Nominal Exchange Rate Misalignment: Is It Particularly Important to Agricultural Trade?

Transcription

1 Agribusiness & Applied Economics Repor No. 516 Augus 2003 Nominal Exchange Rae Misalignmen: Is I Paricularly Imporan o Agriculural Trade? Gue Dae Cho MinKyoung Kim Edwin Sun Hyun Jin Won W. Koo Cener for Agriculural Policy and Trade Sudies Deparmen of Agribusiness and Applied Economics Norh Dakoa Sae Universiy Fargo, Norh Dakoa

2 ACKNOWLEDGMENTS The auhors exend appreciaion o Dr. Rober Hearne, Mr. Richard Taylor, and Mr. Jeremy Mason for heir consrucive commens and suggesions. Special hanks go o Ms. Beh Ambrosio, who helped o prepare he manuscrip. The research was conduced under he U.S. agriculural policy and rade research program funded by he U.S. Deparmen of Homeland Securiy/U.S. Cusoms and Border Proecion Service (Gran No. TC G, ND1301). We would be happy o provide a single copy of his publicaion free of charge. You can address your inquiry o: Beh Ambrosio, Cener for Agriculural Policy and Trade Sudies, Deparmen of Agribusiness & Applied Economics, Norh Dakoa Sae Universiy, P.O. Box 5636, Fargo, ND, , Ph , Fax , beh.ambrosio@ndsu.nodak.edu. This publicaion is also available elecronically a his web sie: hp://agecon.lib.umn.edu/. NDSU is an equal opporuniy insiuion. NOTICE: The analyses and views repored in his paper are hose of he auhor(s). They are no necessarily endorsed by he Deparmen of Agribusiness and Applied Economics or by Norh Dakoa Sae Universiy. Norh Dakoa Sae Universiy is commied o he policy ha all persons shall have equal access o is programs, and employmen wihou regard o race, color, creed, religion, naional origin, sex, age, marial saus, disabiliy, public assisance saus, veeran saus, or sexual orienaion. Informaion on oher iles in his series may be obained from: Deparmen of Agribusiness and Applied Economics, Norh Dakoa Sae Universiy, P.O. Box 5636, Fargo, ND Telephone: , Fax: , or cjensen@ndsuex.nodak.edu. Copyrigh 2003 by Gue Dae Cho, MinKyoung Kim, Edwin Sun, Hyun Jin, and Won W. Koo. All righs reserved. Readers may make verbaim copies of his documen for noncommercial purposes by any means, provided ha his copyrigh noice appears on all such copies. ii

3 TABLE OF CONTENTS Lis of Tables...ii Lis of Figures...ii Absrac... iii Highlighs...iv Inroducion...1 Exchange Rae Misalignmen...3 Relaive Movemen of Exchange Rae Misalignmen...4 Esimaion Model and Daa Sources...5 Variable Consrucion and Daa...6 Esimaion Resuls...7 Furher Invesigaion...11 Conclusion...13 References...14

4 LIST OF TABLES No. Page 1 Esimaion Resuls Esimaion Resuls...12 LIST OF FIGURES No. Page 1 Movemens of Misalignmens in Comparison o he German Mark...5 ii

5 Absrac This paper examines wheher exchange rae misalignmen negaively affecs agriculural rade, compared o oher indusry secors. Nominal exchange rae misalignmen is obained from he percenage deviaion of real exchange raes from heir long-run equilibrium based on he heory of purchasing power pariy. In order o explore his issue, a bilaeral rade marix involving rade flows beween 10 developed counries is consruced. Using panel daa analysis, a graviy model is esimaed for 4 indusry secors over he period The sudy finds ha over-valuaion (under-valuaion) of he nominal exchange rae negaively (posiively) affecs expor performance of he agriculural secor in paricular. In he large-scale manufacuring secors considered in his paper, expors are no significanly affeced by exchange rae misalignmen. Keywords: exchange rae misalignmen, agriculural rade, and graviy model. iii

6 Highlighs The effec of exchange rae misalignmens on inernaional commodiy rade has been one of he mos imporan issues in inernaional economics. Many empirical observaions sugges he exisence of subsanial and persisen exchange rae misalignmen among developed counries under he floaing exchange rae sysem. Agriculural producs are mosly homogeneous and perishable. Due o hese special characerisics, we expec ha he effec of misalignmen on agriculural rade is larger han ha on large-scale manufacuring secors. Relevan lieraure also predics an asymmeric effec of exchange rae shock on differen indusry secors. Using a graviy model, his paper addresses he effec of exchange rae misalignmen on agriculural rade and compares he impac o ha in oher secors. By employing a panel daa se, we can efficienly esimae boh ime-series and cross-secional difference of exchange rae misalignmen on inernaional rade. The daa used in his sudy are comprised of bilaeral rade flows for 10 developed counries during he period of Four differen indusry secors, including he agriculural secor, are considered and heir empirical resuls are compared. The sudy finds ha over-valuaion (under-valuaion) of he nominal exchange rae negaively (posiively) affecs expor performance of he agriculural secor in paricular. In he large-scale manufacuring secors considered in his paper, expors are no significanly affeced by exchange rae misalignmen. In fac, nominal exchange raes have followed heir long-run equilibrium pah so ha hey are cyclically misaligned a bes. Therefore, unlike a shor-run effec, he major problem for inernaional agriculural rade caused by exchange rae movemen is insabiliy in he long-run. Cyclical booms and depressions of agriculural expors by counries could possibly increase resource-wase wihin he agriculural secor, resuling in larger dead-weigh coss compared o oher large-scale indusrial secors. iv

