Exchange Rate Pass-Through into Import Prices. Abstract

Transcription

1 Exchange Rae Pass-Through ino Impor Prices Absrac We provide cross-counry and ime series evidence on he exen of exchange rae pass hrough ino he impor prices of weny-hree OECD counries. Across he OECD and especially wihin manufacuring indusries, we find compelling evidence of parial pass-hrough in he shor-run rejecing boh producer currency pricing and local currency pricing as characerizaions of aggregae behavior. Over he long run, producer-currency pricing is more prevalen for many ypes of impored goods. While we find ha counries wih higher raes of exchange rae volailiy are also hose wih higher pass hrough elasiciies, we also conclude ha macroeconomic variables have played only a minor role in accouning for he evoluion over ime of OECD counry pass-hrough elasiciies. Far more imporan for pass hrough changes have been he dramaic shifs in he composiion of counry impor bundles. Revision: Sepember 28, 2004 (firs version: January 2001) José Manuel Campa IESE Business School and N.B.E.R. Linda S. Goldberg Federal Reserve Bank of New York and N.B.E.R. JEL codes: F3, F4 The views expressed in his paper are hose of he individual auhors and do no necessarily reflec he posiion of he Federal Reserve Bank of New York or he Federal Reserve Sysem. We hank anonymous referees, Rudiger Dornbusch, Richard Marson, Andrew Rose and Alwyn Young for helpful commens, as well as he seminar paricipans a various universiies, he ASSA, NBER, BIS, and Federal Reserve Bank of New York. We also hank Leicia Alvarez and Glenda Oskar for heir research assisance. Address correspondences o Linda S. Goldberg, Federal Reserve Bank of NY, Research Deparmen, 33 Libery S, New York, N.Y Tel: ; fax: ; Linda.Goldberg@ny.frb.org.

2 1. Inroducion While exchange rae pass-hrough has long been of ineres, he focus of his ineres has evolved considerably over ime. Afer a long period of debae over he law of one price and convergence across counries, beginning in he lae 1980s exchange rae pass-hrough sudies emphasized indusrial organizaion and he role of segmenaion and price discriminaion across geographically disinc produc markes. More recenly pass-hrough issues play a cenral role in heaed debaes over appropriae moneary policies and exchange rae regime opimaliy in general equilibrium models. 1 These debaes have broad implicaions for he conduc of moneary policy, for macroeconomic sabiliy, inernaional ransmission of shocks, and effors o conain large imbalances in rade and inernaional capial flows. These debaes hinge on he issue of he prevalence of producer-currency-pricing (PCP) versus local currency pricing (LCP) of impors, and on wheher exchange rae passhrough raes are endogenous o a counry s moneary performance. Low impor price passhrough means ha nominal exchange rae flucuaions may lead o lower expendiure swiching effecs of domesic moneary policy. As a consequence of his insulaion, moneary policy effeciveness is greaer for simulaing he domesic economy. Taylor (2000) also has noed he poenial complimenary beween moneary sabiliy and moneary effeciveness as a policy insrumen. The idea is ha if pass-hrough raes are endogenous o a counry s relaive moneary sabiliy, periods of more sable inflaion and moneary performance also will be periods when moneary policy may be more effecive as a sabilizaion insrumen. Concerns can be raised, hen, abou wheher measured degrees of moneary policy effeciveness are fragile and regime-specific if he degree of exchange rae pass-hrough is highly endogenous o macroeconomic variables. 2 The degree of aggregae exchange rae pass-hrough, and i s deerminans, are herefore imporan for he effeciveness of macroeconomic policy. While pass-hrough of exchange rae movemens ino a counry s impor prices is cenral o hese macroeconomic sabilizaion argumens, o dae only limied relevan evidence on his relaionship has been available. 3 The firs goal of our paper is o provide 1 The implicaions of pass-hrough performance for opimal moneary policy also is explored in Corsei and Peseni (2001), Obsfeld (2000), Devereux (2001), and Devereux and Engel (2000), among ohers. 2 See Taylor (2000). The role of he invoicing decisions of producers in influencing pass-hrough raes is explored in recen work by Devereux and Engel (2001) and Bacchea and vanwincoop (2001). 3 As surveyed by Goldberg and Kneer (1997), mos of he available evidence is from very narrowly defined expor indusries, wih an emphasis ofen placed on he pricing o marke behavior of exporers. Kneer (1993), Marson (1990), P.Goldberg and Kneer (1997), and Kasa (1992) use expor prices or expor uni values from 1

3 exensive cross-counry and ime-series evidence on exchange rae pass-hrough ino he impor prices of weny-hree OECD counries. Using quarerly daa from 1975 hrough 2003, we esimae pass-hrough elasiciies afer appropriaely conrolling for shifs in exporer marginal coss and demand condiions. Our cross-counry evidence is srongly supporive of parial exchange rae pass-hrough in he shor run (defined as one quarer) a he level of he aggregae impor bundle. The unweighed average of pass-hrough elasiciies across he OECD counries is abou 46 percen over one quarer, and abou 64 percen over he longer erm. The Unied Saes has among he lowes pass-hrough raes in he OECD, a abou 25 percen in he shor run and 40 percen over he longer run. Corresponding raes of pass-hrough ino German impor prices are approximaely 60 percen in he shor run and 80 percen in he long run. Wha explains differences across counries in exchange rae pass hrough ino impor prices? A promising recen direcion of research supplemens he earlier microeconomic argumens by focussing on macroeconomic variables. Mos noably, heoreical works argue ha volailiy in moneary aggregaes and exchange raes of counries should influence he choice of invoice currencies in rade [for example, see Engel and Devereux (2001)]. In equilibrium, counries wih low relaive exchange rae variabiliy or sable moneary policies would have heir currencies chosen for ransacion invoicing. The low exchange rae variabiliy counries would also be hose wih lower exchange rae pass-hrough. We find evidence ha counries wih less exchange rae and inflaion variabiliy are likely o have lower raes of pass-hrough of exchange raes ino impor prices. This is a weak sysemaic posiive relaionship beween volailiy over recen decades and passhrough. There are no similar sysemaic relaionships beween counry size and passhrough ino aggregae impor bundles. Anoher issue receiving aenion in he recen macroeconomic debae is he sabiliy of exchange rae pass-hrough raes over ime. Taylor (2001) and Goldfajn and Werlang (2000), among ohers, have argued ha pass-hrough raes may have been declining over ime. The Brazilian experience of he lae 1990s is ofen cied. In his experience, consumer prices responded very lile o a large home currency depreciaion, in sharp conras wih pas depreciaion episodes. The issue posed in hese and relaed sudies is wheher his decline in specific counries o muliple desinaions wih he inen of idenifying price discriminaion or pricing o marke aciviy. While impor prices are by definiion jus he local currency value of anoher producer s expor prices, he impor price series aggregae across producers from all source counries and across a broader array of prices. 2

