The Effects of Devaluation on the Trade Balance and the Balance of Payments: Some New Results Marc A. Miles R~itgti, Colltgr, Rntg<r\-Thr Stnte Lrl?r'er\rt\ This paper examines the statistical relationship between de\raluation ant1 both the trade balance and the balance of paymelits for 16 de\raluatio~lsof 14 countries in the 1960s. Using several tests involv- ing both the seemingly u~lrelated and pooled cross-section time- series regression techniques, the paper tests the effect of devaluation hile sta~~dardizirlg for other variables that map affect the foreign accounts. \Yhile the balance of pa)-merits does seem to improve follo\ving devaluation, no evidence is found to support the hypothe- sis that cievaluation improves the trade balance. The paper con- cludes that the acljustment to devaluation is essentially monetary in nature, involving only a portfolio stock adjust~nent. Within the international trade literature, it is not uncommon to find arguments about lvhether devaluation will improve the trade balance or the balance of payments. Each theoretical approach has its own set of arguments. For example, the proponents of the elasticities ap- proach (e.g., Robinson 1947; Metzler 1948) describe the necessary and sufficient conditions for an improvement in the trade balance in terms of elasticities of demand and supply. If the demand elasticities are sufficiently large and the supply elasticities sufficiently small, devaluation should improve the trade balance. Proponents of the absorption approach (e.g., Alexander 1952; Johiison 1967) describe 11on. devaluation nlay change the terms of trade, increase production, l'he author ~vould like to thank Jacob Frenkel, Harr! Johnson, Arthur Laffer, Stephen Slagee, John Bilson. and an anonvmous referee for helpful aclvice and corn- ments. The\- should not be held responsible. holve~er, for any remaining errors. [Joiir~i(il 01 PoJ~IIc(I/ F~orzo~riv, 1979. xol. Xi, no 11 'G 1979 by 1 he Yni\ersii\ of C:hicago. 0022-380817Y~8703-00Oti$01 58 THE ETFECTS OF DEVAI L ATION 60 1 switch expenditure from foreign to domestic goods, or have some other effect in reducing domestic absorption relative to production and thus improving the trade balance. International niorletarists (e.g., Mundell 197 1 ; Dornbusch 1973n; Frenkel and Rodriguez 1975) argue that derraluation reduces the real value of cash balaiices andlor changes the relative price of traded and nontraded goods, thus im- proving both the trade balance and the balance of payments. This article, however, examines the statistical relationship between devaluation and the two foreign accounts. More specifically, the arti- cle tries to determine if, on the average, devaluation improves the trade balance andior the balance of payments. No attempt is niade to show the merits of one theoretical approach over another. While the theoretical discussion is primarily in terms of a monetarist model, the final empirical 111odel is a reduced-form equation that is not inconsis- tent u.ith the other theoretical models. If devaluation causes a significant improvement in the trade balance, this irnprovernent should he statistically observable regardless of ivhich theoretical ap- proach is used. Section I describes empirical studies of the effects of devaluation by other authors and analyzes why their approaches fail to answer the relevant questions completely. Section I1 describes the functional forms used in this study and summarizes the theory behilid the model. Section I11 describes the various tests and their results. Finally. Section IV summarizes the results and drarvs some conclusions. I. Other Empirical Studies In recent years several papers have appeared tvhich have tried to analyze empirically the effect of devaluation on the trade balance and balance of payments. There are three basic objectio~is that one can niake to these previous studies: (1) They examine only the impact effects and fail to show whether any apparent improvement is tempo- rary or permanent. (2) They do riot compare postdevaluation levels of the accounts with predevaluation levels. It therefore cannot be de- termined if an improvement as compared with the pear of devalua- tion is also an irnproveme~it as compared with predevaluation years. (3) They have not accounted for the effects of other variables such as the government's monetary or fiscal policy. Only the improvement or ~vorsening of the raw account figures follolving devaluation is re- ported. Such a procedure is equivalent to ascribing all the changes in the accounts to devaluation. TYhile all three objections do not apply to each of the previous studies, each study fBils to deal properly with at least one of them. For example, the widely quoted study by Cooper (1971n, 1971b) is subject - - 602 JOURNAL OF POLITICAL ECONOMY to all three objections.' Cooper shows that the impact effect of 15 of 24 devaluations is to "improve" the balance of goods and service^,^ while in 17 of 24 