Printer Friendly

Do devaluations improve the trade balance? The evidence revisited.


This paper reexamines the effectiveness of devaluation in trade balance adjustment. The question is addressed in a framework which improves the previous empirical literature in several respects. The evidence indicates that devaluations have been a successful tool in inducing trade balance adjustment. In particular, nominal devaluations are found to result in significant real devaluations that last for at least three years, and the real devaluation induces significant trade flows that are distributed over a two-to three-year period. The evidence comes from two different samples, 1953-73 and 1975-84, involving twenty-seven countries and sixty devaluation episodes.


Do devaluations affect real magnitudes, in particular the trade balance? A devaluation may effect the trade balance through two channels: devaluation of the real exchange rate and a direct effect on domestic absorption. The traditional approach stresses the first channel. A nominal devaluation is assumed to change the real exchange rate (a relative price) and thus improve competitiveness. In turn, if relative prices (the terms of trade in a two country, two-good model) affect the trade balance, devaluation will be successful--in the sense of improving the trade balance, ceteris paribus.

The absorption effect becomes the sole or most important channel in the monetary approach. In a world in which all goods and assets are perfect substitutes, prices are exogenously given for the small country, and wages and prices are flexible in both nominal and real terms, a devaluation increases the price level by the same percentage. The increase in the price level reduces real balances and thus domestic absorption. Dornbusch [1973, 883] has argued that if such a real balance effect is not present, "then it might stand to reason that the effects of devaluation are negligible, not that there must be other powerful avenues through which it exerts its effects" (emphasis added).

The controversy over the effects of devaluation on the trade balance arises because according to Frenkel and Johnson [1976, 42] "the monetary approach rejects the emphasis given to the role of relative prices in the analysis of devaluation." Global monetarists argue that neither of the links in the chain postulated by the traditional approach is likely to hold in practice (see Laffer [1977]). The major disagreement centers on the question of the effectiveness of nominal devaluation to affect the real exchange rate and the importance of the latter in influencing trade flows. Theoretical arguments have not settled the issue. The question is as strongly debated today as it was forty years ago; see, for example, Branson [1983], Katseli [1983]. Kaldor [1983]. Nashashibi [1983], and McKinnon [1981]. Nor does it seem that the available empirical evidence leads to a consistent answer. The best known studies --Cooper [1971], Laffer [1977], Salant [1977], Miles [1979], Gylfason and Risager [1984]--reach contradictory conclusions.(1) Cooper and Gylfason and Risager find that devaluations improve the trade balance or current account, while Laffer and Miles conclude exactly the opposite. Salant finds that devaluations improve the trade balance but not as often as the overall balance of payments. In the face of such diverse findings, Frenkel [1984] suggests that a reexamination of the issue would be useful. Indeed, Himarios [1985] shows that a close examination of Miles's widely quoted study reveals serious deficiencies that, once addressed, lead to opposite results.

This paper provides new evidence concerning the effectiveness of devaluation in trade balance adjustment. The approach differs from past studies in several respects. It differs from Cooper's, Laffer's, and Salant's studies in that it accounts for the effects of other variables that affect the trade balance. It improves on Miles's study by (1) introducing relative price effects in the analysis of the trade balance, (2) allowing for lagged adjustment of trade flows to exchange rate changes, (3) expressing the trade balance in foreign currency instead of domestic currency (for balance of payments analysis this is the appropriate measure), and (4) expressing the trade balance as a raw figure rather than as a percentage of GNP. Although the ratio of trade balance to GNP is a useful measure in many situations, it is not the measure suggested by theory for the problem at hand.(2) This paper differs from Himarios [1985] in that it (1) examines the relationship between nominal and real exchange rates, (2) extends the fixed exchange rate era sample by 50 percent, (3) studies a large number of devaluation episodes from the more recent 1975-84 period, and (4) introduces the expected rate of devaluation as an additional explanatory variable. Finally, it differs from Gylfason and Risager by examining actual devaluations as opposed to simulated ones.

The question of the effectiveness of devaluation is broken down into two parts. First, is a nominal devaluation successful in significantly altering the real exchange rate? And second, are trade flows sensitive enough to real exchange rate changes so that volume responses are induced? This approach allows us to determine whether an unsuccessful devaluation is caused by its inability to change the real exchange rate or by the inelasticity of trade flows.

To verify the results of this approach and avoid criticisms associated with the construction of the real exchange rate measure, an alternative test is developed in which the construction of a real exchange rate measure is not explicitly required. Both approaches show that devaluation improves the trade balance when all other factors are held constant.

Section II investigates the relationship between nominal and real exchange rate changes. Section III briefly discusses the theoretical specification of a testable trade balance equation. Section IV presents the estimating equation and discusses the data. Section V presents and discusses the results. The final section contains the summary and conclusions.


As mentioned above, a nominal devaluation that fails to alter the real exchange rate will have no effects on the trade balance except through real balance effects. According to Williamson [1983, 156], "The fear that most or all of devaluation will be neutralized by induced inflation is in fact the main current reason for questioning the efficacy of devaluation."

This section examines the relationship between nominal exchange rate changes and for the real exchange rate (or exchange-rate-adjusted relative prices) for the two samples used in this study.(3) According to the monetary approach, real devaluations are "transitory" (see Johnson [1977a, 225]). The term "transitory" is never precisely defined but according to Mussa [1976, 193] if the monetary approach is to have any empirical relevance and be useful for policy discussions, "[Its] advocacy ... necessarily involves the assertion that these 'longer-run' consequences materialize within a time horizon of two or three years." In this section, Mussa's criterion will be used to evaluate the predictions of the monetary approach concerning nominal and real devaluation, and to assess its policy relevance for the time horizon considered.