7 Nominal Exchange Rae Misalignmen: Is I Paricularly Imporan o Agriculural Trade? Gue Dae Cho, MinKyoung Kim, Edwin Sun, Hyun Jin, and Won W. Koo * INTRODUCTION The effec of exchange rae misalignmens on inernaional commodiy rade has been one of he mos imporan issues in inernaional economics. Many empirical observaions sugges he exisence of subsanial and persisen exchange rae misalignmen 1 among developed counries under he floaing exchange rae sysem (e.g., Dornbusch, 1987; Rogoff, 1996; Frankel, 1996). Alhough many agriculural economiss (e.g., Gardner, 1981; Tweeen, 1989) have, in fac, invesigaed he poenial impac of misalignmen on inernaional agriculural rade, here are sill wo remaining quesions relaed o his opic. The firs quesion is how o examine he effec of relaive movemens of exchange rae misalignmens on inernaional rade. For example, when a counry s currency is overvalued compared o pas periods, i does no necessarily negaively affec he counry s expors. If he overvaluaion is relaively less han ha of all compeiors, i is even possible o find a posiive correlaion. According o he heoreical model suggesed by Wolak and Kolsad (1991), he relaive movemen of impor price is one of he imporan facors affecing he choice of an imporer. Because exchange rae is one of he mos imporan facors deermining relaive impor prices, relaive exchange rae movemens should be reaed as imporan facors influencing impors/expors of a counry. A ime-series approach, however, does no incorporae he relaive movemens in he exchange rae; his is a serious omission ha could resul in biased and misleading esimaions. Commodiy rade, especially agriculural rade, is highly volaile due o weaher and poliical condiions in imporing and/or exporing counries. The esimaed coefficiens of he model wih pure cross-secional daa for a paricular year may no provide accurae informaion in evaluaing rade flows of a commodiy (Koo e al., 1994). Therefore, combining ime-series and cross-secional daa is he mos comprehensive mehod o resolve his issue. A second quesion concerns he imporance of exchange rae misalignmen in differen secors of an economy. Is i more vial o inernaional agriculural rade han o he manufacuring secor? If exchange rae misalignmen has he same effec on all indusry secors in a counry, here is no paricular reason we should rea he agriculural secor separaely. However, here are some reasons o believe ha misalignmen affecs inernaional agriculural rade more seriously. For insance, he recen lieraure relaed o sraegic behavior of firms responding o exchange rae shock suggess ha raders sraegically decide heir expor price * Research Assisan Professors, and Professor and Direcor of he Cener for Agriculural Policy and Trade Sudies, Norh Dakoa Sae Universiy, Fargo. 1 Misalignmen, in general, refers o he deparure of nominal exchange raes from long-run equilibrium level or marke fundamenals such as relaive prices and ineres rae differenials beween counries.

8 based upon exchange rae movemens (Goldberg and Kneer, 1997). I has been generally observed ha, when here has been subsanial over- (under-) valuaion of an exporer s currency, exporers do no fully increase (decrease) heir expor price, denominaed in he desinaion counry s currency. 2 In shor, wo heoreical models explaining he moivaions behind exporers sraegic behavior are considered in he relevan lieraure. Baldwin (1988) suggess a supply-side explanaion. Assuming he exisence of an irreversible iniial sunk cos o ener a foreign marke, he shows ha because of he iniial invesmen cos, even a subsanial over-valuaion of an exporer s currency does no in iself compel firms o leave he desinaion marke. If hey decide o say, firms adjus heir price based on he condiion of marke compeiion in he desinaion counry. 3 Froo and Klemperer (1989) emphasize he exporing firm s moivaion o keep heir marke share in a desinaion marke. In he case of a durable good, hey argue ha consumers face subsanial coss in order o swich beween brands of a produc, even if he brands are funcionally idenical. Due o he swiching cos, he curren marke share is an imporan deerminan of he fuure marke share. Exporers are aware of his consumer behavior and, in order o keep heir marke share, hey do no change he produc price as i is denominaed in he desinaion marke currency even when here is an overvaluaion of exporers currency. Many empirical sudies invesigaing he manufacuring secor (e.g., Kneer, 1993) have found exporing firms engaging in price adjusmen behavior. However, i is imporan o noe ha hese cases are no well-fied o explain agriculural rade. This is because agriculural goods are mosly subsiuable and ofen non-durable; moreover, he iniial invesmen cos o ener a foreign marke is no as imporan as i is in he indusrial secor. Therefore, compared o manufacured goods, here is a srong possibiliy ha agriculural rade is more sensiively responsive o exchange rae movemen. Alhough hese are imporan issues, here are few empirical sudies which examine hem. Excepions are Bergsrand (1985, 1989), Koo and Karemera (1991), and Koo e al. (1994). In his earlier papers, Bergsrand invesigaed his issue using a generalized graviy model wih cross-secional daa. His empirical resuls, however, were no consisen wih economic inuiion. Alhough he found weak evidence ha relaive exchange rae movemens have an imporan role in explaining rade flow in he case of he oal rade, resuls were mixed in he case of secoral rade. In 9 ou of 36 cases, he found he expeced sign of he variable; however, in 12 ou of 36 cases, he found exacly he reverse sign of he esimaed coefficien. One of he poenial drawbacks of his empirical research was ha he used relaive nominal exchange rae movemens, which are no economically meaningful indicaors under he floaing exchange rae sysem. 2 This is called incomplee or parial pass-hrough of exchange raes or pricing o marke (PTM) behavior of exporing firms in he relevan lieraure. 3 Dornbusch (1987) also shows ha he price adjusmen of exporing firms should depend on a leas hree facors: he degree of marke concenraion, he relaive marke shares of domesic and foreign firms, and, mos imporanly, subsiuabiliy of he produc. 2