cases the balance of payments improves. However, there is no indication if this improvement is permanent, because figures are not provided for succeeding years. Nor are figures pro- vided for years prior to devaluation, so it is impossible to determine if the "improved" level is an improvement over the levels in the previous pears. In fact, 10 of the 15 "improved" balances of goods and services are still negative follo~ving devaluation, while in contrast 14 of the 17 improved balances of payments are in surplus. Finally, no attempt has been made to separate what portion of the improvement is due to devaluation and what portion is due to changes in other variables. Connolly and Taylor (1972) partially overcome some of the shortcomings of the Cooper paper. Using 16 of the devaluations in the Cooper sample, they try to relate improvements in the balance of payments (defined as net change in reserves) and rates of domestic credit creation. They conclude that (1) the higher the rate of devalua- tion, the greater the improvement in the reserve position; and (2) the higher the rate of domestic credit expansion following devaluation, the smaller the improvement. Laffer (1976) eliminates the first two objections in a test of 15 postwar devaluations. He examines the time path of the trade balance over 'i years, from 3 years before devaluation until 3 years after. He finds that although the trade balance "improves" in the year following devaluation for eight of the 15 cases, in one-half of those cases the trade deficit is still worse than the average balance of the 3 years prior to clevaluation. Furthermore, 10 of the 15 countries have the largest deficit of the 7-year period in the 3 years following devaluation, and two more have the largest deficit in the year of devaluation. Ten of the 14 countries with data for the third year following devaluation have a larger deficit in that year than in the pear after devaluation. Thus, there is little evidence of devaluation causing significant or sustained improvement in the trade balance. An analysis similar to Laffer's is performed by Salant (1976) on 101 devaluations, for both the trade balance and the balance of payments. Salant finds that in about three-quarters of the cases (75 of 101) the In this summary, I have excluded some well-known articles that attempt to explain the effect of devaluation on the percentage change in exports, imports, market shares, o1.5ome other related variables. IVtlile these articles purport to explain the effect on the trade balance (a general equilibriutn concept), their technique is to examine changes in exports or imports in isolation (a partial eq~lilibriuril concept). Since I am investigating the siniultaneous net change in exports and imports in this paper, I have restricted the discussion in this section to studies that have also dealt explicitly with this concept. In each of these studies, an "improvement" is defined as any absolute reduction in a drficit or an increase in a surplus. 603 THE EFFECTS OF DEVALUATIOB average balance of payments for the 3 years following devaluation is improved as compared with the average in the 3 years preceding devaluation. In contrast. in less than one-half the cases (46 of 101) does the average of the trade balance improve. The implication of these studies is that there is considerably more evidence for the balance of payments to improve following devalua- tion than for the trade balance to do so. But, while some of the studies overcome one or two of the previously stated objections, none takes into account all three. Even more important, none of the tests fully accounts for the third objection. Only the Connolly and Taylor (1972) study makes any attempt to take into account the kffects of a variable other than devaluation. The object of this paper is to account for each of the three objec- tions which arise from the previous studies. Evidence of not only the impact but also the longer run effects of devaluation will be examined and compared with the predevaluation behavior of the accounts. Furthermore, an attempt will be made to isolate the effects of devalu- ation alone by standardizing for the effects of other exogenous vari- ables. 11. The Equations The trade balance equation is an absorption model concentrating on factors that affect domestic expenditure relative to domestic output. A/-/ TB = a, + a:-) A(g, - g,) + a:-' A(M, - hIR) Y, (1 ) +ad-' h(G, - GR) + nJr' AEK,. The balance of payments equation is a simple monetary approach model concentrating on factors that affect money supply relative to money demand. where TBi = the level of the trade balance in country i; Bpi = the level of the balance of payments in country i; Yi = the level of output in country i; gi, gR = growth rates of income in country i and the rest-of-world R; Mi, MR = the ratio of the average level of high- powered money (where changes occur only from domestic sources) to output in country i and the rest-of-world R; Gi, GR = the ratio of government consumption to output; and ERL = the exchange rate of country i (see the Appendix for a more detailed description of these variables). 