The real exchange rate is defined as(4)

R = ep(*)/P where

e = the domestic currency price of one unit of the numeraire currency

(the U.S. dollar)

P = an appropriately weighted average of the trading partner's price

levels in terms of the numeraire currency.

P = the domestic price level in domestic currency.

The real exchange rate measure is based on relative price levels as measured by consumer prices indexes. This was a constrained choice imposed by data availability.(5) This measure can be thought of as a proxy for the index of the home country's total unit costs to the competitors' total unit costs. In effect, it attempts to gauge relative producer costs and thus profitability. Although widely used, this measure--along with all others--is subject to several limitations that have been explored in detail in the literature (for example, in Maciejewski [1983]) and should be interpreted with cautions and only as a rough approximation to changes in competitiveness or profitability.

The correlation coefficient between quarterly and annual changes in the nominal and real exchange rates. Under strict purchasing power parity (PPP) theory, changes in nominal exchange rates are not associated with changes in real exchange rates. These estimated coefficients are positive and generally high in magnitude, however, indicating a close association between changes in the nominal and real exchange rates. Although a few countries exhibit a lower correlation coefficient for the 1975-84 period, the overall results do not indicate a significant difference between the two periods. One immediate implication is that the strict PPP theory that underlies the assertions of most monetary models does not perform well. The failure of PPP has been verified elsewhere--by Dornbusch and Jaffee [1978] and by Frenkel [1981]--for both the fixed and flexible exchange rate periods.

The relationship in more detail for the fixed rate period by providing some estimates of the magnitude of the real devaluation achieved during the three years following the date of devaluation (measured at the beginning of period t). Several general conclusions can be drawn from these estimates. First, during the devaluation year (i.e., period t) relative prices in all countries change sharply in the direction predicted by the traditional approach, indicating a significant real devaluation. It gives the magnitude of the nominal devaluation in period t while the other measures the percentage change in the real exchange rate from its predevaluation level for each of the periods t, t+1, t+2. For example, these data indicate that the 12 percent French devaluation of 1969 resulted in a 9.5 percent real devaluation during the first year [1969] but at the end of period t+2 (i.e., 1971) the real exchange rate was at its original level. The average nominal devaluation for the whole sample was 56 percent. The resulting average real devaluation during the same period was 45.5 percent. This implies that during the first year the average "slippage" in the price level from a 100 percent nominal devaluation was 19 percent.

Second, the "slippage" in the domestic price level continues in periods t+1 and t+2 but in most cases it is not large enough to offset the initial real devaluation.In only three out of the twenty-seven episodes was the initially achieved real devaluation completely offset by the end of period t+2. Special factors, however, seem to have contributed to such a total erosion (e.g., a severe drought in Colombia in 1962, the political turmoil and labor unrest in France in 1968, etc.). In two additional episodes (Peru 1958-59, U.K. 1967) a complete erosion occurred but only after four years or more had elapsed. The average erosion in period t+1 and t+2 is 11.73 and 14.9 percent of the initial devaluation. After three years approximately 40 percent of the nominal devaluation is offset by higher prices. If the subsequent inflation is due to direct price effects rather than accompanying policy changes, then one would predict that the more open the economy, the more significant the total erosion would tend to be. The correlation coefficient, however, between the total slippage factor and the openness of the economy, measured as the ratio of imports to GNP, is very low. No clearcut relationship emerges from these numbers.

Third, although the average statistics provide useful information, they should be interpreted with caution. The evidence from both intra- and inter-country suggests that there is no strict rule relating nominal and real devaluation. The "slippage factor" seems to differ both across countries as well as over time within a country. One cannot draw firm conclusions from this analysis, however, concerning the response of prices to devaluation. The analysis of the raw data cannot distinguish between the two major types of reasons for slippage--impact effects caused by arbitrage and induced effects caused by wage responses and monetary accommodation. A discussion of the conceptual issues can be found in Pigott, Rutledge and Willett [1985]. Himarios [1987] has investigated the ceteris paribus effects of devaluation on prices and confirms the difference among countries. There was not enough evidence to indicate instability within a country, however. The difference in the slippage factors observed within a country is the result of other factors working at the time of the devaluation.

It examines this relationship for the more recent period. The countries in this sample fall into a variety of exchange rate systems and usually have had experiences with more than one. Egypt, Norway, Sudan and Thailand are peggers (either to a currency or to a basket). Ecuador, Indonesia, Korea, Mexico and South Africa were peggers for most of the seventies and are currently managed floaters or are adjusting to some set of indicators. The rest of the countries are currently described as floating or adjusting to some set of indicators.

The contemporaneous relationship between nominal and real devaluation shown in columns (1) and (2) is in most cases as high as in the fixed-rate period. The effect of the nominal devaluation on the real exchange rate for a longer time period could be monitored for only a few countries or devaluation episodes. In the majority of the cases, the exchange rate moved very frequently, thus changing the real exchange rate and obscuring the effects of the earlier devaluation. For the few cases where the exchange rate remained constant or roughly constant for at least two quarters, the real devaluation that remained after q quarters had passed has been calculated. For example, in Egypt (a dollar pegger throughout the period) the real exchange rate was still 42.8 percent higher after sixteen quarters had passed. Put differently, less than 50 percent of the initially achieved real devaluation (76.5 percent) was eroded within four years. On the other hand, Sudan's 1978:2 and 1979:3 devaluations were eroded within four or eight quarters, thus making a new large devaluation necessary by 1981. But Sudan's experience seems to be unique among the episodes examined.