9 Using a graviy model, his paper addresses he effec of exchange rae misalignmen on agriculural rade and compares is impac o ha in oher manufacuring secors. By employing a panel daa se, we can efficienly esimae boh ime-series and cross-secional differences of exchange rae misalignmen in inernaional rade. The daa used in his sudy are comprised of bilaeral rade flows for 10 developed counries during he period of Four differen indusry secors, including he agriculural secor, are considered and heir empirical resuls are compared. The paper is organized as follows. Secion 2 oulines he relevan aspecs of exchange rae misalignmen, and Secion 3 presens a discussion of relaive exchange rae misalignmen. In Secion 4, variable consrucion and daa are discussed, while in Secion 5, he economeric specificaion and resuls are repored. The principal resuls are summarized in Secion 6. EXCHANGE RATE MISALIGNMENT Exchange rae misalignmen can be defined as he deparure of he nominal exchange rae from is long-run equilibrium level, where misalignmen can be characerized as eiher over- or under-valuaion of he currency relaive o fundamenals. Measuring misalignmen is difficul and inherenly imprecise, as i requires esimaion of wha is ermed he fundamenal equilibrium exchange rae. Typically in he lieraure, i is assumed ha purchasing power pariy (PPP) is he long-run equilibrium condiion of nominal exchange raes. 4 Essenially, PPP should hold because exchange raes equalize relaive price levels in differen counries. The sandard expression for absolue PPP is: * s = p p, (1) where s is he home currency price of a foreign currency, p is he domesic currency price of a * paricular good(s), p is he foreign currency price of he good(s), and lower case leers denoe logarihmic values. The implicaion of (1) is ha rade of goods will resul in idenical prices across counries. Allowing for facors such as ranspor coss, PPP in is relaive form implies ha a sable price differenial should exis for he same good(s) selling in differen counries, implying ha real exchange raes beween counries should be consan in he long-run, and consequenly here is no misalignmen of exchange raes from relaive PPP, i.e., he real exchange rae should be mean-revering (MacDonald, 1999). In more recen research, he focus has been on he use of co-inegraion mehods applied o he following equaion: * s = β + α 0 p + α1 p + ϕ (2) * If s, p, p are inegraed of order one, I(1), hen a weak form of PPP exiss if he residual erm from esimaion of (2) is saionary, I(0). A sronger form of PPP exiss if homogeneiy is saisfied, i.e., α = 0 1 and α = 1. Using his ype of approach, several early sudies found no 1 4 Alhough PPP is ypically used as he concep agains which o gauge misalignmen, i is no he only measure. There have been more formal aemps o measure he equilibrium exchange rae based on an explici characerizaion of fundamenals. 3

