the stage where the countries were in the business cycle. For this purpose we ... view that crisis are to some extent generated by the recessionary impact of real.
DEVALUATION AND INFLATION AFTER CURRENCY CRISES*
Eduardo Borensztein International Monetary Fund
José De Gregorio Universidad de Chile
This version: February 1999
Abstract This paper studies the effect of large devaluations on inflation. It analyzes a sample of 41 episodes of currency crisis. The data show that about 30% of the devaluation is offset by higher inflation after three months, and the offset climbs to about 60% after two years, with a significant real depreciation present for longer periods. The main determinants of the extent of inflationary pass-through of the devaluations and inflation are the position of output relative to trend, the extent of initial overvaluation, and mainly the initial rate of inflation. Our results explain well the evolution of inflation after currency crisis, except for the EMS crisis of 1992 that has a very low pass-through. We also analyze the inflationary impact of devaluations in the Asian crisis.
____________________________________________ * Part of this paper was prepared while De Gregorio was visiting the Research Department of the IMF. We thank comments by Rodrigo Valdes, and participants at seminars in the Central Bank of Chile, Universidad de Chile, Universidad di Tella, and LACEA meetings. We are also very grateful for the superb research assistanship of Manzoor Gill and Cristobal Huneeus.
I.
INTRODUCTION
The most challenging question when exiting an exchange rate peg is whether the devaluation will be an effective means of changing relative prices (the real exchange rate) in the economy (see, for example, Edwards, 1989). A similar question also applies in the context of more flexible exchange rate regimes prevailing in the post-Bretton Woods era. In particular, in the aftermath of most currency crises there is the concern that the inflationary impact of the large devaluation may greatly complicate the management of the crisis. In this paper we investigate the inflationary effects of large, sudden devaluations. 1 Do changes in the exchange rate of this nature have a large immediate impact on domestic prices? How does the inflationary impact evolve over time? How does the effect depend on the initial conditions of the economy? We investigate these questions by analyzing a large sample of currency crises since 1970. These include some of the most conspicuous cases of large devaluations in recent experience, including the Mexican crisis of 1994, the last European currency realignments, the CFA franc devaluation, and other episodes in Asia and Latin America.
1
In this paper we use the term devaluation to signify either a change in the parity in a pegged exchange rate regime (a devaluation) or a depreciation of the exchange rate in the context of a floating exchange rate regime, perhaps after the abandonment of a peg (a depreciation). In any event, it is often difficult to distinguish exchange rate regimes with precision. In many currency crisis episodes, for example, countries were following a formally floating system but the authorities were de facto pegging the exchange rate.
1
These questions have important policy implications. Large and sudden devaluations usually occur when an exchange rate peg is abandoned, often in the context of strong market pressure or a balance of payments crisis. While in these situations there are always clear signs of overvaluation of the domestic currency, determining the magnitude of the necessary nominal depreciation (or of a desired range for the exchange rate to align policies in a floating exchange rate regime) poses difficult questions. It could be argued that the depreciation should just offset the estimated overvaluation, bringing the exchange rate back in line. However, some degree of pass-through to domestic prices is unavoidable, and thus it is necessary to overshoot the required devaluation initially. The consequences of this decision are often weighty. On the one hand, a large overshooting in the initial nominal depreciation may affect large segments of the population by increasing the prices of basic imported (or tradeable) commodities, create short-run financial strains for banks and corporations with foreign indebtedness, and generally undermine investor confidence. On the other hand, if the devaluation achieves a durable real exchange rate adjustment, the current account will show a substantial improvement, confidence may be regained, and the crisis may be shorter. The speed with which domestic prices rise after a sudden devaluation (and the degree to which they do) thus becomes the critical factor. This issue has become quite relevant with the sequence of emerging market crises in the past few years. In the Asian crises of 1997-98, for example, huge swings in exchange rates were observed after countries were forced to let their exchange rates float. When exchange rates doubled or even quadrupled (in terms of the price of the U.S. dollar) the possibility of triggering an inflation spiral rose strong concerns. Although in some cases, notably Korea and Thailand, the exchange rates rebounded and settled at much more reasonable levels, the apparently small extent and slow speed of the inflation passthrough was a critical factor to permit the return of stability to the foreign exchange markets. The same kind of concerns applies to Brazil in 1999. It may be argued that devaluations of the sort studied in this paper should not have a significant inflationary effect when they are basically correcting a misaligned exchange rate rather than reflecting a sustained monetary imbalance. Simply restoring equilibrium to relative prices while maintaining a broadly appropriate monetary policy
2
stance should not result in generalized inflationary pressures. An opposite argument would be that, with reasonably flexible prices, a policy-induced change in a nominal variable should not have lasting effects on relative prices, and thus devaluations would only produce an increase in the price level in the fairly short run, for example as predicted by purchasing power parity. From this perspective, the view that nominal devaluations have significantly lasting effects on the real exchange rate relies on the existence of sticky prices and slow adjustment. A quick look at the evidence suggests that economies go through prolonged periods of changes of the real exchange rate after changes in the nominal exchange rate, implying that frictions are relevant and that devaluations have real effects. However, it is also true that, in many instances, the inflationary effect of devaluations is quite significant. While changes in the nominal exchange rate help to restore equilibrium in relative prices faster, they also have an impact on inflation, and nominal devaluations have different degrees of effectiveness from country to country and episode to episode. One strand of the empirical literature has focussed on the pass-through from exchange rates to prices in connection with the sharp swings of the U.S. dollar during the 1980s.2 The purpose of those studies was to understand why not all of the appreciation or depreciation of the U.S. dollar was passed on to prices. The explanations were mainly based on imperfectly competitive market structures. Different pass-through coefficients were explained mainly on the basis of market structure in the United States and trading partners. In contrast, this paper focuses on the behavior of the price level, rather than particular markets or sectors. Our work is thus closer to Edwards (1989) who analyzes 39 devaluation experiences between 1970 and 1982, and studies the extent to which whether they were finally undone by inflation. An extension of that work, including more experiences and examining the policy responses, is in Edwards and Santaella (1993). Similarly, Goldfajn and Valdés (1998) examine a similar issue by looking at experiences of “large appreciations” and then uncovering whether the appreciations were corrected by falls in the rate of inflation or by devaluations. Work by Kamin (1988) and Kiguel and 2
See, for example, Dornbusch (1987) and the references in Obstfeld and Rogoff (1996) pp. 711-712.