Overall, the evidence in this section suggests that, after allowing for arbitrage considerations, secondary or induced price effects and other price shocks exogenous to the devaluation, the devaluations in the two samples altered the real exchange rate over a policy-relevant two-to-three-year time horizon.


If the trade balance is not sensitive to relative price changes, then a nominal devaluation that successfully alters relative prices will fail to improve the trade balance. To test the sensitivity of trade flows to relative price changes, I employ a widely accepted specification of the trade balance equation that is generally enough to "nest" the monetarist model but not be limited to it. This trade balance equation can be derived from a two-country Mundell-Fleming type model that includes the real balance effect and the interest rate in the expenditure function.(6)
 TB = f(Y, Y(*), G, G(*), M, M(*), R, r, r(*), ED) (1)
 plus or minus plus or minus minus plus minus plus plus plus or minus plu
s or minus minus

where the variables are defined as follows.

TB = real trade balance

Y(Y(*)) = domestic (foreign) real income

G(G(*)) = domestic (foreign) real government expenditures

M(M(*)) = domestic (foreign) real money balances
 R = the real exchange rate defined as ep(*)/P
 e = the domestic currency price of one unit of foreign currency

P(P(*)) = domestic (foreign) price levels

r(r(*)) = domestic (foreign) opportunity cost of holding money

ED = expected (as of period t) devaluation for period t+1

Increases in domestic real income have an ambiguous effect on the trade balance. An increase in real income increases imports which in turn worsen the trade balance. As real income increases, however, the production of importables also increases and it could increase faster than consumption so as to reduce the volume of imports (Magee [1973]).

Government expenditures in traditional Keynesian macroeconomics increase aggregate demand and, given the level of domestic output, reduce the trade balance surplus. Under alternative assumptions, however, aggregate demand and thus the trade balance may be little affected if the private sector reduces its consumption expenditures in response to the higher government spending (Bailey [1971]).

The monetary variable reflects the major channel of adjustment postulated by the monetary approach, the real balance effect. A devaluation causes an equivalent reduction in real balances, a reduction in expenditures and an improvement in the trade balance.(7) Equivalently, a restrictive monetary policy should have the same effects. Taking the stability of money demand as a maintained hypothesis, a preliminary test of the monetary approach can be conducted by testing whether the coefficient of real money balances is negative and statistically significant (Johnson [1977b, 263]).

The real exchange rate constitutes the core variable in the traditional analysis of devaluation. If the nominal devaluation is successful in altering the real exchange rate or restoring it to its initial equilibrium value, the trade balance will improve provided that excess-demand elasticities are high enough. All the available empirical evidence indicates that these elasticities are high enough not to present any barrier to successful devaluation over a time period of two to three years (Goldstein and Khan [1985]). The total effects of devaluation will rarely occur within a quarter or even a year. Evidence indicates that these effects are spread out over a period of two to three years. Indeed, in the short run the response may even be perverse. Magee [1973], Junz and Rhomberg [1973] and Krueger [1983], among others, have analyzed the reasons for the lengthy adjustment and the possibility of the J-curve phenomenon.

The interest rate in this framework affects the trade balance through its effects on saving and consumption (and hence imports). Since its effects on consumption are ambiguous due to the competing income and substitution effects, its effects on the trade balance are also ambiguous.

All past studies have ignored the effects that an anticipated devaluation might have on the trade balance by assuming that devaluation is unanticipated. And yet, under fixed exchange rates most devaluations are anticipated. Dornbusch [1984] identifies two channels through which such an anticipation can generate an import bulge and thus adversely affect the trade balance. The first channel operates through inventory investment. An anticipated devaluation implies a capital gain on imported goods, be they raw materials or consumer and producer durables. Ceteris paribus, firms would purchase importables prior to the anticipated devaluation and hold them to collect capital gains. The second channel operates through changes in the desired stock of capital. If investment in the tradeable sector has a significant import consent, an anticipated devaluation acts as a subsidy to that investm ent and hence generates an import boom of investment goods. For both of these reasons we would expect the sign of the anticipated devaluation variable to be negative.


Economic theory suggests that the trade balance depends on the current and lagged values of the variables in equation (1). The length of the lags, however, cannot be determined on a priori grounds. Some priors, based on previous econometric research, exist only for the real exchange rate. Studies on import and export equations indicate that, due to consumption and production response lags, exchange rate changes are felt for two to three years after the year of devaluation. Lack of such priors for the other variables suggests that the most appropriate strategy for estimating equation (1) is to start from as general a model as possible and then test down until a more specific model is obtained (see Sargan [1980]). The estimating equation is specified as

TB1 = b0 + a1(L)Y(*) + a2(L)Y + a3(L)M + a4(L)G
 + a5(L)R + a6(L)r + a7ED + a8(L)M(*)
 + a9(L)G(*) + a10(L)r(*). (2)

All the variables are as defined above and a1(L) are polynominals in the lag operator L.

Equation (2) is used to test for the effectiveness of devaluation. Having established the relationship between nominal exchange rate changes and relative prices, a test of the effectiveness of devaluation is a test of whether trade flows respond to relative price changes. a5(L) should thus be positive and significant.

