International Business And Economics Research Journal Volume 2, Number 10

Transcription

1 Inernaional Business And Economics Research Journal Volume 2, Number 10 he Real Exchange Rae Flucuaions Puzzle: Evidence For Advanced And ransiion Economies Amalia Morales-Zumauero, ( Universiy of Málaga, Spain Absrac his paper ries o analyze he sources of he real exchange rae flucuaions for a se of advanced economies and Cenral and Easern European ransiion economies. o address his, in a firs sep, we compue wo measures of he share of he variance of he real exchange rae accouned for movemens in he relaive prices of raded goods beween he counries. One measure is based on R 2 coefficien and he oher one is based on he mean-suared error (MSE) of he changes in he real exchange rae. In a second sep, we esimae srucural (idenified) vecor auoregression (SVAR) models, and decompose real and nominal exchange rae movemens ino hose caused by real and nominal shocks. In a hird sep, we complee previous ones wih an impulse-response analysis. hree cenral messages are derived from resuls: (1) for ransiion economies, under regimes of managed nominal exchange raes, he relaive price of non-raded goods explain a large percenage of he variance of he real exchange rae; (2) here is evidence of insabiliy in he variance decomposiion of he real exchange raes for advanced economies across samples, and (3) as resul of diverse fiscal and moneary policies in ransiion economies, real exchange raes in some economies are driven mosly by real shocks while in ohers are driven mosly by nominal shocks. 1.0 Inroducion R ecen empirical lieraure on real exchange rae flucuaions for advanced economies has focused mainly on hree approaches: he analysis of he real exchange rae volailiy, he compuaion of several measures of he share of he variance of he real exchange rae and he variance decomposiion analysis. In his conex, from papers based on he wo former approaches, he empirical evidence suggess ha real exchange raes movemens could be explained by he relaive price of raded goods beween counries, so he non-raded componen of he real exchange rae accouned for lile of he movemens in he real exchange rae (Engel, 1993; Rogers and Jenkins, 1995; Engel 1999; Engel and Rogers, 2000). his resul provides evidence in favor of sicky price models (Dornbusch, 1976) where nominal shocks would explain he real exchange rae flucuaions. However, from papers based on he relaive imporance of real and nominal shocks for explaining he variance of he real exchange rae he variance decomposiion analysis resuls sugges ha flucuaions in real and nominal exchange raes are due o primarily o real shocks. hus, real shocks dominae nominal shocks for boh exchange rae series over shor and long freuencies (Lasrapes, 1992; Enders and Lee, 1997). his empirical evidence would have an implicaion o model he exchange raes: models focused on he properies of price levels would be adeuae (Balassa, 1964; Samuelson, 1964). On he oher hand, recen empirical papers for ransiion economies ry o explain he srong real exchange rae appreciaion observed in a number of ransiion economies during he, so called, ransiion period: following a sharp iniial depreciaion, real exchange raes have coninuously appreciaed over he course of ransiion (Halpern Readers wih commens or uesions are encouraged o conac he auhor via . 39

