Exchange Rate Pass-Through to Manufactured Import Prices: The Case of Japan

Transcription

1 Exchange Rae Pass-Through o Manufacured Impor Prces: The Case of Japan Gunerane Wckremasnghe and Param Slvapulle Deparmen of Economercs and Busness Sascs Monash Unversy Caulfeld Vcora, 3145 AUSTRALIA Absrac Ths paper examnes he exchange rae pass-hrough o yen based manufacured mpor prces of Japan usng asymmerc un roo and conegraon ess and asymmerc models. Due o scky prces, for example, here are reasons o beleve ha he degree of pass-hrough depends on wheher he exchange rae apprecaes or deprecaes. The sample used n hs sudy covers he perod January 1975 o June Usng wo sae regme swchng models, he esmaed pass-hrough coeffcens correspondng o apprecaon and deprecaon of he currency are found o be 98 percen and 83 percen respecvely; hese coeffcens are shown o be sgnfcanly dfferen, parcularly n he pos recesson perod. Moreover, we have shown ha he recesson n Japan n he 1990s has sgnfcanly affeced he exchange rae passhrough relaonshp parcularly when he yen deprecaes and ha he proposon ha exchange rae deprecaon and apprecaon have sysemac asymmerc effecs on exchange rae pass-hrough coeffcen. Forcng apprecaons and deprecaons o have he same effecs on he mpor prces does no appear o uncover he rue underlyng exchange rae pass hrough relaonshp. JEL Classfcaon: F31 Key words: Exchange rae pass-hrough, Japan, hreshold auoregresson.

2 1. Inroducon Modellng and esmang he relaonshp beween he exchange raes and mpor/ expor prces have been gven consderable aenon durng he las wo decades 1. One of he major reasons whch led o hs ncreased aenon s he adopon of a flexble exchange rae sysem by many counres snce he collapse of he Breon Woods sysem n Counres adop a flexble exchange rae sysem o remedy he balance of paymen crses hey encouner. However, oucomes of he acons aken by he counres o deprecae he currences were dsapponng snce hey were unable o aan he desred resuls, refung he heorecal proposon ha deprecaon of a currency would boos expors and dscourage mpors, leadng o an equlbrum n he balance of rade. In he case of Japan, alhough he value of s currency was apprecang, had a sgnfcan rade surplus agans he US for a long perod of me ndcang ha he changes n exchange raes were no fully passed hrough o mpor prces. The objecves of hs sudy are o () es wheher he exchange rae passhrough o mpor prces s complee, () undersand he facors nfluencng he mpor prces, () examne wheher he mpor prce responds aymmercally o posve and negave changes n exchange raes and oher varables, (v) calculae a long-run measure of exchange rae pass-hrough for posve and negave changes n exchange raes and (v) asceran wheher exchange rae polcy can be used o acheve balance of rade equlbrum of Japan. Accordng o he heory, a flexble exchange rae polcy s expeced o remedy he balance of paymens crses, alhough emprcal evdence s agans. For example, 1 See for example, Spaeller, 1980; Khosla and Teransh,1989; Dwyer, Ken and Pease, 1994; Parsley, 1995, Menon,1996; Kenny and McGeon, 1998; Webber, 2000; Gl-Pareja,

3 Japan, enjoyed a sgnfcan rade surplus for a long me whle he Japanese yen was apprecang agans he US dollar. Earler approaches o exchange rae pass-hrough relaonshp emanaed from esmang he mpor/expor demand and supply elasces (Branson,1972). Several sudes found ha ncomplee pass-hrough s very common n he shor-run and does no carry hrough o he long-run (Blejer and Hllman, 1982). Accordngly, complee exchange rae pass-hrough s a long-run phenomenon. In heory, ncomplee pass-hrough s manly due o marke srucure and produc dfferenaon (Menon, 1996). In a perfecly compeve marke where mpored and domescally produced goods are perfec subsues, he measuremen of pass-hrough s smlar o ha of elasces approach. When mperfec compeon exss, he frms can charge a mark-up on her coss o earn above normal profs even n he long-run. Ths markup can vary dependng on he degree of subsuably beween he domescally produced goods and mpored goods as deermned by he produc dfferenaon and he degree of marke negraon and separaon. In order o examne he heorecal proposons on he degree exchange rae pass-hrough, here has been a plehora of emprcal sudes usng varous mehodologes and daa seres across dfferen counres durng he pas hree decades. We dscuss some of hese sudes and he resuls. Menon (1993) suded he exchange rae pass-hrough o mpor prces of moor vehcles usng Engle-Granger s wo-sep conegraon ess and error correcon models and found ha exchange rae passhrough o mpor prces of moor vehcles s ncomplee even n he long-run. He has pu forward wo possble explanaons for he ncomplee pass-hrough: he presence of quanave resrcons and he prcng pracces on nra-frm sales by mulnaonal frms. Dwyer, Ken and Peace (1994), on he oher hand, found ha n 2

4 he long-run exchange pass-hrough over he docks s complee for boh Ausralan mpors and manufacured expors. However, hey found ha responses o currency movemens dffer sgnfcanly n he shor-run. Pass-hrough o mpor prces s found o be more rapd han ha o manufacured expor prces. See also Dwyer and Lam (1994). Parsley (1995) suded he US mpors of ball bearngs, bols, nus, screws and bcycles and showed ha n general expecaons of fuure exchange raes should be a deermnan of pass-hrough and concluded ha he pass-hrough may change when here are changes n expeced fuure exchange raes. Luz and Relly (1997) used daa for welve European counres and found ha he exchange rae pass-hrough s below 50 percen for all hese counres. Ther sudy furher revealed ha relave marke share of domesc frms or mposon of non-arff barrers has no relaonshp wh he low degree of pass-hrough, whle he sudy by Adolfson (1997) ndcaed he prcng o marke behavour n he majory of ndusres s he reason for lmed passhrough 2. Takng sudes on exchange rae pass-hrough n he leraure and economerc mehodologes developed recenly ogeher, we beleve ha a par of he reason for he lack of emprcal suppor for he exchange rae pass hrough s ha prevous emprcal sudes dd no ncorporae some well-known sylsed facs no modellng and esmang he relaonshp beween he mpor prces and he exchange raes. Our sudy aemps o () address ssues such as asymmerc adjusmen of he ndvdual varables n esng for un roos and conegraon, () examne wheher he response 2 See Lee (1997) for a sudy on he pass-hrough of exchange raes o varous Korean ndusres, Kenny and McGegan (1998) for he relaonshp beween mpor prce and exchange rae n Ireland, and Gross and Schm (1999) for he exchange rae pass-hrough o mpor prces of small and medum szed auomobles n Swss daa. Webber (2000) found sgnfcan long-run esmaes of mpor passhrough n seven ou of nne counres n he Asa-Pacfc regon. 3

5 of he mpor prce ndces o posve and negave changes n exchange raes s asymmerc; and () use a new model for he long-run exchange rae pass-hrough o examne he asymmerc response of mpor prce o changes n exchange raes. See Enders and Granger (1998) for mehodologes for esng asymmerc un roos, conegraon of varables and ohers. Snce he Japanese economy was n recesson snce 1989, we wll examne s effecs on he exchange rae pass hrough relaonshp. Such a sudy wll undoubedly conrbue o he avalable vas leraure on exchange rae pass-hrough relaonshp, and more mporanly, o he debae beween he US and Japan wh regard o Japanese rade surplus agans he US even when s currency was apprecang. The paper s organzed as follows: Secon 2 brefly explans he developmen of he exchange rae pass-hrough model. The mehodology used n hs sudy s dscussed n Secon 3. Measuremen of he varables used n hs sudy s gven n Secon 4. Secon 5 dscusses he daa seres used and her me seres properes. The emprcal analyss and he resuls are repored and dscussed n Secon 6 and he fnal Secon gves he concluson of he sudy. 2. The Basc Model of Exchange Rae Pass-Through Exchange rae pass-hrough can be defned as he exen o whch changes n nomnal exchange raes are refleced n mpor prces. Ths depends on varous facors such as produc dfferenaon and he naure of compeon. Speakng broadly, exporers se her prce (PX * ) a a prof margn (λ) over he cos of producon (C * ). Therefore, he mpor prce (PM) can be defned as follows: PM = PX ER = ( 1 + λ ) C ER (1) 4

6 where (ER) s he exchange rae. Now, se (1+ λ) = ρ, where ρ s he prof mark-up. I has been hypohessed ha exporers base her prcng decsons on compeve pressures n he domesc marke, whch are proxed by he gap beween he prces of domesc mpor compeng goods (PD) and exporers cos n domesc currency (Hooper and Mann, 1989). The prof mark-up can hus be modelled as: ρ { PD /( C ER)} α = (2) Subsung ρ n equaon (2) for (1+ λ) n equaon (1) and akng logarhms of varables (denong hem n lower case leers), we can derve he equaon for he mpor prce as follows: pm = αpd + ( 1 α) c * + (1 α) er (3) The pass-hrough coeffcen (1-α) s expeced o be beween 0 and 1. If he foregn frm s a prce aker n he Japanese compeve marke, hen α = 1 and herefore, he pass-hrough s zero. In hs case, can be seen ha he Japanese mpor prce se by foregn frm s equal o he Japanese domesc prce and changes n exchange raes and foregn coss have no effec on mpor prces. If he foregn frms do no face any compeon n he Japanese marke, boh he changes n he exchange raes and he foregn coss are fully passed hrough o mpor prces leavng he mark-up unchanged. In hs case α = 0. We wll es hese resrcons n Secon 6. A rend varable wll be ncluded n he above model o capure he changes n he varables due o mprovemen n he echnology over me. 3. Mehodology Snce growng evdence suggess ha me seres are nherenly non-lnear, we assume ha adjusmens of varables o her respecve equlbrums are asymmerc, dependng on wheher he shor-erm flucuaons from he long-run equlbrums are posves or negaves. Usng he mehodology developed n Enders and Granger 5

7 (1998), we es for he possbly of asymmerc adjusmens of me seres used n hs sudy. Furher, we compare he me seres properes derved from he asymmerc models wh hose obaned by forcng he varables o have symmerc adjusmens, ha s, symmerc models. Dckey and Fuller (1979) (DF) assume lneary and symmerc adjusmen. In oher words, n such models he null hypohess of un roos s esed agans he alernave hypohess of a symmerc adjusmen of he varable of neres. However, recen sudes have found ha mporan economc and fnancal varables dsplay asymmerc adjusmen pahs (see, for example, Nef, 1984; Falk, 1986 and Enders and Granger,1998). Therefore, here s a possbly ha he well-known DF ype un roo ess rejec he null hypohess of negraon due o an ncorrec alernave hypohess. To overcome hs problem, Enders and Granger (1998) generalzed he DF es by allowng for an asymmerc adjusmen of varables. There are wo major alernave models, whch were proposed by Enders and Granger o deal wh asymmerc adjusmen pahs of varables. These are known as he hreshold auoregressve (TAR) model and he momenum hreshold auoregressve (MTAR) model, dependng on eher he lagged levels or he lagged dfferences of he varables of neres are used n defnng he heavsde ndcaor funcon. Now, we brefly dscuss he adjusmen process of a me seres varable, y. Consder he model: y = I ρ ( + ε (5) 1 y I ) ρ 2 y 1 where I s he heavsde ndcaor funcon, whch can ake one of he followng four forms: 6

8 1f y -1 0 I = (6) 0 f y -1 < 0 or 1f y -1 a0 I = (7) 0 f y -1 < a0 or 1f y -1 a0 + a1( 1) I = (8) 0 f y -1 < a0 + a1( 1) or 1f y -1 0 I = (9) 0 f y 1 < 0 The models wh ndcaor funcons (7) and (8) are known as TAR models. When equaon (7) s used o se he heavsde ndcaor funcon, s assumed ha he longrun equlbrum occurs a pon y = a 0, provded 2 <(ρ 1, ρ 2 )<0. In hs case, f ρ 1 = ρ 2 = 0, hen he seres s a pure random walk. When he heavsde ndcaor funcon s defned accordng o (8) and f 2 <(ρ 1, ρ 2 )<0, he rend lne y = a 0 + a 1 s an aracor such ha he ( y ) me seres s rend saonary. When y -1 s above he rend lne, he me seres ends o decay a he rae of ρ 1 and when s below y -1, he me seres ends o decay a he rae of ρ 2. If eher ρ 1 or ρ 2 les ousde he nerval (- 2,0), hen he sequence may no be rend saonary. In he heavsde ndcaor funcon (6), s assumed ha he y = 0 s he longrun equlbrum value of he me seres. Therefore, f y -1 > y, hen he adjusmen s ρ 1 y -1, and f y -1 < y, hen he adjusmen s ρ 2 y -1,. Ths ype of TAR model can capure aspecs of deep movemens n a seres. The models usng he heavsde ndcaor funcons n equaons (6) and (9) are known as MTAR models. Replacng heavsde ndcaor funcon (6) wh (9) s useful when he adjusmen s asymmerc 7

9 n such a way ha he seres exhbs more momenum n one drecon han he oher (Enders and Granger, 1998). Equaon (5) can be augmened wh he lagged changes n he y o overcome any problem due o seral correlaon n he error erm, ε. In order o selec he opmal lag lengh, he dagnosc checks such as Ljung-Box es and varous model selecon crera such as AIC and/or BIC can be used. If he un roo ess ndcae ha he varables are negraed of he same order, here may exs a conegrang relaonshp beween he varables of neres. If he varables are conegraed, he long run relaonshp arsng from he followng equaon should be saonary or I(0) (Engle and Granger, 1987): pm = β + β er + β c + β c (10) * In order o examne he long-run asymmery of mpor prces o exchange rae changes, a new varable s consruced as n Webber (2000). Beng n lne wh Webber, he exchange rae a any pon n me can usually be expressed as follows: er er + er + er = 0 A D where er s he nal exchange rae, er A = 1 θ ( er er θ er and 1), = 1 for er < 1 D * * θ = 0 for er > er 1, and er θ ( er er 1 ), θ = 1for er > er 1. Accordngly, = 1 he varable A er represens he accumulaed sum of apprecaon epsodes and he varable D er represens he accumulaed sum of deprecaon epsodes. To es for long-run asymmery, only he daa seres relang o one of he epsodes has o be ncluded n he exchange rae pass-hrough equaon. Hence only he seres for deprecaon epsodes s ncluded n he ess for long-run asymmery. Therefore, our new long-run exchange rae pass-hrough equaon s: 8

10 pm = β c + u (11) D 0 + β1er + β 2er + β 3c + β 4 * If pm er, c * c and D er are conegraed, hen here exss a long-run relaonshp among he varables. In he above equaon, he long-run exchange rae pass-hrough coeffcens are ( β 1 + β 2 ) and β 1, correspondng o deprecaon and apprecaon respecvely. Therefore, he resrcon ha he long-run deprecaon pass-hrough s equal o zero ( β 2 = 0 ) s he es agans he long-run asymmery of he mpor prce o exchange rae changes, whch s esed whn boh he Engle and Granger and Johansen (1991,1995) conegraon frameworks. If pm er, c and c * are co-negraed, hen he shor-run asymmery s esed by examnng he followng asymmerc error-correcon model (ECM): pm = φ + k 1 + ω pm + θ er 1 + θ er 1 + = 1 + k 1 = 1 k 1 = 1 λ c * 1 + ϕε 1 k 1 = 1 + υ k 1 = 1 γ c (12) where s he dfference operaor, ε -1 s a one perod lagged error erm from he conegrang equaon and υ s a whe-nose error erm. Wheher he response of he mpor prce changes o posve and negave changes n he exchange rae s asymmerc s esed usng he Wald-ype F-es. For comparson purpose, he above asymmerc error correcon model and a symmerc or a sandard error correcon models are also esmaed. 4. Measuremen of Varables In hs secon, we dscuss how he varables used n he models were consruced. The varables consruced n hs secon are domesc mpor prce, exchange rae ndex, foregn cos ndex and domesc producon cos ndex. 9

11 Domesc Impor Prce The prce ndex for he manufacured mpors s used as a proxy for he mpor prce varable. Several sudes used varous proxes for he mpor prce varable, ncludng un value of mpors and he wholesale prce ndex of radng parner counres. These proxes suffer from varous lmaons: un values are suable only when hey are appled o a sngle produc raher han o a group of heerogeneous producs, hus when he producs are no smlar hs s no a suable proxy for he mpor prce as he un values reflec he changes of prces of dssmlar commodes. Furher, he changes n commody composon of mpors make he un values an napproprae proxy for mpor prce. Wholesale prce ndces hemselves suffer from ceran lmaons. In her consrucon, non-raded goods are also aken no accoun and hey even nclude he changes n coss of goods, whch are no produced for exporng. Furher, hese prce ndces are consruced usng domesc weghs raher han nernaonal weghs. Japan s among he few counres (along wh Ausrala, Germany and he US) for whch mpor prce ndces are avalable for an adequae lengh of me. Snce hese ndces are based on ransacon prces, hey are free from lmaons faced by many prevous sudes ha used aforemenoned proxes for he mpor prce ndex. Beng conssen wh he Sandard Inernaonal Trade Classfcaon (SITC) of he Uned Naons for manufacured commodes, he mpor prce ndex for manufacured mpors of Japan s consruced by combnng he mpor ndces for exles, meals and relaed producs, wood, lumber and relaed producs, chemcals, machnery and equpmen usng he respecve weghs ha were ulsed n 10

12 consrucng he overall mpor prce ndex of Japan. Therefore, he mpor prce ndex for he manufacured goods (pm) s consruced usng he followng formula: pm k = w pm s where, w s s he wegh assgned o secor s, pm s s he mpor prce ndex of secor s. Two mpor prce ndces are consruced as above: one on he bass of yen and he oher on he bass of he conracual currency. Exchange Rae Index In hs sudy, we use a new Exchange rae ndex based on he conracual currency composon of mpors o Japan. Ahukorala and Menon (1994) have been he frs o use a smlar ndex of exchange raes n her sudy on prcng o marke behavour and exchange rae pass-hrough o expor prces of Japan. Ther ndex has been consruced by deflang he expor prce ndex n yen by an ndex represenng he conracual currency prce of expors. A smlar exchange rae ndex reflecng he acual currency composon of mpors s used n hs paper whch s consruced by deflang he domesc currency mpor prce ndex by he conracual currency based mpor prce ndex. An exchange rae ndex based on conracual currency prces beer reflecs real currency composon of mpors/expors han he nomnal effecve exchange rae ndex, whch was exensvely used n many prevous sudes on exchange rae pass-hrough o mpor/expor prces. Moreover, he nomnal effecve exchange rae may dsor he realy by assgnng hgher weghs o ceran currences on he bass of mpor shares alhough such a proporon of mpors may no have been nvoced n hese currences. A sngle conracual currency based exchange rae ndex for he oal manufacured mpors was consruced by usng he followng formula: 11

13 er = k pm w pm C where er s he conracual currency based exchange rae ndex, w s he wegh assgned o conracual currency mpor prce ndex for caegory, pm s he yen based mpor prce ndex for oal manufacured mpors explaned above, and C pm s he conracual currency mpor prce ndex for caegory. Foregn cos ndex The Foregn cos ndex was compued as: c * = k = 1 w c * where c* s he foregn cos ndex, w s he share of manufacured mpors o Japan from counry, and c* s a proxy for cos of producon of counry. To proxy he foregn producon cos, producer prce ndex for manufacurng was used. When producer prce ndex for manufacurng was no avalable for he respecve counry, wholesale prce ndex was used as a proxy for he producon cos. Deals of proxes used o represen foregn producon cos are gven n Appendx C. Index of Domesc producon cos Producer prce ndex for he manufacurng secor of Japan was consdered as a proxy for he domesc cos of producon. 5. Daa seres and he me seres properes The monhly daa seres used n hs sudy spans he perod 1975:01 o 1997:06. In order o examne how he recesson n Japan experenced snce has affeced 12

14 he exchange rae pass hrough relaonshp, he daa seres was dvded no wo subsample perods: 1975:01 o 1989:12 and 1990:01 o 1997:06. We beleve ha hs relaonshp was affeced durng he recesson, parcularly when he yen based mpor prce ndex s used n he emprcal analyss. Fgure 1 n Appendx B shows he me seres behavour of he varables used n hs sudy. Fgure 1 depcs ha exchange rae and he yen based mpor prce ndex follow each oher closely. Overall here was an apprecaon of he exchange rae ll March 1995, hen here s a dp n he exchange rae around Aprl 1995, ndcang a sudden ncrease n he value of yen. I s clear from he graph ha he yen based mpor prce ndex has responded quckly o he change n he exchange rae wh a delay of wo monhs. Subsequenly, he exchange rae was deprecang ll Aprl Ths observed paern s also refleced n he mpor prce ndex. Foregn cos ndex has been on he ncrease durng he whole sample perod. However, he mpor prce has no followed he changes n foregn coss ndcang ha foregn exporers absorb he changes n cos of producon no prof margns. The Japanese cos of producon shows has been sable, ndcang ha over he long-run flucuaons n cos of producon are saonary. Therefore, he exchange rae s expeced o emerge as he man deermnan of he yen based mpor prces. Table 1 repor he resuls of he un roo ess agans TAR/MTAR adjusmens for levels of he varables. As shown n column seven n Table 1, none of he varables was found o be saonary n her levels. However, he frs dfferences were found o be saonary. Furher, none of he varables shows any asymmerc adjusmen owards s equlbrum pon (consan (µ) or rend aracor (T). Noe ha, n he absence of asymmerc adjusmens of he varables, he ADF es was shown o 13

15 be a more powerful es han he asymmerc un roo ess. The ADF es resuls repored n Table 2 are conssen wh hose repored n Table Emprcal Analyss and he Resuls Conegraon es resuls usng he un roo ess agans TAR and MTAR adjusmen are gven n Table 3. The es for a un roo n he resduals from he conegrang equaon (long-run exchange rae pass-hrough equaon) shown n columns wo and four n he able rejecs he null hypohess of no conegraon a he one percen level. However, he resuls repored n columns hree and fve do no rejec he hypohess of symmerc adjusmen of he conegrang resduals owards her long-run equlbrum pon. I s evden ha he convenonal un roo ess fal o deec he long run relaonshp, whle he asymmerc un roo ess uncover such relaonshp. If he long run relaonshp s undeeced, here s a possbly of msspecfcaon of he model n he subsequen analyss. We carred ou he analyss for he full and wo sub-samples wh and whou mposng he asymmery, and he resuls are repored n Tables 4 and 5. The resuls of he symmerc long-run pass-hrough equaon sugges ha all he ndependen varables are sgnfcan a he one percen level excep n one case. Tha s n he second sub-sample perod only he exchange rae ndex s sgnfcan when he Johansen mehod s used. Esmaon resuls for he asymmerc exchange rae passhrough equaon repored n Table 5 ndcae ha he exchange rae varable for deprecaon epsodes s sgnfcan only n he second sub-sample perod. The apprecaon exchange rae pass-hrough coeffcens vary beween 62 percen and 114 percen for he EG mehod and he 98 percen and 157 percen for he Johansen mehod among he hree sample perods used. When deprecaon pass-hrough 14

16 coeffcens are consdered hey vary beween 48 percen and 76 percen for he EG mehod and 18 percen and 124 percen for he Johansen mehod among he hree sample perods used. The hypohess ha he coeffcen of he asymmery varable s zero s repored n Table 7 and s rejeced a one percen level only durng he second sub-sample perod for he EG mehod. However, he Johansen mehod rejecs he same n all he sample perods. The sably of he exchange rae pass-hrough coeffcen was examned by esmang exchange rae pass-hrough equaons recursvely and he me varyng pass-hrough coeffcens are ploed n Fgure 2 n Appendx B wh her wo sandard error bands. The Fgure shows ha he passhrough coeffcens were ncreasng snce Ocober 1980 ll March 1981 and hen decreasng ll he end of June Durng he enre sample perod, he exchange rae pass-hrough coeffcens vared beween and To examne he sgnfcance of he shor-run dynamcs of he exchange rae n he pass-hrough relaonshp, wo error correcon models were esmaed. The resuls for he error-correcon models are repored n Table 6. Accordng o he resuls of he asymmerc error-correcon model, only posve changes n exchange rae (deprecaon) are sgnfcan deermnans of mpor prces durng he oal sample perod whle hs varable and he negave changes were nsgnfcan n he wo subsample perods. When he symmerc error-correcon model s consdered, changes n he exchange rae varable were sgnfcan only durng he oal sub-sample perod. Foregn cos varable and he error-correcon erm are sgnfcan only durng he second sub-sample perod n boh symmerc and asymmerc models. 15

17 7. Concluson Ths paper nvesgaes he exchange rae pass-hrough o yen based manufacured mpor prces of Japan usng asymmerc un roo and conegraon ess and asymmerc models. Due o scky prces, for example, here are reasons o beleve ha he degree of pass-hrough depends on wheher he exchange rae apprecaes or deprecaes. The sample used n hs sudy covers he perod January 1975 o June Snce Japan experenced a recesson snce 1989, he daa seres was dvded no wo sub-sample perods: 1975:01 o 1989:12 and 1990:01 o 1997:06 o fnd evdence of sgnfcanly dfferen exchange rae pass-hrough relaonshps over he pre- and pos-recesson perods. We beleve ha hs relaonshp was affeced durng he recesson, parcularly when he conracual currency based exchange rae ndex s used n he emprcal analyss Usng wo sae regme swchng models, he esmaed pass-hrough coeffcens correspondng o apprecaon and deprecaon of he currency are found o be 98 percen and 83 percen respecvely for he Johansen mehod for he oal sample perod; hese coeffcens are shown o be sgnfcanly dfferen, parcularly n he pos recesson perod. The correspondng exchange rae pass-hrough coeffcens for he EG mehod are 70 percen and 72 percen respecvely whch are no sgnfcanly dfferen from each oher. Moreover, we have shown ha he recesson n Japan n he 1990s has adversely affeced he exchange rae pass-hrough relaonshp; boh EG and Johansen mehods ndcae a sgnfcan asymmerc log-run response of mpor prces o changes n exchange raes. Forcng apprecaons and deprecaons o have he same effecs on he mpor prces, he model does no appear o uncover he rue underlyng relaonshp. The analyss carred ou n hs paper suggess ha one 16

18 can learn much abou he underlyng adjusmen processes n un roo esng, and he possbly of dfferen responses of mpor prces o exchange rae changes. 17

19 APPENDIX A Table 1. Resuls of he ess for asymmerc un roos Sample Perod Varable Model Levels Frs Dfferences Tes for Un roos Tes for asymmery Tes for Un roos Tes for asymmery 1975:01 o 1997:06 er TAR (T) (µ)** MTAR (T) (µ)** pm TAR (T) (µ)** MTAR (T) (µ)** c* TAR (T) (T) MTAR (T) (T) c TAR (T) (T) ** MTAR (T) (T) ** :01 o 1989:12 er TAR (T) (µ)** MTAR (T) (µ)** pm TAR (T) (µ)** MTAR (T) (µ)** c* TAR (T) (T) MTAR (T) (T) c TAR (T) T)** MTAR (T) T)** :01 o 1997:06 er TAR (T) (µ)** MTAR (T) (µ)** pm TAR (T) (µ)** MTAR (T) (µ)** c* TAR (T) (T) MTAR (T) (T) c TAR (T) (T)** MTAR (T) (T)** Noes: er, pm, c*, and c are conracual currency exchange rae ndex, mpor prce ndex n Japanese yen, foregn cos ndex and he domesc cos ndex. Columns four and fve of he able repor he resuls of he es for un roos usng equaon (5) n he ex augmened wh lags of he dependen varable o produce whenose resduals. The number of lags of he dependen varable was deermned by AIC and Ljung-Box-Q sasc. The null hypohess for he un roos usng equaon (5) s ρ 1 =ρ 2 =0 whereas ha for asymmery s ρ 1 =ρ 2. µ and T whn brackes wh a un roo es sasc ndcae wheher he heavsde ndcaor funcon s a consan or a consan and a me rend respecvely. Snce he un roo es sasc n hs case has a non-sandard dsrbuon, crcal values generaed by Enders and Granger whch are gven below were used; Ender and Granger s Crcal Values (1998, Table 1, p.306) Model Sample sze Esmaed Consan Aracor Esmaed Trend Aracor Probably of a smaller value Probably of a smaller value 90% 95% 99% 90% 95% 99% TAR MTAR To make nferences regardng he hypohess for asymmery sandard F-sasc are used (Enders and Granger, 1998, p.307). 18

20 Table 2: Augmened Dckey-Fuller es resuls for Un roos Sample Perod Varable Levels Frs Dfferences Wh Inercep Wh nercep & Trend Wh Inercep Wh nercep & Trend 1975:01 o 1997:06 er (1) (2) (2)** (2)** er D (2) (2) (1)** (1)** pm (1) (1) (2)** (2)** c (3) (3) (2)** (2)** c* (5) (5) (4)** (1)** 1975:01 o 1989:12 er (1) (1) (2)** (2)** er D (1) (1) (2) ** (2)** pm (1) (1) (2)** (2)** c (3) (3) (2)* (2)* c* (1) (1) (2)* (2)* 1990:01 o 1997:06 er (1) (1) (1)** (1)** er D (3) (3) (2)** (2)** pm (1) (1) (1)** (1)** c (4) (4) (1)** (1)** c* (4) (3) (3)** (3)** Noe: See Table 1 for he defnon of he varables excep for er D whch s he exchange rae varable for deprecaon epsodes. ** ( * ) ndcaes sgnfcance a one (fve) percen level. Fgures whn brackes assocaed wh ADF -sascs n columns 3 o 6 ndcae he number of lags of he dependen varable n he ADF regresson o elmnae he seral correlaon from he resduals. Table 3: Conegraon es resuls usng asymmerc un roos Sample Perod Model TAR MTAR Tes for un roos Tes for asymmery Tes for un roos Tes for asymmery 1975:01 o 1997: (µ)*** (µ)*** :01 o 1989: (µ)* (µ)* :01 o 1997: (µ)** (µ)** Noe: ***, ** and * denoe sgnfcance a one, fve and en percen level respecvely. The lags of TAR and MTAR models were seleced usng AIC. Varables ncluded n conegraon regresson are pm, er, c, and c*. See noe for Table 1 and 2 for he defnons of he above varables and crcal values. Table 4: Esmaon resuls for he symmerc long-run exchange rae pass-hrough equaon, Dependen varable: Yen based mpor prce ndex Varable Engle-Granger Mehod Johansen Mehod 1975:01 o 1997: :01 o 1989: :01 o 1997: :01 o 1997: :01 o 1989: :01 o 1997:06 Consan ** Trend ** ** ** ** ** er 0.713** 0.677** 0.769** 0.958** 1.016** 0.633** c* 0.274** 0.571** ** 1.641** ** c 0.773** 0.555** 0.969** ** 2.529** ** sgnfcance a 1% level. In he Johansen conegraon equaon, he dependen varable ha s he yen based mpor prce ndex was normalsed o one. 19

21 Table 5: Esmaon resuls for he asymmerc long-run exchange rae pass-hrough equaon, Dependen varable: yen based mpor prce ndex Varable Engle-Granger Mehod Johansen Mehod 1975:01 o 1997: :01 o 1989: :01 o 1997: :01 o 1997: :01 o 1989: :01 o 1997:06 Consan Trend ** ** ** er 0.704** 0.620** ** 1.574** 1.144** er D ** c 0.230** 0.685** ** ** c* 0.760** 0.422** ** 2.995** 2.130** Noes: ** sgnfcance a 1% level. In he Johansen conegraon equaon, he dependen varable ha s he yen based mpor prce ndex was normalsed o one. Table 6: Resuls of he esmaon of error correcon models, Dependen varable: frs dfferences of he yen based mpor prce ndex Varable Sample perod 1975:01 o 1997: :01 o 1989: :01 o 1997:06 Symmerc Asymmerc Symmerc Asymmerc Symmerc Asymmerc ε ** ** pm (-1) 0.751*** 0.745*** 0.769*** *** ** er (-1) * per (-1) * ner (-1) c (-1) c* (-1) ** ** Noes: ***, ** and * ndcae sgnfcance a one, fve and en percen levels respecvely. The fgures whn brackes assocaed wh a varable n column one denoe he frs lag of he respecve varable. ε -1 denoes he error-correcon erm whch s he one perod lagged resdual from he symmerc conegrang equaon. denoes he frs dfference of a varable. See noe for Table 1 for he defnons of he varables. Table 7: Tes resuls for long-run asymmery Sample Perod Ch-square es sasc Engle-Granger Mehod Johansen Mehod 1975:01 o 1997: * 1975:01 o 1989: * 1990:01 o 1997: ** 2.844* Noes: **, and * ndcae sgnfcance a one and fve percen levels respecvely. Ch-square es s relevan for esng null hypohess ha he coeffcen of he asymmery varable (er D ) s sgnfcanly dfferen from zero. 20

22 APPENDIX B Fgure 1. Tme seres behavour of he log of he varables used n he exchange rae pass-hrough equaon Exchange Rae Domesc Cos n Japan Foregn Cos Impor Prce Index Fgure 2 Recursve Esmae of Exchange Rae Pass-hrough Coeffcen (er) and s Two Sandard Error (SE) Bands for Toal Manufacured Impors from Ocober 1980 o June 1997 usng an nal sample of 70 observaons EPTCoeffcen -2SE +2SE 21

23 APPENDIX C Japan s manufacured Impor shares Counry USA Korea Germany UK Thaland Sngapore France Indonesa Swzerland Sweden Canada Phlpnes Ausrala Ireland Inda Neherlands Span Wegh Daa Sources Exchange raes for all he counres were obaned from Inernaonal Fnancal Sascs (IFS) CD-ROM-2000 of Inernaonal Moneary Fund. Producer Prce Index for manufacurng for Canada, Japan, Korea, France, Germany, Ireland, Swzerland, he USA and he UK were obaned from OECD man Economc Indcaors and hose for he oher counres were obaned from he IFS CD-ROM, Prce ndces for manufacured mpors and her respecve weghs were obaned from he Bank of Japan webse. Value of manufacured mpors accordng o SITC classfcaon used o calculae mpor share were obaned from Foregn Trade by Commodes for publshed by OECD. 22

24 References Adolfson, M. (1997), Exchange Rae Pass-Through o Swedsh Impor Prces, Workng Paper (Augus), Deparmen of Economcs, Sockholm School of Economcs. Ahukorala, P. and Menon, J. (1994), Prcng o Marke Behavour and Exchange Rae Pass-Through n Japanese Expors, Economc Journal, 104, Blejer,M.I. and Hllman, A.L. (1982), A Proposon on Shor-Run Deparures from he Law of One Prce, European Economc Revew, 17, Branson, W.H. (1972), The Effecs of he 1971 Currency Realgnmen, Brookngs Papers on Economc Acvy,1, Dckey, D.A. and Fuller, W.A. (1979), Dsrbuon for he Esmaors for Auo Regressve Tme Seres Wh a Un Roo, JASA,74, Dwyer, J., Ken, C. and Pease, A. (1994), Exchange Rae Pass-Through: Tesng he Small Counry Assumpon for Ausrala, Economc Record, 70, Dwyer, J. and Lam, R. (1994), Explanng Impor Prce Inflaon: A Recen Hsory of Second Sage Pass-Through, Research Dscusson Paper 9407, Reserve Bank of Ausrala. Enders, W. and Granger, C.W.J. (1998), Un-Roo Tess and Asymmerc Adjusmen Wh an Example Usng he Term Srucure of Ineres Raes, Journal of Busness and Economc Sascs,16, Engle, R.F. and Granger, C.W.J. (1987), Conegraon and Error Correcon: Represenaon, Esmaon and Tesng, Economerca, 55, Falk, B. (1986), Furher Evdence on he Asymmerc Behavour of Economc Tme Seres over he Busness Cycle, Journal of Polcal Economy, 94, Gl-Pareja, S. (2000), Exchange Raes and European Counres Expor Prces: An Emprcal Tes of Asymmeres n Prcng o Marke Behavour, Welwschaflches Archv, 136, Gron, A. and Swenson, D.L. (1996), Incomplee Exchange-Rae Pass-Through and Imperfec Compeon: The Effec of Local Producon, AEA Papers and Proceedngs (May), Gross, D.M. and Schm, N. (1996), Exchange Rae Pass-Through and Rvalry n he Swss Auomoble Marke, Welwrschaflches Archv, 132,

25 Hooper, P. and Mann, C.L. (1989), Exchange Rae Pass-hrough n he 1980s: The Case of U.S. Impors of Manufacures, Brookngs Papers on Economc Acvy, (March), Johansen, S. (1991), Esmaon and Hypohess Tesng of Conegraon Vecors n Gaussan Vecor Auoregressve Models, Economerca, 59, Johansen, S. (1995) Lkelhood-based Inference n Conegraed Vecor Auoregressve Models, Oxford Unversy Press. Kenny, G. and McGeon, D. (1998), Exchange Raes and Impor Prces for a Small Open Economy, Appled Economcs, 30, Khosla, A. and Teransh, J. (1989), Exchange Rae Pass-hrough n Expors Prces An Inernaonal Comparson, Hoshubash Journal of Economcs, 30, Lee, J. (1997), The Response of Exchange Rae Pass-Through o Marke Concenraon n a Small Economy: The Evdence from Korea, Revew of Economcs and Sascs,79, Luz, M.G. and Relly, B. (1997), Incomplee Exchange Rae Pass-Through n he European Car Marke: Evdence from Hedonc Prce Regressons, Dscusson Paper 9715, Deparmen of Economcs, Unversy of S. Gallen. Menon, J. (1993), Exchange Rae Pass-Through: Ausralan Impors of Moor vehcles, Inernaonal Economc Journal, 7, Menon, J. (1996), Exchange Raes and Prces: The Case of Ausralan Manufacured Impors, Lecure Noes n Economcs and Managemen Sysems, 433, Berln: Sprnger. Nef, S. N. (1984), Are Economc Tme Seres Asymmerc Over he Busness Cycle, Journal of Polcal Economy, 92, Parsley, D. (1995), Exchange Rae Pass-Through wh Ineremporal Lnkages: Evdence a he Commody Level, Revew of Inernaonal Economcs, 3, Spaller, E. (1980), Shor-run Effecs of Exchange Rae Changes on he Terms of Trade and Trade Balance, IMF Saff Papers, 27(2), Webber, A. (2000), Newon s Gravy Law and Impor Prces n he Asa Pacfc Japan and he World Economy, 12, Yang, W. (1997), Exchange Rae Pass-Through n he U.S. Manufacurng Indusres, Revew of Economcs and Sascs, 79,

26 25