10 evidence of significan mean-reversion of exchange raes oward PPP (Mark, 1990; Fisher and Park, 1991). However, several auhors have argued ha he daa period for he recen floa is oo shor o have any confidence in he power of saisical ess for saionariy of real exchange raes (Lohian and Taylor, 1997). As a consequence, recen research has been based on eiher long-erm pre-floa daa (Lohian and Taylor, 1996) or muli-counry panel daa (Flood and Taylor, 1996; Frankel and Rose, 1996). This more recen evidence rejecs he random walk hypohesis of real exchange raes. Essenially, real exchange raes rever o equilibrium values over he long-run, and, correspondingly, nominal exchange raes and relaive inflaion raes beween wo counries converge. This revives he noion ha PPP is a long-run equilibrium condiion of nominal exchange raes (MacDonald, 1999), alhough consensus esimaes sugges ha he speed of convergence o PPP is very slow, he deviaions appearing o dampen ou a a rae of roughly 15 percen per year (Rogoff, 1996). Specifically, if long-run PPP holds, as shown in (2), he nominal exchange * rae, s = β + α 0 p + α1 p + ϕ, α 0= 1 and α 1= -1, and he underlying innovaion, ϕ, should be a saionary process, which has mean zero and finie long-run variance, σ. The ime-series movemen of he esimaed residuals, ϕˆ, can be hough of as he ime-series movemen of misalignmen. Furhermore, under he assumpion of long-run PPP, we can also express he * equaion as s p + p = β + ϕ. The lef-hand side is simply he log of he real exchange rae, r, so ha i can also be expressed as r = β + ϕ. Since E r ) ( 2 ϕ = β, we can use he movemen of ϕ ( = r β ) as a proxy variable represening movemens for misalignmen. RELATIVE MOVEMENT OF EXCHANGE RATE MISALIGNMENT Figure 1 presens he movemens in four differen misalignmens (franc/dm (FRA), guilder/dm (NET), pound/dm (UK), and dollar/dm (U.S.)) calculaed based on he PPP. Percenage deviaions of exchange raes from heir sample average are calculaed for comparison. By using percenage deviaions from he equilibrium exchange raes, we can normalize differen currency unis and compare movemens of relaive misalignmens wih a unified measuremen. In he figure, we can easily observe ha Germany has faced differen degrees of misalignmen wih each rading parner during he sample period. For insance, in he mid-1980s, he German Mark was undervalued compared o he U.S. dollar, which was expeced o negaively affec U.S. agriculural expors o Germany. The U.S. dollar sared o rever o is equilibrium level afer However, for a German imporer, he U.S. dollar was sill highly overvalued compared o oher rading parners. Even when he U.S. dollar weakened in comparison o he previous period, U.S. agriculural expors o Germany could possibly decrease furher, based on he exchange rae movemens of oher compeiors in he German marke. Anoher example is he movemen in misalignmen beween Germany and he Neherlands. These counries faced a relaively small misalignmen problem during he sample period. However, sable exchange rae movemen beween hese counries alone canno eliminae exchange rae effecs on heir bilaeral rade. For insance, during he mid-1970s, he 4

11 U.S. dollar was undervalued in comparison o he German Mark. In his case, German raders impored producs from he Unied Saes raher han from he Neherlands. Alhough here was no misalignmen problem beween he Germany and he Neherlands, relaive misalignmen among all compeiors srongly affeced rade flows beween counries. If we were o only consider he movemens of misalignmen beween Germany and he Neherlands using ime series analysis, he resuls migh be misleading. Figure 1. Movemens of Misalignmens in Comparison o he German Mark FRA NET UK US ESTIMATION MODEL AND DATA SOURCES The basic economeric specificaion used in his analysis is similar o hose deailed in recen sudies (Rose, 2000; Glick and Rose, 2001; Pakko and Wall, 2001). However, because our focus is on he level of movemen in misalignmen and secor rade, raher han variabiliy 5, i is proper o employ a graviy model used by Bergsrand (1985, 1989) and Feensra e al. (2001). The basic specificaion of he model is 5 In he usual graviy model, we canno include he level of movemen of misalignmen because he dependen variable is he sum or produc of expors (or impors) beween counries. 5

12 ln EX k ij = α + α M 0 1 ij Y j Y i + α Yj α Yi α α POP j POP 2 ln + 3 ln + 4 ln + 5 ln i + α ln( DIST) + α BORDER + α LANG + α EU 6 ij 7 ij 8 ij 9 ij + µ ij + δ + η ij (3) k where EX ij is he real expor value of counry i o counry j in year for secor k, and where k refers o specific expor secors: 1=oal expors, 2=machinery, 3=chemicals, 4=oher k manufacuring, and 5=agriculure. M ij is he proxy variable for he level of misalignmen, which is calculaed as he percenage deviaion of real exchange raes from heir sample mean ( M ij= ln Rij - ln Rij ). Because a posiive (negaive) value means over-valuaion (undervaluaion) of an exporer s currency by daa consrucion, he expeced sign is negaive. Y i and Y j are he annual real U.S. dollar value income of exporing counry i and imporing counry j, respecively, over he sample period. The variable is expeced o include boh demand- and Y Y i j supply-side effecs on bilaeral rade, so ha he expeced signs are posiive. and POP POP represen real per capia income (U.S. dollar value) of he exporing and imporing counries. DIST ij is disance beween he exporing and imporing counries. BORDER ij is a dummy variable which idenifies a common border effec (if here exiss a common border, he variable is 1; oherwise, i is 0). LANG ij is a dummy variable which idenifies a common language effec (if here is a common language, he variable is 1; oherwise, i is 0). The oher dummy variable, EU, is also included o accoun for rade beween members of he European Union (EU), as we would expec membership in a cusoms union o have a posiive impac on bilaeral rade. i j Variable Consrucion and Daa k This sudy uses annual daa from 1974 o The variable EX ij is he real expor value of counry i o counry j in year for secor k, which refers o a specific expor secor and is calculaed in erms of he U.S. dollar and deflaed using he U.S. consumer price index. The variable is consruced as follows: using he OECD bilaeral rade daa se aken from Trade in Commodiies classified by one-digi sandard inernaional rade code (SITC), we collec nominal expor values in U.S. dollars from i o j for each secor k. This is deflaed by he consumer price index in he Unied Saes ( =100) from he Bureau of Labor Saisics (BLS). The secors considered in his sudy are: food and live animals (SITC 0: agriculure), chemical and relaed producs (SITC 5: chemical), manufacured goods classified chiefly by maerial (SITC 6: manufacuring), and machinery and ranspor equipmen (SITC 7: machinery). The variable M ij is he measure of exchange rae misalignmen beween expor counry i and impor counry j a ime. The variable is consruced as follows. Firs, U.S. dollar-based real exchange raes, which are consruced from nominal exchange rae daa from he Inernaional Moneary Fund (IMF) series and deflaed by a U.S./home counry consumer price index (normalized 1990=100), were obained from he Economic Research Service (ERS) of he 6

