflation is based on the average annual percent changes from four quarters before the start of the crash to four quarters after the start of the crash.8 The value of 5.4 in the first row implies that the price level in Italy increased by nearly 11 percentage points more than the average across all industrial countries over the eight quarters surrounding the onset of the 1973 crash. The bottom rows show that the mean of excess inflation in crash episodes was 4.4% and the mean across all countries and time periods was 0 with a standard deviation of 3.3 percentage points. Numbers in bold are more than one standard deviation above or below the sample mean. In about half of the crash episodes, excess inflation was more than one standard deviation above the sample mean and it was never more than one standard deviation below the mean. This finding corroborates the view that currency crashes are often caused by inflationary policies. Despite the small negative value of excess inflation in Australia in 1985, this episode is categorized as an inflationary one because the IMF attributed the crash in part to market concerns about monetary policy and because excess inflation increased markedly over the following two years.9

Table 1. Evidence on Causes of Currency Crashes
EpisodeExcess inflationCurrent accountChange in unemploymentFactors presenta
  • Note: Excess inflation equals the annual inflation rate over the eight quarters centred on the start of the crash minus the average inflation rate in all 20 countries over the same period. Current account is in percent of GDP for the quarter before the start of the crash. Change in unemployment rate is measured over eight quarters ending in the quarter before the start of the crash. Bold entries differ from sample means by more than one standard deviation. Sample means and standard deviations are based on the entire sample period and all 20 countries.

  • a

    I, inflation; C, current account; U, unemployment. Column lists the factors that exceeded the mean by more than one standard deviation in the hypothesized direction. Factors in lower-case were either close to the one standard deviation criterion or were supported by considerations discussed in the text.

  • b

    b For Spain in 1983, the current account is the average over the four quarters of 1982 because there were swings of more than 20 percentage points from quarter to quarter that year.

Italy 1973Q35.4−2.01.5Icu
New Zealand 1975Q43.37.30.2IC 
Italy 1976Q26.2−1.10.6I  
United Kingdom 1976Q23.20.51.7i u
Portugal 1977Q111.87.94.6ICU
Spain 1977Q310.3−3.61.5Icu
Greece 1980Q27.6−3.50.9Ic 
Portugal 1982Q311.812.6−0.5IC 
Spain 1983Q1b3.3−2.53.5I U
Greece 1983Q111.8−3.13.0IcU
Sweden 1983Q10.0−4.91.4 cu
New Zealand 1984Q33.87.61.2ICu
Australia 1985Q2−1.0−4.2−1.0ic 
Greece 1985Q411.57.70.0IC 
United States 1986Q3−2.8−3.2−0.3 c 
Finland 1992Q4−2.8−4.18.9 cU
Italy 1993Q10.5−2.10.1  u
Sweden 1993Q1−0.6−1.95.1  U
Spain 1993Q31.1−1.24.9  U
Crash mean4.4−4.22.0   
Sample mean0.0−0.30.2   
Sample standard deviation3.34.71.7   

The second column of Table 1 lists the current account balance in the quarter before the start of a crash as a percent of GDP. All but one entry is negative, and five are more than one standard deviation below the mean of −0.3%. The table does not plot foreign exchange reserves because the relationship between reserves and currency crashes holds only in the cases of fixed exchange rates.10

The third column of Table 1 shows the change in the unemployment rate over the eight quarters ending in the quarter before the onset of the crash. Six of the crash episodes were associated with a rise in unemployment that exceeded the mean by more than one standard deviation – indeed, four episodes exceeded the mean by more than two standard deviations. Although Italian unemployment had changed little on balance over the eight quarters before the 1993 crash, the episode is associated with rising unemployment because unemployment turned up two quarters before the crash and continued to increase for some time afterwards; also, GDP growth became negative shortly before the crash.

The final three columns of Table 1 summarize the causal factors behind each currency crash. The letter ‘I’ stands for excess inflation, ‘C’ stands for the current account deficit and ‘U’ stands for the change in the unemployment rate. Capital letters are used when the associated factor exceeded its sample mean by more than one standard deviation. Lower-case letters are used when the factor was close to the one standard deviation threshold or when other evidence supports a role for the causal factor.

At least one factor is identified for every crash episode, and all but five episodes are associated with factors that exceeded the one standard deviation threshold.11 Before the mid-1980s, inflation was always an important causal factor. Rising unemployment seems to have become more important later in the sample.