This approach, however, may be subject to a potential criticism. Perhaps it is incorrect to use the real exchange rate to test for the effectiveness of devaluation, since the policy variable available to the authorities is not the real exchange rate but the nominal one. Consequently, it is the nominal exchange rate that should be the focus variable. To address this objection, an alternative test is constructed by decomposing the effects of nominal exchange rate changes from those of relative price levels; the trade balance equation is

TB = Beta0 + Beta1e + Beta2P(*)/P + E (3) where all other variables and lags have been suppressed for expositional simplicity. In flexible-price models, there is a direct relationship between e and P, so that one can write

P(*)/P = s0 + s1e + eta Pe = 1,s = -1 (4)

The error term, eta, in (4) is assumed to capture the effects of all the other excluded relevant variables. Substituting (4) into (3) and rearranging terms yields

TB = b0 + a5e + a'5eta + E (5) where

b0 = Beta0 + Beta2s0

a5 = Beta1 + Beta2 1 Equation (5) can be estimated by substituting the estimated residuals, eta, from equation (4) for the error term eta. a5 now captures the effects of exchange rate changes on the trade balance after allowing for feedback effects on the domestic price level. According to the global monetarist hypothesis, a5 = (Beta1 + Betas2) = 0, since it is assumed that Beta1 = Beta2 and 1 = -1. A test for the effectiveness of devaluation is thus a test of whether a5 is positive and statistically significant.

To carry out this test, the ratio of relative price levels is first regressed on current and past values of the nominal exchange rate. The residuals from this equation, capturing all the other factors that influence relative price levels other than the exchange rate, are then introduced as an independent variable along with the nominal exchange rate in the trade balance equation specified in (2). a5(L) now is the coefficient of the nominal exchange rate and a5(L) is the coefficient of the new variable, the relative-price-levels residuals. This approach has the appealing feature of allowing one to read off directly the effects of a nominal devaluation on the trade balance without having to calculate the real devaluation.

The money supply is M1 or M2, a broad measure of liquidity as defined by the IMF, and the particular definition was chosen on the basis of the best statistical fit. One can justify this procedure by arguing that it amounts to selecting that measure which gives the best demand for money function. In order to satisfy budget constraint requirements (Turnovsky [1977, 39-40]), beginning-of-period values are used for the money supply. This ensures that the money supply is temporally predetermined at the beginning of period t and therefore exogenous to the trade balance, which is measured as of the end of period t.

Expected devaluation for period t+1 is measured as the difference between the black market rate and the official rate at the end of time period t. All but a few developed countries in the first sample had either imposed severe restrictions on international transactions or had inconvertible currencies during the early period. As a result black markets for foreign exchange emerged. Evidence from such markets indicates that movements in the black market are closely associated with changes in the equilibrium exchange rate induced by changes in the purchasing power parity (see Blejer [1978]). According to Himarios [1987], the constructed measure predicted both the direction and magnitude of the exchange rate change with a high degree of accuracy.


The Bretton Woods Era

The shows ordinary least squares estimates of equation (2) for fifteen countries.(8) The estimating period is 1953-73, during which the "adjustable peg" system was in effect. 1953 was chosen because it was the first year for which data were available for all variables and all countries included in the sample.

As in previous research such as Miles [1979] and Bahmani-Oskooee [1985], three of the foreign variables (M(*), G(*), r(*)) were generally insignificant both by a t-test and an F-test. Given the limited number of degrees of freedom and the number of the coefficients to be estimated, established econometric practice (according to Judge, et al. [1980,421]) was followed and these variables were deleted from the final estimation. Although subject to the usual pretesting criticisms, the reestimated equations resulted in almost identical results for the exchange rate variable. This exclusion does not bias the results in favor of one or the other hypothesis.(9)

The results are, on average, consistent with both the traditional theory and previous evidence. Significant positive and negative coefficients for domestic and foreign income and the interest rate occur roughly with the same frequency. The ratios of significant negative coefficients to significant positive coefficients for government expenditures and real money balances are quite high and suggest that increases in these two variables have, ceteris paribus, adverse effects on the trade balance. The cumulative coefficients of relative prices is significant and has the correct sign in all but one case. Finally, an anticipated devaluation, in two-thirds of the cases, has a significant negative impact on the trade balance in the year prior to the actual devaluation.(10) The coefficients of real money balances and relative prices deserve further discussion.

The coefficient of real money balances is positive or statistically insignificant in a number of cases. Several factors may help explain this result. First, one should not expect changes in money supply to have the same effects in all countries regardless of institutional characteristics. Most of the developing countries in this sample made extensive use of import and exchange controls over the sample period. In an exchange-control regime only a small fraction of money supply increases will leak out directly in the form of external deficits. The major effect will be on domestic prices and output (Dervis [1980]). Second, if the money supply increase does not conform to the monetary experiment, i.e., an exogenous once-and-for-all change in the domestic source component of the base (Swoboda [1976, 239]), but instead occurs in order to validate past wage and price increases, its effect on the trade balance might be different, since it does not necessarily lead to an increase in aggregate demand. Third, as far as the deflationary effects of devaluation are concerned, one might argue that in the face of shocks to their income streams, individuals stabilize consumption instead of savings. When a devaluation raises the price level, individuals might engage in dishoarding in order to maintain a stable consumption pattern. Fourth, although expenditure functions may theoretically depend on real balances, hoarding propensities may be quite small in practice. In such a case, the imperceptible effect on the trade balance will not be detected.