3
Ghei (1993) also attempt to examine whether devaluations were effective in changing the real exchange rate, and trace the evolution of a number of macroeconomic variables after the devaluation. These papers, however, have looked mainly at one-time devaluations in pegged exchange rate regimes that set the exchange rate at a new parity. In contrast, we consider all large currency devaluations, regardless of the prevailing exchange rate regime. We study the determinants of the degree of passthrough, and of the inflation rate at various horizons in the period following a large devaluation. We analyze econometrically whether the initial conditions of the economy, such as the inflation rate, the position in the business cycle, the misalignment of the real exchange rate, and current account deficit, affect the degree of pass-trough from devaluation to inflation. We apply the estimation to a sample of 49 episodes of large devaluations, which we reduce to 41 to avoid the distortions created by hyperinflation episodes. The results show that the most important determinants of the degree of passthrough are the initial level of inflation, with a smaller role for the other variables that are, nevertheless, statistically significant, such as the degree of misalignment and deviations of output with respect to trend. In addition, there are no great differences between the results on the determinants of pass-trough at a three-month horizon and a two-year horizon. When we apply our results to out of sample events, we find that the Asian crises of 1997-98 displayed a smaller than predicted passthrough.
II.
DATA AND M ETHODOLOGY
We construct a sample of currency crises that ended with a large depreciation of the domestic currency in the 1970-1996 period. We adopt the definition of crisis of Berg and Pattillo (1998), with one modification. 3 We define a crisis as a depreciation of the domestic currency in a given month that exceeds the sample mean depreciation by at least three standard deviations. Berg and Pattillo define a crisis with reference to an index that is a weighted average of the exchange rate depreciation and the percentage loss in international reserves. For our purposes, only the exchange rate depreciation is relevant,
4
so we adopted a crisis definition that depends exclusively on the exchange rate. The mean and standard deviations of the exchange rate depreciation are computed separately for each country and also separately for periods of high inflation, which is defined as an annual inflation rate higher than 150 percent. The full sample comprised 49 episodes in 26 countries (see Table A.1 in the Appendix). We define the pass-trough index at a k-month horizon as the ratio between accumulated inflation (πk) divided by accumulated depreciation (dk): (1)
PTI k =
πk dk
If the PTI coefficient equals 1, it means that within a period of length k the devaluation was fully offset by an increase in inflation. The closest to zero the index is, the largest the impact of the change in the nominal exchange rate on the real exchange rate. Table 1 presents some summary statistics for devaluation rates, inflation rates and the passthrough index for the episodes included in our sample. For the full sample, the median episode involved a devaluation of 51 percent, with a passthrough of 0.27 over three months and of 0.69 over two years. A few episodes involving extremely high inflation and depreciation rates present some difficulties, as they tend to distort the average values. In these cases, the initial devaluation was the beginning of a high period of instability that led to hyperinflation. In order to focus on less extreme cases we restricted the sample to countries that had a cumulative inflation of less than 100 percent after three months and less than 1000 percent after 2 years. Applying this criterion, eight episodes were excluded: three in Argentina, one in Bolivia, two in Brazil, and one in each Chile and Peru. In this restricted sample, which was used for the econometric analysis, the average depreciation is 54 percent, and the average PTI is 0.25 at three months and about 0.6 at two years. Figures 1 and 2 show the evolution of the annual devaluation and inflation rates, for the full and the restricted sample. In the full sample (Figure 2), both the rates of inflation and of devaluation are very high, and even after two years the average inflation rate rises again to levels above 300%. This pattern is largely determined by countries that were going through a hyperinflationary or near-hyperinflationary process. In those cases, 3
We thank Andy Berg and Catherine Pattillo for providing us with the data. 5
distinguishing between effects running from the depreciation of the exchange rate into prices or, conversely, from price increases into the foreign exchange market becomes nearly impossible. For this reason we concentrate for the rest of the paper on the restricted sample. As can be seen in Figure 1, the average devaluation was largely a onetime event, as can be seen from the fact that the 12-month rate of devaluation decreases sharply after one year. By contrast, inflation shows an more persistent upward trend after the event but, as we will see below, there is great variance across different episodes in this respect. We explore the differences in the pattern of passthrough and inflation in Figure 3, where we consider four groups of countries (a total of 19 events) that show similar profiles. The first panel shows the evolution during the last EMS crisis, comprising the devaluations that took place in September 1992 in Finland, Italy, Spain, Sweden and the United Kingdom. The second panel groups four episodes in Asia in the 1980s: Indonesia in April 1983 and September 1986, Korea in February 1980 and Philippines in October 1983. The next panel aggregates four devaluations in Latin America: Chile in June 1982, Mexico in February of 1982, Venezuela in February 1984, and the Mexican crisis of December 1994. Finally, the fourth panel refers to the January 1994 devaluation in the CFA Franc in Africa and comprises the countries with the larger population and aggregate output within the zone: Benin, Cameroon, Cote d’Ivoire, Mali and Senegal. Figure 3 shows that devaluation rates build up over time in the Europe and Latin America cases to reach a peak almost a year after the event. This is the result of a relatively more spread out depreciation after the first shock occurred. By contrast, in Asia and the CFA Franc Zone the devaluation occurs rather abruptly. This is the same pattern that is displayed by the whole sample as observed in Figure 1. Note that we measure the devaluation with respect to the U.S. dollar in all cases, which certainly affects the shape of this curve that depicts the devaluation rate, especially in the case of the Franc CFA countries. But it would appear that the exchange rate against the U.S. dollar is the most appropriate rate to use in order to measure the impact on prices of an exchange rate change. The rate of inflation starts to increase slowly and reaches a peak between 12 and 15 months after the devaluation in the typical case. Only in the case of Latin America we
6
observe that 3 years after the devaluation inflation remains higher than its pre-devaluation level. Therefore, there is something about Latin America that made the economies to reach a higher inflation plateau after the devaluation. Part of the explanation may stem from the fact that Latin American countries started from already high inflation rates. Hence, the expectation that inflation may increase after a negative shock hits the economy may be more widespread and contribute to faster and persistent price increases. It may also be the result of “artificially” low inflation in the period prior to the event, such as in Chile in 1982, where a tablita and an overvalued exchange rate were used to bring down inflation despite the existing macroeconomic imbalances. In other words, prior to the devaluation, inflation was lower than what would have been its “underlying” level. Indeed, Chile had much higher inflation a year before the devaluation. In all of the other cases inflation returns to its pre-devaluation levels within two years after the event. The notable exception is the European group, that did not display any increase in inflation despite devaluation rates as high as 50 percent. This is good news since it shows that in most countries, especially those starting with low inflation, while devaluation does increase the inflation rate, the effect is not a persistent one. This is not surprising since if fundamental determinants of inflation do not change after the devaluation, the economy should return to its initial level of inflation. This pattern is further confirmed in Figure 4 where we plot inflation one year before the crisis, and three years after. The figure shows that in average there is some increase in inflation, but this is heavily influenced by some Latin American countries, which after the crisis entered periods of high inflation. Most of the other episodes are located close to the 45-degree line.