2 Inernaional Business And Economics Research Journal Volume 2, Number 10 and Wyplosz, 1997; Begg e al. 1999). Par of his lieraure is focused on esing he Balassa-Samuelson effec in hese economies (Halpern and Wyplosz, 2001; Broeck and Sl k, 2001; Éger, 2002). hese papers sugges, in general, ha here is a clear evidence in favor of produciviy-based exchange rae movemens in favor he Balassa- Samuelson effec in he European Union accession counries. his evidence would imply ha real shocks may be he main force o explain movemens in he real exchange rae. However, oher papers sugges ha he similar pah followed by he real exchange rae in ransiion economies is surprising given he differences in moneary and real shocks in differen counries (Brada, 1998; Desai, 1998; Dibooglu and Kuan, 2001). his approach is based on he experience of all ransiion economies wih respec o produciviy growh, raded liberalizaion and capial inflows has no been he same. Moreover, he very differen fiscal and moneary policies among he ransiion economies suggess ha moneary shocks could dominae over produciviy shocks in boh freuency and inensiy. hus, no only real shocks bu nominal shocks oo could explain he real exchange rae movemens in ransiion economies during he ransiion period. he cenral lesson from his previous lieraure is ha here is a mixed empirical evidence on explaining he sources of real exchange rae flucuaions in advanced and ransiion economies. hus, his paper ries o give some ligh on wha we could call he real exchange rae flucuaions puzzle. In oher words, he main goal of his paper is o analyze he sources of real exchange rae flucuaions for a se of advanced economies during he full period 1973:1 o 2000:1, disinguishing he subperiods from 1973:1 o 1990:12 and from 1991:1 o 2000:1 (he ransiion period), and for a se of seleced Cenral and Easern European ransiion economies, during he ransiion period. o address his, we follow hree seps. In a firs sep, we compue wo measures of he share of he variance of he real exchange rae accouned for movemens in he relaive prices of raded goods beween he counries. One measure is based on R 2 coefficien and he oher one is based on he mean-suared error (MSE) of he changes in he real exchange rae. In a second sep, we esimae srucural (idenified) vecor auoregression (SVAR) models, and decompose real and nominal exchange rae movemens ino hose caused by real and nominal shocks. In his sep we ry o idenify he main forces ha explain he behavior of he real exchange rae in advanced and ransiion economies. In a hird sep, we complee previous ones wih an impulse-response analysis. he final goals of he empirical analysis from hese hree seps are o compare resuls on he sources of real exchange raes movemens beween samples for advanced economies, beween advanced and ransiion economies during he ransiion period from plan o marke and, finally, o derive some implicaions on how o model he exchange raes. he remainder of he paper is organized as follows. Secion 2 presens our heoreical saring poin, describes he daa se and repors he empirical resuls on real exchange rae flucuaions in hree subsecions: he R 2 coefficien and MSE measures, he variance decomposiion analysis and he impulse-response analysis. Secion 3 concludes he paper. 2.0 Explaining he Variance of he Real Exchange Rae In his secion we develop he analysis of he sources of flucuaions in he real exchange rae in hree seps. Firsly, we compue wo measures of he share of he variance of he real exchange rae accouned for movemens in he relaive prices of raded goods beween he counries. Secondly, we esimae SVAR models, and decompose real and nominal exchange rae movemens ino hose caused by real and nominal shocks. hirdly, we use he impulseresponse analysis o sudy he effecs of a shock o an endogenous variable on he variables in he SVAR. Our heoreical saring poin is o decompose he real exchange rae in wo componens: he relaive price of raded goods beween counries and a weighed difference of he relaive price of non-raded o raded goods prices in each counry. Le p he log of a general price index for a counry: 40 p (1 a) p ap N (1)

3 Inernaional Business And Economics Research Journal Volume 2, Number 10 where p is he log of raded goods price index, N p is log of he non-raded goods price index and (1-a) and a are he weighs on raded and non-raded goods in he general price index, respecively. Similarly, he foreign general price index can be wrien as where (1-b) and b are he weighs on raded and non-raded goods in he general price index in he foreign counry, respecively. If s denoes he log of nominal exchange rae, he real exchange rae could be wrien as p (1 b) p bp * * N* (2) s p * * N * * N p ( s p p ) b( p p ) a( p p ) (3) z (4) or, similarly, as where z ( s b( p p N * * p p * ) ) a( p N p ) If from euaion (4) we obained ha movemens in he real exchange rae are explained by he relaive price of raded goods beween counries,, his resul would imply ha models focus on nominal exchange rae deerminaion in a framework of sicky prices (sicky price diseuilibrium models) would be a good approximaion o model exchange raes. In oher words, nominal shocks would cause he shor-run excess volailiy in exchange raes (Dornbusch, 1976). On he oher hand, if we obained ha movemens in he real exchange rae are explained by he relaive price of non-raded o raded goods prices in each counry, his resul would sugges ha we could model he exchange raes using models based on he properies of price levels. hus, real shocks would play a cenral role o explain he real exchange rae flucuaions (Balassa, 1964, Samuelson, 1964). o carry ou our empirical analysis, we use hree daa ses. he firs daa se are uarerly daa of he gross domesic produc deflaor, he consumer price index, he aggregae producer price index and end of period nominal exchange raes for Canada, Japan, Unied Kingdom and Unied Saes, from 1973:1 o 2000:1, and for he Euro Zone, from 1982:1 o 2000:1. In paricular, we consruc he general real exchange rae from he gross domesic produc deflaor and we use a crude approximaion o raded an non-raded goods prices: he producer price index as proxy of raded goods prices, and consruc he relaive price of raded goods beween counries from i, and he consumer price index as proxy of non-raded goods prices. he second daa se are monhly daa of he consumer price index, he food consumer price index, he services consumer price index and end of period nominal exchange raes, for a se of advanced economies. In paricular, for Canada, Japan and Unied Saes he sample size covers he period from 1973:1 o 2000:1, for all ime series. For France and Ialy, from 1973:1 o 1998:12 for all series. For Germany from 1973:1 o 1998:12 for all series, excep for he services consumer price index ha covers from 1976:1 o 1998:12, and for Unied Kingdom from 1973:1 o 2000:1 for all series, excep for he services consumer price index ha covers he period from 1975:1 o 2000:1. We consruc he general real exchange rae using he consumer price index. Moreover, we use a beer approximaion o raded and non-raded goods prices: he food price index as proxy of raded goods prices, and consruc he relaive price of raded goods beween counries from i, and he services price index as proxy of nonraded goods prices. 41