VI. Consequences of Crash Episodes

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

Table 2 presents evidence on the consequences of currency crashes. The first three columns repeat the causal factors identified in Table 1 for convenience. The next column shows the change in the inflation rate from the eight-quarter period ending in the quarter before the crash to the eight-quarter period beginning with the start of the crash. Roughly half of the changes were positive and half were negative. As in Table 1, entries in bold differ from the sample mean by more than one standard deviation. The largest changes in inflation, both up and down, occurred in the episodes that were driven by inflationary pressures. When inflationary crashes were followed by tighter macroeconomic policies, inflation often declined after the crash. Conversely, when crashes were not accompanied by tighter policies – and sometimes crashes were themselves a deliberate step to loosen policy – inflation typically increased after the crash.

Table 2. Evidence on Consequences of Currency Crashes
EpisodeCausal factorsChange in inflationChange in net exportsGDP growthChange in GDP growthChange in bond yieldChange in real equity price
  1. Note: Change in inflation is the difference between the annual inflation rate over the eight quarters after the start of the crash and the annual inflation rate over the eight quarters ending the quarter before the start of the crash. Change in net exports is the difference between net exports as a percent of real GDP seven quarters after the start of the crash and net exports in the quarter before the start of the crash. GDP growth is the annual rate for the eight quarters after the start of the crash. Change in GDP growth is the difference between the annual rate of GDP growth in the eight quarters after the start of the crash and GDP growth in the eight quarters ending in the quarter before the crash. Change in bond yield and change in real equity price are measured from the quarter before the start of the crash to the third quarter after the start of the crash. The real equity price is calculated using the GDP deflator. Sample means and standard deviations are based on the entire sample period and all 20 countries.

Italy 1973Q3Icu7.52.20.93.52.1−18
New Zealand 1975Q4IC10.4−0.11.38.32.12
Italy 1976Q2I−0.13.13.52.74.842
United Kingdom 1976Q2iu8.8−0.13.11.50.8−9
Portugal 1977Q1IC U3.81.63.80.65.8.
Spain 1977Q3Icu−0.91.00.7−2.41.946
Greece 1980Q2Ic6.2−1.62.65.9.34
Portugal 1982Q3IC5.16.21.33.23.6.
Spain 1983Q1IU−2.92.02.01.31.78
Greece 1983Q1IcU−2.20.30.31.2.52
Sweden 1983Q1cu0.82.03.72.1−1.051
New Zealand 1984Q3IC u7.42.14.60.38.7−4
Australia 1985Q2ic1.02.12.73.00.426
Greece 1985Q4IC4.01.01.54.72.431
United States 1986Q3c0.61.13.60.00.715
Finland 1992Q4cU0.44.02.07.64.774
Italy 1993Q1u−1.53.32.31.34.826
Sweden 1993Q1U−1.62.83.96.13.154
Spain 1993Q3U−1.41.02.32.61.521
Crash mean 1.01.81.7−0.21.26
Sample mean −0.40.12.6−0.3−0.13
Sample standard deviation 3.02.12.02.61.424

The column labelled ‘Change in net exports’ displays the change in net exports as a percent of GDP from the quarter before the crash to seven quarters after the crash started (eight quarters in total). This change was almost always positive; only for Greece in 1980 was it notably negative. Inflation in Greece around the 1980 crash was so high that the real exchange rate was reduced only briefly and quickly resumed its upward trend; thus, the crash provided no impetus to net exports.

The next column shows the average growth rate of GDP over the eight quarters following the onset of the crash. The average growth rate after the onset of crashes was 1.7%, compared with an average of 2.6% across all countries and time periods. Growth was more than one standard deviation below the sample average in many episodes and only for New Zealand in 1984 did growth come close to exceeding the sample average by more than one standard deviation. All of the low growth episodes were associated with high values of excess inflation in Table 1.

The column labelled ‘Change in GDP growth’ shows the change between the growth rate over the eight quarters ending in the quarter before the crash and the growth rate over the eight quarters after the crash. In this column, there are episodes that differed from the sample average by more than one standard deviation on both the upside and the downside. Decreases in the growth rate only occurred in episodes that had excess inflation as a causal factor. The largest increases in the growth rate occurred in the 1990s episodes that had unemployment, and not inflation, as causal factors.

The next column shows the change in the long-term government bond yield from the quarter before the crash to three quarters after the onset of the crash.12 Bond yields increased sharply in a number of episodes that were associated with high inflation. Bond yields declined sharply in a number of episodes that were associated with rising unemployment and no excess inflation.

