Predicting the Crises of the Greek Drachma

Transcription

1 Review of Development Economics, 6(3), , 2002 Predicting the Crises of the Greek Drachma Costas Karfakis* Abstract The paper investigates the role of fundamentals in the speculative episodes experienced by the Greek drachma during the 1990s, and examines whether the openness of the Greek economy has altered the role of fundamentals in the likelihood of a currency crisis. An interesting aspect of the empirical analysis is that the crises are related significantly to macroeconomic fundamentals of the Greek economy. The analysis shows that the openness of the Greek economy has introduced new elements and concerns in the predictability of speculative attacks on the drachma. 1. Introduction Currency crises in Europe, South America, and East Asia have rekindled the interest of academic scholars and policymakers in the causes and symptoms of such episodes. The theoretical literature distinguishes between crises caused by a deterioration in economic fundamentals and those that result from self-fulfilling speculative attacks caused by fads and the herding behavior of investors fleeing regional markets in fear of financial losses. The seminal work of Krugman (1979) stresses how inconsistencies between domestic economic conditions and an exchange rate commitment can lead to a currency collapse. In this model the excess liquidity stems from the financing requirements of the government, so fiscal deficits and credit expansion could be used as leading indicators of an imminent crisis. The importance of large and growing current account deficits have been stressed as a key determinant of a crisis in Mexico (Dornbusch and Werner, 1994; Dornbusch et al., 1995), while a large appreciation of the real exchange rate, a weak banking system, and a low level of foreign reserves have been identified as leading indicators of a crisis in emerging markets (Sachs et al., 1996). 1 In the early 1990s, the drachma experienced two strong speculative pressures, one associated with the inability of the coalition government to stabilize the economy and pursue a coherent disinflation program, and the other related to the exchange rate turmoil of the ERM. In anticipation of the abolition of all remaining foreign exchange controls on short-term capital movements in July 1994, the drachma came under strong speculative pressure at the beginning of May. In 1995, the authorities announced the use of a hard drachma exchange rate policy as the economy s nominal anchor. Although this continued to pull down domestic inflation by lowering the cost of imported inputs, cost-push pressures worked in an opposite direction. The resulting real exchange appreciation adversely affected the current account deficit, which increased to 4.2% of GDP in 1997; this caused severe pressure on the drachma in May and June Also, the turbulence in the Asian markets, following the collapse of the *Karfakis: University of Macedonia, Thessaloniki , Greece. Tel: ; Fax: ; ckarf@uom.gr. I would like to thank an anonymous referee for useful comments and suggestions that have substantially improved the paper. I have also benefited from the comments of the participants at the International Conference on Globalization: Trade, Financial and Political Economy Aspects (Delphi, May 2000)., 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA

2 330 Costas Karfakis Thai baht in July 1997, pushed many investors to close their positions in the currencies of the emerging markets, including drachma. In the light of these developments, the purpose of this study is twofold. Firstly, it investigates the role of macroeconomic fundamentals in the speculative episodes experienced by the drachma in the 1990s. Secondly, it examines whether the openness of the Greek economy has altered the role of fundamentals in the likelihood of a currency crisis. 2. Theoretical Issues In this section, I present a modified version of the model developed by Sachs et al. (1996), to frame the ensuing analysis. This model is suitable because the variables, which are suggested as alternative signals to watch for, played an important role in the macroeconomic developments of the Greek economy during the 1990s. The model considers a government that is managing a pegged exchange rate, with nominal exchange rate (e) and real exchange rate (e/p), where p is the price level ratio, which is taken as predetermined in the short run and is set equal to 1. The government pegs e as long as foreign reserves (FR) are sufficient to finance a net capital outflow (K), implying there is no devaluation (D = 0). If, however, K > FR, a devaluation occurs. Then the government establishes a new nominal exchange rate (e*) in order to achieve a target real exchange rate. The size of devaluation is denoted as D = (e*/e) - 1. In this model, the target exchange rate reflects the health of the banking sector, which is measured in terms of whether or not the economy has experienced a lending boom (LO). When the banking sector is weak and the lending boom is directed to high-risk areas, such as credit cards and consumer and real-estate loans, it is likely to produce a banking portfolio that is extremely vulnerable to the vagaries of the business cycle. In that case, the government will not use a high interest-rate policy to defend the currency, since the recessionary effects are likely to generate widespread bankruptcies among banks. Therefore, she will choose a real exchange rate more depreciated than its long-run level (e), thus increasing the probability of a currency crisis. If the banking sector is sound and the bank lending is mainly associated with an investment-driven demand, the government will set e* at e, thus reducing the probability of speculative pressure. In addition, e* depends on a number of structural variables, such as the terms of trade (TT) and the capital flows (CM), among others. A deterioration in the terms of trade, which increases the current account deficit, will force the government to choose a real exchange rate more depreciated than its long-run level, as a way of improving the external imbalance. An increase in net capital inflows will force the government to choose a real exchange rate more appreciated than its long-run level, as a way of sterilizing their impact on domestic inflation. A sudden capital inflow reversal will trigger a devaluation of the domestic currency. Thus, the target real exchange rate may be written as e* = ef ( LO, TT, CM), f1 > 0, h< 0, f2 < 0, f3 > 0 (1) and the potential course of the exchange rate can be summarized as D= ( e e) f( LO, TT, CM)- 1 if K> FR. (2) The above expression says that a devaluation occurs when there is a capital outflow in excess of foreign reserves. The size of the devaluation is greater when either the exchange rate is initially appreciated relative to its long-run value, so that e/e is high;