Empirical Approach Our empirical approach is to estimate reduced-form equations for the passthrough index and accumulated inflation, as a function of the size of the devaluation and a number of initial conditions and characteristics of the economies. For the pass-through index we estimate: (2)
PTI k ≡ πk / d k = α0 + α1 X + ε
7
We explore a number of explanatory variables in the set X, with a focus on the conditions that the economy was in at the moment of the currency crisis. In particular, the stylized facts from the various events summarized in Figure 3 suggest to examine whether the degree of misalignment of the real exchange rate and the position of the economy in the business cycle were relevant for the inflationary effects of the devaluation. Note that equation (2) can be rewritten as: π k = α 0 d k + α1 d k X + d k ε which can also be estimated with some additional assumptions on the residual term. In particular if the residual in (2) satisfies the standard OLS assumptions, the residual above will display heteroskedasticity. More in general, if we think that variables in X also have an independent effect on inflation, independent from that stemming from the size of the devaluation, we can estimate the following equation: (3)
π k = β 0 + β1 X + β2 d k + β 3 d k X + ξ
where the effects of a devaluation on inflation depends not only on the magnitude of it, but also on other conditions in the economy. The inflationary effects of a devaluation generally depend on the policy responses after the crisis, perhaps to a considerable degree. It is very difficult, however, to control for the policy response both because it allows for a great degree of arbitrariness in describing the relevant policies and because it will typically have serious endogeneity problems. We do not attempt to control for policy responses in our estimation. We do make an effort to check the robustness of our estimates in regressions (2) and (3), by estimating them at different time horizons, and measuring the explanatory variables in alternative ways. Another general problem is posed by potential omitted variables; however, as long as the omitted variables are uncorrelated with the regressors they will not bias our estimators. In addition, we include initial inflation, in the set of variables X. If the long run determinants of inflation are the same before and after the currency crisis, this variable (initial inflation) could proxy a number of omitted factors which are difficult to measure, such as the degree of central bank independence or extent of indexation in the economy. We apply White’s correction for heteroskedasticity to all of our estimates. The explanatory variables suggested by the four groups of devaluation episodes discussed above are the following. First, we want to control for the output gap, that is,
8
the stage where the countries were in the business cycle. For this purpose we compute the cyclical component of output after detrending GDP with a Hodrick-Prescott filter. We estimate a trend output from yearly data from 1970 to 1997, and apply an HP filter. We define the output gap as the percentage deviation from this trend. Figure 4 shows the evolution of the average output gap in the typical devaluation episode with a two standard deviation band. Year 0 corresponds to the year in which the devaluation occurred, and the other periods represent 3, 2 and 1 year before the devaluation and 1, 2 and 3 years after the devaluation. In the average case, the economy started in an overexpanded situation but decelerated as the time of the crisis approached, to an average of almost 3 percent below trend at the time of the crisis. After the devaluation, output recovered in a fashion consistent with that documented by Kamin (1988) and Kiguel and Ghei (1993), for parity adjustments under pegged exchange rate regimes. Second, the degree of misalignment of the real exchange rate also appeared to be an important consideration. To calculate this variable, we simply take the average monthly real exchange rate since 1970 as a measure of long-run PPP equilibrium, and use the deviations of current real exchange rate from this average as a proxy for misalignment. 4 Consistent with our measurement of real exchange rate as the price of foreign exchange (ep*/p), whenever the real exchange rate is lower (higher) than the average there is an overvaluation (undervaluation). This measure has of course some limitations, especially when productivity differentials or other factors affect the equilibrium real exchange rate over time. As an alternative measure, we examined the behavior of a smoothed series of the real exchange rate using an HP filter. The results were qualitatively similar, but the problem with this latter variable was that, by construction, it forces the period just before the devaluation to appear as periods of overvaluation. Therefore, we adopted the deviation-from-trend measure. Note that this variable measures the equilibrium real exchange rate with reference to the whole sample and not just the period before the crisis. Although there may be reasons for choosing the
4
This is the measure of overvaluation used by Frankel and Rose (1996) and MilesiFerretti and Razin (1998), among others. At the moment of the preparation of this draft we have data on monthly real exchange rate from 1970 for all countries except all Europeans that start in 1979, Benin in 1985, and Mali in 1974.
9
latter alternative, a problem is that many of the early episodes would have too few data points for computing a PPP real exchange rate. In Figure 6 we depict the trajectory of the deviations of the real exchange rate with respect to its average trend. As it is clear from the figure, in average countries entered the period with an overvalued exchange rate, that reached some 10 percent in the year previous to the currency crisis. After the devaluation, the real exchange rate appears undervalued by about 10 percent. This time pattern is common to all experiences, although there is a wide range of values. This strengthens the presumptions that large devaluations result in a real exchange rate depreciation. In addition, this evidence together with the evidence that economies are slowing down before the crisis support the view that crisis are to some extent generated by the recessionary impact of real appreciations. 5 As an alternative measure of misalignment we look at the current account balance, under the assumption that the deficit may be a better proxy for deviations of the real exchange rate from its equilibrium than deviations form PPP. The evolution of the current account balance is consistent with the evolution of the overvaluation measure: in the year prior to the devaluation the current account deficit is in average 5% and narrows to 2.5% in the year of the crisis. In the years after the devaluation the current account deficit remains on average between 1 and 1.5% of GDP. We also use government budget and the degree of openness (trade over GDP) as explanatory variables to capture effects related to the stance of policies and structural characteristics of the economies. All of the variables are measured at the beginning of the period.
III.