4 pass hrough, and a propored more general decline in pass hrough raes, are relaed o improved macroeconomic condiions in he imporing counries. 4 We furher ask wheher hese issues are ones ha exend o he OECD counries. Our work emphasizes he imporance of separaing he analysis of aggregae pass-hrough raes ino wo pars. The firs is a border phenomenon: o wha exen are here changes in pass-hrough raes a he level of impor prices, i.e. a he border? Second, o wha exen are hese border price changes ransmied o consumers or even offse by anicipaed curren or fuure moneary policy changes? Our analysis specifically deals wih he former quesion. 5 Ou of he 23 OECD counries for which appropriae saisical ess could be performed, we confirm ha here has been a weak endency oward declines in exchange rae pass-hrough raes. However, he srengh of his resul should no be oversaed. Low-power in he saisical ess performed and he limied abiliy in deecing changes in pass-hrough over ime requires ha hese resuls are evaluaed wih cauion. Pass-hrough declines were saisically significan in only 4 counries, bu significan increases in pass hrough ino impor prices also were eviden in 2 counries. The Unied Saes is no one of he counries ha have experienced saisically significan declines in he pass-hrough of exchange raes ino is impor prices. We coninue by underaking a direc examinaion of he underlying drivers ha may be causing he change in pass-hough raes ino aggregae impor prices. For any counry, such shifs could arise eiher because of changes in he underlying composiion of producs in a counry s impor bundle, or because of changes in he pass-hrough elasiciies associaed wih hese produc groups. A he level of specific disaggregaed producs, pass-hough elasiciies could evolve because of changes in indusry compeiive condiions or in macroeconomic condiions. We are able o sudy he role of impor composiion a a broad level, since he OECD makes available impor price series, by counry, a he level of five impor caegories: food, manufacuring, energy, raw maerials, and nonmanufacuring. We use hese impor series o furher documen he prevalence of LCP, PCP, or parial pass-hrough, and o underake ess For he purpose of he relevan macroeconomic debae, impor price series wih aggregaion are he appropriae unis for analysis. 4 An alernaive explanaion ress on moneary reacion funcions, as in Gagnon and Ihrig (2002). 5 Our focus should no be confused wih ha of relaed recen papers ha aemp o explain he pass-hrough of exchange raes ino a counry s CPI. In hese papers, exchange rae movemens lead o impor price passhrough. These ener wih weighs ino he aggregae CPI of counries, wih he weighs possibly o be adjused o reflec disribuion coss as in Bursein, Neves, and Rebelo (2001) or cenral bank reacion funcions as in Gagnon and Ihrig (2001). 3

5 for sabiliy in pass-hrough in hese disaggregaed caegories. Once again, here is srong cross-counry evidence on he prevalence of parial pass-hrough ino impor prices. Boh PCP and LCP are srongly rejeced as shor-run descripions of pass-hrough ino Manufacuring and Food impor prices. Since manufacuring rade now dominaes he impors of OECD counries, he parial pass-hrough of overall impor prices is explained. Bu, he issue of sabiliy of pass-hrough remains relevan for he broader debae. Ineresingly, hese pass-hrough raes for disaggregaed impor prices are highly sable over he wo decades of daa examined. We use hese sable pass-hrough elasiciies along wih ime-varying daa on impor composiion o consruc a series ha capures he effec of impor composiion on aggregae pass hrough. We hen run a horse race, conrasing he conribuion o aggregae pass-hrough changes of ime-varying macroeconomic series (counry size, inflaion, and exchange rae variabiliy) agains ha of rade composiion. Despie he fac ha macroeconomic variables especially exchange rae variabiliy maer for he ranking of counry pass-hrough levels (consisen wih Bacchea and van Wincoop (2002) and Devereux and Engel (2001) conjecures), hese variables have no been quaniaively imporan for explaining declining exchange rae pass-hrough ino impor prices across he OECD counries. Far more imporan for overall pass-hrough raes are changes in he composiion of indusries in each counry s impor baske. In paricular, he move away from energy as a high proporion of he impor bundles and he relaed subsanial rise in he share of manufacured producs has been he primary driver behind recen pass-hrough changes ino impor prices among numerous OECD counries. 2. Exchange Raes and Prices The micro-foundaions of pricing behavior by exporers are a useful saring poin for undersanding he dynamics of exchange rae pass hrough ino impor prices. The impor prices for any counry j, P m, j, are a ransformaion of he expor prices of ha counry s rading parners, x, P j, using he exchange rae (domesic currency per uni of foreign) E : P = E P m, j x, j (1) The expor prices, in urn, are a markup ( mkup ) over exporer marginal coss ( mc ). Using lower case leer o reflec logarihms, we rewrie equaion (1) as p = e + mkup + mc (2) m x x x x 4

