+ All Categories
Home > Documents > Exchange rates and the trade balance: the case of Singapore 1970 to 1996

Exchange rates and the trade balance: the case of Singapore 1970 to 1996

Date post: 18-Sep-2016
Category:
Upload: peter-wilson
View: 218 times
Download: 0 times
Share this document with a friend
17
Exchange rates and the trade balance: the case of Singapore 1970 to 1996 1 Peter Wilson,* Kua Choon Tat National University of Singapore, Economics Department, Kent Ridge Crescent, Singapore 119260, Singapore Received 1 November 2000; revised 1 October 2000; accepted 1 December 2000 Abstract The purpose of this paper is to examine the relationship between the real trade balance and the real exchange rate for bilateral trade in merchandise goods between Singapore and the USA on a quarterly basis over the period 1970 to 1996 using the partial reduced form model of Rose and Yellen (1989). We also hope to shed further light on what has become known as the ‘Singapore export puzzle’: the observation that, despite periods of rapid nominal and real appreciation of the Singapore dollar, export growth in aggregate has remained buoyant. Our findings suggest that the real exchange rate does not have a significant impact on the real bilateral trade balance for Singapore and the USA, thus confirming previous work which finds a weak relationship between changes in the exchange rate and changes in export and import prices and volumes for Singapore. We also found little evidence of a J-curve effect. Although positive coeffi- cients linking real exports with lagged values of the real exchange rate might be indicative of ‘small country’ pricing by exporters in U.S. dollars, it is not clear that this is masking J-curve effects from an initial rise in import values as the home currency depreciates. © 2001 Elsevier Science Inc. All rights reserved. Keywords: Trade balance; Exchange rate; J-curve The greatest tragedy in all history is the murder of a beautiful theory by a gang of brutal facts. (Anonymous) * Corresponding author. Tel.: 165-874-3997; fax: 65-234-4985. E-mail address: [email protected] Journal of Asian Economics 12 (2001) 47– 63 1049-0078/01/$ – see front matter © 2001 Elsevier Science Inc. All rights reserved. PII: 1049-0078(01)00072-0
Transcript
Page 1: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

Exchange rates and the trade balance: the case ofSingapore 1970 to 19961

Peter Wilson,* Kua Choon Tat

National University of Singapore, Economics Department, Kent Ridge Crescent, Singapore 119260,Singapore

Received 1 November 2000; revised 1 October 2000; accepted 1 December 2000

Abstract

The purpose of this paper is to examine the relationship between the real trade balance and the realexchange rate for bilateral trade in merchandise goods between Singapore and the USA on a quarterlybasis over the period 1970 to 1996 using the partial reduced form model of Rose and Yellen (1989).We also hope to shed further light on what has become known as the ‘Singapore export puzzle’: theobservation that, despite periods of rapid nominal and real appreciation of the Singapore dollar, exportgrowth in aggregate has remained buoyant.

Our findings suggest that the real exchange rate does not have a significant impact on the realbilateral trade balance for Singapore and the USA, thus confirming previous work which finds a weakrelationship between changes in the exchange rate and changes in export and import prices andvolumes for Singapore. We also found little evidence of a J-curve effect. Although positive coeffi-cients linking real exports with lagged values of the real exchange rate might be indicative of ‘smallcountry’ pricing by exporters in U.S. dollars, it is not clear that this is masking J-curve effects froman initial rise in import values as the home currency depreciates. © 2001 Elsevier Science Inc. Allrights reserved.

Keywords:Trade balance; Exchange rate; J-curve

The greatest tragedy in all history is the murder of a beautiful theory by a gang of brutalfacts.

(Anonymous)

* Corresponding author. Tel.:165-874-3997; fax: 65-234-4985.E-mail address: [email protected]

Journal of Asian Economics 12 (2001) 47–63

1049-0078/01/$ – see front matter © 2001 Elsevier Science Inc. All rights reserved.PII: 1049-0078(01)00072-0

Page 2: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

1. Introduction

It is part of the folklore of economics, transmitted to generations of students, that a realcurrency depreciation tends to improve the real trade balance (or current account) in thelong-run but that the response of the trade balance to a change in the real exchange ratefollows a ‘J’ tilted to the right. In other words, a real depreciation worsens the trade balancein the short-run as the value of imports measured in home currency rises by more than exportrevenues, but is successful in improving it over the longer run.

The rationale behind this process is rooted in the ‘elasticities’ approach to balance ofpayments adjustment, which focuses on the magnitude of the supply and demand priceelasticities of exports and imports as enshrined in the Bickerdike–Robinson–Metzler (BRM)imperfect substitutes model with the familiar Marshall–Lerner condition as a special case.2

The improvement of the trade balance in the long run is fundamental to the stabilizationpolicies of the International Monetary Fund, including those enacted in the wake of the recentfinancial crises in Asia, even if devaluation/depreciation is not in itself to be regarded as apanacea for restoring external competitiveness and the window of opportunity it opens maybe short-lived.

The J-curve phenomenon has also been invoked to explain the persistence of the U.S.trade deficit following the fall in the U.S. dollar from its peak in 1985. See, for example,Krugman and Baldwin (1987) and Krugman (1989). According to this perspective, thoughthe U.S. price elasticities of export and import demand are clearly large enough to satisfy theMarshall–Lerner condition in the longer term (although they are not very far above unity),the substantial lags in the adjustment of both prices and quantities to exchange rate changesare seen as important contributing factors to the sluggish response of the U.S. deficit.3

The related “pass-through” literature was also, to some extent, stimulated by the apparentresilience of the trade balances of the major industrial countries to changes in their exchangerates and the fact that this could not be explained by elasticity pessimism. The focus of thepass-through approach (See the survey by Menon, 1995) is on the complex relationshipsbetween exchange rate changes and the prices of internationally traded goods and the extentto which exchange rate changes are ultimately reflected in the destination currency prices oftraded goods. Low pass-through would make it possible for trade flows to stay relativelyinsensitive to currency changes even if export and import demand is highly elastic over theshort and long run. The weak transmission of exchange rate changes to prices, together withsubsequent lags in the quantity response to these price changes, as suggested by themainstream J-curve literature, could thus significantly impede overall balance of paymentsadjustment.