Turning to the real exchange rate coefficients, the results indicate that in over 60 percent of the cases one can reject the null hypothesis, at the 5 percent level or better, of no contemporaneous relationship between the trade balance and the real exchange rate. The estimates imply that a real devaluation during period t affects the trade balance positively during the same period, i.e., any initial J-curve worsening has been reversed by the end of the first year. For the U.K., however, the negative effects of the J-curve dominate for the first year.

a5(L) shows the dynamic response of the trade balance to the contemporaneous and two lagged values of the real exchange rate. As mentioned earlier, the maximum lag of the real exchange rate was determined by prior information. Amemiya's PC criterion (see Judge et al. [1980, 420]) was used to determine whether individually insignificant coefficients should be retained or dropped. An interesting but not surprising result is the number of statistically significant lagged coefficients. This suggests that relative price changes affect trade flows long after they have occurred. In many cases, the third-year effect is quantitatively larger than the first-year effect and statistically significant or highly significant (see, for example, the equations for Costa Rica, Ecuador, Finland, India, and Sri Lanka).(11) This finding is consistent with both casual observation and theoretical considerations (Kenen [1975]) and Dornbusch and Krugman [1976]). A joint t-test on the sum of the real exchange rate coefficients (Ea5i). Reveals that in over 80 percent of the cases the cumulative real exchange rate coefficient is significant at the 5 percent level or better. In the majority of the cases during the Bretton Woods period, a real devaluation had a statistically significant net positive effect on the trade balance.

It presents the results of using the nominal exchange rate in the trade balance equation. Under column a5(L), the coefficients of the contemporaneous and lagged values of the nominal exchange rate are shown. In most cases, the lag structures track closely. The only exceptions are Finland, India and Spain, where the lag structure becomes shorter or longer. In over 85 percent of the cases, the cumulative nominal exchange rate coefficient, is significant at the 5 percent level or better. The advantage of this method, as mentioned earlier, is that it allows us to calculate the effects of the nominal exchange rate change on the trade balance directly, without having to calculate "slippage factors." For example, the 17 percent nominal devaluation of the British pound in 1967 should have improved the U.K. trade balance by 44.34 million dollars in real terms by 1970. Remarkably, the U.K. trade deficit went from -41.5 million dollars in 1967 to -1.81 million dollars in 1970 (in real terms). In nominal terms, the devaluation should have caused an improvement of 1892 million dollars. The actual trade balance went from a -1566 million dollars deficit in 1967 to -82 million dollars in 1970. The effects of the devaluation were thus very significant. Similar large ceteris paribus effects can be observed for most other countries.

The Recent Experience: 1975:1-1984:4

The current policy relevance of the results in the previous section may be questioned on the grounds that they are based on a past era and a different world monetary system. It has been argued that the greater integration of national economies and the increased international competition has diminished the effectiveness of devaluation (for example, see McKinnon [1981]). The results in this section indicate that such concerns are overstated. As mentioned in section II, the fifteen countries included in this sample cover a variety of exchange rate systems but they have all experienced single or multiple large devaluations. The availability of quarterly data allowed the use of a more complicated lag structure for the exchange rate.(12) Following Bahmani-Oskooee [1985], an Almon polynomial lag without restrictions has been employed to study the effects of devaluation in these countries. The lag length and the degree of the polynomials were determined as suggested in Johnston [1984, 356-58]. The highest degree polynomial experimented with was four, and the maximum lag was sixteen. It presents the estimates of equation (2).(13) Only the estimates with the nominal exchange rates are presented, since, as in the previous sample, no significant differences appear between the two specifications, and the second specification has an intuitive advantage over the first.

The fit of the equations is quite satisfactory. The overall explanatory power is generally less than in the fixed-rate period, but this is the result of using more "noisy" quarterly data instead of the annual data employed in the fixed-rate period. The cumulative exchange rate coefficients are given under column a5(L). In twelve out of fifteen cases, this cumulative exchange rate coefficient is significant at the 5 percent level or better and has the "correct" sign. It is significant at the 10 percent level in Ecuador and it is insignificant in Thailand. Finally, for Sri Lanka the exchange rate coefficient is significant but has the "wrong" sign. The size of the coefficients imply very significant effects of the trade balance. For example, the 78.8 percent devaluation of the Egyptian pound in 1979:4 should have improved the merchandise trade balance after two years by 1933 million dollars. The actual improvement was 1725 million dollars (from a deficit of 1651 million dollars in 1978:4 to a surplus of 74 million in 1980:3). Similar large effects can be seen for other countries as well. In conclusion, in over 80 percent of the cases, devaluation is shown to be an effective means of trade balance adjustment. This figure is similar to the one for the fixed-rate period.

The dynamic adjustment of the trade balance to devaluation is of particular interest here since quarterly observations are available. These are lags as long as three years (Zambia),(14) although in most countries most of the adjustment occurs within eight quarters. There is clear evidence of J-curve effects in four countries: Ecuador, France, Greece and Zambia. For the first three or four quarters, the trade balance actually deteriorates but this is more than offset later on. In the case of Thailand, the initial positive response is offset by subsequent negative effects, while in Sri Lanka the devaluation has perverse effects. Such perverse effects are theoretically possible and have been explored in the literature (see Katseli [1893]) but, as these results indicate, they are the exception rather than the rule.


A decade ago, Mundell [1975, 50] argued that

After seeing the failure of past devaluations, many people now take

refuge in the argument that it takes three years for devaluation to

work itself out. Where did this magic number come from? All I can

conclude is that devaluation has been shown empirically not to affect

trade balances the way people thought it did, so they invented a time

horizon of three years and told us, in effect, that if we waited until

that beautiful day three years from now, we would get a favorable

delayed reaction. Well, there is no empirical foundation for that idea


The evidence presented in this paper contradicts such assertions and provides support for the alternative hypothesis, showing that over a relevant policy horizon (three years) a nominal devaluation changes relative prices and affects the trade balance positively and significantly. In over 80 percent of the cases, devaluation causes a long-run net improvement (ceteris paribus) in the trade balance in both the fixed-rate period and the recent past.