The final column shows the change in a broad equity index deflated by the GDP deflator from the quarter before the crash to three quarters after the onset of the crash. Falling equity prices tend to be associated with inflationary currency crashes and rising equity prices tend to be associated with non-inflationary currency crashes.

A. Relating Causes to Consequences

Table 3 presents the results of regressions that relate the consequences of currency crashes to the causal factors behind crashes.13 In each column, one of the variables of Table 2 is regressed on the variables of Table 1.14 For GDP growth and the change in GDP growth, the value of GDP growth over the eight quarters ending in the quarter before the onset of the crash is added to the regression to control for cyclical dynamics that may not be related to crashes. The columns labelled ‘full’ include all independent variables. The columns labelled ‘restricted’ show the results of a sequential test-down procedure that successively eliminates the non-significant independent variables (except the constant).

Table 3. Relating Causes to Consequences
 GDP growthChange in GDP growthChange in bond yieldChange in real equity price
FullRestrictedFullRestrictedFullRestrictedFullRestricted
  • Note: This table presents OLS regressions of consequences, as defined in Table 2, on causes, as defined in Table 1. To control for potential dynamics, GDP growth over the eight quarters ending with the start of the crash are initially included in the regressions of GDP growth and the change in GDP growth after the onset of the crash. Restricted regressions reflect the outcome of sequentially testing down and eliminating variables that are not significant at the 10% level; for the change in real equity price, all variables are significant at the 10% level, and so the restricted regression applies a 5% criterion. In the equity price regression, the current account is not significant at the 5% level when the unemployment rate is dropped. Values in parentheses represent standard errors.

  • *, ** and ***

    Significance at the 10%, 5% and 1% levels, respectively.

Constant1.91*2.69***0.18−1.12−0.30−0.54−3.426.4***
(1.04)(0.58)(1.50)(1.03)(1.42)(0.88)(12.7)(8.1)
Lagged GDP growth0.30 −0.23     
(0.21) (0.30)     
Excess inflation−0.22**−0.22**−0.16−0.24*0.36*0.45***−5.6***−5.7***
(0.10)(0.09)(0.15)(0.13)(0.17)(0.14)(1.2)(1.4)
Current account0.08 0.22 −0.16 −5.9** 
(0.16) (0.23) (0.26) (2.5) 
Change in unemployment0.32 0.83**1.03***−0.29 4.5* 
(0.23) (0.33)(0.26)(0.29) (2.3) 
R20.400.270.640.600.460.400.700.53
Number of observations1919191917171717

For GDP growth after the onset of a currency crash, excess inflation during the crash is a significant explanatory factor. Each percentage point of excess inflation lowers GDP growth roughly one quarter of a percentage point. For the change in GDP growth after the onset of a currency crash, both excess inflation and the lagged change in unemployment are significant predictors. Excess inflation has an effect similar to that observed in the previous regressions, but now the change in unemployment is an even more important factor. Each percentage point rise in the unemployment rate before a crash implies a percentage point increase in the GDP growth rate after a crash begins. This finding likely results from policy easing in response to rising unemployment, which appears to be successful in stimulating growth.

For bond yields, the only causal factor that matters is excess inflation. Higher excess inflation is associated with rising bond yields. For real equity prices, all factors appear to matter. The effect of excess inflation is highly significant and robust to the other variables; higher excess inflation is associated with falling real equity prices. Current account deficits and rising unemployment are associated with rising real equity prices, but these effects are less significant and less robust to inclusion of the other variables than the inflation effect.

An important factor excluded in these regressions is the stance of macroeconomic policy. Quarterly data on fiscal deficits are not available for most of these countries in the 1970s and 1980s. Moreover, calculating cyclically adjusted deficits and the effective monetary stance are fraught with pitfalls, given the uncertainty about potential output, the effects of price, wage and capital controls, expectations about future inflation and future fiscal policies and differences across countries in the relative importance of interest-rate versus exchange-rate channels for monetary policy. The real short-term interest rate (based on trailing four-quarter inflation) is not significant when added to any of the regressions in Table 3. However, the analysis of case studies (below) supports the view that the subsequent stance of macroeconomic policy is critical in explaining the consequences of inflationary currency crashes for GDP growth. When governments shifted to fighting inflation, GDP growth typically suffered in the short run. When governments did not fight inflation, GDP growth was generally maintained over the near term. For non-inflationary currency crashes that were associated with rising unemployment, governments always eased macroeconomic policies and GDP growth always rebounded.