By and large the empirical evidence amassed over the last three decades has supported the“elasticity optimists” but also the existence of significant J-curve effects. Testing mostlyrevolves around variants of the two-country imperfect substitutes model to obtain thenecessary estimates of the price elasticities by using structural demand functions. See, forexample, the survey by Goldstein and Khan (1985), Noland (1989) on Japan, and Bahmani–Oskooee (1985) for Greece, India, Korea, and Thailand.

Recent empirical work, however, utilizing the nonstructural partial reduced form approachof Rose and Yellen (1989), which models the real trade balance directly as a function of the

48 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 3: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

real exchange rate and real home and foreign output variables, has found very little evidenceof a reliable long-run relationship between the real trade balance and the real exchange rateand no evidence at all for J-curve effects for bilateral trade flows between countries or inaggregate. See, for example, Rose and Yellen (1989) for the U.S. with respect to her otherG7 partners and in aggregate; Rose (1991) for five OECD countries in aggregate; and Rose(1990) for 30 developing countries annually and 19 quarterly. A key feature of these studies,compared to earlier work, is the application of a time-series econometric methodology thattests for unit roots and cointegration. Typically, unit roots are detected but no cointegration,so estimation proceeds with first differenced data. A variety of estimating techniques areused, including nonparametric estimation and simultaneous equation methods. The applica-tion of a batch of robustness tests has also generally failed to reverse these negative results.4

The purpose of this paper is to examine the relationship between the real trade balance andthe real exchange rate for bilateral trade in merchandise goods between Singapore and theU.S. on a quarterly basis over the period 1970 to 1996 using the partial reduced form modelof Rose and Yellen (1989). As the data in Table 1 suggests, the U.S. accounts for asubstantial proportion of Singapore’s exports over this period, especially “domestic” exports(excluding re-exports). The advantage of the bilateral approach is that it avoids the asym-metric response of trade flows to exchange-rate changes across countries. We also hope toshed further light on what has become known as the “Singapore export puzzle”: theobservation that, despite periods of rapid nominal and real appreciation of the Singaporedollar, export growth in aggregate has remained buoyant. See, for example, Abeysinghe andTan (1998) and Tongzon and Menon (1995). Two specific questions are addressed: is the realexchange rate a significant determinant of the real bilateral trade balance between Singaporeand the U.S.? Does the J-curve exist for Singapore?

Our findings suggest that the real exchange rate does not have a significant impact on thereal bilateral trade balance for Singapore and the U.S. and we can find little evidence of aJ-curve effect. Our results thus generally confirm previous work that finds a weak relation-ship between changes in the exchange rate and changes in export and import prices andvolumes for Singapore.

We begin in Sections 2 and 3 with some background on the relationship between trade and

Table 1Singapore U.S. bilateral trade 1970 to 1996

U.S. as a percent of total Singapore:

Exportsa Domesticexportsb

Totaltradec

1970 11 29 111980 13 20 131990 21 32 181996 18 31 17

Source:Economic Survey of Singapore, Ministry of Trade and Industry, Singapore (1996).a Total merchandise exports.b Total merchandise exports less re-exports.c Total merchandise exports and imports.

49P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 4: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

exchange rates for a small open “reexport” economy such as Singapore, including the J-curvephenomenon. This is followed in Section 4 by the specification of an estimating equationbased on the Rose and Yellen (1989) partial reduced form model. Preliminary tests arecarried out to detect unit roots in the data and evidence of cointegrating relationships betweenthe variables and the estimating equation is then suitably transformed into a differenceformulation to convert the unit root variables into stationary processes. Our empirical resultsare presented in Section 5 where we also carry out some robustness checks. Our key findingsare then brought together in the form of a brief conclusion.

2. The trade balance and real exchange rate for a very open re-export economy

In common with many other countries Singapore moved to a floating exchange rateregime in June, 1973. In September 1975 the Monetary Authority of Singapore (MAS) beganto actively manage the Singapore dollar in relation to an undisclosed basket of the currenciesof her major trading partners motivated, in part, by the desire to stop an appreciation fromnegatively impacting on export competitiveness. Since 1981, however, managed floating hasbeen linked explicitly to a strong dollar policy to neutralize the effects of imported inflation.5

Fig. 1 plots the nominal and real bilateral exchange rate between the Singapore dollar andthe U.S. dollar and Fig. 2 plots the real bilateral trade balance. A rise (fall) in the exchange

Fig. 1. The Singapore Dollar-U.S. dollar nominal and real exchange rate

50 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 5: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

rate signifies a depreciation (appreciation). Table 2 presents some corresponding summarystatistics for the key trade and exchange rate variables.

In terms of annual percentage changes (Table 2) there has been a clear tendency for theSingapore dollar to appreciate over trend, both in nominal and real terms. The onlysignificant period of nominal depreciation was in the early to mid-1980s. In real terms therewas a sharp depreciation between the first quarter of 1975 and the third quarter of 1976, butan extended period of fairly flat behavior in the 1980s until a peak in the first quarter of 1985,followed by a period of sustained appreciation.