13 U.S. Deparmen of Agriculure (USDA). Bilaeral real exchange raes beween exporing and imporing counries are based on aking he U.S. dollar-based real exchange rae for he imporing counry j and dividing by he U.S. dollar-based real exchange rae for he exporing counry i, giving he cross-rae R ij. The measures of misalignmen are based on heory and recen empirical evidence of PPP, which sugges ha he real exchange raes among developed counries are mean-revering. Therefore, deviaion of real exchange raes from heir sample averages could be reaed as a measuremen of movemen in misalignmens. For each pair of real exchange raes, we calculaed heir sample averages, ln Rij, and hen calculaed he percenage deviaion of real exchange raes from heir sample averages ( M ij = ln Rij - ln Rij ), which we rea as measures of misalignmen. The advanage of his measure is ha by using a percenage deviaion, we have a unified measure wih which o examine he effec of relaive movemens of misalignmen on inernaional rade. The gross domesic producs and per capia domesic producs daa for each counry are given in heir nominal value in U.S. dollars from he World Economic Oulook Daabase (IMF, 2001), and are deflaed by he U.S. consumer price index ( =100). Finally, he disance daa beween counries are obained from Rose s daa se. 6 Given he sample of 10 counries, here is a cross-secion of 90 bilaeral rade flows (10 x 9), wih annual daa covering 26 years ( ) for each rade flow, generaing a complee panel of 2340 observaions (90 x 26) for each secor k. ESTIMATION RESULTS Alhough he choice of a proper economeric specificaion o esimae he graviy model wih panel daa is no simple, as indicaed by Mayas (1997, 1998), Egger (2000), and Pakko and Wall (2001), one/wo-way fixed or random effec models are popularly used (e.g., Koo and Karemera, 1991; Glick and Rose, 2001). For his sudy, we used random effec models. There are wo reasons why we employ he random effec model raher han he fixed effec model. Firs, as Glick and Rose (2001), and Head and Ries (2001) discussed, he fixed effec model ignores he common cross-secional variaion of he daa by adding a se of counry-pair specific inerceps or forming deviaions from individual means; herefore, he resuls should be inerpreed as ime-series evidence. 7 The main focus of his paper is o esimae he effec of relaive exchange rae misalignmen on he expor flow, as well as heir ime-series movemens. Thus, if we employ he fixed effec model, he choice of esimaor conradics he primary economic quesion of he paper. Alernaively, by using he random effec model, we can uilize boh ime-series and cross-secional informaion of he daa. 6 Rose s daa can be found on his websie hp:/hass.berkeley.edu/~arose. 7 More deailed heoreical discussions abou his economeric issue are discussed by Maddala (1971) and Hausman (1978). 7

14 The second reason for using a random effecs model is ha he fixed effecs model is quie sensiive o errors in variables (Hausman, 1978). Since much of he variaion in he daa is removed, especially cross-secional variaion, by adding a se of counry-pair specific inercep or forming deviaions from individual means, he amoun of inconsisency would be greaer for he fixed effecs esimaes if errors in variables are presen. If here are errors in he variables, he fixed effec esimaor is biased and inconsisen, so ha he Hausman es is unreliable. 8 Table 1 summarizes he regression resuls for each indusry secor. 9 Before we inerpre he esimaed coefficien of our main variable, exchange rae misalignmen, i is desirable o check he coefficiens of oher variables o examine wheher our empirical model evidence is consisen wih previous sudies. The firs pair of variables is exporer s and imporer s GDP. According o he recen paper by Feensra e al. (2001), he parameer values on exporer s and imporer s GDP give us useful informaion o es he so-called home marke effec proposed by Krugman (1980). Specifically, hey demonsrae ha, if expors are differeniaed goods, he esimaed coefficien of he exporer s GDP should be larger han ha of he imporer s GDP under he assumpion of monopolisic compeiion. If expors are homogeneous, he esimaed coefficiens of he exporer s GDP should be smaller han ha of he imporer s GDP under he assumpions of oligopoly and segmened markes. Furhermore, hey show ha he esimaed coefficien on he exporer s GDP rises as we move from homogeneous o differeniaed goods, if Krugman s hypohesis is correc. In our sample, we considered he machinery secor as an indusry ha produces differeniaed producs, while he agriculural secor was considered as an indusry producing relaively homogeneous producs. Our empirical evidence is, in fac, consisen wih his inerpreaion of he esimaed coefficiens of GDP. In he case of he machinery secor, he esimaed coefficien of exporers GDP is 1.141, which is larger han ha of he imporers GDP, By conras, in he case of he agriculural secor, he esimaed coefficien of exporer s GDP is , which is far smaller han ha of he imporer s GDP, In our case, he Hausman es suggess a fixed effec model. However, we found ha he resuls of a fixed effec model are unreasonable compared o he resuls of previous sudies (i.e., Bergsrand (1989) and Feensra e al. (2001)) due o he mulicollineariy problem beween imporer s and exporer s income. These variables are no correlaed in a cross-secional sense bu highly correlaed in a ime-series sense, which produce he unreasonable esimaed coefficiens. The resuls of fixed effec models are available from he auhors on reques. 9 We use TSCSREG procedure of SAS 8.2 for he regression. 8