Correcting for the deficiencies of past studies reveals that the evidence is not as controversial as it might appear at first. The negative results of some of the previous studies appear to be more the consequence of misinterpretation or misspecification than of a true pattern in the data.

The results of this paper support the traditional view that devaluation can be a useful tool, under the right circumstances, in effecting changes in real variables and the structure of the economy. They also lend support to the well-known proposition that the positive effects of devaluation will be offset if devaluation is accompanied by expansionary macro policies. Further, postponement of a devaluation, the timing of which is uncertain but which markets have come to expect firmly, has negative effects on the trade balance. Making the devaluation effective improves the trade balance not only by stopping speculative behavior but by realigning relative prices as well.



The following annual and quarterly data series come from IMF's International Financial Statistics, Yearbook, 1980 (data for 1953-73) and International Financial Statistics, monthly issues 1973-85 (data for 1973-84).
 Trade balance in dollar terms : line 77a,d (or 77aed)
 Consumer price indices : line 64
 Exchange rates : line ae

Government expenditures (or consumption
 when expenditures is not available) : line 82 (or 91f)
 Nominal income measured as GDP : line 99b
 Money supply : line 34 (or 34 + 35)
 Interest rate (where available) : line 60
 Wholesale price indices : line 63
 Real income : line 99b/64

Foreign variables were constructed by taking a trade-weighted average of the relevant variables for the main trading partners of the home country. Constant weights (those obtained for 1966 for the first sample and 1980 for the second) were used and they were constrained to sum to unity. The source of these weights is IMF's Direction of Trade Statistics, yearbooks for 1966 and 1980.

Individual price indices were first converted into dollar terms and then the weighted average was multiplied by the relevant home country/U.S. dollar exchange rate. The exchange-rate-adjusted relative prices are thus the ratio of foreign prices in the home country's currency to the home country's prices. A decrease in the ratio indicates that domestic prices are rising faster than foreign prices expressed in domestic currency.

Black market rate were obtained from Pick's World Currency Report, various issues. The expected devaluation for U.K. was obtained by annualizing the three month forward discount on the pound prior to the devaluation date.

Quarterly figures for GNP for the whole 1975-84 period were available only for Italy and Korea. The quarterly figures for the rest of the countries were generated by interpolation and, where available, supplemented by actual figures. Quarterly figures for government expenditures were not available for the following countries: Ecuador, Egypt, India, Norway, Sudan, Thailand.

Percentage changes, where applicable, are defined as (Xt-Xt-1)/Xt-1

REFERENCES Bahmani-Oskooee, M. "Devaluation and the J-Curve: Some Evidence from LDCs." The Review

of Economics and Statistics, August 1985, 500-504. Bailey, M. J. National Income and the Price Level. New York: McGraw-Hill, 1971. Blejer, M. "Black-Market Exchange Rate Expectations and the Domestic Demand for Money:

Some Empirical Results." Journal of Monetary Economics, November 1978, 767-73. Branson, W. H. "Economic Structure and Policy for External Balance." Staff Papers, International

Monetary Fund, March 1983, 39-66. Buiter, W.H. and J. Eaton. "Keynesian Balance of Payments Models: Comment." American

Economic Review, September 1981, 784-95. Connolly, M. and D. Taylor, "Exchange Rate Changes and Neutralization: A Test of the Monetary

Approach Applied to Developed and Developing Countries." Economica, August 1979.

281-94. Cooper, R.N. "An Assessment of Currency Devaluation in Developing Countries," in Government

and Economic Development, edited by G. Ranis. New Haven: Yale University Press, 1971,

472-513. Dervis, K. "Analyzing the Resource Pull Effects of Devaluation under Exchange Control. "Journal

of Development Economics, March 1980, 23-47. Dornbusch, R. "Devaluation, Money and Non-Traded Goods." American Economic Review,

December 1973, 871-80. _____. "External Debt, Budget Deficits and Disequilibrium Exchange Rates." National Bureau

of Economic Research Working Paper No. 1336, 1984. Dornbusch, R. and P. Krugman, "Flexible Exchange Rates in the Short Run." Brookings Papers

on Economic Activity 3, 1976, 537-75. Dornbusch, R. and D. Jaffee. "Purchasing Power Parity and Exchange Rate Problems:

Introduction," Journal of International Economics, May 1978, 157-61. Edwards, S. "Are Devaluations Contradictory?" The Review of Economics and Statistics, August

1986, 501-8. Frenkel, J.A. "The Collapse of Purchasing Power Parities During the 1970s." European Economic

Review, May 1981, 145-65. _____. "Comment-Does Devaluation Improve the Current Account?" European Economic

Review, June 1984, 65-68. Frenkel, J.A. and H.G. Johnson. "The Monetary Approach to the Balance of Payments: Essential

Concepts and Historical Origins," in the Monetary Approach to the Balance of Payments,

edited by J. Frenkel and H. Johnson, London: George Allen and Unwin, 1976, 1-21. Goldstein, M. and M. S. Khan, "Large vs. Small Price Changes and the Demand for Imports,"

Staff Papers, International Monetary Fund, March 1976, 200-25. _____. "Income and Price Effects in Foreign Trade," in Handbook of International Economics,

edited by R. W. Jones and P. B. Kenen. Vol. II. Amsterdam: North Holland, 1985. Gylfalson, T. and O. Risager, "Does Devaluation Improve the Current Account?" European