Notwithstanding this appreciation, the real bilateral trade balance became positive for thefirst time in 1983, with short-lived negative values in the last quarters of 1983 and 1984, andagain in the first quarter of 1991 (Fig. 2). Real exports, on the other hand, grew very fast inthe 1970s and 1980s at 24 and 18% per annum, respectively, but slowed in the 1990s to about4% (23% 1987–1996). This export performance was very robust with few periods ofnegative change on a quarterly basis and only three times on a calendar year basis (1973,1974, and 1996). Moreover, changes in both the real trade balance and real exports to theU.S. display low correlations with changes in the real exchange rate (measured as the costof a U.S. dollar in Singapore dollars) over the whole period, and are negatively correlated inthe 1990s.

The absence of strong positive coefficients linking real exports and the real trade balanceto the real exchange rate is not uncommon in the exchange rate literature. Meese and Rogoff(1988), for example, found it very difficult to identify variables which exhibited a stable

Fig. 2. The Singapore-U.S. real bilateral trade balance

51P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 6: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

correlation with exchange rates. We test this relationship between the real exchange rate andthe real trade balance more formally below.

A number of explanations have been put forward to explain this failure to establish a closerelationship between exchange rates and the trade balance for Singapore. It is possible thatmovements in the nominal exchange rate are determined largely by purchasing power paritydifferentials rather than by trade flows per se, and that this relationship is reinforced by theregime of managed floating in Singapore as the MAS offsets imported inflation by active orpassive adjustments to the nominal effective exchange rate. Although Abeysinghe and Lee(1992) and Chou and Shih (1997) found some support for the purchasing power parityhypothesis for the Singapore dollar-U.S. dollar relationship, the theory was not confirmed forother currencies relative to the Singapore dollar or when more flexible fractional cointegra-tion tests were applied. U.S.-Singapore bilateral results might also not generalize to effectiveexchange rates. Robinson et al. (1998)6 found little support for either the absolute or therelative version of purchasing power parity for Singapore between 1980 and 1996 using threemeasures of the real effective exchange rate based on the consumer price index, wholesaleprices, and unit labor costs. It appears that the Singapore dollar real effective exchange ratepersistently deviates from a mean-reverting time path.

An alternative explanation focuses on the very high import content of Singapore’s exportsand the strong pass-through effect that changes in import prices have on domestic costs andprices and subsequently on export prices.7 This, together with the MAS policy of managedfloating, may have acted to cushion the translation of exchange rate changes into domesticcosts and prices. In fact, Abeysinghe and Tan (1998) argue that this high import content of

Table 2Exchange rates and the trade balance for Singapore 1970 to 1996

1970–79 1980–89 1990–96

Annual average percentage change

Nominal exchange rate 23.36 21.04 24.57Real exchange rate 21.36 20.38 25.12Real merchandise exports 24.04 18.04 4.16

Correlation coefficient with the real exchange rate (changes)1970–79

Real exports 0.26Real trade balance 0.14

1980–89Real exports 0.24Real trade balance 0.16

1990–96Real exports 20.01Real trade balance 20.33

1970–96Real exports 0.13Real trade balance 0.04

The exchange rate variables are defined as the nominal or real cost of a U.S. dollar measured in Singaporedollars. All real variables are based on 19905 100.

Source: appendix.

52 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 7: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

exports in tandem with the policy of managed floating is the most persuasive explanation ofthe “export puzzle.” Using cointegration techniques and a quarterly error-correction model,they find a close correlation between export prices and import prices (except for food andchemicals), so a Singapore dollar appreciation lowers import prices measured in domesticcurrency and, hence, the domestic price of exports, effectively cushioning the conventionalelasticities relationship between changes in the exchange rate and changes in export vol-umes.8

An alternative, but not mutually exclusive, explanation of the puzzle focuses on thepossibility of a low pass-through from exchange rate changes to export prices as a directconsequence of “pricing to market” practices by multinational corporations, which accountfor a substantial portion of production for export in Singapore. Tongzon and Menon (1995),for example, have calculated the aggregate pass-through coefficient for Singapore as 0.31over the period 1978 to 1993, with a corresponding pricing to market coefficient of 0.69.

This would also be consistent with a view that the trade balance is essentially driven byexternal demand rather than by currency or price factors or that exchange rate induced lossesof competitiveness are compensated for by improvements in productivity as producers learnto anticipate long periods of currency strength. Singapore is not unique in achieving rapidexport growth in the face of persistent nominal and real exchange rate appreciation—Japanmanaged the same feat between 1985 and 1995.

There remains, however, the possibility that the pass-through effects on the prices oftraded goods are complete but that these changes are not translated into volume effectsbecause of low export and import price elasticities of demand, both in the short and long run.We test this relationship between the real exchange rate and the real trade balance explicitlybelow using a partial reduced form equation from the generalized elasticities approach. Butwhat about the J-curve effect?

3. The elasticities approach and the ‘J’ curve phenomenon

The J-curve describes the lag with which a real currency depreciation improves the realtrade balance and can be rationalized in terms of a currency contract effect and/or insuffi-ciently high short-run price elasticities of export and import demand, which results in asluggish response of trade values to changes in relative prices brought about by currencychanges. The currency contract effect arises if import and export orders reflect decisionsmade in advance of the depreciation at the old exchange rate. If imports are invoiced inforeign currency and exports in domestic currency, prices will be sticky in sellers’ currencyso there will be an increase in the value of precontracted imports measured in domesticoutput terms, but the domestic value of exports remains unchanged. Only in the longer runwhen new contracts are signed will the volume of home imports fall and home exports risesufficiently to improve the trade balance so the effect of the depreciation is cumulativelypositive.