6 where for simpliciy we have dropped he counry superscrip j We furher allow markups o have boh an indusry-specific fixed effec and a componen ha is sensiive o macroeconomic condiions, expressed for simpliciy a his poin as a funcion only of he exchange raes, mkup x = φ +Φ e (3) and specify exporer marginal coss as rising wih expor marke wages, marke demand condiions y. 6,7 x x 0 1 x w, and desinaion mc = c y + c w (4) so ha impor prices are wrien in general form as ( 1 ) 0 1 p = φ + +Φ e + c y + cw (5) m x This srucure permis exchange rae pass-hrough β = 1+Φ o depend on he srucure of compeiion in he indusry. This is consisen wih he large lieraure on explaining crosssecional differences on exchange rae pass-hrough, as has been exposied simply and eloquenly in Dornbusch (1987) and Marson (1990), among ohers, and suppored empirically by Kneer (1993) and Yang (1997). This srucure also has a direc analogue in he discussion of producer versus local currency pricing. If Φ = 0, producer currency pricing akes place; if Φ= 1here is local currency pricing and exporers fully absorb he flucuaions in exchange raes in heir own markups. 3. Exchange Raes Pass-Through ino Aggregae Impor Prices: The Evidence A. Daa and Esimaion Mehods. We capure he argumens of equaion (5) hrough a loglinear regression specificaion similar o ha esed hroughou he exchange rae passhrough lieraure: 8 6 More precisely, one should include as he appropriae demand variable an index of income levels across he producer s home marke and he desinaion marke for is expors. Since we do no have informaion on he composiion of demand facing exporers in differen counries, our proxy here is he GDP of he imporing counry. 7 The exchange rae can also be an argumen in he exporer s cos funcion o he exen ha he exporer relies on impored inpus or has oher coss ha move wih he relaive value of he desinaion marke currency. See Campa and Goldberg (1997), Feensra (1998), and Hummels, Ishii and Yi (2001) for evidence on increasing reliance on impored inpus and verical inegraion of producion across counries. 8 P.Goldberg and Kneer (1997) overview he relaionships beween hese sudies. Beyond he indusrial organizaion hemes, here also are sudies ha allow for pass-hrough elasiciies o differ beween appreciaion and depreciaion periods (Swamy and Thurman 1994) or o be disinc for anicipaed versus unanicipaed exchange rae changes (Marson 1990). 5

7 p = α + δw + βe + ϕy + ε (6) where p are local currency impor prices, e is he exchange rae, w is a primary conrol variable represening exporer coss, and y is a vecor of oher conrols, including real GDP of he desinaion marke. Biased esimaes of he pass-hrough coefficien could arise if foreign wages or GDP are correlaed wih exchange raes bu omied from he regression. We have used quarerly daa on impor price indices, from he OECD, compiled quarerly for 23 OECD counries, wih he series commencing around 1975 and ending in Nominal exchange raes are from he Inernaional Financial Saisics (series neu), defined in our specificaions as domesic currency per uni of foreign currencies (1/neu), so ha home currency depreciaions appear as increases in he nominal exchange rae series. Real exchange raes also are from he Inernaional Financial Saisics (series reu). The real GDP series used are hose of he imporing counries (source: Inernaional Financial Saisics). I is more difficul o find a primary conrol variable ha capures he shifing relaive coss of a counry s aggregaed rading parners. We consruc a consolidaed expor parners cos proxy by aking advanage of he IFS reporing of boh real reu and nominal neu exchange rae series and compuing W = neu P reu by counry in our sample. This x, j j j j gives us a measure of rading parner coss (over all parners x of imporing counry j), wih each parner weighed by is imporance in he imporing counry s rade. 9 We limi our sample o he OECD counries because we also need corresponding informaion on he impor prices of more disaggregaed caegories of impor. These disaggregaed series are no consisenly available ouside of his OECD daabase. A deail descripion of he daa is provided in he Daa Appendix. 6

8 For each of he 23 aggregae impor price indices, he firs sage of our analysis enails esimaing shor-run (one quarer) and long-run pass-hrough elasiciies, ˆ β, from equaion (5). Expressed in firs-differences, wih he addiion of lagged exchange rae and foreign producion cos erms o allow for he possibiliy of gradual adjusmen of impor prices o exchange raes, 10 he esimaion equaion is: 4 4 j j j j j j j j α i i i i ϑ i= 0 i= 0 p = + a e + b w + c gdp + (7) The shor-run relaionship beween exchange raes and he impor prices of counry j is given j by he esimaed coefficien a 0. The long run elasiciy is given by he sum of he coefficiens on he conemporaneous exchange rae and four lags of exchange rae erms j a i. The i= 0 esimaion mehodology applied is ordinary leas squares on variables in log differences, seleced afer we performed exensive checks on he saionariy of series and on appropriaeness of a coinegraion approach B. Esimaes of Exchange Rae Pass-Through ino Aggregae Impor Prices. Esimaes of exchange rae pass-hrough ino impor prices for he OECD counries are presened in Table 1. Taking unweighed averages across counries, we find ha average pass-hrough ino 10 We include up o four lags of exchange raes and foreign prices/producion coss in he regression. Mos of he pass-hrough response occurs over he firs and second lags afer an exchange rae change, so he inerpreaion of four quarers as long run is empirically validaed. An alernaive specificaion, which used a lagged dependen variable and relied on a parial adjusmen model generaed very similar empirical resuls (no repored in his version of he paper). However, he lagged dependen variable model imposed a se of consrains on model coefficiens ha were rejeced in he majoriy of cases. Consequenly, we repor he resuls of only he gradual adjusmen specificaion depiced in equaion (7). 11 We were unable o rejec he hypohesis ha he (log) series of impor prices, foreign coss, and effecive exchange raes were nonsaionary. Dickey Fuller Uni roo ess on he logarihmic values of he impor price, foreign coss, and exchange rae series in an economeric specificaion wih ime rends rejec he uni roo hypohesis a he 5% level in only 2 of 69 insances (3 series for 23 counries). We herefore accep ha he (log) series of impor prices, foreign coss, and effecive exchange raes are nonsaionary, wih he srong cavea ha hese saionariy ess have low power. We performed addiional ess o deermine wheher hese hree variables were coinegraed, i.e. wheher a linear combinaion of hese variables resuled in a saionary process. Absracing from he issue of low power of hese ess, and despie predicions of heory, we rejeced he coinegraion hypohesis and consequenly do no apply an error correcion model. We reached his conclusion by firs rejecing ha he log real exchange rae is saionary and rejecing he vecor (1,1,-1) as a coinegraing vecor as suggesed by he heory on he real exchange rae. We also esed for he possibiliy ha a coinegraing vecor exised bu was differen from wha exchange rae heory predics. Specifically, we run a model where p() = a + b*e() + c*w() + u(), and compue û() = ρ* û(-1) + e(). We es wheher he esimaed coefficien, ρ, is differen from uniy, and rejeced for only 3 cases he hypohesis ha ρ is differen from uniy a he 5% level. This is slighly higher han he 1.23 insances ha saisical error would sugges, bu sill very low. 7