15 Table 1. Esimaion Resuls Agriculure Machinery Chemicals Manufacuring ln Y i a (-4.46) a (15.3) a (6.97) a (4.97) ln Y j a (3.42) a (12.9) a (9.67) a (8.79) Yi ln POPi a (7.85) a (-7.02) b (2.45) a (-3.69) Y j ln POPj a (-5.19) (-1.13) a (-2.82) (1.42) ln DIS ij (1.34) a (-7.73) a (-6.47) a (-5.29) BORD ij a (4.23) (0.54) c (1.71) a (2.81) LANG ij (0.66) (1.19) (0.78) (0.88) EU ij a (5.57) (0.61) (1.70) b (2.68) M ij a (-8.27) (0.92) (0.01) (0.15) Consan b a (2.53) (11.0) (8.32) Noes: -raios are in parenhesis; a, b, and c denoe significance a he 1, 5, and 10 percen level a (10.3) The mehod of inerpreaion for he esimaed coefficien of exporers and imporers per capia GDP was inroduced by Bergsrand (1989). In his heoreical model, he exporer s per capia GDP represens a proxy variable for he capial-labor raio of a counry. Therefore, a posiive (negaive) value of he esimaed coefficien of he exporer s GDP implies ha he indusry is capial-inensive (labor-inensive). Conversely, he esimaed coefficiens of he imporer s GDP could be inerpreed as usual income elasiciy. 9

16 In our empirical resuls, he esimaed coefficiens for exporers per capia GDP are posiive and saisically significan in he case of he agriculure and chemical secors, while hey are negaive in he case of he machinery and manufacuring secors. The resuls sugges ha producs in he agriculure and chemical secors are capial-inensive, while in he case of he machinery and manufacuring secors, producs are labor-inensive. The esimaed coefficiens for imporers per capia income are negaive and saisically significan for he agriculure and chemical secors, suggesing ha hese producs are necessiies. However, in he case of machinery and manufacuring secors, he esimaed coefficiens are posiive, alhough no saisically significan, which implies ha producs of hese indusries are luxuries. In erms of he addiional conrol variables, almos all he ime-invarian variables have he expeced signs. In he case of he disance variable, all of he esimaed coefficiens show he expeced negaive sign and are saisically significan a he 1 percen level, excep for he agriculural secor. In he case of he common border variable, he esimaed coefficiens are all he expeced posiive signs and are saisically significan a he 1 percen level in he agriculure and manufacuring secors and a he 10 percen level in he chemical secor. The resuls sugges ha, in he case of he agriculural secor, counries ha have a common border rade abou imes ( e =5.08) more han counries wihou a common border. The amoun of rade for counries wih a common border is 1.4 imes ( e =1.43), and 1.8 imes ( e =1.81) more han ha of non-coniguous counries for chemical and manufacuring secors, respecively. 10 In he case of he EU dummy variable, he esimaed coefficiens show he expeced posiive sign and are saisically significan a he 1 and 5 percen levels for agriculural and manufacuring secors, respecively. The esimaed coefficien for agriculural rade is The resul implies ha, when boh counries are members of he EU, rade is 10.7 imes ( e =10.7) greaer han if one of he counries is no a member of he EU. The resuls demonsrae ha he rade inegraion of he EU has been he sronges in he case of he agriculural secor. Finally, he esimaed coefficien of he misalignmen measure is saisically significan a he 1 percen level only in he case of agriculural rade. The esimaed coefficien is , which implies ha a one-percen over- (under-) valuaion of a currency compared o he long-run equilibrium level reduces (increases) agriculural expors by around 1.1 percen. In conras, we canno find any saisically significan relaionship beween variables in oher secors. 10 The inerpreaion of he coefficiens of ime-invarian variables follows Engel and Rogers (1996), and Rose (2000). 10