Economic Review, June 1984, 37-64. Himarios, D. "Devaluation, Devaluation Expectations and Price Dynamics," Economics, August

1987, 299-313. _____. "The Effects of Devaluation on the Trade Balance: A Critical View and Re-examination

of Mile's 'New Results'". Journal of International Money and Finance, December 1985,

553-63. International Monetary Fund. International Financial Statistics. Washington, D.C., monthly,

1973-85. _____. International Financial Statistics, Direction of Trade Statistics. Washington, D.C.,

annually, 1966, 1980. _____. International Financial Statistics, Supplement on Price Statistics. Washington, D.C.,

1981. _____. International Financial Statistics, Yearbook, 1980. Washington, D.C. Johnson, H. G. "The Monetary Approach to Balance-of-Payments Theory," in The Monetary

Approach to the Balance of Payments, edited by J. Frenkel and H. Johnson. London: Allen

and Unwin, 1976. _____. "The Monetary Approach to Balance of Payments Theory and Policy: Explanation and

Policy Implications," Economica, August 1977a, 217-29. _____. "The Monetary Approach to the Balance of Payments: a Non-technical Guide." Journal

of International Economics, August 1977b, 251-68. Johnston, J. Econometric Methods, 3rd ed. New York: McGraw Hill, 1984. Judge, G. G. et al. The Theory and Practice of Econometrics. New York: John Wiley and Sons,

1980. Junz, H. and R. Rhomberg. "Price Competitiveness in Export Trade Among Industrial Countries."

American Economic Review, May 1973, 412-18. Kaldor, N. "Devaluation and Adjustment in Developing Countries." Finance and Development,

June 1983, 35-37. Katseli, L.T. "Devaluation: A Critical Appraisal of the IMF's Policy Prescriptions." American

Economic Review, Papers and Proceedings, May 1983, 359-63. Kenen, P. "International Monetary Reform and Exchange-Rate Issues: Discussion," in Key Issues

in Monetary Reform, edited by R. Hinshaw, New York: Marcel Dekker, 1975. Krueger, A. O. Exchange-Rate Determination. Cambridge Surveys of Economic Literature.

London: Cambridge University Press, 1983. Laffer, A.B. "Exchange Rates, the Terms of Trade and the Trade Balance," in The Effects of

Exchange Rate Adjustments, edited by P. B. Clark, D. E. Logue and R. J. Sweeney. OASIA

Research Department of the Treasury, 1977. Maciejewski, E. B. "Real Effective Exchange Rate Indices: A Reexamination of the Major

Conceptual and Methodological Issues," Staff Papers, International Monetary Fund,

September 1983, 491-541. Magee, S. P. "Currency Contracts, Pass Through and Devaluations," Brookings Paper on

Economic Activity 1, 1973, 303-23. McKinnon, R.I. "The Exchange Rate and Macroeconomic Policy: Changing Postwar Perceptions."

Journal of Economic Literature, June 1981, 531-57. Miles, M. A. "The Effects of Devaluation on the Trade Balance and the Balance of Payments:

Some New Results," Journal of Political Economy, June 1979, 600-20. Mundell, R. "International Monetary Reform and Exchange-Rate Issues: Discussions," in Key

Issues in Monetary Reform, edited by R. Hinshaw, New York: Marcel Dekker, 1975. Mussa, M. "Tariffs and the Balance of Payments: A Monetary Approach," in The Monetary

Approach to the Balance of Payments, edited by J. A. Frenkel and H. G. Johnson. London:

George Allen and Unwin, 1976. _____. in "General Discussion: What Have We Learned?" Where are the Fundamental Conflicts

of Opinion? Scandinavian Journal of Economics 78(2), 1976, 386-412. Nashashibi, K. "Devaluation in Developing Countries: The Difficult Choice," Finance and

Development, March 1983, 14-17. Pigott, C., J. Rutledge, and T. Willett. "Estimating the Inflationary Effects of Exchange Rate

Changes," in Exchange Rates, Trade and the U.S. Economy, edited by S. Arndt, R. Sweeney

and T. Willett. Cambridge, MA: Ballinger Publishing Co., 1985. Salant, M. "Devaluations Improve the Balance of Payments Even If Not the Trade Balance," in

The Effects of Exchange Rate Adjustments, edited by P. B. Clark, D. E. Logue and R. J.

Sweeney, OASIA Research Department of the Treasury, 1977. Sargan, J. D. "The Consumer Price Equation in the Post War British Economy: An Exercise in

Equation Specification Testing." Review of Economic Studies, January 1980, 113-35. Swoboda, A. K. "Monetary Policy under Fixed Exchange Rates," in The Monetary Approach to

the Balance of Payments, edited by J. A. Frenkel and H. G. Johnson. London: George Allen

and Unwin, 1976. Turnovsky, S. J. Macroeconomic Analysis and Stabilization Policy. London: Cambridge University

Press, 1977. Williamson, J. The Open Economy and the World Economy. New York: Basic Books, Inc., 1983.

(*) Associate Professor, Department of Economics, The University of Texas at Arlington. I would like to thank Richard Ashley, Daniel Orr, Richard Sweeney and two anonymous referees of this Journal for their criticisms and suggestions that have greatly improved both the style and substance of the paper. Responsibility for omissions or errors is solely min e. Suan Chew provided able research assistance.

(1). Following Miles [1979], I exclude from this list some well-known articles that attempt to explain the effects of devaluation on exports or imports in a partial equilibrium framework. Instead, I concentrate on studies that adopt a general equilibrium concept bylooking at the net change in the trade balance.