3 CRISES OF THE GREEK DRACHMA 331 or there has been a deterioration in the terms of trade, a net capital inflow, and an increase in bank lending directed to high-risk areas, so that f(lo, TT, CM) is large. 3. Data and Methodology The empirical analysis is carried out using monthly data from January 1990 to March 1998 for the following variables: nominal (e) and real (RER) effective exchange rates of the drachma, foreign exchange reserves (FR), overnight interest rate (i), broad money supply (M4), total bank loans to the private sector (DC), net capital movements (NCM), and terms of trade (TT), defined as the ratio of export prices to the prices of final foreign goods. We measure the extent of speculative pressures on the foreign exchange (FX) market with an index (denoted IND), which is defined as the percentage change in the nominal effective exchange rate plus the percentage change in the foreign reserves minus the percentage change in the overnight interest rate. 2 Since the three components of the index have different volatilities, the weights that we apply to each series are given by the relative precision (the inverse of the variance) of each series over the sum of precisions during the sample period. As the index decreases with a depreciation of the currency, a loss of international reserves and a rise in the interest rate, a fall in the index reflects strong selling pressures on domestic currency. In the empirical analysis, a currency crisis is identified as an episode where the value of the index exceeded a specified threshold. This threshold was set equal to the mean of the index minus 1.5 times the standard deviation of the calculated index, so that 5% of the monthly index values would exceed that threshold if the values were distributed normally. Relative to that threshold, the major speculative episodes identified over the two sample periods are reported in Table 1. The break point of the sample is set at April 1994, which coincides with the last period of maintaining capital controls on short-term capital movements. 3 It is worth noting that if the attacks are anticipated, the size of the jump in the components of IND is reduced at the attack time. Therefore, selecting only extreme values of IND as a measure of crisis may reduce the share of predictable crises in the sample. On the other hand, if the crises are not predictable, then IND is more likely Table 1. Identifying Crises of the Drachma Period: 1990(1) 1994(4) Period: 1994(5) 1998(12) M/Y IND De DFR Di M/Y IND De DFR Di 3/ / / / / / / / / / / / / / / Notes: The sample size for the variables e, FR, and i extends to December The entries are the values of IND, De, DFR, Di which exceed the calculated thresholds -75.2, , , for the first period, and , , , for the second period. M/Y stands for month/year.