4 Inernaional Business And Economics Research Journal Volume 2, Number 10 he hird daa se are similar o previous one bu for a seleced se of Cenral and Easern ransiion economies. In paricular, he series cover he period from 1991:1 o 2000:1, he ransiion period, for Czech Republic, Hungary, Poland and Romania, and from 1991:12 o 1999:12 for Slovenia 1. As in previous daa se, we consruc he general real exchange rae from consumer price index and consider he food price index as proxy of raded goods prices and he services price index as proxy of non-raded goods prices. he daa have been obained from he OECD daabase, in paricular, from he Main Economic Indicaors daabase for all counries, excep for he aggregae uarerly daa for he Euro Zone ha have been obained from he Eurosa daabase, paricularly, from Eurosaisics he R 2 and MSE Measures In his subsecion, we compue wo measures of he share of he variance of he real exchange rae,, due o he relaive price of raded goods beween counries,. he firs one is a very simple measure based on he R 2 coefficien from he ordinary leas suare regression on he changes of on a four differen horizons. he second measure is based on he MSE of he change in he real exchange rae 2. We ry o compue how much of he MSE of changes of is aribuable o changes in, a four differen horizons. In paricular, we use he nex decomposiion MSE( MSE( n n ) ) MSE( z z n ) (5) MSE( n ) var( n ) 2 mean( ) (6) n where he MSE is defined as he sum of he suare drif and he variance, wih where N is he sample size and n is he horizon. n mean( n ) ( N 1 ) N 1 N var( n ) ( N n 1)( N n) N n j 1 j n j mean( j n j ) 2 We have obained he empirical resuls regarding he simple R 2 measure a four differen horizons, for he full sample, 1973:1-2000:1, and for he subsamples 1973:1-1990:4 (or 1990:12 for monhly daa) and 1991:1-2000:1 3. Resuls are presened for all en possible bilaeral relaionships beween he advanced economies Canada, Japan, Euro Zone, Unied Kingdom and Unied Saes, using uarerly daa and employing he crude approximaion o raded and non-raded goods prices: he aggregae producer price index and he consumer price index, 1 hese counries have been seleced depending on he daa availabiliy. 2 See Engel (1999). 3 All he numerical resuls of his subsecion are available upon reues. 42

5 Inernaional Business And Economics Research Journal Volume 2, Number 10 respecively. Resuls sugges ha average R 2 values for horizons of 1 uarer, 2 uarers, 4 uarers and 12 uarers are 0.95, 0.94, 0.94 and 0.92, respecively, for he full sample. Similar resuls are obained for he wo subsamples. In addiion, we have obained he values of R 2 measure for he advanced economies Canada, France, Germany, Ialy, Japan, Unied Kingdom and Unied Saes, using a beer approximaion o raded and non-raded goods prices: he food price index and he services price index, respecively. We observe ha, for example, he average value of he R 2 measure is 0.86 for horizon 1, during he subperiod 1973:1-1990:12 while for he ransiion period i increases o I would sugges some evidence in favor of a more imporan role of sicky prices in he subsample 1991:1-2000:1. In oher words, his resul inuiively would sugges ha nominal shocks are more imporan in he period 1991:1-2000:1 for advanced economies. Finally, he resuls of R 2 measure for ransiion economies Czech Republic, Hungary, Poland, Romania and Slovenia, during he ransiion period, show ha he relaive prices of non-raded goods would explain he movemens in he real exchange rae in a greaer percenage han in advanced economies. hese preliminary resuls are confirmed when he measure of he share of he variance of he real exchange rae, aribued o movemens in he raded real exchange rae based on he MSE, is used. Resuls for advanced economies sugges ha, in general, using he crude approximaion o raded and non-raded goods, explains he majoriy of he variance of, during he full sample and subsamples. However, using monhly daa, we observe ha, for he full sample and subsample 1973:1-1990:12, he relaive price of non-raded o raded goods prices plays some role o explain he movemens in. For example, i could be observed in he MSE measure values of he bilaeral relaionships beween Canada and Unied Saes, Germany and France, Japan and Canada, Japan and he hree Euro Zone counries and hose bilaeral relaionships ha include o Unied Kingdom. However, when we analyze he ransiion period from 1991:1 o 2000:1, causes he majoriy of he variance of. Finally, we obain he measure of he share of he variance of he real exchange rae aribued o movemens in he raded real exchange rae based on he MSE, for ransiion economies. We observe he resuls change srongly: in general, he relaive prices of non-raded goods explain much more han in advanced economies. hen, i would sugges an imporan resul: in hese economies, under managed nominal exchange raes, he relaive prices of non-raded goods could explain for 50% o 70% of he variabiliy of he real exchange rae. For example, for he bilaeral relaionship beween Czech Republic and Unied Saes we observe ha he relaive prices of nonraded goods cause he 50%, 51%, 57% and 70% for 1, 6, 12 and 36 monhs, respecively, of he variabiliy of he real exchange rae. In summary, wo cenral messages are derived from his firs approximaion o explain he flucuaions in he real exchange rae: (1) for advanced economies, using he food and services price indexes as proxies of raded and non-raded goods prices, respecively, we observe ha he relaive price of raded goods explain he majoriy of he variance of he real exchange rae for he ransiion period. In oher words, i looks as if he nominal shocks play a much more cenral role during he ransiion period han along he full sample and subsample 1973:1-1990:4 (or 1990:12 for monhly daa). (2) For ransiion economies, he evidence would sugges ha under regimes of managed nominal exchange raes, he relaive price of non-raded goods explains a large percenage of he variance of he real exchange rae he Variance Decomposiion Analysis In his subsecion we esimae srucural (idenified) vecor auoregression SVAR models, and decompose real and nominal exchange rae movemens ino hose caused by real and nominal shocks. In his sep, we ry o idenify he main forces ha explain he behavior of he real exchange rae in advanced and ransiion economies, and compare i beween hem during he ransiion period from plan o marke. 43