The theoretical basis of the ‘J’-curve effect is the classical elasticities approach, which isa partial equilibrium micro-oriented analysis focusing on the effects on import and exportvalues of changes in relative prices brought about by changes in exchange rates, disregarding

53P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 8: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

macro effects on variables such as wages, prices, and economic activity (or implicitlyassuming these effects are neutralized by policy). Although there is no a priori reason whya currency devaluation/depreciation should have any particular effect on the trade balance,a set of necessary and sufficient conditions for an improvement can be derived from thetwo-country imperfect substitutes model (imports and exports are imperfect substitutes fordomestic goods) in the form of the generalized BRM solution in which the outcome dependsprimarily on the relative elasticities of the demand and supply for imports and exports.9 Thereal trade balanceB will improve when the real exchange rateq, defined as the nominalexchange rate (domestic currency price of foreign exchange) multiplied by the ratio of anindex of the prices of foreign produced goods to an index of the prices of domesticallyproduced goods, increases if:

dB/dq5 (1 1 Es) [Ed*/Ed* 1 Es] Dm* rpx 2 Dm q rpx* (1 2 Ed) [Es*/Es* 1 Ed] .0 (1)

WhereDm (Dm*) is the quantity of imports demanded by domestic (foreign) residents,rpx

(rpx*) is the relative price of exportables at home (abroad),Ed (Ed*) are the absolute priceelasticities of demand for home (foreign), andEs (Es*) are the corresponding price elasticitiesof supply.

The sign of Eq. (1) is indeterminate but if trade is initially balanced (B 5 0) andEs andEs* are infinite this reduces to the familiar Marshall–Lerner condition that a real depreciationimprovesB so long as the sum of absoluteEd andEd* exceed unity.

The assumption of infinite supply elasticities implies that the foreign price of imports andthe domestic price of exports are fixed. In this sense, the J-curve is more plausible foradvanced developed countries which invoice predominantly in sellers’ currency, but not forsmall open economies, such as Singapore, which are price-takers in international markets(cannot influence the foreign price of their imports since they are negligible buyers in foreignmarkets, but can sell any volume of exports abroad at a given foreign price) so changes inthe real exchange rate have no perceptible effects on world prices and thus on the foreigncurrency prices of their exports and imports. Hence, if Singapore’s imports and exports areboth predominantly invoiced in foreign currencies such as the U.S. dollar, the value of bothrise during the currency contract period. In this case, both the foreign demand and supplyelasticities will be infinite so the price of imports and exports in foreign currency will beconstant. The condition for a depreciation to improveB becomes:

Es 1 Ed (Dm q rpx*)/(Dm* rpx) . (Dm q rpx* 2 Dm* rpx)/(Dm* rpx). (2)

If there is trade balance initially,Es 1 Ed . 0. In other words, the outcome of a currencydepreciation is always positive for the trade balance so there is no J-curve. The worst casescenario here is when both the elasticity of demand for imports and the elasticity of supplyof exports are zero soEd 5 Es 5 0, but even in this case the trade balance remainsunchanged. This implies that the J-curve effect cannot occur for a small open economyinitially in trade equilibrium.

If, however, there is initially a trade deficit and the short-run supply and demandelasticities in the home country are sufficiently low, Eq. (2) will not hold and a J-curve maybe observed.10 Note that if there is an initial trade surplus Eq. (2) holds regardless of the

54 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 9: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

values of the elasticities. In Singapore’s case, as noted earlier, the bilateral trade balance withthe U.S. has switched over time, so either outcome could be plausible.

Thus, whether the BRM condition holds in the long-run, sodB/dq . 0, and “perverse”J-curve effects are present in the short run, due to currency contract effects or low tradeelasticities, are ultimately empirical questions which can be tested with an appropriate modeland econometric procedure. We shall also analyze export and import data separately to seeif the failure to detect a J-curve in the short-run can be attributed to the small countryassumption.

4. Model specification

Our starting-point is a partial reduced form equation for the real trade balanceB, derivedfrom the two-country imperfect substitutes model,11 which assumes that imports and exportsare imperfect substitutes for domestic goods:

B 5 B (q,y,y*) (3)

According to this formulation,B, measured in terms of units of home output, depends onlyon real domestic income y measured in domestic output terms (negatively), foreign incomey* measured in foreign output terms (positively), and on the real exchange rateq defined as:

q 5 e 3 p*/p (4)

wheree is the domestic currency price of foreign exchange andp (p*) are indexes of theprices of domestically (foreign) produced goods.12

We are interested here in whether a real depreciation will improve the trade balance ifyand y* are held constant, and whether it worsens in the short-run but improves in thelong-run, thus producing the J-curve effect. In Eq. (3)B will improve if y* increases ory fallsbut the effect ofq given bydB/dq is ambiguous. According to the BRM condition discussedin the previous sectiondB/dq will be positive if:

dB/dq 5 (1 1 Es) [Ed*/Ed*1 Es] Dm* rpx 2 Dm q rpx* (1 2 Ed) [Es*/Es* 1 Ed] .0(5)

The special cases discussed above can then be derived from Eq. (5) by making appropriateassumptions about the trade balance and the magnitude of the elasticitiesEd (Ed*) and Es

(Es*).The next step is to formulate an empirical equation to generate estimates of the coeffi-

cients of the contemporaneous and lagged values ofq as a direct measure ofdB/dq.We begin by specifying a log-linear13 dynamic approximation to Eq. (3), and form a

general autoregressive distributed lag (ADL) model for the real trade balanceBt conditionalon the real exchange rateq, real income in domestic output termsy, and real income inforeign output termsy*. Because the trade balance can take on both positive and negativevalues it was left in nonlog form and a constant is added to allow for deterministic drift. Fulldetails of the data sources and construction of the variables are relegated to an appendix.

Bt 5 a0 1 S a1j Bt 2 j 1 S b0i qt 2 i 1 S b1j yt 2 j 1 S b2j y* t 2 j 1 et (6)

55P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 10: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

wherei 5 0 to r with current andr 5 8 lags ofq; j 5 0 to s with current ands 5 4 lags onB, y, andy*; et 5 a well-behaved random error term;a0 5 a constant term.