4 332 Costas Karfakis to reach extreme values and is less likely to be significantly correlated to fundamentals (Kaminsky et al., 1998). The analysis of the behavior of IND indicates that the extreme values of the index do not select the EMS crisis in September 1992 and the Asian crisis in July However, the cumulative value of IND from July to October 1997 was equal to -34%, indicating a persistent pressure on the drachma throughout the period. Episodes of significant speculative pressures in the FX market were identified by focusing on the behavior of each of the three components of the index. I refer to those months in which changes in the exchange rate were at least 1.5 standard deviations below the sample mean as episodes of currency crises, and these are selected first. From the remaining observations, episodes where changes in the foreign reserves were 1.5 standard deviations below the mean are selected next. Finally, the list of speculative episodes was completed by selecting periods where changes in the interest rate were 1.5 standard deviations above the mean. All the remaining observations are identified as periods of tranquility. 4 The results, which are also reported in Table 1, show that all major crises are now selected. Having identified the crisis episodes of the drachma, we can set up and estimate the following probit-type econometric model, which tries to model the probability p of speculative attacks on the drachma as a function of macroeconomic fundamentals discussed in section 2: p = a1( DRARt-j* DUM1)+ a2( DRARt-j* DUM2) + a3( DDCt-j* DUM1)+ a4( DDCt-j* DUM2) + a5( DNCMt-j* DUM1)+ a6( DNCMt-j* DUM2) + a7( DRERt-j* DUM1)+ a8( DRERt-j* DUM2) + a9 ( DTTt -j * DUM1)+ a10( D TTt-j * DUM2), (3) where RAR = M4/FR denotes the reserve adequacy ratio, 5 and = 1, 90(1) 94(4) DUM1 { = 0, elsewhere = 1, 94(5) 98(3) DUM2 { = 0, elsewhere. In this model, the FX market decision whether or not to attack the drachma is treated as a discrete variable, which takes on only two values: unity, when there is a period of speculative pressure, and zero when there is a period of tranquility. To prevent the continuation of a speculative episode from being identified as a new episode, a one-month window was created by dropping one observation around previously identified episodes. The two dummies, DUM1 and DUM2, account for the potential effects of the openness of the Greek economy on the likelihood of a currency crisis. The maximum-likelihood results obtained by using the Newton Raphson iterative algorithm are reported in Table 2. 6 In the first period, only the credit expansion affects significantly the probability of a currency crisis, with the right sign. Its coefficient is negative, suggesting that a rising credit growth is associated with an investment-driven credit demand, which reduces the probability of speculative pressure. 7 Also, the terms of trade significantly affected the probability of a crisis, but its sign was not consistent with theory. It is worth mentioning that the net capital inflows were not significantly affecting the likelihood of a currency crisis, reflecting the effectiveness of capital controls, which were in force until The insignificance of the real exchange rate may

5 CRISES OF THE GREEK DRACHMA 333 Table 2. Probit Maximum-Likelihood Results Dependent variable: p Regressors Coefficient Standard error t-ratio DRAR t - 1 *DUM DRAR t - 1 *DUM *** DDC t - 1 *DUM ** DDC t - 1 *DUM *** DNCM t - 1 *DUM DNCM t - 1 *DUM *** DRER t - 4 *DUM DRER t - 4 *DUM ** DTT t - 4 *DUM ** DTT t - 4 *DUM ** Factor for calculation of marginal effects = 0.21 Goodness of fit = 0.80 Pesaran Timmermann statistic = *** Notes: *** and ** indicate significance at 1% and 5%, respectively. indicate that the exchange rate policy pursued before 1994 did not lead to a significant real overvaluation of the drachma. The results of the second period are more interesting and support the theory discussed in section 2. A rise in the growth rate of money supply relative to the growth rate of foreign exchange reserves is associated with a rising probability of a currency crisis.the impact of this result is augmented by the fact that a rise in net capital inflows increases the probability of a currency crisis. One implication is that the probability of a currency collapse increases in a time of massive capital inflow reversal, unless the domestic monetary base is fully backed by foreign reserves. A real appreciation of the drachma increases the probability of a currency crisis, implying that the hard drachma policy pursued during the second half of the 1990s could have increased the probability of a currency collapse at a time of massive capital inflow reversal. The sign of the terms of trade turns out to be negative, suggesting that a deterioration in the terms of trade increases the probability of a speculative attack. Also, the analysis shows that a rising credit growth is associated with an investment-driven credit demand, which reduces the probability of speculative pressure. The goodness-of-fit statistic is equal to 0.80, which implies that a significant proportion of the observations are correctly predicted by the model. Figure 1 plots the actual and fitted values of the regression. The Pesaran Timmermann (1992, PT hereafter) test statistic measures the extent to which the fitted values of p and the actual values are correlated. Under the null hypothesis that the actual and fitted values are independently distributed, the PT statistic is asymptotically distributed as standard normal variate. The PT test statistic is equal to , which is highly significant. Thus, there is real evidence that the likelihood of a currency crisis is predictable using macroeconomic fundamentals. To estimate the marginal effect of a unit change in one variable, computed at sample means, on the probability of a currency crisis, we must multiply the factor 0.21 by the coefficient of that variable. Specifically, the marginal effect of the RAR variable on the probability of a currency crisis is given by 0.21 times 0.18, which is 0.04.