6 Inernaional Business And Economics Research Journal Volume 2, Number 10 We assume ha here are wo ypes of shocks affecing nominal and real exchange raes. he firs shock has effec on boh exchange raes and we will call i real shock. he second one, ha we will call nominal shock, has no long-run effec on real exchange rae. his assumpion is consisen wih he noion of long-run money neuraliy. hus, he real shock affecs he real and nominal exchange rae while he nominal shock affecs he nominal exchange raes, bu i has ransiory bu no permanen effec on real exchange rae. hen, he neuraliy resricion forces real shock o explain all he variance of real exchange rae a an infiniive forecas horizon, so we are ineresed on explanaory power of real and nominal shocks over shor horizons. his assumpion allows us o idenify he model and decompose he exchange rae series. he wo disurbances are uncorrelaed a all leads and lags. We presen now he join process followed by he real exchange rae,, and nominal exchange rae, s, for advanced economies 4. We will call o his empirical model he SVAR(, s ) model. In addiion, as nominal exchange raes are managed during ransiion period for ransiion economies, we will compue he decomposiion on real exchange raes and prices levels, insead of nominal exchange raes, for ransiion economies 5. We will refer o his empirical model as he SVAR(, p ) model. As series are non-saionary in levels, saionary in differences 6 and are no coinegraed series, using he Johansen s (1988, 1992) coinegraion mehodology 7, hen a bivariae auoregression model in firs differences is adeuae. Le u r and u n denoe he real and nominal shocks in, respecively. Since he vecor of he firs differences in real and nominal exchange raes ' average represenaion, x, s is saionary, i has he nex bivariae moving or, similarly x C s C ' s C( L) u (7) 11 21, ( L) ( L) C C ( L) u ( L) u r n (8) where is he firs-difference operaor and C ij (L), for i,j=1,2, are polynomials in he lag operaor, L. o idenify he shocks, we have assumed ha nominal shocks have no long-run effec on he real exchange rae. In erms of euaion (8), his resricion is euivalen o impose ha he sum of coefficiens in C 12 ( L) sum o zero. hus, where c 12 ( k) is he kh coefficien in C ij (L). he empirical model in euaion (8) is compleed for each ransiion economy by inroducing a dummy variable o capure he effecs on endogenous variables of changes in he exchange rae regimes during he ransiion period. c12 ( k) 0 (9) k 0 Finally, we carry ou our empirical analysis for advanced economies from 1973:1 o 2000:1, and for he subperiods 1973:1-1990:4 (or 1990:12 for monhly daa) and 1991:1 o 2000:1, and for ransiion economies for he period 1991:1 o 2000:1. Moreover, we calculae he real exchange raes aking o Unied Saes as reference counry 4 See Blanchard and Quah (1989), Lasrapes (1992) and Enders and Lee (1997). 5 As Dibooglu and Kuan (2001) sugges. 6 he resuls on uni roo analysis, based on he Dickey-Fuller and Phillips-Perron uni roo ess, are available upon reues. 7 Coinegraion resuls are available upon reues. We noice ha here is only a marginal evidence of coinegraion beween real and nominal exchange raes for Ialy when monhly daa are used. 44