B. Inflation, GDP Growth and Crashes

The preceding subsections show that currency crashes are followed by weak or declining GDP growth only when they appeared to be caused by inflationary policies. A related question is whether crashes themselves retard GDP growth or whether other aspects of inflationary periods tend to damp growth. To answer this question, consider the growth outcomes of the most inflationary episodes across all countries and years. There are 25 episodes in which four-quarter inflation in one of the 20 countries exceeded the average across all countries by more than two standard deviations (7.2 percentage points).15 The average GDP growth rate over the eight quarters after this inflationary threshold was breached was 2.2%, slightly lower than the sample average growth rate of 2.6%. Seven of these 25 episodes were associated with a currency crash.16 In the crash episodes, GDP growth averaged 2.9%. In the non-crash inflationary episodes, GDP growth averaged 2.0%. The difference between the crash and the non-crash growth means is not statistically significant, but it does suggest that currency crashes may help to ameliorate some of the harmful after-effects of inflationary episodes.

VII. Case Studies

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

The discussion paper version of this paper (Gagnon 2009a) includes an examination of key macroeconomic events surrounding each currency crash to bolster the conclusions reported above and to better understand the individual circumstances associated with currency crashes. The primary sources are IMF documents written shortly after each crash episode. The following is a brief summary of the case study evidence. Figures 2–11 display key macroeconomic variables in each country that experienced at least one crash.

image

Figure 2.  Australia

Download figure to PowerPoint

image

Figure 3.  Finland

Download figure to PowerPoint

image

Figure 4.  Greece

Download figure to PowerPoint

image

Figure 6.  New Zealand

Download figure to PowerPoint

image

Figure 7.  Portugal

Download figure to PowerPoint

image

Figure 9.  Sweden

Download figure to PowerPoint

image

Figure 10.  United Kingdom

Download figure to PowerPoint

image

Figure 11.  United States

Download figure to PowerPoint

A common theme among many of the inflationary crash episodes is that governments underestimated the natural rate of unemployment (sometimes called the non-accelerating inflation rate of unemployment, or NAIRU). The rise of the NAIRU in the 1970s and early 1980s is discussed in Blanchard (2006, p. 2), who describes it as a ‘tough learning experience, both for economists and for policy makers’. Attempts by policy-makers to stimulate economic activity and reduce unemployment below the NAIRU spilled over into high and rising rates of inflation, and appear to have been a major causal factor behind the IT1973, NZ1975, IT1976, UK1976, PT1977, SP1977, GC1980, PT1982 and GC1983 currency crashes.17

Declining terms of trade, especially due to rising prices of energy imports, appear to have contributed to the NZ1975 and GC1980 crashes. The ERM crashes (FI1992, IT1993, SD1993, SP1993) came after central banks had maintained restrictive monetary stances for several quarters to defend currency pegs despite rising unemployment. In several crashes (SP1983, NZ1984, GC1985), governments deliberately engineered a sharp depreciation to increase exports, but offset the resulting macroeconomic stimulus through other measures such as tight fiscal policy. Of course, there are idiosyncratic factors in almost all of these crash episodes. Finland, in 1992, was suffering from the collapse of export demand from the former Soviet Union; Finland, in 1992, and Sweden, in 1993, were also suffering from the hangover of a banking crisis; and the United States, in 1986, had an unsustainable current account deficit that was driven by what was probably a bubble in the exchange value of the dollar.

Policy responses to crashes (or to the inflation, unemployment or current account deficit that caused them) vary dramatically. Tightening of monetary and/or fiscal policy typically slowed GDP growth, whereas growth often held up when policy did not tighten. Nevertheless, episodes of sustained high inflation always required a serious policy tightening and a period of economic slack at some point before the economy could return to sustainable non-inflationary growth.

VIII. Conclusions

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References

Currency crashes in industrial countries have always been associated with at least one of the following causal factors:

  • Inflationary macroeconomic policies that place upward pressure on all prices, including the price of foreign currency.
  • Weak aggregate demand and rising unemployment that encourage policy-makers to stimulate growth through expansionary monetary policy, including devaluation in the case of a fixed exchange rate.
  • Large capital outflows or current account deficits that run into financing difficulties. In some cases, these deficits may reflect either of the above forces, but they may also reflect exogenous shifts in the terms of trade or in financial market sentiment.