A priori we expectb1 , 0 due to the inverse relationship between real domestic outputand the real trade balance, andb2 . 0 as an increase in foreign output improvesB. The keyparameter isb0, which represents the sum of the effects of the real exchange rate on the realtrade balance. Becauseq is defined as the domestic (Singapore dollar) currency price of theforeign currency (U.S. dollar), a positive change inq implies a real depreciation of the homecurrency. Consequently,b0 . 0 would satisfy the “generalized Marshall–Lerner condition”and a J-curve effect would be verified only if there are initially negative values ofb0

followed by positive values.The exact dynamic structure of Eq. (3) is unknown but we initially included lags forq of

eight quarters to cover a two year span for the J-curve effects to work themselves out, butonly four quarters forB, y, andy* because the responses here are expected to be quicker. Thisis still somewhat arbitrary but necessary to retain sufficient degrees of freedom in ourestimation. Variations in lag length are tested below under the robustness checks.

As a preliminary step all the variables in Eq. (6) were tested for the presence of unit rootsand cointegration tests were carried out to see if there are any long-run steady-staterelationships between the variables that also contain unit roots. Table 3 lists the results of theunit roots tests onB, q, y, andy*, together with details of the exact test procedure used. Asignificant test statistic rejects the null hypothesis of a unit root in favor of stationarity. Apartfrom y*, and in line with previous studies such as Rose (1990), the null cannot be rejected.

Given the presence of unit root nonstationarity raises the possibility of cointegration.Table 4 reports the results of the Engle–Granger and Johansen and Juselius (1990) cointe-gration tests (excludingy*). There is no evidence of cointegration from the Engle–Grangertests or the maximum eigenvalue and trace tests at the optimum lag based on minimizationof the Akaike Information Criterion. Again, in line with previous studies, there is little

Table 3Testing for unit roots

Variable Test statistic

Real trade balance (B) ADF 21.86(0.68)WS 21.91(0.71)

Real exchange rate (q) ADF 21.63(0.78)WS 21.86(0.74)

Real domestic output (y) ADF 22.68(0.24)WS 21.77(0.79)

Real foreign output (y*) ADF 24.02(0.01)**WS 23.71(0.01)**

All variables exceptB are in logs. The tests are carried out over the period 1970(1) to 1996(2), include aconstant and a time trend and are augmented up to a maximum of 10 lags. The choice of optimum lag for the ADFand WS tests was decided on the basis of minimizing the Akaike Information Criterion. The probability valuesare in parentheses with an asterisk indicating significance at the 5% level, and two asterisks at the 1% level. Thetest statistics and critical values are from the COINT procedure in TSP version 4.3. ADF is the augmentedDickey–Fuller (tau) test with critical values based on MacKinnon (1994). WS is the augmented WeightedSymmetric (tau) test based on Pantula et al. (1994).

56 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 11: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

evidence for the existence of a stable linear steady-state relationship between the levels of thereal trade balance, the real exchange rate, and real domestic output.

The presence of unit roots and absence of cointegration implies that estimation of Eq. (6)in level form would be inappropriate. Consequently, Eq. (6) was transformed into a differ-ence model to convert the unit root variables into stationary processes with time lags tocapture the short-run dynamics. Eq. (7) is the transformed version of Eq. (6) in differenceform in which the first difference of the real trade balance is determined by current andlagged values of the first differences of the logs of the real exchange rate and real domesticand foreign income. No restrictions are imposed by this transform. AD signifies a firstdifference operator:

DBt 5 a0 1 Oj51

3

aljDBt-j 1 Oi50

7

b0iDqt-i 1 Oj50

3

bljDyt-j1 Oj50

3

b2jDy* t-j1et (7)

5. Results and robustness checks

Eq. (7) was estimated using quarterly data from 1972(1) to 1996(1). The results arepresented in Table 5. Of particular interest are the signs and statistical significance of the lagcoefficientsS b0i (standard errors are in parentheses), which measure the sum of the effectsof the change in the real exchange rateq on the real trade balanceB. The sum of the effectsof q on B, y on B, andy* on B are correctly signed, theF-test of the significance of eachvariable under the null that all the coefficients are jointly zero is not rejected in all cases atthe 5% level, although both income variables are significant at 10%. Thus, as with previousstudies, including Rose and Yellen (1989), Rose (1990),14 and Hsing and Savvides (1996),

Table 4Engle–Granger and Johansen–Juselius cointegration tests

Engle–Granger:

Regressand Test statistic

Real trade balance (B) 22.16Real exchange rate (q) 21.18Real domestic output (q) 23.05

Johansen–Juselius:

lmax Trace

H0 r 5 0 18.75 30.83

All variables exceptB are in logs. The tests are carried out over the period 1970(1) to 1996(2), include aconstant and a time trend and are augmented up to a maximum of 10 lags. The choice of optimum lag was decidedon the basis of minimizing the Akaike Information Criterion. An asterisk indicates significance at the 5% level,and two asterisks at the 1% level. For the Engle–Granger procedure the test statistic is the Engle–Granger (tau)from the COINT procedure in TSP version 4.3, with critical values from MacKinnon (1994). The maximumeigenvalue (lmax) and trace tests are from the COINT procedure in TSP version 4.3 and include a finite samplecorrection. Critical values are from Osterwald–Lenum (1992).

57P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 12: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

which use a similar model and econometric methodology, we are unable to reject the null thatcurrent and lagged values of the first difference of the real exchange rate are jointlyinsignificant determinants of the real trade balance at the 5% probability level.