6 334 Costas Karfakis Figure 1. Actual and Fitted Probabilities I have also split the sample period into two subperiods, with the break point set at April 1994, and have estimated the model (3) for each period separately. The maximum-likelihood results presented in Table 3 are consistent with those obtained from the whole period. When the growth rate of the real exchange rate was replaced by the real overvaluation of the drachma, computed from the residuals of a cointegrating equation of the real effective exchange rate on the terms of trade, the results were similar. This suggests that not only the changes in the real exchange rate, but also the deviations of the drachma from its long-run level, signal a FX market pressure. The above results suggest that the openness of the Greek economy has introduced new elements in the predictability of crises of the drachma. The implication of these results is that any policy aimed at increasing the rate of growth of money supply in line with the rate of growth of foreign reserves, and reducing the real overvaluation of the drachma and the growth rate of net capital inflows, should reduce the likelihood of a currency crisis. 4. Concluding Remarks This study has investigated the role of fundamentals in the speculative episodes experienced by the Greek drachma during the 1990s, and has examined whether the openness of the Greek economy altered the role of fundamentals in the likelihood of a currency crisis. An interesting aspect of the empirical analysis is that the crises of the drachma are related significantly to a number of macroeconomic fundamentals of the Greek economy, such as the reserve adequacy ratio, domestic credit, the real exchange

7 CRISES OF THE GREEK DRACHMA 335 Table 3. Probit Maximum-Likelihood Results Dependent variable: p Period: 1990(1) 1994(4) Period: 1994(5) 1998(3) Regressor Coeff. SE t-ratio Coeff. SE t-ratio DRAR t *** DDC t ** *** DNCM t *** DRER t ** DTT t ** ** Factor G-fit PT statistic *** *** Notes: Factor refers to the number which is used to calculate marginal effects. G-fit stands for the goodness of fit. PT statistic is the Pesaran Timmermann test statistic. See also notes to Table 2. rate, the terms of trade, and net capital inflows. Also, the analysis shows that the openness of the Greek economy has introduced new elements and concerns in the predictability of speculative attacks on the drachma. References Calvo, Guillermo A., Variates of Capital-Market Crises, manuscript, University of Maryland (1995). Dornbusch, Rudiger and Alejandro Werner, Mexico, Stabilization, Reform, and No Growth, Brookings Papers on Economic Activity 1 (1994): Dornbusch, Rudiger, Ilan Goldfajn, and Rodrigo O. Valdes, Currency Crises And Collapses, Brookings Papers on Economic Activity 2 (1995): Eichengreen, Barry, Andrew K. Rose, and Charles Wyplosz, Exchange Market Mayhem: The Antecedents and Aftermath of Speculative Attacks, Economic Policy 21 (1995): Flood, Robert and Nancy Marion, Perspectives on the Recent Currency Crisis Literature, International Journal of Finance and Economics 4 (1999):1 26. Funke, Norbet, Vulnerability of Fixed Exchange Rate Regimes: The Role of Economic Fundamentals, OECD Economic Studies 26 (1996): Garber, Peter M. and Lars E. O. Svensson, The Operation and Collapse of Exchange Rate Regimes, in G. Grossman and Kenneth Rogoff (eds.), Handbook of International Economics, vol. III, North Holland (1995). Kaminsky, Graciela, Saul Lizondo, and Carmen M. Reinhart, Leading Indicators of Currency Crises, IMF Staff Papers 45 (1998):1 48. Krugman, Paul, A Model of Balance-of-Payments Crises, Journal of Money, Credit, and Banking 11 (1979): Moreno, Ramon, Macroeconomic Behaviour During Periods of Speculative Pressure or Realignment: Evidence from Pacific Basin Economies, Federal Reserve Bank of San Francisco Economic Review 3 (1995):3 16. Pesaran, M. Hashem and Allan Timmermann, A Simple Nonparametric Test of Predictive Performance, Journal of Business and Economic Statistics 10 (1992): Pesaran, M. Hashem and Bahram Pesaran, Working with Microfit 4.0, Oxford, UK: Oxford University Press (1997).

8 336 Costas Karfakis Sachs, Jeffrey D., Aaron Tornell, and Andres Velasco, Financial Crises in Emerging Markets: The Lessons from 1995, Brookings Papers on Economic Activity 1 (1996): Notes 1. For three excellent surveys on currency crises, see Garber and Svensson (1995), Kaminsky et al. (1998), and Flood and Marion (1999). 2. This speculative index has been proposed by Eichengreen et al. (1995). 3. The Greek authorities fully liberalized short-term capital flows on 14 May Moreno (1995) follows the same approach to identify speculative episodes in the Pacific Basin economies. 5. In a time of capital inflow reversal, the central bank must be prepared to cover all its liquid liabilities with reserves, unless the fixed exchange rate adjustment is abandoned. Thus, the appropriate yardstick with which to evaluate the abundance of reserves is a broad measure of liquidity compared with the stock of foreign exchange reserves, since at a time of currency crisis the larger the stock of privately held domestic liquid assets, the larger the contingent demand for foreign assets (Calvo, 1995). 6. See Pesaran and Pesaran (1997). 7. A similar result was obtained by Funke (1996), for the member states of the EU which participated in the exchange rate mechanism (ERM).