9 impor prices is 0.46 in he shor-run and 0.64 in he long-run. These averages mask ineresing cross-counry differences in pass-hrough ino impor prices. The Unied Saes has relaively low pass-hrough, 23 percen wihin one quarer and 42 percen over he longer run. Pass hrough esimaes for counries such as France, Germany, and Swizerland are closer o 60 percen in he shor run and 80 o 90 percen over he longer run. Smaller European counries ypically have noisier and less sable pass-hrough raes, bu a precise relaionship beween pass-hrough and counry size is no empirically significan. Table 1: Elasiciies of Exchange Rae Pass-hrough ino Aggregae Impor Prices Shor Run Pass-Through Long-Run Pass-Through Ausralia.56*+.67*+ Ausria Belgium Canada.75*+.65*+ Czech Republic.39*+.60* Denmark.43*+.82* Finland.55*.77* France.53*+.98* Germany.55*+.80* Hungary.51*+.77* Ireland Ialy.35*+.35+ Japan.43*+ 1.13* Neherlands.79*+.84* New Zealand.22*+.22+ Norway.40*+.63* Poland.56*+.78* Porugal.63*+ 1.08* Spain.68*+.70* Sweden.48*+.38*+ Swizerland.68*+.93* Unied Kingdom.36*+.46*+ Unied Saes.23*+.42*+ Average Noes: *, + imply ha an elasiciy is significanly differen from zero or one a a 5 percen level. 8

10 A recurren issue in he recen macroeconomics lieraure is he prevalence of local currency price sabiliy (LCP) versus producer currency pricing (PCP). In our specificaions, LCP represens a null hypohesis of zero pass-hrough while PCP implies a pass-hrough of uniy. Noaion included in Table 1 highlighs our ess for he exisence of local currency pricing, producer currency pricing, or parial pass-hrough ino impor prices. LCP can be rejeced for 20 of he 23 counries in he shor run and 18 of 23 in he long run. PCP can also be overwhelmingly rejeced in he shor run (for 22 ou of he 23 counries) while in he long run is much harder o rejec (only for 7 of he 23 counries). For counries in he OECD, we overwhelmingly rejec complee pass-hrough (or PCP) and zero pass-hrough (or LCP) as a descripion of aggregae impor prices in he shor run. In he longer run, pass-hrough elasiciies are larger and closer o one, hus PCP is beer suppored as a longer run characerizaion. 12 C. Are here differences across counries in aggregae pass hrough? We have esed for he saisical differences across counries in pass-hrough elasiciies shown in Table 1 by re-esimaing equaion seven wih he daa pooled for all counries and imposing he resricion ha esimaed coefficiens be he same across counries. We rejeced his hypohesis a he one per cen level. We also re-esimaed equaion (7) for he pooled sample allowing coefficiens in he non-exchange rae erms o vary by counry and we also rejeced he hypohesis of equaliy of exchange rae pass hrough across counries. There are various heoreical argumens for cross-counry differences in exchange rae pass hrough raes. Among hese is a role for he sabiliy of local moneary policy, as in Devereux and Engel (2001). If exporers se heir prices in he currency of he counry ha has he mos sable moneary policies, impor prices in local currency erms would be more sable in counries wih more sable moneary policy. All else equal, exchange rae passhrough would be higher for counries wih more volaile moneary policy. Exchange rae 12 The resuls of hese ess will be sensiive o wheher he pass-hrough regression coefficiens are unconsrained, as in he specificaion repored, or consrained o lie beween zero and one. Alhough heoreically i is possible o jusify pass-hrough raes greaer han one, such raes are unlikely o be observed. If our esimaed coefficiens are resriced o he finie inerval [0,1], we can correc he sandard errors of he esimaed coefficiens using he Fisher ransformaion. Using his ransformaion we esed for he significance of he ransformed number z, where z =.5[ln(1+ βˆ )-ln(1- βˆ )]. This ransformaion ends o rejec equaliy o zero and one slighly more frequenly. For insance, in Table 1, 18 ou of 23 counries rejeced a coefficien equal o one in he shor run. In he long-run his happened for 17 ou of 23. We have performed similar ess for he disaggregae impor price daa wih similar resuls. 9

11 variabiliy and local moneary volailiy could also ener hrough exporer compeiion for marke share, as discussed in Froo and Klemperer (1989): exchange rae pass-hrough may be lower when nominal exchange rae variabiliy is high and exporers o a counry ry o mainain local marke share. Counry size may be anoher imporan facor in ranking passhrough elasiciies of counries. As iniially exposied by Dornbusch (1987), exchange rae pass-hrough may be higher if he exporers are large in number relaive o he presence of local compeiors. One approximaion o his poin is ha pass-hrough elasiciies migh be inversely relaed o counry real GDP. An alernaive approach would be o also consider measures of secor-specific openness for counries. We es for he imporance of hese alernaive hypoheses by re-esimaing he model in equaion (7) and allowing he coefficiens on shor-run and long-run exchange rae passhrough o be a funcion of observable macro economic variables so ha he esimaed coefficiens for pass-hrough are subsiued in equaion (7) by he following expression a j i j j = α + β x + ε, for i = 0,, 4 i i (8) where x j is a vecor represening all he exogenous regressors ha may explain cross-counry differences in exchange rae pass hrough. We have used as exogenous variables: counryspecific quarerly inflaion raes, money growh raes, exchange rae volailiy, and real GDP during he sample period. The ime series variables used in consrucing he righ-hand-side macro variables are all measured quarerly over he sample period 1975:1 o 1999:4. These variables include: Money measured as he annualized growh rae of he money supply (in logs); Inflaion is annualized inflaion rae, based on consumer price indices (in logs). Exvol is he average of he monhly squared changes in he nominal exchange rae during he previous year; GDP: is he nominal value in naional currency deflaed using he CPI deflaor and convered ino U.S. dollar a he average 1996 nominal exchange rae. The resuls of his specificaion exploring he macroeconomic deerminans of exchange rae pass-hrough are presened in Table 2. These resuls show ha counry-specific raes of exchange rae pass-hrough ino impor prices are significanly correlaed wih inflaion, money growh, and nominal exchange rae volailiy. Inflaion and money growh are highly correlaed variables, and when included joinly as deerminans of exchange rae pass-hrough we find ha inflaion raes are saisically significan, while he rae of money growh is no (las column of Table 2). This resul is inuiive. Counries wih higher raes of 10