RESULTS
The results of the estimations for the pass-trough index, equation (2), are presented in Tables 2 and 3. In the first table we present results for the 3- and 6-month horizon, and in the next one we present the results for the 1- and 2-year horizon. All
5
All time profiles have the same pattern for the extended sample of 49 episodes. 10
equations where estimated using OLS and the standard errors where computed using the White heteroskedasticity-robust procedure. At the three-month horizon, the PTI is not well explained by the regressions. The only significant variables (regression 2.1) are inflation and the current account. The effect of the current account appears to be very small. For example, a current account deficit equivalent to 5 percent of GDP would have reduced the impact of a 10% devaluation by only 0.7%. Results for the six-month (Table 2) and one-year (Table 3) horizons are more favorable. In both cases, the output gap and the degree of misalignment are statistically significant. If output were 5 percent below trend, the PTI would be approximately 0.12 lower than otherwise. This is a sizeable magnitude, as the average PTI is 0.38 at 6 months and 0.55 at 1 year. An initial misalignment of 10 percent reduces the PTI by 0.05 at 6 months and 0.1 at 1 year. Initial inflation does not appear to have effects on the PTI at these horizons. The contemporaneous budget balance appears to have a negative effect on the PTI at 6-months. This result, which is not robust across specifications, would suggest that economies that started with a budget balance higher by two percent of GDP would experience a PTI lower by 0.03 over six months. This effect is small in magnitude and it may be the reason why it does not appear to be significant in other specifications. Openness was not found to be significant in any of the specifications. At the 2-year horizon the results become weaker again. To examine whether economic conditions within the two year period affect the evolution of inflation we include some variables measured in the year after the devaluation. This is illustrated in regression 3.4, where the coefficient is not statistically significant. Initial inflation is the only variable that appears to have significant effects on the PTI at the two-year horizon. The reason may be that two years is a long enough period so that the effect of different policies pursued in different countries and the different shocks that hit the economies become the most important determinants of the PTI. Results of estimating equation (3), with inflation as the le ft-hand side variable, are presented in Tables 4 to 6. In Table 4 we show the simple regression between inflation and devaluation in our sample of 41 observations. A 10% devaluation, on average, results in 2.6%, 3.5%, and 4.8% inflation after 3 months, 6 months and one year,
11
respectively. That is, only about half of the devaluation is offset by higher inflation after one year according to these results. At 2 years, the impact declines slightly, but this result is probably influenced more strongly by omitted variables as discussed above. In Tables 5 and 6 we present the results including interaction effects. After controlling for other variables, the impact of a 10% devaluation on inflation is approximately 3%, 4%, 5%, and 5.5%, after 3 months, 6 months, 1 year and 2 years, respectively. Except for the regressions at the two-year horizon, initial inflation is an important determinant of post-devaluation inflation. This implies basically that the inflation is a persistent phenomenon. Two results regarding the coefficients for variables not interacting with devaluation are important to note. First, in several regressions, in particular at 6 months and 1 year, the cyclical component of GDP is a significant determinant of inflation. When output is below its trend, inflation is lower, consistently with some version of the Phillips curve in a cross-section dimension. For example, if there is no devaluation, each percentage point of output below trend will result in 2% lower yearly inflation. Second, and similarly, at the one-year horizon we see that economies with a real exchange rate overvalued (real exchange rate more appreciated than PPP) will have lower inflation, which also coincides with the sticky prices notion that a real appreciation reduces inflation. The interaction effects of devaluation with initial inflation, misalignment and output gap, are statistically significant. The effects of current account and openness are weaker. For the latter variable, in particular, although the impact of a devaluation on inflation is larger in more open economies, but the effect seems to be small and not significantly different from zero. For example, at the three-month horizon, a 10% devaluation would generate 0.6% of additional inflation in a country 10 percentage points more open. 6 This effect appears very small. In fact, one could expect inflation to be proportional to the product of the share of traded goods times the increase in the exchange rate, at least in the short run. In contrast, the long-run adjustment of the real exchange rate, and hence the inflationary impact of a devaluation, should be less dependent on the degree of openness. Perhaps there could be a relationship between the
12
dynamic adjustment of the nominal exchange rate and the price level with the degree of openness, an issue that is not explored in this paper. For example, for a given adjustment of the real exchange rate in a more open economy, the required change in of the nominal exchange rate, and hence, inflation, could be larger. 7 The interaction term between devaluation and deviations of the real exchange rate from PPP is always significant, except at the one-year horizon, where we find only a direct relationship between misalignment and inflation. In the other regressions the interaction effect implies that the effects of a devaluation are smaller the more overvalued is a currency (negative difference between actual real exchange rate and long run PPP). This is consistent with the idea that the closer is the exchange rate to its equilibrium the more neutral a devaluation is. Consider, for example, equation 5.3. A 20% devaluation implies 8% inflation after six months as a direct effect. But, if the currency is 20% overvalued, which happens with 11 data points in our sample, the effect declines modestly to 7.2%. Therefore, despite the coefficient’s significance, the effects do not appear to be too strong. To some extent, our result may be weakened by the fact that we consider only episodes where there has been a currency crisis, implying that many of the countries did have a significant misalignment to start with. The interaction between the magnitude of the devaluation and initial inflation is negative, and hence, countries starting with high inflation will have a less than proportional impact on inflation from devaluation. That is, the negative sign reflects a non-linear relationship between initial inflation and future inflation for a given devaluation. The same happens with the output gap. The negative sign of the interaction effect, with a positive sign for the direct effect implies that the relationship is non-linear. But the interaction coefficient implies that this nonlinear component is quite small. For example, in equation 6.2, for an economy where the output gap is zero, a devaluation of 20% is expected to generate inflation of 10.2% after one year. But if the initial output gap were 3% (below trend), inflation would be 6% after one year. For a larger value of the initial devaluation, the dampening impact of the cyclical position of the economy would be marginally smaller. 6 7
The average for this variable in our sample is 47%. See Romer (1993) and Terra (1998) on openness and inflation. 13
Summarizing, our main findings are the following: •
The strongest and most robust results are obtained at the six month and one year horizons
•
The pass-through from devaluation to inflation after a currency crisis is not 1:1 even after two years. This implies that currency crisis, in general, produce a long lasting change in the real exchange rate.
•
The level of inflation after devaluation depends crucially on the initial level of inflation.
•
Economies that start with an overvaluation or output below trend tend to have lower passthrough. In addition, the relationship tends to be nonlinear, being less important for higher values of the initial devaluation rate. However, these nonlinear effects are quantitatively small.
IV.
ASSESSMENT AND R ECENT EXPERIENCES
In order to evaluate the results of this paper, in this section we use them to analyze two issues. First we assess the ability of the model to explain actual inflation of the episodes of our sample, and the causes for actual performance. The second issue we analyze is the implications of our results for recent currency crises. The first issue is addressed in Table 7, where we compute the forecast of inflation and compare it with actual inflation for different groups of countries. In addition, in order to isolate the effects of different variables we compute a “partial” forecast. The partial forecast is calculated assuming that the deviations of the real exchange rate from PPP and of output from trend are both equal to zero. Therefore, the difference between the actual forecast and the “partial” forecast represents the contribution of the misalignment and the output gap variables to the actual inflationary effect of the devaluation. We have grouped countries in Europe, CFA-zone, Latin America, and Asia. 8 In addition to get sharper results from well-known crises we also calculate forecasts for the 8
To include all countries in some category, Israel and Turkey were included in Latin America, since they have shown some similarity in their macroeconomic behavior. 14