604 JOURNAL OF POLITICAL ECOSOhIY Following Laffer (1969, 1975), increases in a country's growth rate relative to the average growth rate in the rest of the world are expected to cause a deterioration in the trade balance but an im- provement in the balance of payments. These anticipated relation- ships are consiste~lt with the theory that commodities such as money and goods tend to flow from areas of the ~vorld where there is excess supply to areas of excess demand. An increase in the relative growth rate of a country increases demand for money and goods relative to elsewhere, increasing the flow of these con~modities into the country and causing the trade balance to Ivorsen and the balance of payments to impr~ve.~ The nlonetary variable is not the usual money supply variable such as M 1 or M2. Rather, it represents changes in the level of the domestic portion of high-powered money, the portion of high-powered money directly under the control of the monetary authorities. The effect of this variable on the trade balance is uncertain, depending upon se\,- era1 factors. According to some theorists (Johnson 1972; Dornbusch 19730, 1975; Frenkel and Rodriguez 1975), as the government in- creases the money supply, the level of real balances in the economy increases. Individuals perceive their wealth to rise, causing the level of expenditure to increase relative to income and the trade balance to deteriorate. However, a negative relationship between changes in the money supply and changes in the trade balance may not be observed for at least three reasons. First, nominal money balances may be only a small fraction of total wealth. In this case, a devaluation will not cause a significant change in real wealth. Second, expenditure may respond only slightly to changes in wealth. Thus, even if the reduction in the real value of money significantly reduces real ~vealth, an impercepti- ble change in the trade balance will be observed. Finally, money may not be perceived as net wealth by the private sector. In this case, there "he hypothesized relationship bettveen the trade balance and growth rates ausumes that demand shocks dominate supply shocks. If ari increase in the relative growth rate represents an exogenous supply shock, then output increases Inore than demand, creating an excess supply of goods and a positive relationship between changes in relative growth rates and changes in the trade balance. However, if the increased relative growth is caused b\ an exogenous dernand shock, domestic demand rises by more than domestic supply, creating the negative relationship between changes in relatibe growth rates and changes in the trade balance. A general equilibrium model of ho~ demand shocks produce the negative relationship is provided in Laffer 11975) and Laffer and hliles (1980), chap. 8. Empirical cvidence that demand shocks tend to dominate is found not only in this paper but also in Laffer (1969), Laffer and Ranson (1975), and Miles (1977). An alternative explariation of the negative relationship in terms of permanent income is found in Miles (1977). It' current economic conditions strongll irlfluerice expectatiorls about the future, a rise in the current growth rate raises the expected increase in income not only in the current year but in fl~ture years as well. Permanent incomc theretor-e rises b! more than present income, causing current demand to rise by more than current supply and a worsening in the trade balance. 605 THE EFFECTS OF DEV.4LUATIOS is no real balance effect and the trade balance does not deteri~rate.~ If any of these three conditio~ls holds, the real balance effect loses its empirical relevancy. Increases in the money supply are expected to cause the balance of payments to worsen regardless of whether the real balance effect is relevant. Where a real balance effect exists, increasing the money supply increases the supply of real balances relative to demand; and this excess supply ~vill be alleviated by an outflolv of money through the balance of payments. 111 the absence of a real balance effect, increasing the money supply raises the ratio of money to bonds above the desired level, necessitating a portfolio adjustmeilt that occurs exclusively through a balance of payments-capital account deficit. Notice that it is not just the changes in the domestic ratio of money to output which is used as the independent variable but, rather, changes in the difference between the domestic and rest-of-world values. For example, as long as there is no reserve creation in the ~~orld, the sum of all changes in the balance of payments must be zero. If all governments are increasing the domestic ratios of money to output, not all balances of pavments can be deteriorating. Instead, the account will be expected to