7 Inernaional Business And Economics Research Journal Volume 2, Number 10 for advanced economies, and we calculae real exchange raes for ransiion counries aking as reference counry no only o Unied Saes bu o Germany oo. Empirical resuls on variance decomposiion analysis are repored in ables 1 o 3. able 1, panel (a), shows he resuls of variance decomposiion of real and nominal exchange raes for advanced economies, for he full sample 1973:1 o 2000:1, using uarerly daa, and aking o Unied Saes as he reference counry o consruc he real exchange raes. As we can observe, for Canada real shocks cause almos all of forecas error variance of he real exchange rae. I occurs a any forecas horizon. Moreover, nominal shocks explain abou 10% of he variance of nominal exchange rae a a 1 uarer forecasing horizon, bu i falls abou 5% afer one uarer. Resuls sugges ha for he full sample real shocks explain almos all he variance of real and nominal exchange raes. Similar resuls are obained for he Euro Zone, alhough he evidence shows ha nominal shocks are a bi more imporan han in Canada o explain he variance of he real exchange rae. For example, nominal shocks explain abou 21% of he forecas variance of real exchange rae a a horizon of 12 uarers. For Japan real shocks dominae o nominal shocks, bu he evidence suggess ha nominal shocks play an imporan role o explain he variance of he nominal and real exchange raes over shor horizons. Finally, for Unied Kingdom he real shocks look o cause almos all he variance of he real exchange rae, and nominal shocks explain a subsanially par of he variance of he nominal exchange rae. Similar resuls are obained for he subperiod 1973:1-1990:4 (able 1, panel (b)). For he subperiod 1991:1 o 2000:1, he evidence from able 1 panel (c) changes subsanially. In general, for all economies nominal shocks play a cenral role o explain he variance, no only of he real exchange rae bu he nominal exchange rae oo. In paricular, nominal shocks dominae o real shocks o explain he variance of he real exchange rae for Canada, Japan and Unied Kingdom. For he Euro Zone nominal shocks explain, for example, abou 30% of he variance of he real exchange rae a a 12 uarers forecasing horizon. In addiion, nominal shocks dominae o real shocks o cause he variance of he nominal exchange rae for Canada a any horizons (excep for one uarer), for Japan a 8 and 12 uarers forecasing horizons, and for Unied Kingdom over shor forecasing horizons. able 2 illusraes he resuls of he variance decomposiion analysis for advanced economies, using monhly daa. From panels (a) and (b) resuls hardly differ from previous ones: we observe ha, for all counries, real shocks look o cause almos all forecas error variance of he real and nominal exchange raes. Resuls for he ransiion period are repored in panel (c). We noice ha he imporance of nominal shocks increases subsanially o explain he variance of he real and nominal exchange raes. In paricular, for Canada we observe ha nominal shocks dominae o real shocks and hey cause almos all he variance of he real and nominal exchange raes. For example, nominal shocks explain abou 85% of he variance he real exchange rae a a horizon of 1 monh and his percenage increases abou 95% a an horizon of 9 monhs. For France, nominal shocks explain a large percenage of he variance of he real exchange rae and are cenral o explain he variance of he nominal exchange rae a any horizon. For Germany, nominal shocks explain an imporan percenage of he variance of he real exchange rae, overall, in shor run horizons. Similar resuls are obained for he nominal exchange rae. Finally, for Ialy, Japan and Unied Kingdom he imporance of he nominal shocks increases by comparison wih he full sample and subsample 1973:1-1990:12, bu i is weaker han in Canada, France and Germany. In general, for Ialy nominal shocks explain somewha more of he variance of he real exchange rae a 12, 24 and 36 horizons, and for Japan and Unied Kingdom nominal shocks explain somewha more of he variance of he real exchange rae a shor horizons. In summary, resuls using uarerly daa sugges ha for he full sample and subsample 1973:1-1990:4 real shocks cause almos all of he variance of he real exchange rae for Canada, Euro Zone and Unied Kingdom, and for Japan real shocks dominae o nominal shocks bu nominal shocks explain an imporan par of he variance of he real and nominal exchange raes, over shor horizons. However, for he ransiion period nominal shocks are cenral o explain he variance of he real and nominal exchange raes. On he oher hand, resuls using monhly daa sugges ha, for he full sample and he subsample 1973:1-1990:12, for all counries, real shocks cause almos all forecas error variance of he real exchange raes while for he ransiion period he imporance of nominal shocks increases subsanially o explain he variance of real exchange raes. 45