Latin American countries during the debt crisis, and for the European countries during the EMS crisis of 1992. There are a number of results that are worth noting. Only for Europe, and in particular during the crisis of the EMS, is the forecast poor. In the latter case, after one year, and with a 45% depreciation against the dollar, inflation was 3.5%, almost the same as before the crisis, while our regression results predict a 15% inflation. Note that at the six-month and the one-year horizons, partial is between 2 and 4 percentage points larger than the forecast. That was the contribution to lower passthrough of the initial misalignment and unemployment of resources (output below trend.) The effect is not negligible, but much smaller than the residual, which is about 5% and 12% at six-month and the one-year horizons, respectively. There must have been factors not considered in our regressions that explain this result. The policy reaction, in particular, may have played some role, and some have emphasized the importance of using an inflation target in countries like the United Kingdom (and Sweden after the crisis). 9 Another possibility has to do with measurement error. Since we have been using, for convenience, the exchange rate with respect to the US dollar, and not the effective exchange rate, we could be overestimating the actual devaluation. In fact, the effective exchange rate depreciated less than the bilateral rate. However this could be also a problem in other countries where the explanatory power of our regressions is much better. For the other groups of countries, the results are more encouraging since residuals are relatively small. In the case of the CFA Franc Zone, from six months to 2 years the forecast is very accurate. In addition, the deviation of the real exchange rate from PPP as well as the deviation of output from trend do not explain a relevant part of the actual effect of devaluation on inflation. The most important factor explaining why those countries were able to bring inflation down was the fact that they started with low inflation, and after a couple of years it returned to its original level. In Latin American countries, in particular during the debt crisis, the devaluations were massive, and so was the rise in inflation. The forecast is precise, and although the
9
This has been argued by King (1996). On the European experience see also Eichengreen and Wyplosz (1993), and for a view that questions the success of countries that left the EMS see Gordon (1998). 15
economies started in a boom, the large overvaluation of most of the currencies dampened the inflationary effects of the devaluation. However these effects are small when compared to the effects of very large devaluations that, contrary to most of other experiences, led to rising inflation. Finally, for the Asian countries of our sample we slightly underestimate inflation. Overall, our forecasts work relatively well, except for the EMS crisis. They trace well the evolution of inflation after controlling for the path of the exchange rate. The effects of the output gap and the misalignment of the domestic currency may have played some role in dampening inflation in the EMS and the Latin American countries during the debt crisis. However, the quantitative effect is relatively small when compared to the direct role of the magnitude of the devaluation and initial inflation. We apply the results of our empirical model to the most important recent currency crises in emerging markets to try to explain some apparent anomalies in the extent of passthrough from the devaluations. In Table 8 we present the inflation forecasts at the six-month and one-year horizons obtained from the specifications labeled as 5.3 and 6.2 in Tables 5 and 6. The crises that we study comprise Mexico in 1994, and five Asian economies in 1997: Indonesia, Korea, Malaysia, Thailand, and the Philippines. The Mexican crisis of 1994 is part of the sample over which the models were estimated but the Asian crises were not included in that sample. We use the actual values of the righthand side variables to generate forecasts for the rate of inflation over the corresponding horizon (in the column labeled “explained”) and compare those forecasts to the actual value of the inflation rate. The model tracks developments in Mexico fairly closely, as forecast inflation rates do not deviate much from actual inflation, at both the six-month and one-year horizons. By contrast, the Asian countries all register inflation rates lower than those predicted by our regressions, sometimes by a substantial margin. In fact, with the exception of Indonesia, all countries had fairly low inflation given the significant depreciation rates that they experienced. In addition, it is interesting to note that the partial forecast is in most of the cases below the effective forecast (the “explained” column). As the partial forecast does not include the effect of the output gap and the misalignement of the currency, this means that these two variables contributed to a higher
16
predicted value for inflation. In fact, of the Asian countries entered the crisis with output above trend, which contributes to a higher predicted inflation rate. Also, with the exception of the Phillipines, the Asian countries entered the crisis with a slight undervaluation in their currencies, which also contributes to a higher predicted inflation rate. A number of factors could explain the over-prediction of the inflationary impact. First, in many cases, exchange rates were believed to have overshot initially, which may have delayed price increases in the expectation that the domestic currency would promptly strengthen. Indeed, this strengthening is most noticeable in the case of the Thai Baht, and it is also the case that the overprediction for Thailand is a lot smaller at the oneyear horizon, when the Baht had stabilized at a more credible level. Similarly, although the Asian economies were still booming when the crises started, expectations quickly turned sharply around, and the anticipated recession and increase in unemployment may have had a dampening influence on wage and price increases. Second, most of the Asian economies observed sharp drops in their dollar export prices, and, to a lower extent, in their dollar import prices. One possible explanation is that these economies are not price takers in international markets and might have cut export prices in response to the devaluation. In any event, this drop in U.S. dollar prices would dampen the inflationary passthrough of the exchange rate depreciation. Third, our calculation of misalignment is based on an implicit assumption of long-term PPP without allowance for a long term trend. As the Asian countries experienced large productivity gains over the sample period, one can presume that the equilibrium real exchange rate appreciated in a way consistent with the Balasa-Samuelson hypothesis, and hence that we have underestimated the extent of misaligment was smaller. This problem may be more important in this case because the Asian crises happened at the end of the sample, where the deviations from a trend would be at their maximum values. Finally, it is possible that formally or informally controlled or administered prices may have contributed to reduce the observed inflation rates.
17
V.
CONCLUSIONS
In this paper we have examined the response of inflation after currency crisis. As evidenced in this paper there is no a full pass-through from devaluation to inflation and about 30% passes in the first three months to reach 55% after two years. This implies that changes in the real exchange rate are long lasting. This does not imply that devaluations per se will be that effective since we are selecting episodes of currency crises when some disequilibrium existed. It is likely that in more normal conditions devaluations are more neutral. Therefore, this paper does not address the opportunity of devaluations, rather their effects. In particular, we find that at six months and one year we can track well the effects of devaluations on inflation. The episodes we analyze are characterized, on average, by a deteriorating current account balance and a real exchange rate below the long term average. Inflation, in general, was falling before the crisis, which to some extent may be explained by countries using the exchange rate to disinflate. Output comes from above trend output to fall below trend in the year of the crisis. Then, after the devaluation output recovers, the current account balance too, and the real exchange rate depreciates persistently. Inflation, measured on a yearly basis, peaks between 12 and 18 months after the devaluation, to then return, in general, to levels similar to those pre-crisis. Only in Latin America, specifically during the debt crisis, we find that inflation remains higher than pre-crisis, and even in some cases economies move into extreme inflation. Regarding the determinants inflation after the crisis, beyond the size of the devaluation, we find that the initial level of inflation is key in predicting future performance. In addition, we find that inflation and the pass-through depend on the degree of misalignment and the level of output with respect to trend. An overvaluation tends to reduce the inflationary impact of the devaluation, as expected because a larger real exchange rate adjustment is required. When output is below trend, inflation after the crisis tend to be smaller. We have found, however, that in our sample these effects are small in magnitude. The reason is probably the fact that we focus exclusively on currency crisis, and perhaps compared to countries that do not have such crises the effects of output trend and real exchange rate deviation from PPP could be larger.
18
Finally, our results explain relatively well inflation within the sample, with the exception of the European countries after the EMS crisis, where inflation did not rise at all after the substantial devaluations. The model also does not match the out-of-sample experience in the Asian crises, although a number of special factors may account for this fact.