12 inflaion should have higher raes of pass-hrough of exchange raes ino impor prices. Counries wih more nominal volailiy have higher pass-hrough raes. The resul ha lower nominal volailiy is associaed wih lower pass-hrough is consisen wih he main heoreical resuls of Devereux and Engel (2001). The role of counry size, however, is insignifican in he rankings of pass-hrough raes across counries. Despie he observaion ha U.S. pass-hrough raes are quie low, across he OECD here is no sysemaic relaionship beween pass-hrough and a counry real GDP. In fac, he poin esimae is insignifican, reflecing he fac ha some large counries have high pass-hrough (Japan) while some small counries have low pass-hrough (Czech Republic). Table 2 Deerminans Pass-Through Elasiciies ino Impor Prices: Alernaive Cross-Counry Panel Regression Specificaions Exogenous Variable (mean) Money Growh (0.010) Inflaion (0.004) Exchange Rae Volailiy (0.111) Real GDP (2.261) A. Deerminans of Shor Run Elasiciies 2.566** (0.284) ** (1.054) 3.900** (0.251) B. Deerminans of Long Run Elasiciies (0.340) (0.340) 7.394** (1.422) 2.370** (0.465) (0.007) Money Growh (0.010) Inflaion (0.004) Exchange Rae Volailiy (0.111) Real GDP (2.261) 5.129** (0.693) ** (1.567) 7.010** (0.482) (0.718) (0.719) ** (1.836) 2.100** (0.727) 0.069** (0.013) Adj. R ***, **, * indicae saisical significance a he 1, 5 and 10 percen levels, accordingly. All regressions are weighed leas squares. 11

13 D. Sabiliy of Aggregae Pass hrough Elasiciies As noed in our inroducion, an ousanding issue is wheher pass-hrough raes have been declining over ime, and if so, figuring ou if such declines are relaed o changes in macroeconomic policy variables. We can confron he firs par of his issue direcly by performing srucural change ess on he pass-hrough elasiciies, alhough such ess will have limied saisical power in he small daa sample available for analysis. One sandard es is a Chow es, wherein we firs assume an exogenously imposed break poin in he passhrough relaionship and perform associaed ess for parameer sabiliy. A second se of ess we perform has a similar flavor, bu insead allows for endogenously deermined srucural break poins. 13 In he process of doing hese ess, we furher idenify he dominan direcions of pass-hrough changes. The ess are implemened for all counries excep for Hungary and he Czech Republic, for which he available daa samples are shorer han for he oher counries. In our implemenaion of he Chow-ess we compare elasiciies esimaed over he firs half of he sample, 1975 hrough 1989, wih hose over 1990 hrough The resuls from his spli sample approach are ha here is a mix of increases and decreases in exchange rae pass-hrough elasiciies across counries. Shor-run and long-run exchange rae passhrough elasicies declined for 15 of he 21 counries, and increased for he oher 6 counries. While declines appear more prevalen, he Chow ess deec significan changes in only 4 of hese cases. The second se of sabiliy analyses es for he presence of a srucural break in passhrough using he mehods proposed by Andrews (1993) and Andrews and Ploberger (1994). These mehods es for he exisence of a srucural break poin in he saed relaionship a some unknown dae wihin he sample period. These ess have he advanage ha he researcher does no need o specify a priori he dae in which he srucural break akes place. However, hese ess are asympoic and heir power in our conex is quie limied by he number of observaions in our impor price series (generally around 100 quarers per series). While shor run pass-hrough sabiliy is also rejeced for 7 counries, i is difficul o assign he iming of insabiliy o a paricular break dae, suggesing ha he insabiliy is gradual raher han associaed wih a disinc poin in ime. We can never rejec sabiliy of long run pass-hrough according o hese ess. 13 Hansen (2001) provides a good criique of differen ypes of srucural change ess. 12