As far as J-curve effects are concerned, despite the correct sign on the sum of the realexchange rate coefficients, inspection of the signs and significance of the individual coeffi-cients for current and lagged values ofq (see Table 6) suggest a cycling between positive andnegative values rather than negative values in earlier quarters followed by positive valueslater as the lag length increases, which would signify the presence of a J-curve.

In order to check the robustness of our results to changes in the estimation procedure ormodel composition, a number of additional estimates were produced. For reasons of brevitythese are not presented here but can be obtained from the authors on request. Initial checkscentered on varying the lag lengths on the explanatory variables. We also changed thestarting date of the regressions to 1975 to omit the turbulent last years of the adjustable pegsystem and the transition to a managed floating regime in Singapore, and to 1982 coincidingwith the change of exchange rate regime to a strong S$ policy. The results were unchangedexcept that the foreign output variable was now no longer significant even at 10%.

A third exercise involved running the equation in logs of levels. Although this is strictlyinappropriate here given the evidence for nonstationarity in the underlying time-series, thiswas done for comparison with previous studies, which use a structural approach and foundevidence to support J-curves, such as Krugman and Baldwin (1987) and Noland (1989). Roseand Yellen (1989) also found some evidence of a significant effect ofq on B when theyre-estimated their partial reduced form model in level form, but found J-curve effects only

Table 5Significance tests for the exchange rate and output coefficients 1972(1) to 1996(1)

(7) Levels

Sb0i 1.51(7.32) 249.10(122)F(8,77) 0.56 0.47Probability 0.81 0.89Sb1j 28.35(6.38) 299.40(117)F(4,77) 2.18 1.93Probability 0.08 0.09Sb2j 21.10(10.20) 428(358)F(4,77) 2.08 2.09Probability 0.09 0.08SE 93.63 93.86DW 1.93 1.94AR 1–5 (5) 2.17 1.01Normality x (2) 7.45* 9.09*Wald x2 (3) 4.82 9.38*

TheF test is a test of the significance of each variable under the null that all the lag coefficients are jointly zero.Standard errors for the sum of the lag coefficients are in parentheses. An asterisk indicates significance at the 5%probability level, and two asterisks at the one percent level. SE is the standard error of the regression, DW is theDurbin–Watson statistic for first-order autocorrelation. AR is the Lagrange multiplier test forrth order autocor-relation, distributed asx2 (r ) under the null that the errors are white noise. The test is for lags 1 to 5. Normalityx is ax2 test for the normality of the residuals. Wald is a test that all the long-run coefficients except the constantare zero.

58 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 13: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

for U.S. trade with Germany and France. Hsing and Savvides (1996) found some cases of asignificant long-run effect ofq on B for Korean and Taiwanese trade with the U.S., Japan,and the rest of the world using the data in level form, but with one exception (Korea’sbilateral trade with the U.S.) the individual coefficients were insignificant and the J-curvephenomenon was not observed at all. For Singapore (Table 5) we could find no evidence ofa significant positive relationship betweenq andB or of J-curve effects (Table 6) in the logsof levels version of our estimating equation.

Our two final robustness checks centered on testing for “small country” effects (discussedabove) and the validity of the assumption of zero homogeneity of the trade balance withrespect to the components of the real exchange rate.

The initially negative response of the trade balance to changes in the exchange rateaccording to the J-curve hypothesis stems essentially from movements in the value ofimports rather than exports. An increase in the value of imports is expected since these aredenominated in foreign currency, but in so far as exports are denominated in domesticcurrency, a depreciation leaves their value in domestic currency unchanged. In the smallcountry case, on the other hand, both exports and imports are denominated in foreigncurrency and would increase in value measured in domestic currency. Thus the need to treatexports and imports as separate regressors and test whether exports increase in the short-runso part of the reason we fail to find a J-curve is because the country is small.

Accordingly, we ran further regressions in differences for the log of real exports onq andy* and the log of real imports onq andy. Although the sum of the lag coefficients on the realexchange rate variable was positive and significant in the export equation (Table 6), the signsfor the individual coefficients indicative of small country behavior cycled after the second lagand the expected positive sign for the import coefficients was true only for the current timeperiod. Thus, although it is possible that the positiveq coefficients in the real export equationare a reflection of pricing by exporters in U.S. dollars, it is not clear that this is maskingJ-curve effects from an initial rise in import values.

As far as the zero homogeneity restriction is concerned, we re-estimated the modelallowing for both short-run and long-run nonhomogeneity of the real exchange rate byincluding separatelyD log (p*/p) and D log e, wheree is the domestic currency price of

Table 6Signs and significance of the real exchange rate coefficients

Lag 0 1 2 3 4 5 6 7 8 Sb0i F

(7) 21.62 3.42 24.47 7.40 23.91 0.32 2.35 21.99 — 1.51 0.56(0.43) (0.85) (1.12) (1.80) (0.92) (0.08) (0.53) (0.48)

Levels 2226 516 2745 1130 2965 211 336 2535 229 249.1 0.47(0.58) (0.80) (1.10) (1.70) (1.40) (0.31) (0.47) (0.76) (0.60)

Exports 0.59 0.56 0.08 20.52 0.19 0.72 20.86 0.85 — 1.61 2.95**(1.62) (1.50) (0.21) (1.45) (0.55) (2.05)* (2.29)* (2.32)* —

Imports 0.83 20.70 0.57 20.66 20.26 0.12 20.73 0.57 — 20.26 1.33(1.94) (1.55) (1.27) (1.44) (0.55) (0.26) (1.55) (1.29)

The t values are in parentheses beneath each coefficient. TheF test is a test of the significance of each variableunder the null that all the lag coefficients are jointly zero. An asterisk indicates significance at the 5% probabilitylevel, and 2 asterisks at the 1% level.