19
REFERENCES Berg, A. and C. Pattillo (1998), “Are Currency Crises Predictable? A Test,” IMF Working Paper WP/98/154. Dornbusch, R. (1987), “Exchange Rates and Prices”, American Economic Review, 77: 93-106. Edwards, S. (1989), Real Exchange rate, Devaluation and Adjustment, Cambridge, Mass.: MIT press. Edwards, S. and J. Santaella (1993), “Devaluation Controversies in the Developing Countries: Lessons from the Bretton Woods Era,” in Eichengreen, B. and M. Bordo (eds.), A Retrospective on the Bretton Woods System: Lessons for International Monetary Reform, Chicago, Ill.: The University of Chicago Press. Eichengreen, B. and C. Wyplosz (1993), “The Unstable EMS,” Brooking Papers of Economic Activity, vol. 24, No. 1, pp. 51-124. Frankel, J. and A. Rose (1996), “Currency Crashes in Emerging Markets,” Journal of International Economics, 41: 351-366. Goldfajn, I. and R. Valdés (1996), “The Aftermath of Appreciations”, NBER Working Paper No. 5650. Gordon, R. (1998), “The Aftermath of the 1992 ERM Breakup: Was there a Macroeconomic Free-Lunch?,” presented at the NBER Currency Crisis Conference. Kamin, S. (1988), “Devaluation, External Balance, and Macroeconomic Performance: A Look at the Numbers,” Princeton Studies in International Finance, No. 62. Kiguel, M. and N. Ghei (1993), “A Note on Devaluations in Low Inflation Economies,” mimeo, The World Bank. King, M. (1996), “How Should Central Banks Reduce Inflation?--Conceptual Issues,” in Federal Reserve bank of Kansas Symposium, Achieving Price Stability. Milesi-Ferretti, G. and A. Razin (1998), “Current Account Reversals and Currency Crisis-Empirical Regularities,” IMF Working Paper WP/98/89. Obstfeld, M. and K. Rogoff (1996), Foundations of International Macroeconomics, Cambridge, Mass.: MIT Press.
20
Romer, D. (1993), “Openness and Inflation: Theory and Assessment,” Quarterly Journal of Economics, Vol. CVIII, pp. 869-903. Terra, C. (1998), “Openness and Inflation: A New Assessment,” Quarterly Journal of Economics, Vol. CXIII, pp. 641-648.
21
Table 1: Summary Statistics
D3M
D2
P3M
P2
PTI3M
PTI2
2794 52 78787 4 11997
0.32 0.27 1.58 -0.01 0.30
0.63 0.69 1.81 -5.31 0.95
119 50 835 4 190
0.25 0.21 0.73 -0.01 0.18
0.57 0.62 1.81 -5.31 1.02
Full Samplea Average Median Max Min St. Dev.
83 51 1268 10 177
2554 74 61812 -3 9491
28 16 411 0 60 Restricted Sampleb
Average Median Max Min St. Dev
54 46 159 10 37
189 70 2310 -3 396
14 9 46 0 13
D3M corresponds to accumulated devaluation in 3 month, D2 in two years. P3M corresponds to accumulated inflation in three month and P2 in two years. PTI3M corresponds to pass-through index (PTI) in three month and PTI2 in two years. a 26 countries, 49 observation. b 26 countries, 41 observation. Restricted to inflation 3 month after ≤ 100, 2 years ≤ 1000.
22
Table 2: Pass-Trough Indices 3 And 6 Months
PTI3M 2.1 Output gap (t-1) Misalignment (t-1) Inflation (t-1) Openness (t-1) Budget
0.0009** (0.0005)
Dependent Variable PTI3M PTI6M 2.3 2.4 0.012* 0.024* (0.007) (0.011) 0.002 0.004* (0.001) (0.002) 0.0006 (0.0006) 0.157 (0.166)
PTI3M 2.2 0.011 (0.007) 0.001 (0.001) 0.0007 (0.0005) 0.062 (0.144)
PTI6M 2.5 0.024** (0.010) 0.005** (0.002)
PTI6M 2.6 0.026** (0.009) 0.005** (0.001)
-0.013** (0.004)
CurrentAccount -0.013** (t) (0.005) R2 Observations
0.18
0.16
0.12
0.30
0.28
0.35
41
41
41
41
41
41
Standard errors in parenthesis t-(+)x corresponds to x years before (after) the devaluation. Output gap is the deviation of output from trend, and hence a negative number implies that output is below trend. Analogously, misalignment is measured as the deviation of the real exchange rate (ep*/p) from long-run average, and hence a negative number represents an overvaluation. All estimations compute White heteroskedasticity-consistent standard errors and covariance. * Significant at 10% ** Significant at 5%
23
Table 3: Pass-Trough Indices 1 And 2 Years
Output Gap (t-1) Misalignment (t-1) Misalignment (t+1) Inflation (t-1)
PTI1 3.1 0.022** (0.011) 0.009** (0.003)
Budget
Dependent Variable PTI1 PTI1 PTI2 3.2 3.3 3.4 0.025** 0.024** 0.033 (0.010) (0.010) (0.024) 0.010** 0.009** (0.003) (0.003) 0.010 (0.016) -0.0002 0.0016** (0.0006) (0.0005) -0.010* -0.010* (0.006) (0.005)
PTI2 3.5 0.019 (0.017)
0.0019** (0.0007) -0.009 (0.009)
R2
0.21
0.23
0.23
0.08
0.02
No. Obs.
41
41
41
41
41
See notes to table 2.
24
Table 4: Inflation and Devaluation
Same Period Devaluation1 R2 N obs. 1
Dependent Variable: Acc. Inflation 3 Month 6 Month 1 Year 2 Year 4.1 4.2 4.3 4.4 0.26** 0.35** 0.48** 0.43** (0.03) (0.05) (0.06) (0.08) 0.59 41
0.61 41
0.77 41
Corresponds to accumulated devaluation during the same period.
25
0.81 41
Table 5: Regression Results For Inflation With Interaction Effects Dependent Variable: Inflation 3 months 6 months 5.2 5.3 Same period 0.31** 0.39** devaluation 1 (0.03) (0.03) Inflation (t-1) 0.41** 0.74** (0.09) (0.10) Gap (t-1) 1.41* (0.76) Interaction Effect: Same Period Devaluation Times: Inflation (t-1) -0.004** -0.004** -0.005** (0.001) (0.001) (0.001) Misalign. (t-1) 0.002* 0.002** 0.002* (0.001) (0.001) (0.001) Gap (t-1) 0.006 0.005 -0.013 (0.004) (0.004) (0.010) C. Acc. (t) -0.001 (0.005) Openness 0.06 (0.10) N: 3 months 5.1 0.29** (0.07) 0.39** (0.10)
R2 No. Obs. a
0.77 41
0.77 41
0.89 41
Corresponds to accumulated devaluation during the same period. See notes to table 2.