14 E. Exchange Rae Pass hrough ino Disaggregaed Impor Prices. In addiion o he counry aggregaes on impor prices, he OECD compiles daa on disaggregaed impor prices a he counry level for he same counries in he sample (excep Iceland) for five produc caegories: Food, Energy, Raw Maerials, Manufacuring, and Non- Manufacuring producs. We reesimaed equaion (7) for his sample of disaggregaed price daa. 14 As deailed in Appendix Table 1 15 and summarized in Table 3, mos indusries exhibi a sriking degree of parial pass-hrough. For each produc caegory excep Energy, we rejec he hypohesis of zero exchange rae pass-hrough (LCP) for more han half of he counries. For Manufacuring and Food, we similarly rejec complee pass hrough (PCP). The evidence in suppor of parial pass-hrough is sronges for Manufacuring impors, for which shor run pass-hrough differs significanly from boh zero and one in 19 ou of 22 counries. Food also exhibis parial pass-hrough in he shor run. Local currency pricing is ofen rejeced for Non-Manufacuring and Raw Maerials, wih rejecions of producer currency pricing are more mixed across counries. We have explored furher he possibiliy of mis-specified equaions on pass-hrough of exchange raes ino energy prices by using only bilaeral exchange raes agains he U.S. dollar and alernaive exporer cos series (he U.S. dollar price of energy). In hese specificaions, while shor-run pass hrough is less han one for mos counries, long run pass hrough is essenially one in all cases (only rejeced in one case) We also performed ess for nonsaionariy of each of hese price series and, by counry, for he exisence of a coinegraing relaionship beween hese series, he exchange rae, and he foreign price. The resuls of hese ess were similar o hose for he aggregae impor price series. Mainly, we could no rejec nonsaionariy of impor price levels and we could rejec he exisence of a coinegraing relaionship among he hree variables. One shor-coming of our esimaion sraegy is ha he same weighed foreign-wage variable is used in regressions for he differen impor aggregaes. This assumpion resuls in measuremen error ha may bias he foreign wage variable o he exen ha counry impors of differen goods are sourced from counries of differen imporance as rade parners. We suspec he errors associaed wih his assumpion o be sronges for he energy and raw maerial impors of counries, were supply is likely o be concenraed in a se of counries unrepresenaive of he mix in he counry s aggregae impor bundle. However, here is a significan daa consrain in building impor marginal cos indices ha are indusry and counry specific. We insead use a mehodologically homogeneous approach for any counry, applying an index measure of foreign uni coss, by counry, in he regressions. 15 Appendix Table 1 provides hese esimaes, by counry. Anoher imporan issue wih respec o moneary policy is he pass-hrough comparison for final goods prices versus impored inermediae goods prices (Obsfeld 2000). Energy and Raw Maerials can be viewed as being closer o classificaion as impored inermediae goods han Food, Manufacuring, and Non-manufacuring Producs. 16 We also performed pass-hrough esimaion for more disaggregaed produc caegories wihin he energy secor. As repored in he working paper version of his paper [Campa and Goldberg (2002)], pass-hrough ino he impor prices of hree more disaggregaed energy impors coking coal, seam coal, and oil were more precisely esimaed. Coking coal and seam coal, viewed as more heerogeneous producs han oil, ofen exhibied lower degrees of exchange rae pass hrough han oil prices. 13

15 We conduc furher analysis of he pass hrough elasiciies using ime-series panel specificaions wihin counry, and es wheher he pass hrough elasiciies of any indusries differ saisically from a base indusry. Taking prices of manufacuring impors as a base, we find somewha regular paerns in he rejecions of equaliy of pass-hrough coefficiens across indusries. The pass hrough of exchange rae changes ino food and agriculural producs is no saisically differen from pass-hrough ino manufacuring in any of he 22 counry regressions. The mos exreme conrass are beween he pass hrough ino manufacuring goods prices versus ino he prices of energy producs: in 9 (8) of 22 counries we see saisically differen coefficiens on energy price sensiiviy o exchange raes in he shor run (long run). 17 Table 3: Exchange Rae Pass Through Tendences in Disaggregaed Bundles of Impor Prices Shor run Rejec local currency pricing (β=0) Rejec producer currency pricing (β=1) Enries in able show number of counries for which each hypohesis is rejeced. Toal number of counries is 23 for all impors, 22 for disaggregaed producs. Food Energy Raw Maerials Manufacuring Non- Manufacuring Rejec boh LCP and PCP Average Pass-Through elasiciy Long run Rejec local currency pricing (β=0) Rejec producer currency pricing (β=1) Rejec boh LCP and PCP Average Pass-Through elasiciy These findings of differen pass hrough elasiciies across indusries, and in paricular via he role of energy impors, sugges a possible moivaion for changing pass-hrough ino aggregae bundles of impor prices over ime. The counries of our sample have seen large 17 For raw maerials impors, in 5(4) of 22 counries we see saisically differen coefficiens on price sensiiviy o exchange raes in he shor run (long run). In non-manufacuring, he number is 2 (4) of

16 changes in he composiion of heir impor bundles since he 1970s, and coninuing hrough ino he early 2000s. The main forces a work have been remendous increases in he relaive imporance of manufacuring impors, and declines in he relaive imporance of raw maerials, and especially energy. Figure 1 shows he evoluion in impor composiion for he counries in our sudy, specifically depicing he share of manufacured goods in counry impor bundles for he years 1980, 1992, and The firs bar shown for each counry depics manufacuring share for In 1980 manufacuring impors ypically were less ha 50 percen of he overall (merchandise) impor bill for mos counries. Counries heavily relian on impored energy had much smaller shares of manufacuring impors in oal impors: noably, for Japan, Ialy and Spain hese shares were below 30 percen. Due o lower energy prices, changes in energy policies, and he dramaic growh of manufacuring rade, by he 1990s here was a sriking cross-counry shif in he composiion of impors. By 1992 manufacured producs accouned for 65 percen of impors in he OECD counries, wih many counries having shares of manufacuring impors of more han 70 percen of oal impors. A he same ime, hese counries experienced a clear decline in he share of energy and raw maerial producs in oal impors and an almos idenical increase in he share of manufacuring producs. This rend coninued during he nex decade. By 2002, he average share of manufacuring impors ino oal impors for he OECD was 70 percen. Japan coninued o have he lower share of manufacuring impors wih 46.7 percen bu his share had more han ripled since As repored in Table 4, sabiliy ess applied o he exchange rae pass-hrough raes ino hese impor prices seldom find saisically imporan evidence of insabiliy. Long run insabiliy is only observed in 4 of 100 impor price regressions according o Chow Tess, and in 1 of 100 cases by he Hansen ess. Insabiliy in shor-run pass hrough is observed in 13 of 100 cases according o Chow ess, and 15 of 100 cases using Hansen ess. Differences over ime in pass-hrough poin esimaes a he produc level, however, are small compared wih differences observed in he aggregaed impor price series. Moreover, many of hese insances of produc caegory insabiliy are aribued o daa from New Zealand, he Neherlands, and Japan. Togeher, hese observaions sugges greaer sabiliy in componen series han in pass-hrough elasiciies for aggregae bundles of impors Insabiliy ess were also performed on he energy impor price regressions bilaerally esimaed agains dollar exchange raes. There is essenially no evidence of insabiliy in hese relaionships (2 rejecions of 42 cases). 15