The consequences of currency crashes depend critically on the causes. Poor outcomes have occurred only after inflationary currency crashes. The responses of macroeconomic policy-makers after inflationary currency crashes have had important implications for GDP growth. Tighter policies to fight inflation generally reduced GDP in the short run. When authorities did not fight inflation, GDP growth generally held up in the near term. Bond yields usually increased and real equity prices usually declined during and immediately after inflationary currency crashes. Non-inflationary currency crashes uniformly led to good outcomes: GDP growth was average to above average, bond yields declined and real equity prices increased.

Overall, the lessons for policy-makers are clear. A currency crash may be a warning that the stance of policy is too loose and inflationary. In such a case, policy-makers should correct the source of the problem – the stance of macroeconomic policy – and not solely the symptoms in the foreign exchange market. But not all currency crashes reflect inflationary policies; when inflation is low and stable before and immediately after a crash, policy-makers need not fear harmful outcomes. Indeed, a sharp currency depreciation may be a helpful part of the economy's transition to sustainable growth with low inflation.

Footnotes
  1. 1This paper does not shed light on whether inflation, by itself, is harmful to growth. Indeed, there appear to be episodes in which growth was lacklustre after high inflation even in the absence of tight macroeconomic policies. However, the differences in growth outcomes are clearly accentuated by differences in policy stance.

  2. 2Five of the 26 countries with BIS exchange rate data (Hong Kong, Mexico, Singapore, South Korea and Taiwan) are not classified as industrial countries by the IMF. Most of these countries are now considered to be ‘advanced’, but they were still viewed as developing countries in the 1970s and 1980s. Austria is also among the countries in the BIS data set, but quarterly national accounts data for Austria are not available over most of our sample period. Austria did not experience a currency crash under the definition used here.

  3. 3Greek equity prices from 1979 to 1984 were obtained from the Bank of Greece Reports for the years 1982–85. Greek unemployment rates are available only at an annual frequency. The Greek current account for 1998 was obtained from Eurostat. Greek current account data were not available on a seasonally adjusted basis, and were adjusted by regressing on quarterly dummies. Spanish current account data for 1982 contained extremely large quarterly fluctuations that appear to be spurious. They were replaced by a flat value equal to the average for the year.

  4. 4The balances for New Zealand from 1971 to 1985 refer to fiscal years (ending 31 March) and were obtained from the IMF Recent Economic Developments reports for New Zealand dated January 1978 and June 1986. The balances for Portugal from 1974 to 1977 were obtained from the IMF Recent Economic Developments report for Portugal dated July 1984.

  5. 5The depreciation is measured as the difference in the logarithm of the exchange rate.

  6. 6Only six of these episodes would also meet a stricter standard of at least a 20% depreciation. The episodes that meet the stricter standard are AL1985, GC1983, GC1985, IT1976, IT1993 and SP1977.

  7. 7The dashed vertical line for the United States represents the dollar crisis of 1978, which did not meet these criteria, but was widely viewed as a period of intense weakness of the US dollar.

  8. 8Inflation before the crash is included to capture inflationary pressure in fixed exchange rate regimes. Inflation after the crash is included to capture inflationary pressure in floating exchange rate regimes or cases in which a devaluation of the fixed exchange rate was one of the primary inflationary policy actions.

  9. 9See the IMF Recent Economic Developments report for Australia in February 1986, p. 2.

  10. 10Under a fixed exchange rate, reserve depletion often precedes a currency crash for both high inflation and high unemployment episodes. Under a floating exchange rate, reserves generally have little relationship with currency crashes.

  11. 11For UK1976, both excess inflation and the change in unemployment were very close to the one standard deviation threshold. For SD1983, the current account was very close to one standard deviation below the sample mean. For AL1985, the current account deficit was close to the threshold and, as mentioned above, IMF reports indicate that the crash was spurred by financial market concerns over future inflation. For US1986, the current account deficit was unprecedented both in terms of US experience and relative to the volume of global trade. For IT1993, as mentioned above, unemployment started to increase rapidly less than a year before the crash and continued to increase for several quarters afterwards.

  12. 12A shorter window is used for the financial variables because they adjust to the crash faster than the other variables adjust.

  13. 13These regressions use data only from the 19 crash episodes. When the regressions are run on all countries and time periods, the coefficients have the same signs (except for the current account in the equity price regression) and are always much smaller than those in Table 3. This result suggests that the qualitative nature of the effects presented here holds in general, but that these effects are much stronger during crashes.