59P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 14: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

foreign exchange, andp*/p is the price differential fromq 5 e (p*/p). A priori, a rise in eandp*/p should be associated with an improvement inB. As with Rose and Yellen (1989),we found the data to be consistent with the homogeneity assumption that the current andlagged values of both price differentials and the nominal exchange rate (in logged differ-ences) are jointly insignificant in the trade balance equation. Neither the nominal exchangerate not the relative price variables were significant overall and only one of the 16 individualcoefficients (for the price variable at lag 2) was significant at the 5% level.

6. Conclusion

The purpose of this paper has been to examine the relationship between the real tradebalance and the real exchange rate for bilateral trade in merchandise goods betweenSingapore and the U.S. over the period 1970 to 1996 using the partial reduced form modelof Rose and Yellen (1989), derived from the two-country imperfect substitutes model.Singapore is particularly interesting in this respect since, despite periods of rapid nominaland real appreciation of the Singapore dollar, export growth in aggregate has remainedbuoyant.

Our findings suggest that the real exchange rate does not have a significant impact on thereal bilateral trade balance for Singapore and the U.S. and we can find little evidence of aJ-curve effect. Although positive coefficients linking real exports with lagged values of thereal exchange rate might be indicative of “small country” pricing by exporters in U.S. dollars,it is not clear that this is masking J-curve effects from an initial rise in import values as thehome currency depreciated. Our results thus generally confirm previous work that finds aweak relationship between changes in the exchange rate and changes in export and importprices and volumes for Singapore.

Notes

1. We would like to thank Tilak Abeysinghe and an anonymous referee for some helpfulcomments on a first draft and Colin Kirkpatrick and the Staff at IDPM Manchester forputting up with me during my Study Leave in August 1999.

2. See, for example, Goldstein and Khan (1985) and Argy (1994).3. The rapid improvement in the Mexican trade balance with the U.S. following the

sharp depreciation of the peso against the dollar in December 1994 is not seen to beinconsistent with this view. The absence of any obvious J-curve complications isexplained by Krugman and Obsfeld (1997, p. 470–1) in terms of the massive size ofthe devaluation, the financial crisis which accompanied the devaluation leaving nodoubt that the change was permanent, so exporters and importers adjusted rapidly, theimprovement in the trade balance brought about by a cut in domestic spending, andprior trade liberalization, which made import and export volumes more sensitive torelative price changes.

4. A rare exception is Hsing and Savvides (1996), who found some evidence of a J-curve

60 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 15: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

for bilateral quarterly trade between Korea and the U.S. when a polynomial distrib-uted Almon lag structure was imposed on the exchange-rate coefficients and themodel is estimated in level form, even though pre-testing confirmed the presence ofunit roots in the data. However, when a less restricted distributed lag structure isimposed and the data are differenced, there is no evidence of J-curves for eithercountry.

5. For some background on the foreign exchange regime in Singapore over this period,see Peebles and Wilson (1996, ch. 6).

6. This study was done by the authors in their personal capacities and should not beattributed to the MAS.

7. This exceptionally high import content of exports led Lloyd and Sandilands (1986) todub Singapore as a “very open reexport economy.” The term “reexport economy”refers here not so much to traditional entrepot exports, which have declined inimportance for Singapore in recent decades, but to the high proportion of importedintermediate inputs, which find their way into Singapore’s manufactured exports.

8. Earlier evidence for this effect can be found in Low (1994). On the basis ofsimulations using the MAS (unpublished) quarterly model for Singapore, he foundthat a one percentage point rise in import prices led to an equi-proportional increasein wholesale prices and a 0.7% increase in the consumer price index within 2 years.This is precisely why targeting the nominal effective exchange rate is regarded by theMAS as the most effective use of monetary policy to achieve low and stable inflationdespite the short-run negative effects an appreciating currency might have on exportcompetitiveness.

9. See, for example, Argy (1994, ch. 12).10. The intuitive reason is that ifEs 5 Ed 5 0, a 10% devaluation will increase the import

bill in domestic currency by 10% and export receipts will also rise by the sameamount. Because imports exceed exports initially the absolute deficit increases. Thismay, however, be partly or more than offset by some import response and someexport response.

11. See, for example, Goldstein and Khan (1985, pp. 1044–50) and Argy (1994, ch. 12).12. The advantage of the partial reduced form approach is that it is easier than obtaining

the structural form (volume and pass-through) parameters by estimating a set ofstructural supply and demand equations for exports and imports. Because the focushere is on the effects on the net trade balance, it is simpler to get the estimates directlyfrom Eq. (3) and then test for the restriction of homogeneity of degree zero inp, p*,ande using one equation.

13. There is no particular reason to take logs of the right-hand side variables, but becausethis has been the practice in previous work, it is done here for comparison.

14. In the Rose (1990) case, out of a sample of 19 developing countries for whichquarterly data existed, nine had a (wrong) negative sign on the real exchange-ratevariable, but Singapore was the only country for which this was also significant at thefive percent level.

61P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 16: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

Appendix. Data sources and manipulation

Most of the data came from the International Monetary FundInternational FinancialStatistics(IFS) from the first quarter of 1970 to the fourth quarter of 1996 and was extractedvia Dbank Release 97.1 online at the National University of Singapore, or from CD-Rom inthe Central Library at the University of Warwick. The exactIFS series selected is indicatedbelow in parentheses after each variable. All real variables are based on 19905 100. Thestatistical and econometric work was carried out by using Doornik and Hendry’s (1994)PcFiml (8.0) and PcGive (8.0), and Time Series Processor (TSP) version 4.3 (see Hall, 1995).