8 Inernaional Business And Economics Research Journal Volume 2, Number 10 From he evidence of he variance decomposiion analysis of he real and nominal exchange raes for advanced economies we derive several cenral messages: (1) he evidence suggess ha, in general, for he full sample and subsample 1973:1-1990:4 (or 1990:12 for monhly daa) real shocks play a cenral role o explain he movemens of he real exchange rae, so changes in he real exchange rae would be dominaed by real shocks. Nominal shocks are lighly more imporan for explaining he forecas variance of nominal exchange raes; (2) for he ransiion period, resuls sugges ha a large proporion of real exchange raes movemens is due o nominal shocks; (3) hese resuls sugges evidence in favor of insabiliy in he variance decomposiion of real exchange raes across samples. In oher words, he sources of flucuaions of exchange raes depend on he sample ha we consider. I would have a clear implicaion for modelling exchange raes in advanced economies. For he period 1973:1 o 1990:4 (or 1990:12 for monhly daa) real shocks look o dominae nominal shocks and hen, his resul would sugges ha models ha emphasize he imporance of real shocks o explain he sources of flucuaions of he real exchange rae, such as he Balassa s (1964) and Samuelson s (1964) models, would be adeuae. However, for he subperiod 1991:1 o 2000:1, in general, nominal shocks play a cenral role o explain flucuaions in exchange raes, so he sicky price models, such as Dornbusch s (1976) model, would be more adeuae o model he exchange rae behavior. able 3 presens he resuls of he variance decomposiion analysis for ransiion economies, during he ransiion period from 1991:1 o 2000:1. he analysis has been done from our empirical model in euaion (8) bu using he real exchange rae and he price levels as endogenous variables, he SVAR(, p ) model. We ake o Unied Saes and Germany, alernaively, as reference counries o consruc he real exchange raes. able 3 illusraes resuls for he SVAR(, p ) model where Unied Saes is he reference counry. he dummy variable ha capure he specific regime changes in each ransiion economies, has been significan only for Hungary and Romania. Resuls sugges ha for Czech Republic real shocks dominae he changes in he real exchange raes. For example, real shocks explain abou 79% of he variance of he real exchange rae a a 1 monh forecasing horizon. Moreover, nominal shocks explain almos all he error variance of prices levels, alhough is imporance decreases along he horizons. he resuls are similar for Hungary: real shocks explain almos all he variance of he real exchange rae, overall, over shor horizons. For example, real shocks explain 97% of he variance of he real exchange rae a an horizon of 1 monh. Nominal shocks cause a large percenage of he variance of he prices over shor horizons. However, resuls change from previous ones for Poland. We observe ha nominal shocks dominae real shocks o explain he variance of he real exchange rae a shor horizons. For example, nominal shocks explain abou 73% of he variance of he real exchange rae a an horizon of 1 monh. However, prices are dominaed by real shocks a shor horizons. For Romania resuls show as changes in he real exchange rae are dominaed by nominal shocks over shor horizons. For example, nominal shocks explain abou he 92% of he variance of he real exchange rae a a 1 monh horizon. However, real shocks look o explain almos all of he variance of prices. Finally, for Slovenia resuls show a clear dominance of real shocks o explain he variance of he real exchange rae. In addiion, if we consider o Germany as reference counry o consruc he real exchange raes 8, resuls hardly differ from previous ones: for Czech Republic, Hungary and Slovenia real shocks mosly explain he flucuaions of he real exchange rae while nominal shocks are he main force o explain he flucuaions in he real exchange raes for Poland and Romania 9. 8 Resuls are available upon reues. 9 Resuls regarding he SVAR (, s ) model for ransiion economies are available upon reues. he real exchange rae resuls do no differ from he SVAR (, p ) model. In paricular, for Czech Republic, Hungary and Slovenia real shocks explain a large percenage of he forecas error variance of he real exchange rae, while nominal shocks are he main force o explain he flucuaions of he real exchange rae for Poland and Romania. In addiion, real shocks dominae nominal shocks o explain he variance of he nominal exchange rae in Romania. For he Czech Republic resuls show ha real shocks explain almos all of he variance of he nominal exchange rae a any horizon. For Hungary real shocks explain he variance of he nominal exchange rae over shor horizons. However, nominal shocks dominae real shocks o explain he variance of he nominal exchange rae, over shor horizons, for Poland. Finally, real shocks dominae nominal shocks o cause he movemens in he nominal exchange rae in Slovenia. However, hese resuls could be affeced for he managed naure of he nominal exchange raes during he ransiion period. 46