17 Furher Invesigaion In he previous secion, we used he sample averages of real exchange raes o normalize he misalignmen for each counry pair. According o he heory of PPP, i is proper o use his normalizaion. To check he robusness of he regression resuls, we will use anoher mehod of normalizaion in his secion. In his case, he measure of misalignmen is calculaed by M = ln R ij - ln R ij, where R ij is he level of real exchange rae in 1973 for each counry pair. Two imporan poins should be emphasized here. Firs, by equalizing all he real exchange rae measures in 1973, we acually resric he nominal exchange raes among sample counries in 1973 o he properly aligned nominal exchange raes based on PPP. This resricion also means ha real exchange raes in 1973 were long-run equilibrium raes among sample counries. This choice o use 1973 as he base year follows Williamson (1985) and De Grauwe (1988). 11 The underlying raionale of he choice is ha, a he saring year of he floaing exchange rae sysem, mos developed counries decided heir exchange raes using bilaeral agreemens. Therefore, nominal exchange raes in 1973 could represen properly aligned exchange raes. 12 The esimaion resuls wih he measure of misalignmen are presened in Table 2, and he resuls are similar o hose in Table 1. The only excepion is a saisically significan posiive relaionship beween exchange rae misalignmen and expor in he case of he machinery secor. As indicaed by Frankel and Romer (1999), his posiive relaionship migh be due o he simulaneiy beween exchange rae and rade. For insance, if he paern of expor in he machinery secor is imporan o deermine he expecaion of foreign exchange marke paricipans, an increasing level of expors in he machinery secor can cause an appreciaion of he real exchange rae, resuling in a posiive relaionship beween he variables. However, in he case of he agriculural secor, we sill find he esimaed coefficien of he misalignmen measure is negaive and saisically significan a he 1 percen level, indicaing he relaionship beween he variables is robus ij 11 In his paper, Bergsrand (1985, 1989) chose 1960, 1965, and 1966 as base years. 12 However, excep for his inuiive reason, here is no heoreical reason why we believe real exchange raes among he sample counries are a heir long-run equilibrium level in In fac, no economiss know when nominal exchange raes have been perfecly aligned, and his is he reason why measuring misalignmen is inrinsically imprecise. 11

18 Table 2. Esimaion Resuls Agriculure Machinery Chemicals Manufacuring ln Y i a (-2.91) a (16.5) a (6.94) a (4.98) ln Y j a (3.80) a (13.8) a (9.62) a (8.84) Yi ln POPi a (7.39) a (-8.01) b (2.50) a (-3.85) Y j ln POPj a (-6.19) (-0.54) a (-2.85) (1.55) ln DIS ij (0.71) a (-8.43) a (-6.44) a (-5.31) BORD ij a (4.43) (0.57) c (1.71) a (2.81) LANG ij (0.80) (1.31) (0.78) (0.88) EU ij a (5.80) (0.64) (1.70) b (2.68) 1973 M a ij (-9.35) b (2.04) (-0.04) (0.39) Consan a a (3.73) (11.8) (8.30) Noes: -raios are in parenhesis; a, b, and c denoe significance a he 1, 5, and 10 percen level a (10.3) 12

19 CONCLUSION This paper has focused on wheher exchange rae misalignmen negaively affecs agriculural rade, compared o oher secors. Nominal exchange rae misalignmen was obained from he percenage deviaion of real exchange raes from heir long-run equilibrium based on he heory of purchasing power pariy. Unlike he usual ime-series analysis, we have explored he poenial impac on rade associaed wih relaive misalignmen using panel daa. Moreover, we have used more disaggregaed daa since he effec of misalignmen may vary by secor, depending on secoral characerisics such as he role of sunk coss and durabiliy. In order o explore his issue, we have consruced a bilaeral rade marix involving rade flows beween 10 developed counries. Using panel daa analysis, a graviy model was esimaed for 4 secors over he period We have found ha overvaluaion (undervaluaion) of he nominal exchange rae negaively (posiively) affecs expor performance of he agriculural secor in paricular. In he large-scale manufacuring secors considered in his paper, expors are no significanly affeced by exchange rae misalignmen. In fac, nominal exchange raes have followed heir long-run equilibrium pah so ha hey are cyclically misaligned, a bes. Therefore, unlike a shor-run effec, he major problem for inernaional agriculural rade caused by exchange rae movemen is insabiliy in he long-run. Cyclical booms and depressions of agriculural expors by counries could possibly increase resource-wase wihin he agriculural secor, resuling in larger dead-weigh coss compared o oher large-scale indusrial secors. 13