26
6 months 5.4 0.38** (0.03) 0.74** (0.11) 0.54 (0.33) -0.005** (0.001) 0.002** (0.001)
0.88 41
Table 6: Regression Results For Inflation With Interaction Effects Dependent Variable: Inflation N: 1 year 1 year 2 years 2 years 6.1 6.2 6.3 6.4 Same period 0.50** 0.51** 0.59** 0.53** Devaluation1 (0.07) (0.06) (0.11) (0.03) Inflation (t-1) 0.88** 0.87** 1.14** 0.68** (0.28) (0.28) (0.51) (0.07) Gap. (t-1) 2.01** 1.86** (0.88) (0.80) Misalignment (t-1) 0.28** 0.20 (0.12) (0.14) Interaction Effect: Same Period Devaluation Times: Inflation (t-1) -0.003* -0.003** -0.001 (0.001) (0.001) (0.001) Misalign. (t-1) -0.001 0.004** 0.005** (0.001) (0.001) (0.001) Gap. (t-1) -0.019** -0.017** -0.003 (0.009) (0.007) (0.005) C. Acc(t) -0.006 (0.013) Openness (t-1) 0.14 0.29* (0.23) (0.16) R2 No. Obs. 1
0.93 41
0.93 41
0.98 41
Corresponds to accumulated devaluation during the same period. See notes to table 2.
27
0.98 41
Table 7: Forecasting Inflation*
Devaluation
Inflation 1
Initial
Actual
Forecast
Residual 2
Partial
6 Month Europe
26.7
3.3
4.1
5.7
7
-1.6
EMS 3
37.1
3.8
2.2
6.9
8.9
-4.7
90.4 78.1
0.5 22.3
29 32
26.1 31.8
27.8 33.1
2.9 0.2
101.0
41.7
35.7
35.1
40.1
0.6
31.4
3.6
10.7
8.5
8.8
2.3
CFA L. America4 Debt Crises5
Asia
1 Year Europe
32.0
6.8
7.5
12.3
15.1
-4.8
3
44.6
3.8
3.5
15.0
18.9
-11.5
CFA L. America4
85.2 126.8
1.0 49.6
40.5 66.1
34.9 67.1
37.4 71.4
5.6 -1.0
205.8
41.7
85.0
86.3
98.7
-1.3
37.9
7.4
21.7
14.2
18.2
7.5
Europe
31.8
14.0
15.1
26.0
27.1
-10.9
EMS 3
37.2
3.8
6.7
26.4
28.4
-19.7
69.7 340.9
2.0 123.8
50.0 213.8
48.5 207.7
49.9 241.7
1.4 6.1
751.0
41.7
346.8
341.3
466.6
5.5
41.8
15.3
32.5
30.2
35.0
2.4
EMS
Debt Crises5
Asia
2 Years
CFA L. America4 Debt Crises5
Asia
* For 6 month we use equation 5.3, for 1 year 6.2 and for 2 years 6.4. 1 Initial annual inflation adjusted to the relevant period. 2 We assume that the countries have no misalignment and output is equal to trend. 3 Includes Finland, Italy, Spain, Sweden, UK. 4 In addition to Latin American countries it includes Israel and Turkey. 5 Includes Argentina, Brazil, Chile, Mexico, Uruguay, Venezuela in the early 1980s.
28
Table 8: Devaluation and Inflation in Recent Crises* Devaluation
Inflation
Residual
Initial
Actual
Explained
Partial1
6 Month Indonesia Korea Malaysia Philippines Thailand Mexico
217.1 36.5 71.5 54.2 77.5 73.2
2.7 2.1 1.5 2.2 2.2 2.6
32.2 6.4 2.7 3.6 6.0 30.0
78.0 13.8 26.4 17.3 26.1 23.4
74.4 8.9 21.0 15.6 24.2 23.4
-45.8 -7.4 -23.7 -13.8 -20.1 6.6
Average
88.3
2.2
13.5
30.8
27.9
-17.4
Indonesia Korea Malaysia Philippines Thailand Mexico
254.9 26.5 61.5 51.0 35.9 122.5
7.3 4.3 2.1 4.8 4.9 6.9
82.4 6.8 5.8 10.1 10.0 48.5
111.9 18.0 34.8 24.7 19.1 56.0
124.1 10.3 26.2 22.8 15.4 59.3
-29.5 -11.2 -28.9 -14.6 -9.1 -7.5
Average
92.1
5.1
27.3
44.1
43.0
-16.8
1 Year
*
In the first 5 countries the devaluations took place during the second semester of 1997 (out of sample), while in Mexico the devaluation took place at the end of 1994 (in sample). 1 Excluding the effects of misalignment and the output gap.
29
Figure 1: Devaluation and Inflation Restricted Sample (41 obs.)
12 months rate of change
90 80
Inflation
70
Devaluation
60 50 40 30 20 10
month from devaluation
Figure 2: Devaluation and Inflation Full Sample (49 obs.)
800
Inflation
700
Devaluation
600 500 400 300 200 100
month from devaluation
30
36
33
30
27
24
21
18
15
12
9
6
3
0
-3
-6
-9
0 -12
12 month rate of change
900
36
33
30
27
24
21
18
15
12
9
6
3
0
-3
-6
-9
-12
0
Figure 3: Devaluation and Inflation Selected Experiences
Europe
Asia
28
32
36
28
32
36
24
20
-20
16
-12
36
32
28
24
20
16
8
4
0
-4
-8
12
Months from devaluation
Months from devaluation
Latin America
CFA Franc Zone
160
140 120
36
32
28
24
20
16
12
8
4
0
-4
-8
-12
Months from devaluation
24
0 -20
20
0
20 12
20
Inflation
40
8
40
60
0
Inflation 60
-4
80
80
-8
100
Devaluation
100
-12
12 months rate of change
Devaluation
120
16
140
4
-10
-12
0
12
10
Inflation
25 20 15 10 5 0 8
Inflation
4
20
Devaluation
0
30
12 months rate of change
12 months rate of change
Devaluation
-4
12 months rate of change
40
50 45 40 35 30
-8
50
-40 Months from devaluation
31
Figure 4 Inflation Before and After Devalulation
3 Years After
100 90 80 70 60 50 40 30 20 10 0 -10
y = 1,4291x R2 = 0,5
0
10
20
30
40
50
60
70
80
90
100
1 Year Before
Figure 5: Output Gap average +/- 2s.d.
15 10 5 0 -5
-3
-2
-1
0
-10
32
1
2
3
Figure 6: Real Exchange Deviation from PPP average +/- 2s.d
80 60 40 20 0 -20 -3
-2
-1
1
2
3
-40 -60
Figure 7: Current Account Balance average +/- 2s.d.