deteriorate in those countries where the domestic ratio increases the most. Government consunlption represents one source of direct expen- diture in the economy. Thus, if the government increases the level of its corlsumption by a dollar, as long as individuals do not immediately view government expenditures as merely replacing their private con- sun~ption,total expenditure will rise. Given the level of output, a rise ill total expenditure causes the trade balance to deteriorate. As in the case of the monetary variable, the change in the domestic ratio rela- tive to the rest-of-~vorld ratio is used as the independent variable. Since devaluation is a monetary phenomenon, the expected effects depend upon the importance of the real balance effect. If the real balance effect is empirically relevant, devaluation is expected to im- prove both the trade balance and the bala~lce of pavments. The Eeduction in real balances caused by devaluation both ;-educes expen- diture and creates an excess demand for money, ivhich has the effect of improving both accounts. Hon~ever, if the real balance effect is not empirically relevant, only the balance of payments is expected to improve. The absence of a real balance effect means that both expen- diture and the trade balance are unaffectecl. Instead. devaluation Qank-created or "inside" monev is certainly recognired not to be net wealth to the private sector. The onlv remaining rnone! asset is fiat or high-powered mane\. Fiat monev is the liability of the guvernment, To the extent that the private sector recog- nizes that governmental liabilities represent tax liabilities on the PI-ivate sector, fiat money is si~nultaneousl\ both a private asset and liability. 606 JO~RNAL OF POLITICAL ECONOMY reduces the ratio of real money assets to real net bond assets below the desired level. The necessary adjustment occurs entirely as a portfolio reacljustment, as bonds are exchanged for money. 111. Empirical Tests In this section the functional forms of the trade balance and the balance of payments equations described above are used in several tests to try to determine if devaluation has a positive effect on either account. As an interesting by-product of the procedure, the impor- tance of the individual variables in affecting the accounts can also be measured. By regressing country equations separately using ordinary least squares (OLSs), it is implicitly assumed that the errors of the indi- vidual countr?. equations are unrelated. However, such an assumption may not be valid for the trade balance and the balance of payments. For the world as a whole, the sum of all trade balances or the sum of all balance of payments is constrained to be zero. For each export in the world there must be an equivalent import. Thus, the trade balance and balance of payments errors for all countries in the world are not independent. The remaining question is what effect any covariance of errors will have on the estimated coefficients. The effect of the covariance is measured through the use of seemingly unrelated re- gressiorls (Zellner 1962; Kmenta 1971). The first step of the proce- dure is to estimate the individual country equations by OLSs. At this stage, the equations are assumed unrelated. However, in the second step the errors of these equations are used to create a variance- covariance matrix. The final step is to use Aitken's generalized least squares to reestimate the individual regression coefficients incor- porating the information about the covariance of the errors. Incor- porating the information increases the efficiency of the estimators. The sample of countries for the seemingly unrelated tests consists of 14 countries. One problem is the possible first-order autocorrela- tion of the disturbances of individual countries. To handle the au- tocorrelation, estimates of the autocorrelation coefficient p obtained from the iterative Cochrane-Orcutt technique (CORC) on OLSs are used to transform the data before the present test is performed. The transformation leaves each variable in each of the 14 countries with 17 observations for the period 1956-72. 607 THE EFFECTS OF DEVALUATIOh' The seemingly unrelated regressions are used for both a residuals test and a direct test of the exchange rate, both of which are described below :' Residuals Test This test utilizes equations (1) and (2) ~~itho~itthe exchange-rate variable. Excluding the exchange rate in the estimated equation means that the residuals of the equation represent the effects of devaluation and all other variables. Using a test similar to one em- ployed by Fama, Fisher, Jensen, and Roll (1969), the path of the residuals around devaluation is examined to determine if they be- have differently after devaluation occurs as compared with prior behavior. Examination of these residuals around the period of devaluation provides information about the effectiveness of devaluation. For example, if devaluation causes a permanent