17 Figure 1 Manufacuring Goods Share of Impors Counry 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Unied Saes Japan Germany France Ialy Unied Kingdom Canada Ausria Belgium Denmark Finland Ireland Percenage oal impors Mexico Neherlands Norway Porugal Spain Sweden Swizerland Ausralia New Zealand Table 4: Pass-Through Parameer Sabiliy Disaggregaed Price Series Enries in able show he number of counries for which sabiliy is rejeced for each ype of impor price series. The oal number of counries is 20 for disaggregaed impor price caegories Raw Maerials Non- Manufacuring Food Energy Manufacuring Chow Tes Shor run insabiliy Long run insabiliy Hansen Tes Shor run insabiliy Long run insabiliy Undersanding he Evoluion of Aggregae Pass Through A. Theoreical underpinnings. Various explanaions could be offered for changes over ime in he counry raes of exchange rae pass hrough. In his secion we disinguish beween hese general macro-based explanaions and an alernaive explanaion based on changes over ime in he composiion of impors. Recall from equaion (5) ha any impor price series is m given by φ ( 1 ) 0 1 p = + +Φ e + c y + cw. This equaion is derived direcly from a firs order x condiion of a firm and i holds a he individual produc level. In he previous secion we 16

18 used his equaion as a jusificaion for he esimaion of a pass-hrough rae for he counry using an aggregae impor price series. Obviously, his aggregae impor price is composed of a weighed average of indusry specific impor price indices. If N producs are wihin a counry s impor bundle, we can rewrie in equaion (5) for an aggregae index as N N N m = i i + i 1+Φi + i i= 1 i= 1 i= 1 i, x α φ α ( ) α (9) p e mc where αi represens he weigh of any produc caegory i in a counry s impor bundle. Shorrun aggregae pass-hrough β and changes in aggregae pass hrough hen can be expressed as N αi ( 1 i) (10) β = +Φ i= 1 N i ( 1 i) i i (11) β = α +Φ + α Φ i= 1 i= 1 N Equaion 11 saes ha changes in aggregae pass hrough can arise due o changes in he weighs of differen ypes of producs in he overall impor bundle, or due o changes over ime in he markup sensiiviies o exchange raes for paricular indusries. We can easily nes in his model he macroeconomic hypohesis formulaed in he pass-hrough macro lieraure (Engel, Devereux, Taylor) by specifying his markup response as having an indusry fixed effec relaed o he indusry s compeiive condiions and a imevarying componen relaed o macroeconomic variables. Φ i = µ i + µ Χ so ha Φ i = µ Χ (12) Combining equaions (11) and (12), N n i i i i= 1 i= 1 (13) β = α Φ + α µ Χ Equaion 13 saes ha aggregae impor-price pass hrough can change because of he impor composiion effec and because of he effecs of macroeconomic condiions on markups. B. Composiion versus macro variabiliy as deerminans of evolving pass-hrough. Aggregae pass hrough elasiciies, impor composiion, and macroeconomic ( exogenous ) variables over he full period are no represenaive of behavior over shorer inervals. Consequenly, o es he ype of relaionship given by equaion (13) we spli he 17

19 full sample period ino four sub-periods: 1975:1 o 1981:4, 1982:1 o 1988:4, 1989:1 o 1995:4, and 1996:1 o 2003:4. For each sub-period, we run a firs-sage regression of he ype shown by equaion (7) and generae four esimaes of he shor- and long-run passhrough elasiciies of aggregaed impor prices for each counry. We also inroduce a imeseries panel version of equaion (8) as he second sage specificaion, wih macroeconomic variables measured over he respecive sub-periods, 19 and add an impued rade composiion variable. The second-sage specificaion over he esimaed elasiciies (4 per counry, 23 counries) akes he form: ˆ β = γ ln money + γ lninflaion + γ ln exchvol + γ lngdp j j j j sr orlr j + γ lnimpued + γ 5 (14) We apply a weighed leas squares procedure in order o reduce he imporance of he noisier parameer esimaes in driving overall conclusions (he weighs are he inverse of he esimaed sandard error each pass-hrough). Wihin his ime-series panel approach, he second sage regressions include ime dummies in order o accoun for oher period-specific fixed effecs ha are no capured by he exogenous righ-hand-side variables. For each counry and ime period he impued aggregae pass-hrough elasiciy capures he changes in a counry s aggregae pass-hrough elasiciy ha are aribuable exclusively o changes in is composiion of impors. The consruc uses he ime-invarian (full sample period) esimaes of pass-hrough elasiciies for each of he five indusry groupings for each counry. The impued elasiciy is consruced by varying each period he weighs of each ype of impor in each counry s oal impor bundle. We use as weighs he impor share values a 1980, 1986, 1992, and Two furher adjusmens need o be done o his variable. Firs, in is compuaion we use esimaed pass-hrough elasiciies for each of he five produc caegories. Some of he poin esimaes for hese elasiciies are ouside he inerval [0,1] wih large sandard errors. Given ha pass-hrough elasiciies beyond [0,1] are hard o jusify, we compue he impued rade elasiciies resricing he esimaed elasiciies a he produc level o lie wihin his inerval. 20 Second, he impued rade elasiciy variable is likely o be correlaed wih he error erm, since i has been esimaed 19 The GDP variable reflecs he 1996 U.S. dollar value of each counry s GDP in 1978, 1984, 1990 and We also esimaed he resuls from his analysis wihou resricing he esimaed elasiciies a he produc level o lie wihin he [0, 1] inerval. The resuls in his case were acually sronger han hose repored in Table 18