15 10 5 0 -5 -3
-2
-1
0
-10 -15
33
1
2
3
APPENDIX
Table A.1 Sample COUNTRY
D A T E
D3M
D2
P3M
P2
PTI 3 M
PTI 2
ARGENTI NA
MAR_75
5 1 . 0
2 9 2 2 . 0
2 2 . 7
2 0 0 8 . 3
0 . 4 5
0 . 7
ARGENTI NA
NOV_76
1 0 2 . 9
5 3 2 . 8
3 3 . 4
6 3 1 . 0
0 . 3 2
1 . 2
ARGENTI NA
J U N _ 8 1
5 8 . 7
2 3 1 0 . 0
3 0 . 1
8 3 5 . 0
0 . 5
0 . 4
ARGENTI NA
J U L _ 8 2
1 5 7 . 0
3 0 3 8 . 6
5 6 . 1
2 8 9 4 . 2
0 . 4
1 . 0
X
ARGENTI NA
APR_89
1 2 6 8 . 4
6 1 8 1 1 . 9
4 1 0 . 6
7 8 7 8 7 . 1
0 . 3
1 . 3
X
B E N I N
J A N _ 9 4
9 7 . 1
6 9 . 7
3 1 . 8
5 8 . 8
0 . 3
0 . 8
B O L I V I A
OCT_72
6 8 . 4
6 8 . 4
2 2 . 4
1 2 4 . 0
0 . 3
1 . 8
B O L I V I A
F E B _ 8 2
7 6 . 0
1 9 5 9 . 1
4 1 . 4
1 7 1 5 . 6
0 . 5
0 . 9
B R A Z I L
F E B _ 8 3
6 5 . 4
1 1 7 8 . 1
2 4 . 1
7 3 6 . 1
0 . 4
0 . 6
B R A Z I L
F E B _ 8 7
5 1 . 5
5 6 6 9 . 1
5 6 . 1
5 9 7 9 . 5
1 . 1
1 . 1
X
B R A Z I L
MAR_90
1 2 0 . 7
6 1 4 9 . 4
1 2 6 . 7
5 1 4 0 . 8
1 . 0
0 . 8
X
CAMEROON
J A N _ 9 4
9 7 . 1
6 9 . 7
2 8 . 6
5 8 . 7
0 . 3
0 . 8
CHI L E
MAY_73
3 7 . 7
1 1 5 7 1 . 6
5 9 . 5
4 0 8 7 . 3
1 . 6
0 . 4
CHI L E
J U N _ 8 2
4 1 . 9
1 3 1 . 3
6 . 0
5 6 . 8
0 . 1
0 . 4
CHI L E
J U L _ 8 5
1 4 . 9
4 0 . 7
3 . 4
3 9 . 9
0 . 2
1 . 0
COLOMBI A
F E B _ 8 5
1 2 . 5
9 1 . 9
9 . 2
4 9 . 6
0 . 7
0 . 5
COLOMBI A
AUG_95
1 0 . 4
2 3 . 4
2 . 2
4 0 . 6
0 . 2
1 . 7
COTE
J A N _ 9 4
9 7 . 1
6 9 . 7
2 0 . 3
4 4 . 1
0 . 2
0 . 6
F I N L A N D
OCT_82
1 1 . 4
3 0 . 6
1 . 9
1 6 . 6
0 . 2
0 . 5
F I N L A N D
S E P _ 9 2
2 6 . 9
2 9 . 2
0 . 9
4 . 0
0 . 0
0 . 1
I N D O N E S I A
NOV_78
5 0 . 6
5 0 . 8
3 . 8
4 2 . 7
0 . 1
0 . 8
I N D O N E S I A
APR_83
3 8 . 7
5 6 . 6
5 . 0
1 7 . 5
0 . 1
0 . 3
I N D O N E S I A
S E P _ 8 6
4 5 . 8
5 0 . 1
5 . 5
2 0 . 9
0 . 1
0 . 4
I S R A E L
NOV_74
4 2 . 9
1 0 0 . 2
2 3 . 0
9 1 . 8
0 . 5
0 . 9
I S R A E L
NOV_77
5 3 . 3
1 9 0 . 1
1 6 . 6
2 0 1 . 0
0 . 3
1 . 1
I T A L Y
S E P _ 9 2
2 3 . 8
4 5 . 5
1 . 3
8 . 5
0 . 1
0 . 2
KOREA
J A N _ 8 0
2 0 . 6
4 3 . 7
1 2 . 1
5 0 . 3
0 . 6
1 . 2
MALAYSI A
J U L _ 7 5
1 1 . 8
8 . 8
1 . 1
7 . 2
0 . 1
0 . 8
MALI
J A N _ 9 4
9 7 . 1
6 9 . 7
1 5 . 7
4 3 . 1
0 . 2
0 . 6
MEXICO
S E P _ 7 6
9 5 . 0
8 2 . 7
1 4 . 2
5 7 . 2
0 . 1
0 . 7
MEXICO
F E B _ 8 2
7 4 . 7
4 5 2 . 4
1 3 . 6
2 6 4 . 2
0 . 2
0 . 6
MEXICO
DEC_94
6 5 . 2
1 3 0 . 0
9 . 1
8 9 . 7
0 . 1
0 . 7
PERU
J U N _ 7 6
4 4 . 4
2 1 6 . 7
1 8 . 8
1 2 8 . 9
0 . 4
0 . 6
PERU
OCT_87
7 6 . 5
2 4 2 8 0 . 4
2 4 . 8
3 1 4 2 5 . 2
0 . 3
1 . 3
PHI L I P P I N E
OCT_83
2 7 . 3
6 9 . 2
1 6 . 0
8 4 . 4
0 . 6
1 . 2
SENEGAL
J A N _ 9 4
9 7 . 1
6 9 . 7
2 6 . 0
4 5 . 1
0 . 3
0 . 6
S P A I N
F E B _ 7 6
1 2 . 7
3 5 . 0
5 . 5
5 2 . 0
0 . 4
1 . 5
S P A I N
S E P _ 9 2
2 2 . 1
3 9 . 3
1 . 0
9 . 6
0 . 0
0 . 2
SWEDEN
AUG_77
1 0 . 3
- 3 . 0
2 . 7
1 5 . 9
0 . 3
- 5 . 3
SWEDEN
OCT_82
1 8 . 2
3 7 . 7
2 . 4
1 7 . 8
0 . 1
0 . 5
SWEDEN
S E P _ 9 2
1 7 . 6
4 6 . 3
1 . 3
7 . 4
0 . 1
0 . 2
T H A I L A N D
NOV_84
1 8 . 8
1 3 . 5
- 0 . 2
4 . 9
0 . 0
0 . 4
TURKEY
MAR_94
8 9 . 6
2 6 1 . 8
4 4 . 2
2 9 6 . 5
0 . 5
1 . 1
UK
S E P _ 9 2
2 7 . 0
2 5 . 7
0 . 6
4 . 2
0 . 0
0 . 2
URUGUAY
NOV_82
1 2 6 . 8
3 5 9 . 8
2 4 . 6
1 6 0 . 2
0 . 2
0 . 4
VENEZUELA
F E B _ 8 4
7 4 . 4
7 4 . 4
4 . 1
2 8 . 7
0 . 1
0 . 4
VENEZUELA
DEC_86
9 3 . 3
9 3 . 3
5 . 6
7 9 . 9
0 . 1
0 . 9
VENEZUELA
MAR_89
1 5 9 . 2
2 7 0 . 1
4 6 . 4
1 4 6 . 5
0 . 3
0 . 5
VENEZUELA
MAY_94
5 4 . 4
2 1 0 . 6
2 1 . 9
2 1 9 . 0
0 . 4
1 . 0
D' I VO
34
E x c l u d e d X
X
X
X