9 Inernaional Business And Economics Research Journal Volume 2, Number 10 In shor, wo cenral messages could be derived from he variance decomposiion analysis carried ou for ransiion counries: (1) he sources of flucuaions of he real exchange raes depend on he ransiion economy considered. In paricular, real shocks explain mosly he movemens in he real exchange raes for Czech Republic, Hungary and Slovenia while nominal shocks play a cenral role o explain he variance of he real exchange rae for Poland and Romania. hus, our resuls give suppor o previous lieraure (Brada, 1998; Desai, 1998; Dibooglu and Kuan, 2001) ha suggess ha as resul of diverse fiscal and moneary policies, real exchange raes in some accession economies could be driven mosly by real shocks and in ohers could be driven mosly by nominal shocks. So, hese resuls sugges ha differen moneary and fiscal policies in ransiion economies imply differen resuls from he variance decomposiion analysis. (2) From his analysis we could derive implicaions for modelling he real exchange rae in ransiion economies. For Czech Republic, Hungary and Slovenia real shocks dominae and i would sugges models such as Balassa s (1964) and Samuelson s (1964) models would be adeuae. hus, during he ransiion period, as conseuence of ransformaion of sizable indusrial secors, whose capial socks have proven largely obsolee, hese economies have presened an imporan increase of he produciviy growh in heir indusrial secors, along he ransiion from plan o marke. I could explain he srong appreciaion of real exchange rae in hese ransiions economies. However, nominal shocks would explain mainly he movemens in he real exchange raes for Poland and Romania, so diseuilibrium models would be adeuae o model he real exchange rae. hus, i looks ha moneary policy has had an imporan role in influencing he real exchange rae in Poland and Romania Impulse-Response Análisis In his subsecion, we use he impulse-response analysis o sudy he effecs of boh ypes of shocks on he endogenous variables in he SVAR models. In general, we observe several feaures: 1. For advanced economies, during 1973:1 o 2000:1 10 using uarerly daa, Figure 1 (available from he auhor), panel (a), shows ha here is evidence of a smooh increase in he real and nominal exchange raes in response o a real shock. For he ransiion period, Figure 1, panel (b), nominal shocks look o cause an immediae increase in he real exchange rae, wih srong evidence in favor of overshooing in Canada. 2. For advanced economies, during he full period 11 and using monhly daa, resuls in Figure 2 (available from he auhor) are similar o pervious ones: real shocks cause a smooh increase in he real and nominal exchange raes. Moreover, nominal shocks cause an unnoiceable effec on real exchange raes, hus, here is no evidence of overshooing. However, during he ransiion period, nominal shocks ends o cause, in general, an increase on real exchange raes: here is some evidence of overshooing. 3. For ransiion economies, in general, he real exchange rae raises due o a real shock. For he Czech Republic and Poland real exchange rae depreciaes smoohly in response o a nominal shock. For Romania we observe a more imporan depreciaion of he real exchange rae in response o a nominal shock. hus, for hese accession economies here is evidence in favor of overshooing. 3. Concluding Remarks his paper has analyzed he sources of real exchange rae flucuaions for a se of advanced economies and Cenral and Easern ransiion economies. he cenral messages derive from he resuls sugges ha: 1. For advanced economies, previous empirical evidence, ha suppor ha flucuaions in he real exchange rae are explained by he relaive price of raded good beween counries, does no hold for ransiion economies: under regimes of managed nominal exchange raes, he relaive price of non-raded goods explains an high percenage of he variance of he real exchange rae. 2. here is evidence in favor of insabiliy in he variance decomposiion of he real exchange raes across samples for advanced economies. In oher words, he sources of flucuaions of he real exchange raes depend on he sample ha we consider: for he period 1973:1 o 1990:4 (or 1990:12 for monhly daa) real shocks look o dominae nominal shocks and for he period 1991:1 o 2000:1, in general, nominal shocks 10 Similar resuls are obained for he subperiod 1973:1-1990:4. 11 Similar resuls are obained for he subperiod 1973:1-1990:12. 47