  14. 14The change in inflation is not included because it is strongly affected by the exchange rate regime and the policy environment surrounding the crash. The change in net exports is not included because it is almost always positive and because GDP is a broader gauge of the effects on economic activity.

  15. 15Episodes are deemed to last until inflation declines below the two standard deviation threshold for four consecutive quarters.

  16. 16An inflationary episode is associated with a currency crash if a crash started in any of the four quarters before or the eight quarters after the inflation criterion was met. These episodes were NZ1975, IT1976, PT1977, SP1977, GC1980, PT1982 and NZ1984.

  17. 17The only measure of the stance of monetary policy that is available for all of these countries and time periods is the real short-term interest rate, which is displayed in Figures 2–11. Measures of money and credit have severe breaks in most countries and are missing for some of the episodes. To the extent it is available, credit growth does appear to have been rapid before and during most of the inflationary crash episodes but not before and during the non-inflationary episodes.

References

  1. Top of page
  2. Abstract
  3. I. Introduction and Summary
  4. II. Previous Studies
  5. III. Data Sources
  6. IV. Identification of Crashes
  7. V. Causes of Crash Episodes
  8. VI. Consequences of Crash Episodes
  9. VII. Case Studies
  10. VIII. Conclusions
  11. References
  • Algieri, B., and T. Bracke (2007), ‘Patterns of Current Account Adjustment: Insights from Past Experience’, European Central Bank Working Paper No. 762.
  • Berg, A., E. Borensztein and C. Pattillo (2004), ‘Assessing Early Warning Systems: How Have They Worked in Practice?’, IMF Working Paper No. 04/52, International Monetary Fund.
  • Blanchard, O. (2006), ‘European Unemployment: The Evolution of Facts and Ideas’, Economic Policy, 21, 559.
    Direct Link:
  • Croke, H., S. Kamin and S. Leduc (2006), ‘An Assessment of the Disorderly Adjustment Hypothesis for Industrial Economies’, International Finance, 9, 3761.
  • Debelle, G., and G. Galati (2007), ‘Current Account Adjustment and Capital Flows’, Review of International Economics, 15, 9891013.
  • Dooley, M., and J. Frankel (2003), Managing Currency Crises in Emerging Markets. Chicago: University of Chicago Press.
  • Edison, H. (2003), ‘Do Indicators of Financial Crises Work? An Evaluation of an Early Warning System’, International Journal of Finance and Economics, 8, 1153.
  • Frankel, J., and A. Rose (1996), ‘Currency Crashes in Emerging Markets: An Empirical Treatment’, Journal of International Economics, 41, 35166.
  • Freund, C., and F. Warnock (2007), ‘Current Account Deficits in Industrial Countries: The Bigger They Are, the Harder They Fall?’, in R. Clarida (ed.) G7 Current Account Imbalances. Chicago: University of Chicago Press, pp. 13367.
  • Gagnon, J. (2009a), ‘Currency Crashes in Industrial Countries: What Determines Good and Bad Outcomes?’, International Finance Discussion Paper No. 966. Available at http://www.federalreserve.gov/pubs/ifdp/2009/default.htm
  • Gagnon, J. (2009b), ‘Currency Crashes and Bond Yields in Industrial Countries’, Journal of International Money and Finance, 28, 16181.
  • Kumar, M., U. Moorthy and W. Perraudin (2003), ‘Predicting Emerging Market Currency Crashes’, Journal of Empirical Finance, 10, 42754.
  • Milesi-Ferretti, G. M., and A. Razin (1998), ‘Current Account Reversals and Currency Crises: Empirical Regularities’, National Bureau of Economic Research Working Paper No. 6620.
  • Osband, K., and C. Van Rijckeghem (2000), ‘Safety from Currency Crashes’, IMF Staff Papers, 47, 23858.
  • The Financial Times (2008), ‘The Dollar Danger Is Not Over Yet’, Editorial on 8 May.
  • The Wall Street Journal (2008), ‘The Buck Stops Where?’, 17 March, p. A16.
  • Tudela, M. (2004), ‘Explaining Currency Crises: A Duration Model Approach’, Journal of International Money and Finance, 23, 799816.
  • Volcker, P. (2005), ‘An Economy on Thin Ice’, The Washington Post, 10 April, p. B7.
  • Wright, J., and J. Gagnon (2006), ‘Predicting Sharp Depreciations in Industrial Country Exchange Rates’, International Finance Discussion Paper No. 881, Board of Governors of the Federal Reserve System.