improvement in the trade balance, there should be large positive residuals at or near the year of devaluation and only small or random residuals in subsequent years. The large positive residuals indicate a large positive change in the trade balance that is not explained by government monetary or con- sumption policy or by growth rates. The subsequent small residuals would indicate little change in the trade balance in the future that is not due to the two policies or growth rates. The monetary approach suggests that there should be only a tem- porary improvement in the trade balance as real cash balances return to the desired level. A temporary improvement ~vould be indicated by positive residuals at or near the year of devaluation and negative residuals in subsecluent years. This pattern of residuals would indicate a positive change in the trade balance not explained 11y the stan- dardized variables, followed by eventual negative changes as the trade balance returns to its original level. If the trade balance is to return to its original level, the sum of the positive residuals should approximate the sum of the negative residuals. A third possible path of' behavior is the so-called J-curve. Under this hypothesis, devaluation causes the trade balance to worsen initially, due to price effects, but eventually to improve as demand for the countr-y's goods increases. This hypothesis would be consistent with initially large negative residuals followed by large positive residuals. If This paper reports only the results of the seemingly unrelated regression equations. However, a comparison of the results for both OL.Ss and seeminglv unrelated regres- sions is reported in Miles (1978). The primary effect of using the seemingly unrelated technique is to increase the frequency of significant coefficients of the expected sign. 608 JOURNAL OF POLITICAL ECONOMY the trade balance eventually improves, the sum of the positive re- siduals must exceed the sum of the negative residuals. The residuaIs for each country are divided bx the standard error of the seemingly urlrelated regreosion for that country. This procecture prevents the residuals of countries with large 1 ariame in the residuals from having a disproportionate n eight when residuals are averaged across countries." In each country, the year of devaluation is called !,ear t. Residuals (divided by the standard errors) are collected for each country for the 7 years t - 3 tot + 3. The residuals of similar periods (~vith respect to devaluations) are then summed across countries, and average residu- als are created by dividing by the number of countries in the group. The resulting number represents the average residual as a fraction of the standard error for all the countries in the year of devaluation. The procedure is repeated for the other six time periods, permitting comparison of the averages from 3 years before until 3 years after devaluation. If devaluation improves the trade balance or balance of payments, the improvement will surely occur within 3 years. The results of this procedure are shown for the trade balance in table 1 and for the balance of payments in table 2. Both tables contain the residuals for 16 devaluations of the 14 countries, as well as the average values. Examine first the results for the trade balance. The average residuals are snlall but positive in the 2 years prior to devalu- ation. This result suggests that any deterioration of the trade balance in those years appears, on the average, to be caused by government policies and grolvth. In the year of devaluation, however, there is a very large negative residual. The residual is larger in absolute size than the residual in any of the other 6 years. In the following 3 years, there is a positive residual only in year t + 1. But this positive residual is only about three-fifths the absolute size of the negative residual in the year of devaluation. In other words, there is no evidence of devaluation improving the trade balance even temporarily. Devalua- tion has a large negative effect in year t and continues to have a net negative effect in the follo~ving 3 years. In fact, the trade balance deteriorates even more follolving year i + 1. " The purpose of dividing the residual\ b) the standard error of the corresponding equation is to include deviations that are not due to just a poorl) fittecl regression equation. The procedure is an attempt to tackle the problern of weighting outl)ing observations. However, an alternative vielv might be that the residuals should not be weighted. Tlle residuals test has been performed on both seemingly unrelated and OLSs for (i) residuals ~eighted hy the inverse of the standard errors. (ii) un\ieighted residuals, and (iii) unweighted residuals, but excluding countries ~\.ith "large" or outly- ing residuals. In each case, the results are esentiallv the same for both the trade balance and the balance of payments. The results, therefore, do not appear sensitive to either the weighting wheme or the tvpe of regression analysis. For thc I-esults using un- weighted residuals of the OLSs equatiom, see hIiles (1078). m~?