10 Inernaional Business And Economics Research Journal Volume 2, Number 10 play a cenral role o explain flucuaions in exchange raes. his resul would imply ha models ha emphasize he imporance of real shocks o explain he sources of flucuaions of he real exchange rae, would be more adeuae for he subsample 1973:1-1990:4 (or 1990:12 for monhly daa), while sicky price diseuilibrium models would be more adeuae for he ransiion period. 3. As resul of diverse fiscal and moneary policies in ransiion economies, real exchange raes in some economies are driven mosly by real shocks while in ohers are driven mosly by nominal shocks. References 1. Balassa, B. (1964): he purchasing power pariy docrine: a reappraisal, Journal of Poliical Economy, 72, Begg, D., Halpern, L. and Wyplosz, C. (1999): Moneary and exchange rae policies, EMU and Cenral and Easern Europe, Forum Repor of he Economic Policy Iniiaive, 5 (London: CEPR). 3. Blanchard, O. and Quah, D. (1989): he dynamic effecs of aggregae demand and supply disurbances, American Economic Review 79, Brada, J.C. (1998): Inroducion: exchange raes, capial flows, and commercial policies in ransiion economies, Journal of Comparaive Economies 26, Broeck, M.D. and Sl k,. (2001): Inerpreing Real Exchange Rae Movemens in ransiion Counries, IMF Working paper, Desai, P. (1998): Macroeconomic fragiliy and exchange rae vulnerabiliy: A cauionary record of ransiion economies, Journal of Comparaive Economies, 26, Dibooglu, S. and Kuan, A. (2001): Sources of real exchange rae flucuaions in ransiion economies: he case of Poland and Hungary, Journal of Comparaive Economies 29, Dornbusch, R. (1976): Expecaions and exchange rae dynamics, Journal of Poliical Economy, 84, Éger, B. (2002): Esimaing he impac of Balassa-Samuelson effec on inflaion and real exchange rae during he ransiion, Economic Sysem 26, Enders, W. and Lee, B.S. (1997): Accouning for real and nominal exchange rae movemens in he pos-breon Woods period, Journal of Inernaional Money and Finance, 16, Engel, C. (1993): Real exchange raes and relaive prices. An empirical invesigaion, Journal of Moneary Economics 32, Engel, C. (1999): Accouning for U.S. real exchange rae changes, Journal of Poliical Economics 107, Engel, C. and Rogers, J. (2000): Violaing he law of one price: should we make a federal case ou of i, Journal of Money, Credi and Banking, 14. Halpern, L. and Wyplosz, C. (1997): Euilibrium exchange raes in ransiion economies, Inernaional Moneary Fund Saff Paper 44, Halpern, L. and C. Wyplosz (2001): Economic ransformaion and real exchange raes in he 2000 s: he Balassa-Samuelson connecion, Manuscrip. 16. Johansen, S. (1988): Saisical Analysis of Coinegraed Vecors, Journal of Economic Dynamic and Conrol 12, Johansen, S. (1992): Esimaion and Hypohesis esing of Coinegraion Vecors in a Gaussian Vecor Auoregressive Models, Economerica 59, Lasrapes, W.D. (1992): Sources of Flucuaions in real and nominal exchange raes, Review of Economics and Saisics 74, Rogers, J. and Jenkins, M. (1995): Haircus and hyseresis: sources of movemens in real exchange raes, Journal of Inernaional Economies 38, Samuelson, P.A. (1964): heoreical noes on rade problems, Review of Economics and Saisics 46, able 1: Variance Decomposiion of Real and Nominal Exchange Raes (a) Sample 1973:1-2000:1 b (uarerly daa) Advanced Economies a (b) Sample 1973:1-1990:4 b (uarerly daa) 48 (c) Sample 1991:1-2000:1 (uarerly daa)

11 Inernaional Business And Economics Research Journal Volume 2, Number 10 RER c NER c RER NER RER NER CAN Horizon d u r e u n u r u n u r u n u r u n u r u n u r u n EZ JAP UK Noes a. CAN: Canada, EZ: Euro Zone, JAP: Japan, UK: Unied Kingdom. b. Sample period for he Euro Zone: 1982:1-2000:1. c. RER: real exchange rae, NER: nominal exchange rae. d. Percenage of forecas error variance accouned for by real shocks. e. Percenage of forecas error variance accouned for by nominal shocks. able 2: Variance Decomposiion of Real and Nominal Exchange Raes (a) Sample 1973:1-2000:1 b (monhly daa) Advanced Economies a (b) Sample 1973:1-1990:12 b (monhly daa) (c) Sample 1991:1-2000:1 (monhly daa) 49

12 Inernaional Business And Economics Research Journal Volume 2, Number 10 RER c NER c RER NER RER NER CAN Horizon d u r e u n u r u n u r u n u r u n u r u n u r u n FRA GER IA JAP UK Noes a. CAN: Canada, EZ: Euro Zone, JAP: Japan, UK: Unied Kingdom. 50

13 Inernaional Business And Economics Research Journal Volume 2, Number 10 b. Sample period for he Euro Zone: 1982:1-2000:1. c. RER: real exchange rae, NER: nominal exchange rae. d. Percenage of forecas error variance accouned for by real shocks. e. Percenage of forecas error variance accouned for by nominal shocks. able 3: Variance Decomposiion of Real and Nominal Exchange Raes ransiion Economies a (USA reference counry) Sample 1991:1-2000:1 b (monhly daa) 51

14 Inernaional Business And Economics Research Journal Volume 2, Number 10 Real Exchange Raes Prices CREP Horizon c u r d u n u r u n HUN Horizon u r u n u r u n POL Horizon u r u n u r u n ROM Horizon u r u n u r u n SLOV Horizon u r u n u r u n Noes. a. CREP: Czech Republic, HUN: Hungary, POL: Poland, ROM: Romania, b. SLOV: Slovenia. c. Sample period for Slovenia: 1991: :12. d. Percenage of forecas error variance accouned for by real shocks. e. Percenage of forecas error variance accouned for by nominal shocks. 52