20 using he full sample period. Therefore, we insrumen his variable using he impued elasiciy using daa from he previous period. This variable is highly correlaed wih he impued rade elasiciy for he period, and is pre-deermined, making i a valid insrumen. The resuls from hese specificaions are repored in Table 5. Consisen wih Taylor s (2000) argumens, shor run pass-hrough is lower when a counry achieves lower inflaion, or less money growh. Lower and more sable moneary condiions induce producers o pass on a smaller percenage of cos shocks ino final goods prices. These resuls, however, are never saisically significanly differen from zero. Exchange rae volailiy is highly noisy and does no have any clear effec on pass-hrough raes. Finally, he measure of passhrough elasiciy impued from he evoluion of he pass-hrough elasiciies esimaed from disaggregaed daa is always posiive and saisically significan. Table 5 Macroeconomic Variables versus Composiion as Deerminans of Pass-Through Changes over Time: Time Series Panels Esimaion Mehod Weighed Leas Squares Shor-Run Pass-Through (log Levels) Weighed Leas Squares Insrumenal Variables Long-Run Pass-Through (log Levels) Weighed Leas Squares Weighed Leas Squares Time dummies Money (1.156) Inflaion (2.252) Exchange rae volailiy (x100) (1.836) Trade Impued Elasiciy 0.934** (0.476) Real GDP (0.024) (1.960) (1.825) 0.945** (0.485) (0.021) (5.035) (9.881) (2.236) 1.386*** (0.376) (0.028) (1.931) (3.958) (3.564) 1.053*** (0.545) (0.036) (3.268) (3.316) 1.146*** (0.310) (0.035) Insrumenal Variables (8.878) (15.30) (4.040) 1.961** (0.671) (0.041) Adj. R Adj. R2 from specificaion w/only Macro variables Adj.R2 from rade impued elasiciy only # obs ***, **, * indicae saisical significance a he 1, 5 and 10 percen levels, accordingly. All regressions are weighed leas squares, ime series panel specificaions. Repored regressions exclude counry dummies bu include ime dummies. 5. The impued rade elasiciy was always highly significan and had greaer explanaory power han he resuls repored in able 5. 19

21 Despie he saisical significance of inflaion in hese specificaions, he included macroeconomic variables accoun for a negligible amoun of he variaion over ime in passhrough elasiciies across counries. The join explanaory power ha hese macro variables have in explaining he evoluion of pass-hrough is basically zero (he adjused R2 saisic is negaive). F-ess canno rejec he hypohesis ha hese macro variables have no explanaory power for long-run pass hrough raes across our OECD counry sample. Common ime dummies, macro variables and impued rade shares explain abou 30 percen of he observed differences over ime in he shor-run pass-hrough elasiciies of counries. Almos all of he explanaory power of he regressions comes from he impued rade elasiciy variables, even hough our composiion argumens have been made wih only he coarsely disaggregaed series ha are available in he impor price daa. Trade composiion effecs are he clearly dominan explanaions for movemens over ime in he shor-run and long-run sensiiviy of impor prices o exchange raes. Furher evidence for he role of he impued measure racks comes from direc ess agains he changes observed in he acual pass-hrough esimaes for he sample of 20 counries for which comparisons are possible. The impued measure generaes declines for 21 of he 27 cases where declines were observed in he acual daa. The impued measure generaes pass-hrough increases in 9 of he 14 cases where increased pass-hrough was observed in he acual daa. The main reason for his decline in he aggregae impor price elasiciy is due o he decline in he relaive weigh in overall impors of energy and raw maerials. These are he wo producs for which he impor price elasiciies were ofen highes. According o his calculaion, he aggregae pass-hrough elasiciy for he Unied Saes would have declined from 0.36 o 0.27 beween 1980 and 2002 (a 25% decline) solely due o he change in he produc composiion of impors. For Ialy, he decline would have been of a larger absolue magniude, from 0.86 o Conclusions In his paper we have provided cross-counry, ime-series, and indusry-specific evidence on he pass-hrough of exchange raes ino impor prices across a large sample of OECD counries. As a cross-counry average, impor prices in local currencies reflec 46 percen of exchange rae flucuaions in he shor run, and nearly 65 percen over he longrun. By conras, exchange rae pass-hrough ino U.S. impor prices is abou 23 percen in 20

22 he shor run and 42 percen over he long run. For he OECD as a whole, parial passhrough is overwhelmingly he bes descripion of impor price responsiveness shorly afer an exchange rae movemen. In he longer run, pass-hrough elasiciies are closer o one, alhough complee pass-hrough or producer currency pricing is sill rejeced for many counries. Macroeconomic variables play a significan bu limied role in explaining crosscounry differences in levels of pass-hrough elasiciies. Mos noably, pass-hrough ino impor prices is lower for counries wih low average inflaion and low exchange rae variabiliy. While here is evidence ha pass-hrough raes have been declining over ime in some counries, his paern of pass-hrough decline has no been a common or robus feaure of all OECD counries. Shor-run exchange rae pass-hrough elasiciies rise wih price inflaion (or higher money growh raes). Despie saisical correlaions, he quaniaive imporance of hese macroeconomic effecs have been small in he OECD. Recen argumens for viruous cycles beween inflaion, money policy effeciveness and pass-hrough have no been of firsorder imporance wihin he OECD counries. Observed changes in pass-hrough raes ino aggregae impor prices more closely reflec changes over ime in he composiion of impor bundles of OECD counries. Passhrough elasiciies for manufacuring producs and food producs are generally parial, so ha boh local currency price sabiliy and producer price sabiliy are rejeced for mos counries. By conras, energy and raw maerial impors appear o have pass-hrough elasiciies closer o one. The shif in he impor composiion of rade ha has aken place over he las wo decades oward manufacures and away from energy and raw maerials impors have conribued significanly o pass-hrough declines in abou half of he OECD counries examined. These ypes of changes of pass-hrough ino impor prices -- associaed wih widespread changes in he composiion of indusrial aciviy and rade --- are likely o be more durable han hose associaed wih he ypes of changes in macroeconomic policy environmens observed in he OECD in recen decades. Our findings inform recen discussions of he exchange rae disconnec puzzles, wherein exchange rae movemens have been shown o have a much smaller effec on consumer prices han would generally be expeced. By focussing on he impor prices, we have shown ha he border prices of goods are in fac very sensiive o exchange rae flucuaions, even for he Unied Saes. This evidence leads o he implicaion ha he focus of he disconnec is likely no o depend on wheher inernaional prices for goods are se in 21