-bcm~q~,m,ma-bbmq'cm, mmamrnc~~~~pbmcb~a~qb~ P~lr,r:cqf+cq~5!xq-D!Lyq-CI!D~cc:- - I i-i I 1x1 I I l [...]... positive and even larger in value In year t + 2, the residual again becomes negative, but since none of the improvement in the balance of payments is offset for at least 2 years, devaluation has clearly inlproved the balance of payments T h e contrasting pattern of response to devaluation for the trade balance and the balance of payments is also apparent within the individual countries In five of the 16 devaluations,... the 16 devaluations, the residuals of the balance of - payments equation in the year of devaluation are positive and at least one standard error in size None of the negative residuals are that large In contrast, for the trade balance only one residual is positive and one standard error in size, while five negative residuals are that large Even in t + 1, the trade balance equations have only three positive... predictions of the monetary approach that devaluation will produce a stock adjustment of the money supply On the other hand, there continues to be no evidence of devaluation producing even a temporary improvement in the trade balance IV Summary and Conclusion While the preceding tests imply informatior1 about the effects of several variables, this section will be confined primarily to the implications about devaluation. .. 61 1 T H E EFFECTS OF DEVALUATION T h e pattern of the I-esiduals for the trade balance contrasts sharply with the pattern for the balance of payments The average balance of payments residuals are negative for the 3 years prior to devaluation, peaking in t - 1, the year before devaluation However, in the year of devaluation, the average residual suddenly beco~nes positive In year t + 1, the residual... Examining the residuals of the equation without the exchange rate, however, provides a different result T h e residuals indicate a small improvement in the trade balance in the year following devaluation But this improvement is small compared with the deterioration of the trade balance in the year of devaluation or succeeding years On the other hand, there is clear evidence of the balance of payments. .. except Spain and Sweden T h e monetary variable excludes Italy and SH'itzerland T h e final variables used in the equations are the first differences of the trade balance and balance of payments ratios and the first differences of the difference between the domestic and foreign ratios for the other variables References Alexander, Sidney S "The Effects of a Devaluation on a Trade Balance. " 1.1M.F Staff Papers... in the value of monetary wealth In any case, since devaluation does not improve the trade balance but improves the balance of payments, by definition the capital account must be improving Devaluation therefore seems to cause only a simple portfolio adjustment Rather than affecting the size of the portfolio, and thus net wealth, devaluation causes a simple excess demand for money and excess supply of. .. method of testing the effects the exchange rate directly into the trade payments equations, as in equations (1) significance and sign of the exchange-rate of devaluation is to enter balance and balance of and (2) T h e statistical coefficient then gives an Ten of the 16 devaluations occurred in 1967 This simultaneity of devaluation coulcl possibly affect the results since some of the 10 countries may... here are the negative ones between the trade balance and growth and between increases in the money supply and the balance of' payments As a further attempt to test the importance of the variables the coefficients of the variables are constrained to be the same in all 614 JOURNAL O F POLITICAL ECONOMY countries T h e procedure involves performing a pooled cross-section time-series regression on the 14... residuals of one standard error in size, compared with four for the balance of payments Furthermore, the improvement in the balance of payments in years t and t + 1, followed by a deterioration in year t 2 , is consistent with the pattern of behavior associated with a "stock adjustnlent of the money supply following devaluation While the negative value in t + 2 does not completely offset the combined . I l l THE EFFECTS OF DEVALUATION 61 1 The pattern of the I-esiduals for the trade balance contrasts sharply with the pattern for the balance of payments. The average balance of payments. years, devaluation has clearly inlproved the balance of payments. The contrasting pattern of response to devaluation for the trade balance and the balance of payments is also apparent within the. countries. In five of the 16 devaluations, the residuals of the balance of payments equation in the year of devaluation are positive and at least one standard error in size. None of the negative