20 References Baldwin, R., Hyseresis in Impor Prices: The Beachhead Effec, American Economic Review, 78 (1988): Bergsrand, J. H., The Graviy Equaion in Inernaional Trade: Some Microeconomic Foundaions and Empirical Evidence, The Review of Economics and Saisics 67 (1985): Bergsrand, J. H., The Generalized Graviy Equaion, Monopolisic Compeiion, and he Facor-Proporions Theory in Inernaional Trade, The Review of Economics and Saisics 71 (1989): Bureau of Labor Saisics. Consumer Price Index Daabase. hp:// De Grauwe, P., Exchange Rae Variabiliy and he Slowdown in Growh of Inernaional Trade. IMF Saff Papers, 24 (1988): Dornbusch, R., Exchange Raes and Prices, American Economic Review, 77 (1987): Economic Research Service, U.S. Deparmen of Agriculure. Exchange Rae Daabase. hp:// Egger, P., A Noe on he Proper Economeric Specificaion of he Graviy Equaion, Economics Leers 66(2000): Engel, C. and J. Rogers, How Wide is he Border? American Economic Review, 86 (1996): Feensra, R. C., J.R. Markusen, and A.K. Rose (2001), Using he Graviy Equaion o Differeniae among Alernaive Theories of Trade, Canadian Journal of Economics, 34: Fisher, E. and J.Y. Park, Tesing Purchasing Power Pariy under he Null Hypohesis of Co- Inegraion. Economic Journal, 101 (1991): 1-8. Flood, R.P. and M.P. Taylor, Exchange Rae Economics: Wha s Wrong Wih he Convenional Macro Approach? In J.A. Frankel, G. Galli, and A. Giovannini (Ediors), The Microsrucure of Foreign Exchange Markes. Chicago, IL: Chicago Universiy Press, Frankel, J.A. Recen Exchange-Rae Experience and Proposals for Reform. American Economic Review, 86 (1996): Frankel, J.A. and D. Romer, Does Trade Cause Growh? American Economic Review, 89 (1999):

21 Frankel, J.A. and A.K. Rose, A Panel Projec on Purchasing Power Pariy: Mean Reversion wihin and beween Counries. Journal of Inernaional Economics, 40 (1996): Froo, K.A. and P. D. Klemperer, Exchange Rae Pass-Through When Marke Share Maers, American Economic Review, 79 (1989): Gardner, B. On he Power of Macroeconomic Linkages o Explain Evens in U.S. Agriculure. American Journal of Agriculural Economics 63 (1981): Glick, R. and Rose, A.K., Does a Currency Union Affec Trade? NBER working paper, No (2001). Goldberg, P.K. and M. Kneer, Goods Prices and Exchange Raes: Wha Have We Learned? Journal of Economic Lieraure, 35 (1997): Hausman, J. A., Specificaion Tess in Economerics, Economerica, 46 (1978): Head, K., and J. Ries, Increasing Reurns versus Naional Produc Differeniaion as an Explanaion for he Paern of US-Canada Trade, American Economic Review, 91 (2001): Inernaional Moneary Fund. Inernaional Financial Saisics Yearbook. Inernaional Moneary Fund (Various issues), Kneer, M.M., Inernaional Comparisons of Pricing-o-Marke Behavior American Economic Review, 83 (1993): Koo, W.W. and D. Karemera, Deerminans of World Whea Trade Flows and Policy Analysis, Canadian Journal of Agriculural Economics 39 (1991): Koo, W.W., D. Karemera, and R. Taylor, A Graviy Model Analysis of Mea Trade Policies, Agriculural Economics 10 (1994): Krugman, P., Scale Economies, Produc Differeniaion, and he Paern of Trade, The American Economic Review 70 (1980): Lohian, J.R. and M.P. Taylor, Real Exchange Rae Behavior: The Recen Floa from he Perspecive of he Pas Two Cenuries. Journal of Poliical Economy, 104 (1996): Lohian, J.R. and M.P. Taylor, Real Exchange Rae Behavior. Journal of Inernaional Money and Finance, 16 (1997): MacDonald, R. Exchange Rae Behavior: Are Fundamenals Imporan? The Economic Journal, 109 (1999):

22 Maddala, G. S. The Use of Variance Componens Models in Pooling Cross Secion and Time Series Daa. Economerica, 39 (1971): Mark, N.C. Real and Nominal Exchange Raes in he Long Run: An Empirical Invesigaion. Journal of Inernaional Economics, 28 (1990): Mayas, L., Proper Economeric Specificaion of he Graviy Model, World Economy 20 (1997): Mayas, L., The Graviy Model: Some Economeric Consideraions, World Economy 21 (1998): Organizaion for Economic Cooperaion and Developmen. Trade in Commodiies. CD-Rom and various issues. Pakko, M. R. and H. J. Wall, Reconsidering he Trade-Creaing Effecs of a Currency Union, Federal Reserve Bank of S. Louis, Quarerly Review, Sep/Oc (2001): Rogoff, K. The Purchasing Power Pariy Puzzle. Journal of Economic Lieraure, 34 (1996): Rose, A. K., One Money, One Marke: The Effec of Common Currencies on Trade. Economic Policy, 30 (2000): Tweeen, L. Farm Policy Analysis. Wesview Press, Williamson, J. (ed.) Esimaing Equilibrium Exchange Raes, Washingon: Insiue for Inernaional Economics, Wolak, F. A. and C. D. Kolsad, A Model of Homogeneous Inpu Demand Under Price Uncerainy, American Economic Review, 81 (1991):