1. Real trade balanceB

Monthly bilateral imports and exports for Singapore to the U.S. in U.S. dollars were takenfrom the OECD, Monthly Statistics of Foreign Trade(Paris: OECD), converted into quar-terly figures and into domestic currency by using the market exchange rate (IFS, rf) anddeflated by the home country’s consumer price index (IFS, 64). The real trade balancemeasured in real home currency units was then calculated as the net of the two series.

2. The real bilateral exchange rateq

The bilateral nominal exchange rate (IFS, rf) for Singapore with respect to the U.S. dollarwas converted into a real exchange rate by multiplying the nominal rate by the ratio of theforeign price level to the domestic price level. The domestic price level was the consumerprice index (IFS, 64) and the foreign price level was proxied by the wholesale price index(IFS, 63).

3. Real incomey, y*

The U.S. index of real industrial production (IFS, 66c) was used to obtain real foreignincome in foreign output terms. For Singapore real home income in domestic output termswas analogously defined by a manufacturing production index (IFS, 66ey).

References

Abeysinghe, T., & Lee, C. (1992). Singapore’s strong dollar policy and purchasing power parity.The SingaporeEconomic Review, 37,70–78.

Abeysinghe, T., & Yeok, T. L. (1998). Exchange rate appreciation and export competitiveness: the case ofSingapore.Applied Economics, 30,51–5.

Argy, V. (1994).International macroeconomics: theory and policy. New York: Routledge.Bahmani–Oskooee, M.(1985). Devaluation and the J-curve: some evidence from LDCs.The Review of Economics

and Statistics, 67,500–4.Chou, W. L., & Shih, Y. C. (1997). Long-run purchasing power parity and long-term memory: evidence from

Asian newly industrialised countries.Applied Economics Letters, 4,575–578.Doornik, J. A., & Hendry, D. F. (1994).PcFiml 8.0: interactive econometric modelling of dynamic systems.

Oxford: University of Oxford Institute of Economics and Statistics.

62 P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63

Page 17: Exchange rates and the trade balance: the case of Singapore 1970 to 1996

Doornik, J. A., & Hendry, D. F. (1994).PcGive 8.0: an interactive econometric modelling system. Oxford:University of Oxford Institute of Economics and Statistics.

Goldstein, M., & Khan, M. S. (1985). Income and price effects in foreign trade. In R.W. Jones and P. B. Kenen(Eds.).Handbook of international economics(pp. 1041–1105). Amsterdam: North Holland.

Hall, B. (1995).Time-series processor version 4.3. Palo Alto, CA: TSP International.Hsing, H.-M., & Savvides, A. (1996). Does a J-curve exist for Korea and Taiwan?Open Economies Review, 7,

127–145.Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration.Oxford

Bulletin of Economics and Statistics, 52,162–210.Krugman, P. (1989). The J-curve, the fire sale, and the hard landing.The American Economic Review, 79,31–35.Krugman, P., & Baldwin, R. E. (1987).The Persistence of the US Trade Deficit. Brookings Papers on Economic

Activity, 1–2, 1–43.Krugman, P., & Obsfeld, M. (1997).International Economics. Reading, MA: Addison-Wesley.Lloyd, P., & Sandilands, R. (1986). The trade sector in a very open reexport economy. In L. C. Yah and P. Lloyd

(Eds.).Singapore resources and growth(pp. 182–219). Singapore: Oxford University Press.Low, V. (1994). The MAS model: structure and some policy simulations. In A. Chin and N. Kee Jin (Eds.).

Outlook for the Singapore economy(pp. 19–32). Singapore: Trans Global Publishing.MacKinnon, J. G. (1994). Approximate asymptotic distribution functions for unit-root and cointegration tests.

Journal of Business and Economic Statistics, 2,167–176.Meese, R. A., & Rogoff, K. (1988). Was it real?Journal of Finance, 43,933–948.Menon, J. (1995). Exchange rate pass-through.Journal of Economic Surveys, 9,197–231.Noland, M. (1989). Japanese trade elasticities and the J-curve.The Review of Economics and Statistics, 71,

175–179.Osterwald–Lenum, M. (1992). Practitioners’ corner: a note with quintiles for the asymptotic distribution of the

maximum likelihood cointegration rank test statistic.Oxford Bulletin of Economics and Statistics, 3,461–471.Pantula, S. G., Gonzalez–Farias, G., & Fuller, W. A. (1994) A comparison of unit-root test criteria.Journal of

Business and Economic Statistics, 4,449–459.Peebles, G., & Wilson, P. (1996).The Singapore Economy. Cheltenham: Edward Elgar.Phillips, P. C. B., & Perron P. (1988). Testing for a unit root in time-series regression.Biometrika, 75,335–346.Robinson, E., Surendran, T., Wang, Y., & Foo, L. E. (1998).What lies behind Singapore’s real exchange rate?

An empirical analysis of the purchasing power parity hypothesis. Singapore: Monetary Authority of Singa-pore.

Rose, A. K. (1990). Exchange rates and the trade balance.Economics Letters, 34,271–275.Rose, A. K. (1991). The role of exchange rates in a popular model of international trade: does the Marshall-Lerner

condition hold?Journal of International Economics, 30,301–316.Rose, A. K., & Yellen, J. L. (1989). Is there a J-curve?Journal of Monetary Economics, 24,53–68.Sargan, J. D. (1964). Wages and prices in the United Kingdom: a study in econometric methodology. In P. E.

Hart, G. Mills, & J. K. Whitaker (Eds.), Econometric analysis for national economic planning. London:Butterworth, 22–54.

Tongzon, J. L., & Menon, J. (1995). Exchange rates and export pricing in a small open NIC: the Singaporeanexperience.The Singapore Economic Review, 38,201–211.

63P. Wilson and K.C. Tat / Journal of Asian Economics 12 (2001) 47–63


Recommended