(2). Consider the case where the trade balance is insensitive to an exchange rate change. The trade balance to GNP ratio will rise if devaluations are contractionary and fall if they are expansionary. A regression of this ratio on the exchange rate would thus give misleading information. Evidence that devaluations have output effects at least over a two-years period is provided in Edwards [1986].

(3). The first sample is a modification of the one used by Miles and is larger by one country. Miles has excluded Colombia, Ghana, India, Korea and Peru for reasons that are unclear. My sample has excluded four countries from Miles's sample: Denmark, Ireland, New Zealand and Guyana. All countries devalued only as a result of the British devaluation and not because of balance of payments problems. Their exclusion does not bias the sample since for two of them Miles finds the exchange rate coefficient significant or nearly significant. Connolly and Taylor [1979] also exclude countries that belonged to the sterling block and devalued with the U.K. The second sample was chosen from a large number of countries that followed adjustment programs supported by the IMF. Availability of data was the main criterion used to select these countries. In addition, some industrial countries are included that experienced sizable devaluations during the recent years.

(4). A different concept of the real exchange rate, employed by many monetary models, is constructed by taking the ratio of prices of traded to nontraded goods. Devaluation in these models can affect the balance of trade by changing the sectoral composition of output. This ratio is usually approximated by the ratio of export unit values to wholesale prices. It could be constructed for only a few countries. As predicted by the theory, the ratio exhibits a sharp increase during the year of devaluation but starts eroding after two to three years. The measure of the real exchange rate used in the text is the one implied by a Mundell-Fleming model where each country produced a single differentiated composite output.

(5). Availability of data was determined on the basis of the information inIMF's Supplement on Price Statistics and International Financial Statistics. In some cases, I was able to construct a real exchange rate measure base on wholesale price indexes. These are used as an alternative measure in some of the test below. The two indexes exhibited similar behavior.

(6). Such a model was developed in Buiter and Eaton [1981]. To save space the derivation is not presented here, since there is wide agreement in the field about the factors that influence the trade balance. The expected devaluation variable (ED) does not follow from the Buiter and Eaton model but it is introduced here since there is enough theory to suggest that it could be an important explanatory variable.

(7). A more eclectic view within the monetary approach would allow for temporary effects on the real exchange rate. Since such changes occur only in the process of adjustment toward long run equilibrium, most monetary models do not explicitly analyze them (Mussa [1976, 411]).

(8). Two stage least squares was also employed in order to take into account the possible simultaneity that exists between income and the trade balance. Since these estimates did not differ significantly from ordinary least squares, which is more efficient, they are not presented here. These results justify the practice in all past studies of ignoring the question altogether. I have also avoided pooling cross-section and time-series data because both theory and the OLS estimates suggest that there are wide differences in the response of each country to a change in a particular variable. Although, in principle, these differences could be dealt with, the existence of different lags and opposite signs makes it very doubtful that the pooling would result in higher efficiency.

(9). The exchange rate coefficients for the unrestricted model are as follows: Colombia 0.17 (2.1), Costa Rica 4.5 (3.2), Ecuador 0.21 (2.8), Finland 7.6 (6.2), France 0.04 (2.4), Ghana -2.6 (0.7), Iceland 0.03 (5.6), India 26.9 (2.8), Israel 26.1 (3.2), Korea 0.02 (6.7), Peru 0.19 (2.6), Philippines 4.86 (3.4), Spain 1.54 (2.8), Sri Lanka 6.2 (1.2), U.K. 1481.5 (7.1). The estimates have been rounded off and the numbers in parentheses are t-statistics. These are almost identical to the restricted estimates presented in the last column of Table IV. (10). A possible simultaneity problem may be associated with the result that anticipated devaluation has a negative impact on the trade balance in the year prior to devaluation since a large trade deficit may contribute to expectations of devaluations. I do not think that this is a serious problem, however. A regression of expected devaluation on the currentand lagged value of the trade balance reveals that in most cases the coefficients are insignificant. In two cases, Spain and Finland, their sum is significant at the 10 percent level, and in one case, Peru, at the 5 percent level. In all three cases, however, it is the lagged variable that has most of the explanatory power. (11). It could be argued that this is the result of using a CPI-corrected exchange rate. The lags could be the result of slow-adjusting consumer prices indices. A more trade-oriented price index might show less variation and would probably result in insignificant coefficients. To address this objection, the equations for those countries for which the WPI-corrected exchange rate could be constructed were reestimated. The results for the cumulative real exchange rate coefficients are as follows: Colombia 0.287 (2.264), Costa Rica 2.116 (1.854), France 0.030 (5.267), India 12.919 (2.427), Korea 0.050 (3.682), Philippines 1.832 (2.779), Spain 2.903 (2.191), U.K. 996.2 (2.506). It is obvious that the results do not depend on the real exchange rate used. More detailed estimates of the lag structure are available from the author. (12). Quarterly GNP and government expenditure (or consumption) figures were not available for all the countries in the sample (see Data Appendix). They were interpolated from annual data using the technique in Goldstein and Khan [1976]. (13). I was not able to compile consistent black market rate series for the developing countries in this sample. The expected devaluation variable is thus omitted. (14). Zambia exhibits significant coefficients at lags 10 and 11 as well, which are not shown in the Table due to space limitations. These coefficients are 3.836 (4.118) and 4.696 (3.863) respectively.
COPYRIGHT 1989 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1989 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Himarios, Daniel
Publication:Economic Inquiry
Date:Jan 1, 1989
Previous Article:Co-movements in relative commodity prices and international capital flows: a simple model.
Next Article:Grading software programs accompanying selected Principles texts.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters