The International Journal of Economic Policy Studies

Transcription

1 The Inernaional Journal of Economic Policy Sudies Volume Aricle 3 PASSTHROUGH OF EXCHANGE RATE INTO DOMESTIC PRICES: THE CASE OF FOUR EASTASIAN COUNTRIES 1 Siok Kun SEK Add: Olshausensr, 66A007, 24118, Kiel Tel.: +49(0) siokksek@bwl.unikiel.de Zhanna KAPSALYAMOVA Add: Olshausensr., 68001, 24118, Kiel Tel.: +49 (0) kapsalyamova@bwl.unikiel.de ABSTRACT The paper underakes a comparaive empirical analysis of he effecs of shocks on domesic prices in four Asian counries before and afer he financial crisis of 1997 in SouhEas Asia. We apply wo differen esimaion mehodologies, namely srucural VAR and a single equaion approach. The resuls of he wo mehods are consisen, alhough he magniude of he elasiciies of he exchange rae passhrough are differen due o he inclusion of differen variables, lag erms and differen assumpions made in boh mehods. The resuls show ha he degrees of exchange rae passhrough in hese counries are differen across counries and over ime. In mos cases, he passhrough raes are incomplee. The degree of exchange rae passhrough is highes on impor prices, moderae on PPI and lowes on CPI. In some cases, he passhrough raes on CPI are even negaive. The effec of he impor price shock is sronger compared o ha of he exchange rae shock in deermining he movemen of domesic prices in hese counries. Trade openness has a weak correlaion wih he degree of exchange rae passhrough. Keywords: domesic prices, exchange rae passhrough, SVAR, single equaion approach JEL classificaion: C22, C32, F41 1 We are graeful o Prof. Yoshino from he Takushoku Universiy (Tokyo), Bjorn van Roye from he Kiel Insiue for he World Economy for heir invaluable commens on he earlier drafs of his paper.

2 The Inernaional Journal of Economic Policy Sudies PASSTHROUGH OF EXCHANGE RATE INTO DOMESTIC PRICES: THE CASE OF FOUR EASTASIAN COUNTRIES 1. Inroducion One of he main issues in inernaional macroeconomics is he relaionship beween exchange rae movemens and price adjusmens of raded goods, he socalled exchange rae passhrough. Exchange rae passhrough is defined as he percenage change in domesic/impored prices led by a one percenage change in he exchange rae beween he imporer and exporer currency. According o Sahminan (2002) and An (2006), exchange rae movemens ransmi o domesic prices hrough hree channels: impored consumpion goods, impored inermediae goods and domesic goods priced in foreign currency prices. A oneoone response of domesic prices o exchange rae changes is known as a full/complee passhrough while a parial/incomplee passhrough occurs when here is a less han oneoone response in prices as a resul of exchange rae changes. In a real siuaion, passhrough of nominal exchange rae changes o domesic prices is incomplee. For insance, Campa & Goldberg (2001) esimae exchange rae passhrough equaions for 25 OECD counries for a period beween 1975 and They find ha a hypohesis of complee shor run passhrough can be rejeced in 22 ou of 25 counries. Bu in conras, hey are only able o rejec complee long run passhrough in 9 of 25 counries. More evidence of imperfec passhrough is shown in Campa & Gonzalez (2002). A lower degree of passhrough implies ha he nominal exchange rae leads o lower expendiure swiching effecs of domesic moneary policy. Hence, moneary policy may deal more effecively wih real shocks. Conversely, one may quesion he effeciveness of he moneary policy if he passhrough rae has endogenous effecs on moneary sabiliy (Campa & Goldberg, 2002). Therefore, i is very imporan o know and undersand he deerminans of he exchange rae passhrough as well as he ransmission of shocks under differen degrees of passhrough o he economy. In general, he lieraure on exchange rae passhrough can be divided ino wo srands as discussed in An (2006): micro and macro level. The firs srand of he lieraure focuses on he analysis of exchange rae passhrough ino domesic prices based on he micro level such as foreign firms pricing behavior, disaggregaed produc bundles/indusries and marke srucures. The second srand of he lieraure, on he oher hand, sudies exchange rae passhrough a he macro level. I invesigaes exchange rae passhrough from he moneary policy view. I esimaes exchange rae passhrough ino producer prices index (PPI), impor price index (IMP) and consumer prices index (CPI). Our sudy conribues o his srand of he lieraure. 46

3 Vol Alhough here exiss a body of research on exchange rae passhrough, is primary focus is on he indusrial counries. Analysis of i based on he emerging Asian economies is limied (Sahminan, 2002, Sao e.al, 2005). To fill his gap his sudy focuses on some emerging economies of Asia, namely Korea, Malaysia, Singapore and Thailand. To have a complee picure on exchange rae passhrough in he Asian economies we should have included oher Asian counries. However his analysis was no possible due o he unavailabiliy of daa for oher counries. Korea, Malaysia, Singapore and Thailand are srongly affeced by developed counries such as he US and Japan, given ha he US and Japan are main rade parners. Mos rade is in US dollars even among Asian counries. For insance, in 1980, 96.1% of expors and 93.2% of impors in Korea were invoiced in US dollars. This figure remained high in 2000 when 84.8% of expors and 80.4% of impors correspondingly were invoiced in US dollars. A similar rade invoicing condiion perains for Thailand. In 2000, 87% of Thai expors and 79% of Thai impors were invoiced in US dollars. This implies ha he marke is imperfec, where he US dollar dominaes he marke, according o Kamps (2006). Anoher feaure of he counries in he sample is ha rade componens feaure a high proporion of inermediae goods. For insance, counry specific repor/daa (IMF, 2004) show ha more han 74% of impors ino Malaysia are represened by inermediae goods. This implies ha prices in hese counries migh be srongly affeced by exernal shocks hrough impored inflaion of inermediae goods. Korea, Malaysia, Singapore and Thailand are among hose counries which were hi by he financial crisis of , which firs sared in Thailand when he bah was floaed on July 2, The financial crisis promped he crisishi Asian counries o aler heir moneary policy and exchange rae regimes. Before he crisis, hese counries adoped narrow, rigid exchange rae regimes and policy auhoriies focused on moneary base argeing. Afer he crisis, drasic acions were aken o reconsruc moneary policy implemenaion. Mos of hese counries have moved o more flexible or floaing exchange rae regimes. A he same ime, few counries have adoped inflaion argeing regimes. Korea ook he firs move o implemen an inflaion argeing regime in April 1998, followed by Indonesia in 2000, hen Thailand in May 2000 and he Philippines in January Whereas Malaysia and Singapore have no adoped inflaion argeing bu moved o managed floaing regimes. These drasic changes in he moneary policy and regimes hrow up some implicaions regarding he economies in hese counries. By comparing he daa before and afer he financial crisis in hree Asian counries, Osawa (2006) finds ha exchange rae volailiy in hese counries has increased over 2 Korea officially adoped an inflaion argeing regime in April 1998 wih headline CPI as inflaionary arge bu swiched o core CPI argeing from January

4 The Inernaional Journal of Economic Policy Sudies ime. A he same ime, foreign exchange reserves and ineres raes in hese counries have declined. These changes are due o he move in moneary policy and regimes from a rigid o a more flexible one. Turning o economic indicaors, inflaion raes in hese counries have declined slighly afer he financial crisis or afer he implemenaion of new moneary policy. M2 growh in hese counries also declines over ime. Addiionally, hese counries have improved heir curren accoun balance from negaive o posiive balances. These counries also show an increase in rade openness over ime alhough he degrees of rade openness in hese counries are quie differen. Malaysia and Singapore reain heir high degrees of rade openness over ime. According o Mishkin & Savasano (2001), an inflaion argeing counry should le he exchange rae floa. This requiremen can be explained by he heory of he Impossibiliy of he Holy Triniy where capial mobiliy and moneary policy independence canno coexis wih a pegged exchange rae regime. Mishkin (2004) discusses he danger of focusing srongly on limiing exchange rae movemens. The firs danger is ha he economy faces he risk of ransforming he exchange rae ino a nominal anchor ha akes over he inflaion arge. The second danger is ha he impac of he exchange rae on inflaion and oupu may be deermined by he source of shocks. Due o he above reasons, i is arguable ha he adopion of an inflaion argeing regime generaes a cos in he form of higher exchange rae volailiy. However, Edwards (2006) in his sudy on he relaionship of exchange rae and inflaion argeing shows ha here is no evidence ha he adopion of inflaion argeing leads o higher volailiy in he exchange rae. Previous sudies show ha exchange rae shocks in hese economies end o passhrough ino aggregae inflaion a a much faser rae han in he indusrial economies. Exchange rae passhrough is very rapid for emerging markes bu slow for advanced economies according o Devereux and Lane (2001). How well do hese saemens apply o Asian counries? Does exchange rae passhrough change in hese counries afer moving o a more flexible exchange rae regime and inflaion argeing? Are here any differences in passhrough raes among Asian counries? This sudy seeks o answer he above quesions and has hree main objecives. Firs, we seek o compare he degrees of exchange rae passhrough ino differen domesic prices (impor price, PPI and CPI) before and afer he financial crisis of 1997 (or afer moving o new moneary policies) in he counries in he sample. Second, we compare he effecs of exchange rae shock wih oher shocks on domesic prices. Previous sudies show ha he effec of impor price is much larger han ha of exchange rae shock. For insance Hahn (2003) finds ha he passhrough of impor price shock is larges and faser on domesic prices. I is imporan o invesigae how large he effec of exchange rae shock is relaive o oher shocks. If he passhrough rae is high bu he relaive effec of he exchange rae is small, hen 48

5 Vol exchange rae passhrough will no have a significan and large effec on deermining domesic price sabiliy. Third, using he same range of daa hrough wo differen approaches and empirical mehodologies, we seek o esimae he degrees of exchange rae passhrough in hese counries. We use he srucural vecor auoregressive model (VAR) and single equaion approaches. The resuls from boh mehods are consisen wih each oher. Our findings are in line wih he resuls of previous sudies where exchange rae passhrough is incomplee boh in he shorrun and he longrun in mos cases. The passhrough rae is he highes on impor prices, moderae on PPI and lowes on CPI. Exchange rae passhrough does no decline in all counries considered in his sudy. Addiionally we find a weak correlaion beween rade openness and he degree of exchange rae passhrough. In general, he effec of exchange rae shock on domesic prices is lower compared o ha of impor price shock. The paper is organized as follows. In secion 2, we describe he mehodology and daa. Secion 3 discusses esimaed resuls from SVAR and single equaion approaches. Secion 4 concludes. 2. Theoreical Framework This secion discusses some heoreical issues in modeling exchange rae passhrough and also he relaionship beween domesic prices (impor price, producer price and consumer price) and exchange rae movemens Modeling Exchange Rae Passhrough Following he sep of many previous sudies in modeling exchange rae passhrough, he exchange rae passhrough equaion consruced in his sudy is based on he concep of he law of one price (LOOP). Under he assumpion of LOOP, he price of impor denominaed in he domesic imporing counry s currency ( im P ) is equal o he impor price denominaed in he foreign exporing ex counry s currency ( P ) muliplied by he exchange rae of he imporing counry ( E ). im ex P P. E = (1) ex The exporing producer ses P based on he markup ( λ ) over marginal cos of producion * ( C ). Subsiuing equaion (2) o (1): P P C ex * = λ. im * = λ. C. E (2) (3) 49

6 The Inernaional Journal of Economic Policy Sudies where * C is he exporer s producion cos and λ is he markup. The markup depends on he demand pressure in he desinaion marke. I can be represened by he real GDP of he imporing counry. Equaion (3) is ransformed ino log form (denoed in lowercase leers): p = α λ + α c + α e (4) im * The exchange rae passhrough is capured by he coefficien of α 3 which is he parial elasiciy of impor price wih respec o exchange rae. Exchange rae passhrough is complee when α 3 = 1bu zero when α 3 = 0. In realiy, exchange rae is parial or incomplee where 0< α3 < 1. Equaion (4) is modified by including lagged erms and is esimaed using single equaion mehod (see secion 3.2). Besides esimaing he passhrough ino impor price, esimaions also include passhrough ino producer and consumer prices Sensiiviy of Domesic Prices o Exchange Rae Movemens Previous sudies on exchange rae passhrough show ha passhrough is higher on impor price han consumer price (for example Edwards (2006) and Campa & Goldberg (2006)). In general, passhrough is highes on impor price, moderae on producer price and lowes on consumer price (for example, McCarhy (2000), Hahn (2003) and Io e.al (2005)). This phenomenon can be explained by several facors. Focusing on he role of disribuion margin and impored inpus on he sensiiviy of domesic prices o exchange rae movemens, Campa & Goldberg (2006) show ha producion coss are more sensiive o exchange rae and impor prices as hey rely on impored componens, domesic suppliers and disribuors on impored inpu. High disribuion coss and low reliance on impored inpus induce low passhrough ino consumer price. However, disribuion margin can lead o more sensiiviy in consumer price if impored inpus are used in producion of nonradables. Bursein, Eichenbaum & Rebelo (2002) seek o invesigae he low inflaion in consumer price compared o producer price afer large devaluaions. As in Campa & Goldberg (2006), hey explain ha passhrough ino CPI is low due o disribuion coss and nonradable inpus. Besides, here are inferior radable goods which are only produced for he domesic marke. These inferior goods may subsiue for impored goods which leads o low passhrough ino consumer price. Third, domesic consumers may swich heir demand from impored goods o local radable (inferior) goods during crisis periods or devaluaions of domesic currency. 50

7 Vol Daa and Esimaion Mehods This secion of he paper describes he daa used in esimaion and mehodology. As he financial crisis sared in July 1997, he daa is divided ino wo subperiods: 1991M1 1997M7 (period before he crisis) and 1999M1 2007M5 (period afer he crisis). We apply wo differen approaches: single equaion approach and srucural VAR esimaion. In doing so, we aemp o compare he robusness of he resuls o wo differen mehodologies. We esimae exchange rae passhrough ino domesic prices (impor price, PPI and CPI) and analyze dynamic effecs of shocks on he economy in he Asian counries Daa All he monhly daa was obained from Inernaional Financial Saisics (IFS), IMF. This daa includes oil price index, money or M1, nominal effecive exchange rae, impor price, producer price index (PPI), consumer price index (CPI) and indusrial/manufacuring producion index (IP). All series are seasonally adjused using he Census X11 program and are ransformed ino logarihmic form (apar from he oupu gap variable). The oupu gap is consruced as he log difference beween acual oupu (IP) and poenial oupu (HP filer adjused indusrial producion index 3 ). The main problem in his sudy is he availabiliy of daa. Mos Asian counries do no have long enough series. Malaysia does no have he impor price series and herefore we only apply he 6 variable VAR model (excluding he impor price variable). Thailand has shorer series, saring from 1999 M1, hus he analysis is made only for he second subperiod. All he series (in logarihmic form) are esed wih he uniroo saionariy es and become saionary afer he firs differenced ransformaion, excep for he oupu gap series. Mos of he oupu gap series are saionary in level. We compare he resuls of wo uniroo ess, namely he Augmened DickyFuller (ADF) and Schmid Phillips (SP) ess. All he series are divided ino wo subperiods. In general, he resuls of he uniroo es for he firs and second periods do no change much. For all ime series we work wih, boh ADF and SP ess have agreed on he order of inegraion o be 1 (I(1) process) Single Equaion Approach This secion seeks o esimae he exchange rae passhrough model consruced in secion 2.1. The mehodology employed in his par of he paper applies LS and IV (insrumenal variables) 3 Thailand uses GDP insead of he Indusrial Producion index series due o he unavailabiliy of he corresponding series. 51

8 The Inernaional Journal of Economic Policy Sudies echniques in order o esimae long run and shor run exchange rae passhrough on domesic prices. There is a huge debae around he mehods of esimaion of long run passhrough. In heory, impor prices, he exchange rae and foreign price level should have a long run Engle and Granger coinegraing relaionship. However in realiy and in he empirical lieraure his fac is no always confirmed (Campa and Goldberg, 2005). De Brand, Banerjee and Kozluk (2007), using ime series and upodae panel daa echniques, es for coinegraion wih he possibiliy of srucural breaks and show ha he long run may be resored in he esimaion. The lag order and inroducion of srucural breaks may significanly change he resuls from he coinegraion ess, according o De Brand e.al, (2007). Given ha some Asian counries experienced financial crisis in lae 1997 we apply wo differen coinegraion ess: Johansen race es and Saikkonen & Lükepohl ess for wo subperiods separaely using differen lag orders based on he SC, AIC, FPE and HQ informaion crieria. Since none of he ess ouperforms he oher, our conclusions are based on he resuls from he wo ess applied. We es for coinegraion beween CPI (PPI, impor price index), exchange rae and foreign price level as well. If he coinegraion relaionship is revealed we apply errorcorrecion model (ECM). When i is no he case we use he LS procedure suggesed by Campa and Goldberg (2005). ECM is esimaed using sandard LS echniques in wo seps. Firs we esimae ECM as follows: p 1 q r z Δ k = pk + sk + b jδpk j a jδsk j + c j yk j + 1, ( 1 θ1), 1 θ2, 1,,, j= 1 j= 0 j= 0 j= 0 p d Δp + u j * us, j (4 ) where pk, represens home CPI (home PPI or impor price index) for counry k, sk, is he nominal effecive exchange rae, yk, is he oupu gap, pus,* is he PPI of he US. All variables are expressed θ2 in logs, Δ indicaes firs difference. Long run passhrough is β = Coefficien â 0 is inerpreed 1 θ as shor run exchange rae passhrough. The lag orders of y k,, Δ p k,, y k, and Δ p us,* are deermined on he basis of he AIC and SC crieria and significance of he parameers. Afer he final model is formulaed is residual is checked again for saionariy by means of he ADF es wih he lag order based on he AIC and SC informaion crieria. Given ha he esimaed residual follows a nonsandard disribuion, we use criical s for he ADF coinegraion ess. Inclusion of he deerminisic variables rend or inercep depends on he behavior of he imes series. If he variables exhibi rend behavior we include rends in he coinegraing equaion, if no we don. The disadvanage of his model is ha i does no provide he sandard errors of he long run esimae direcly. To calculae hem we apply Bewley ransformaion of he ECM: 1 52

9 Vol bˆ aˆ cˆ dˆ u p s p s y p p 1 q 1 r 1 z 1 j j j j * k, = β k,1 Δ k,1 Δ k, j k, j Δ us, j + j= 11 θ1 j= 01 θ1 j= 01 θ1 j= 01 θ1 1 θ1 (4 ) p wih bˆ j = b i, aˆ j = a i, cˆ j = c i dˆ j = d i, i= j+ 1 q i= j+ 1 r i= j+ 1 z i= j+ 1 aˆ 0 = a for j 1 0 Given ha Δp k, is correlaed wih disurbance u we esimae his equaion using insrumenal variables (IV) esimaion wih insrumen pk,1 for Δ p k,. We laer drop he subscrip k for simpliciy. However one should ake ino accoun ha he second sep is possible only if 1 θ1 is significanly differen from zero. Thus before coninuing wih he second sep we es wheher 1 θ1 is zero. If i is he case we apply sandard wo sep Engle Granger procedure. If evidence of coinegraion is no found we apply he mehod suggesed by Campa and Goldberg (2005) which in our case is defined as follows: 12 r 1 z 1 Δ = c + a jδs j + c j y j + j= 0 j= 0 j= 0 p d Δp + u j * j (5) The lag order of he oupu gap and foreign price index as before are deermined on he basis of he AIC, SC informaion crieria and he significance of parameer esimaes. Given ha we have monhly daa in our analysis we include exchange rae series up o lag of order 12 in order o ge an approximae measure of long run passhrough. Given ha his assumpion is ad hoc, one should no compleely rely on hese esimaes; hey can be inerpreed as benchmark esimaes of long run passhrough only. Shor run passhrough in he presen case is a parameer which deermines he â conemporaneous effec of he exchange rae on price level, equal o 0 in our case. Given ha we migh have an endogeneiy problem in he regression we reesimae our model using he wo sages leas squares mehod (TSLS) and use as insrumens 13 lags of he exchange rae, 12 lags of oupu gap, 13 lags of foreign price index and 13 lags of he dependen variable. The number of lags for insrumens is subjec o variaion (Mihailov, 2005). For comparaive purposes we presen obained passhrough esimaes from OLS and TSLS in Table Srucural Vecor Auoregression Model (SVAR) Following he consrucion of he SVAR model in analyzing he effecs of shocks in previous sudies (Io and Sao, 2006 and McCarhy, 2006), he SVAR model consiss of seven variables: ( ) x = ΔOIL GAP ΔM ΔNEER ΔIMP ΔPPI ΔCPI 53

10 The Inernaional Journal of Economic Policy Sudies OIL where GAP sands for he oil price index, he oupu gap, NEER he nominal effecive exchange rae and IMP he impor price, M he moneary aggregae or M1, PPI producer price index CPI and consumer price index. Δ denoes he firs differenced operaor. All variables excep he oupu gap are in logarihms and are seasonally adjused using he Census X11 program. All series are esed wih he uniroo saionariy es and become saionary afer he firs differenced ransformaion. The analysis of exchange rae passhrough in his secion akes a differen approach by including differen variables compared o he single equaion approach ha applies he concep of LOOP. I follows he seup of a sandard open economy macroeconomics model which consiss of he variables/equaions of supply (oil price index), demand (oupu gap), moneary policy (M1), exchange rae and domesic prices. The seup of he model is based on he ordering of variables and resricion imposed on he srucure of shocks where decomposiion of he variance covariance marix of reduced form residuals is wrien in a lower riangular marix. n(n1)/2 resricions are imposed globally on he riangular marix in order o idenify srucural shocks where some of he srucural shocks do no have conemporaneous impacs on oher variables. e ε ε OIL S GAP e S21 S M e S31 S32 S NEER e = S41 S42 S43 S IMP e S51 S52 S53 S54 S PPI e S61 S62 S63 S64 S65 S66 0 CPI e S 71 S72 S73 S74 S75 S76 S 77 OIL GAP M ε NEER ε ε ε ε IMP PPI CPI (6) The ordering of he variables deermines he srucure of he shocks. The firs variable has influenial effecs on all variables below i bu is no affeced by hese variables. The second variable only receives impac from he firs variable. I does no have any impac on he firs variable bu i can influence all he variables below i. This rule applies o all subsequen variables. Following Io and Sao (2006), he oil price index is ordered firs, as supply or oil price shocks may affec he oher variables bu conemporaneously are unlikely o be affeced by oher shocks. The second variable is he oupu gap. This variable is likely o be affeced by oil price shocks only and is assumed o affec all variables in he sysem excep oil price shocks. The moneary policy variable is ordered nex before he exchange rae. This variable is ordered below he exchange rae in he second model for robusness checking. The hree domesic prices are ordered hus as previous sudies show passhrough is highes 54

11 Vol on impor price and PPI bu lowes on CPI (see discussion in secion 2.2). Such ordering allows inerpreaion on how he price of impored inpus is ransmied ino consumer final goods. Io & Sao (2006) use he nominal effecive exchange rae o represen he exchange rae variable and argue ha i is no appropriae o use he bilaeral exchange rae wih he US Dollar in his sudy as mos Asian counries had de faco adoped he US Dollar. In his sudy, we run he baseline SVAR model using he nominal effecive exchange rae series. The advanages of applying he srucural VAR in his paper are: firs i solves he endogeneiy problem ha arises under he single equaion mehod; second, his echnique applies resricions o idenify srucural shocks; and hird, i enables us o invesigae he effec of exchange rae and passhrough rae on he chain of domesic prices (impor price, PPI and CPI) in he same sysem equaion. For he purpose of robusness, we run he VAR model using wo differen ordering schemes such ha: ( ) x = ΔOIL GAP ΔM ΔNEER ΔIMP ΔPPI Δ CPI, (7) which is Model I in our case and ( ) x = ΔOIL GAP ΔNEER ΔM ΔIMP ΔPPI Δ CPI, (8) which is Model II in he presen case. Exchange rae is assumed o influence he decision of moneary policy in Model II. The wo differen ordering variables of model are esimaed separaely for each counry for boh subperiods. The number of lags is deermined based on he SC, AIC and HQ informaion crieria. However, he final decision is made based on he resuls from he diagnosic ess for he residuals. The resuls of he firs ordering are compared wih he resuls of he second ordering. According o Mihailov (2005), under he srucural VAR esimaion, he exchange rae passhrough a horizonal lag ( ) is obained hrough he accumulaed impulse response of domesic prices (P) wih respec o an innovaion of one sandard deviaion in he exchange rae (NEER) equaion of he VAR. The parial derivaives can be wrien as: ( dln P ), >0 NEER ε On he oher hand, he exchange rae passhrough under he single equaion esimaion is: ( dln P ) ln NEER, >0 According o Mihailov (2005), here is no direc comparison of he elasiciy of exchange rae passhrough beween he wo mehods. In order o make he passhrough raes, obained from srucural VAR and he single equaion esimaions, comparable, ransformaion or normalizaion of he impulse response o an innovaion in NEER VAR equaion is made in he following way: 55

12 The Inernaional Journal of Economic Policy Sudies ( dln P ) ε ln NEER ε NEER NEER, >0 4. Empirical Resuls This secion presens empirical resuls from he single equaion and SVAR esimaions. Addiionally i conducs comparaive analysis on he exchange rae passhrough obained using hese wo approaches Single Equaion Esimaion Resuls This par of he secion focuses on he comparison of exchange rae passhrough across counries and across periods. The coinegraion ess (Johansen and S & L ess) show evidence of a coinegraed relaion beween exchange rae and domesic prices in he case of CPI in Malaysia (period I) and Thailand (period II) and PPI in Korea (period I) 4. In hese cases, he passhrough equaion is esimaed using he ECM mehod. In he cases where no coinegraed relaion is found, OLS and TSLS mehods are applied. The resuls of coinegraion ess are summarized in he Appendix (Table I(ac)). The resuls for exchange rae passhrough can be summarized as follows: (1) Exchange rae passhrough on impor prices Korea and Singapore have incomplee passhrough on impor prices, unlike Thailand which has complee passhrough on ha in he second period. Shor run exchange rae passhrough on impor prices is in general lower han long run counerpars, excep ha of Korea in he second period. Exchange rae passhrough ino impor prices in Korea is very rapid in he shor run. 4 The coinegraing rank is seleced using he principle where he null hypohesis canno be rejeced for he firs ime. 56

13 Vol Time horizonal 1 Shor run IMP Korea Malaysia Singapore Thailand PPI Korea Malaysia Singapore Thailand CPI Korea Malaysia Singapore Thailand Time horizonal 12/ LR IMP Korea Malaysia Singapore Thailand PPI Korea Malaysia Singapore Thailand CPI Korea Malaysia Singapore Thailand Table 1: Exchange Rae PassThrough Raes Across Mehodologies and Trade Openness OLS TSLS VAR Average Trade openness Period I Period II Period I Period II Period I Period II Period I Period II * * 0.073* 0.013* 0.089* 0.06* 0.012* 0.056* * * 0.142* 0.262* 0.182* 0.027* 0.020* 0.016* 0.012* * * 0.073* 0.026* 0.088* 0.06* 0.012* 0.057* * * * 0.116* 0.02* 0.162* 0.028* 0.025* 0.02* 0.012* Period I Period II Period I Period II Period I Period II Period I Period II * * * * * 0.052* * * * * * denoes ha elasiciy is significanly differen from zero a 1% level * denoes ha elasiciy is significanly differen from 1 a a 1% level The following cases apply o ECM esimaion as coinegraion relaions are found: passhrough ino CPI in Malaysia (period I) and Thailand (period II); passhrough ino PPI in Korea (period I) The exchange rae passhrough s in he VAR are obained from he normalized IRF, model I Trade openness is defined as he oal rade divided by GDP. The figures are calculaed by he auhors using he annual daa from ADB: period I ( ) and period II ( ) 57

14 Inernaional Journal of Economic Policy Sudies However impor prices in Korea ge adjused in he long run so ha he effec of exchange rae changes has declined over ime. This is in line wih he empirical findings obained by Io, Sasaki and Sao (2005). Comparing exchange rae passhrough elasiciies across LS and TSLS we find no large difference beween hem. Given ha we migh have an endogeneiy problem in he OLS esimaion we rely more on he TSLS esimaes. Comparing passhrough on impor prices across periods in he long run, we observe i decline in Korea, bu increase in Singapore. The drop in he passhrough rae in Korea may be due o he change in exchange rae regime o inflaion argeing in (2) Exchange rae passhrough on PPI There is an overall incomplee passhrough on PPI across counries boh in he shor and long run. Exchange rae passhrough on PPI is lower ha ha on impor prices. This is a quie sylized empirical fac, widely suppored in he empirical lieraure such as McCarhy (2000) and Hahn (2003). Shor run exchange rae passhrough on PPI is lower han is long run counerpar. Comparing exchange rae passhrough across periods we observe a decline in he PPI in Korea and Malaysia bu a rise in Singapore. (3) Exchange rae passhrough on CPI Exchange rae passhrough on CPI is he lowes, and negligible compared o ha on impor prices and PPI. The esimaes obained are similar in magniude o hose obained by Choudhri and Hakura (2006). As before, shor run elasiciies are lower han heir long run esimaes. Comparing long run passhrough on CPI across periods we observe i increase in Korea and Singapore, bu decline in Malaysia in he second period, alhough by a negligible magniude. We are no able o conduc similar analysis for Thailand due o he daa availabiliy problem. However since mos of he analysis made in his par of he paper is based on an ad hoc esimae of long run exchange rae passhrough raes hese esimaes should be considered wih cauion. Furher analysis should be made by applying more sophisicaed echniques 4.2. SVAR empirical resuls This par of he secion presens he resuls from SVAR esimaion. Firs, he responses of impor prices wih respec o various shocks are discussed. The responses of oher variables impaced by he exchange rae are also presened (IRF). Second, he relaive explanaory power of shocks on domesic prices is compared using forecas error variance decomposiions (FEVD). Finally, he degrees of passhrough before and afer he crisis are compared. 58

15 Vol (1) (Accumulaed) Impulse response funcions (IRF) In he SVAR model, i is assumed ha here are 7 shocks in he economy: oil price shocks, oupu gap shocks, exchange rae (NEER) shocks, impor price shocks, producion cos shocks (PPI) and nonoil price (CPI) shocks. The impulse response funcion shows he response of each variable o one posiive sandard deviaion of each shock. The middle line represens he responses while he upper and lower dashed lines are wo sandard error bands. The verical axis shows he percenage poin change in he domesic price index or he percenage of passhrough and he horizonal axis shows he ime (in monhs). However we are no going o discuss he effecs of all shocks assumed; we will presen here only he responses of domesic prices wih respec o NEER and impor price shocks. The resuls of he cumulaive impulse response funcions are summarized in he appendix. Figures 1 and 3 show he graphs of he responses in he domesic prices (IMP, PPI and CPI) o a one percenage increase (appreciaion) in nominal effecive exchange rae beween he wo periods using models I and II. In general, he resuls are consisen wih he resuls of previous sudies: an appreciaion in nominal exchange rae leads o a decline in domesic prices. Or equivalenly, depreciaion in he exchange rae causes an increase in domesic prices. This resul holds in all counries beween he wo periods wih he excepions of Malaysia, Singapore and Thailand. In hese counries, depreciaion in he exchange rae leads o a decline in CPI in cerain periods. This in line wih he resuls repored in he paper by Choudhri and Hakura (2006). In order o have a beer comparison on he responses of domesic prices under exchange rae shock beween he wo periods, we summarize he numerical s capured when running he impulse response funcions, i.e he responses of domesic prices under a 1% exchange rae shock (see Table 2). In general, he percenage changes of one sandard deviaion in he innovaion in he NEER equaions are differen across counries. Impor price in Thailand shows he highes response o exchange rae shock as compared o oher counries. The response of domesic prices o exchange rae shock is highes on impor price, moderae on PPI and lowes on CPI. The passhrough of exchange rae changes o impor prices is higher in Korea and Thailand bu lower in Singapore. The cumulaive impulse responses of domesic prices o impor price shock are summarized in Table 2 as well. Comparing he resuls of Table 2, we observe ha in general, he percenage changes in domesic prices led by impor price shock are higher han hose of exchange rae shock, wih he excepion of Thailand. As in he case of exchange rae shock, we observe ha he effec of impor price shock is highes on impor prices, moderae on PPI and lowes on CPI, wih he excepion of Singapore. In Singapore, he effec of impor price shock is highes on PPI, followed by impor prices and CPI. The response of impor price o a one percenage change in impor price is highes in 59

16 Inernaional Journal of Economic Policy Sudies Thailand, followed by Korea and Singapore. However, he effec of impor price shock on PPI is highes in he case of Singapore, followed by Korea and Thailand. In order o obain exchange rae passhrough raes which are comparable wih he passhrough raes esimaed using he single equaion mehod, we follow he ransformaion suggesed by Mihailov (2005). The resuls are summarized in Table 1. The resuls show ha in mos cases, exchange rae passhrough ino domesic prices is incomplee, wih he highes passhrough on impor price, followed by PPI and CPI. Comparing he resuls beween wo subperiods across prices, we observe ha differen price indices respond differenly o he exchange rae changes. Addiionally, changes in exchange rae passhrough in domesic prices differ across counries. This migh be due o counryspecific characerisics and percenage change in commodiies composiion of price indices over ime. I is necessary o noe ha in general he magniude of changes is large in Singapore and moderae in Korea and Malaysia. (2) Forecas error variance decomposiion (FEVD) The FEVDs show he s of he percenage share of variance of he nsep forecas error of a variable ha can be explained by innovaion in anoher variable (Billmeier, 2002, p.13). Table 3 shows he maximum effecs of seven shocks on domesic prices ha are obained from he FEVD. The resuls show ha impor price shock is he main deerminan of he flucuaions of impor price in boh periods. Exchange rae shock in is urn can explain quie well he movemen of impor prices especially in he case of Korea and Thailand. I can explain a mos 24% and 41% of flucuaions in impor prices in Korea and Thailand correspondingly (in period II). PPI shock can mosly explain he movemen of PPI in all counries for boh periods, wih he excepion of Singapore. In Singapore, he movemen of PPI is mainly deermined by impor price shock. Impor price shock can explain a mos 75% and 85% of he movemen of PPI in Singapore for periods I and II correspondingly. In general, he explanaory power of impor price shock on domesic prices is higher han ha of exchange rae shock, wih he excepion of Thailand. Exchange rae has a very low effec in deermining he movemen of domesic prices in Malaysia and Singapore. These resuls hold in boh periods. 60

17 Vol Table 2: Impulse Response Funcions: NEER Shock and Impor Prices Shock (IMP) o Domesic Prices Time Period I NEER o impor price Period II NEER o impor price Period I IMP o impor price Period II IMP o impor price Kor Mal Sp Thai Kor Mal Sp Thai Kor Mal Sp Thai Kor Mal Sp Thai Time Period I NEER o PPI Period II NEER o PPI Period I IMP o PPI Period II IMP o PPI Kor Mal Sp Thai Kor Mal Sp Thai Kor Mal Sp Thai Kor Mal Sp Thai Time Period I NEER o CPI Period II NEER o CPI Period I IMP o CPI Period II IMP o CPI Kor Mal Sp Thai Kor Mal Sp Thai Kor Mal Sp Thai Kor Mal Sp Thai Noes: Firs half of he able: NEER o domesic prices (impor prices, PPI, CPI) show he changes in domesic prices led by one percenage depreciaion in exchange rae (he s are subjec o muliplicaion by 10 2 ). Second half of he able: IMP o domesic prices (impor prices, PPI, CPI) indicae he changes in domesic prices led by one percenage increase in impor prices (he s are subjec o muliplicaion by 10 2 ) The s are obained by running he impulse response funcions 61

18 Inernaional Journal of Economic Policy Sudies Table 3: Forecas Error Variance Decomposiions: Maximum Effecs of Shocks on Domesic Prices Period I (I) Impor price OIL GAP M NEER IMP PPI CPI Korea Malaysia Singapore (II) PPI OIL GAP M NEER IMP PPI CPI Korea Malaysia Singapore (III) CPI OIL GAP M NEER IMP PPI CPI Korea Malaysia Singapore Period II (I) Impor price OIL GAP M NEER IMP PPI CPI Korea Singapore Thailand (II) PPI OIL GAP M NEER IMP PPI CPI Korea Malaysia Singapore Thailand (III) CPI OIL GAP M NEER IMP PPI CPI Korea Malaysia Singapore Thailand Noes: All he s are obained from he forecas error variance decomposiions 4.3. Discussion and comparisons In his par of he secion, we check he consisency and robusness of he resuls by firs, comparing he resuls of differen ordering of variables in he srucural VAR model; second, by comparing he resuls of SVAR wih ha of he single equaion mehod. We also discuss he link beween rade openness and he degree of exchange rae passhrough and he reasons ha drive he differences in overall resuls across counries. (1) Comparing resuls wihin SVAR models For he purpose of robusness, we run again he SVAR model using wo differen ordering schemes, Models I and II as described in secion

19 Vol Comparing he resuls of boh orderings (see Figure 1 o 4), one can observe ha he responses of domesic prices o exchange rae and impor price shocks are very similar o each oher in all cases. The effec of he exchange rae changes remains low in Malaysia and Singapore. The responses of domesic prices in general are higher under impor price shock as compared o ha of exchange rae shock. The consisency of he resuls implies ha he resuls obained from srucural VAR are robus. For he double check, we compare he resuls of srucural VAR wih ha of he single equaion resuls (see he relevan secion below). (2) Comparing he resuls of SVAR wih single equaion approach Comparison of he resuls obained from SVAR and single equaion approach echniques on exchange rae passhrough show ha he magniude of he exchange rae passhrough elasiciies obained from he wo differen mehods differs o some exen. However in general he wo mehods exhibi similar rends on he behavior of exchange rae passhrough ino domesic prices over ime. Similar resuls are repored in Mihailov (2005), who focuses on analysis of exchange rae passhrough in US, Germany and Japan in wo subperiods. Mihailov (2005) seeks o compare he resuls obained using generalized VAR and he single equaion mehod. He finds ha he resuls from boh mehods show he same general rends in exchange rae passhrough bu he precisions of he elasiciies of passhrough depend crucially on he economeric mehod, daa frequency and variable proxy employed. As in Mihailov (2005), he dissimilariies here may be due o he differen mehodologies employed, number of lags and variables included in he model. In SVAR we use 7 variables: nominal effecive exchange rae, money base, oupu gap, PPI, CPI, impor price index and oil price index, wih all hree price indices joinly included ino one model. Shocks are idenified based on a cerain ordering of he variables. Number of lags for regressors differs across counries and periods. Whereas, in he single equaion approach we regress CPI, PPI and impor prices on exchange rae, oupu gap and foreign price level independenly. We use he same number of lags for he exchange rae variable, whereas he lags of oupu gap and foreign price level are subjec o change. In general he resuls from boh mehods applied are consisen wih each oher, alhough here are some dissimilariies. The exchange rae passhrough ino impor price is highes, moderae on PPI and lowes on CPI across mehodologies applied. Behavior of passhrough raes differs across price indices and counries. As menioned above hey migh be due o counryspecific characerisics and he composiion of he commodiies in he price indices. Resuls from boh mehods show ha appreciaion in exchange rae leads o an increase in CPI (in some periods) in he case of Malaysia and Singapore. This resul is also found in previous sudies (Choudhri and Hakura, 2006). 63

20 Inernaional Journal of Economic Policy Sudies There are many facors which conribue o passhrough rae and he explanaory power of he effecs of hese facors varies across counries. In order o ge more consisen esimaes of passhrough raes furher research is needed. Inclusion of regressors ino he model should be made based on he specific characerisics of he counries. (3) Discussion Why is he degree of exchange rae passhrough differen across counries? Does rade openness maer? Do exchange rae passhrough raes decline over ime across differen price indices? We aemp o compare he degree of rade openness wih he degree of exchange rae passhrough across counries for boh periods. Table 1 displays he summarized resuls. In general, he resuls show ha here is a weak correlaion beween rade openness and he degree of exchange rae passhrough. Counries ha have higher rade openness do no have higher exchange rae passhrough. For insance, Singapore has he highes rade openness as compared o he oher hree counries; however i does no have he highes degree of exchange rae passhrough. The degrees of rade openness have increased over ime in all counries. However, no all counries have experienced increase in he degree of exchange rae passhrough. For insance, long run exchange rae passhrough on PPI has declined in Malaysia and ha on CPI increased, bu by negligible magniude, alhough rade openness has increased in his counry. This resul implies ha rade openness is no he main facor which deermines he degree of exchange rae passhrough in hese counries, a leas in his sudy. Exchange rae passhrough raes ino domesic prices exhibi differen rends over ime, across price indices and counries. There is no simulaneous change in passhrough raes over ime across counries. This migh be due o facors which are specific o he paricular counry. In general hese changes are affeced by possible change in he rade srucure afer he financial crisis of So far we have analyzed he passhrough of exchange rae ino domesic prices. A his sage i is naural o ask wha are he main facors ha drive exchange rae passhrough? Wha are he counryspecific characerisics which lead o differences in exchange rae passhrough across counries? How big is heir role in undersanding his mechanism? These ypes of quesions need o be answered. However we leave his kind of analysis for fuure research. Addiionally our research can be exended by expanding he ime period used in he analysis in he paper, given ha his sudy only covers he laes periods of he 1990s and 2000s, excluding he crisis periods of For fuure sudy, one can include daa from he earlier years (for example he 1970s or 1980s), when he Asian economies flucuae more as compared o he periods before and afer he crisis covered in he paper. Expansion of he ime period is expeced o have an effec on he magniude of exchange rae passhrough raes. However his exension can be problemaic, given he daa availabiliy problem. 64

21 Vol Conclusions In his paper, we underook an empirical analysis of exchange rae passhrough ino domesic prices for Korea, Malaysia, Singapore and Thailand. Given ha hese counries were hi by he financial crisis in 1997 o a differen exen, we conduced analysis before and afer he crisis in isolaion. Several findings have been obained, which can be summarized as follows. Firs, in general, our resuls parially suppor he previous resuls in his area, where he degree of exchange rae passhrough is differen across counries and over ime. The changes in passhrough raes across counries may be due o he change in he srucure of rade and moneary policy. As in he case of Singapore, he increase in exchange rae passhrough may be due o he very high degree of rade openness and low volailiy in he exchange rae (as Singapore implemens exchange rae argeing policy). Besides, Singapore is a manufacuringbased and nonresource producing counry. This implies ha Singapore may impor producs ha have higher passhrough raes such as raw maerials and fuel/ oil. Second, he degree of exchange rae passhrough is highes on impor price, moderae on PPI and lowes on CPI. Third, passhrough ino CPI remains very low in hese counries. In some periods, appreciaion of he exchange rae leads o an increase in CPI in Malaysia and Singapore. Due o he low passhrough ino CPI and he effec of exchange rae shock on domesic prices being relaively low in hese counries, he inflaion rae in hese counries remains sable and does no increase much over ime. Fourh, impor price shocks have higher explanaory power on movemen in domesic prices as compared o exchange rae shocks. The effec of exchange rae shock is very small in he case of Malaysia and Singapore, bu he effecs of impor price shock on domesic prices are very high in all counries, especially in he case of Singapore. Fifh, comparing he degrees of passhrough ino domesic prices over ime, we find ha here is weak correlaion beween rade openness and exchange rae passhrough. Sixh, exchange rae passhrough on CPI across periods and across counries is small in magniude and even changes sign from posiive o negaive in some cases. This migh be due o he pricing o marke behavior of monopolisic firms, which aemp o se lower prices in he counries where hey expor in order o have higher compeiive power relaive o local firms. Our analysis gives a clear indicaion ha exchange rae passhrough ino domesic prices is counry specific, given ha i is differen across counries. However, counryspecific characerisics are no revealed in his paper and his is lef for fuure research. 65