Monetary Policy and Exchange Rate Interactions in an Open Economy

Transcription

1 Moneary Policy and Exchange Rae Ineracions in an Open Economy Hilde C. Bjørnland* Universiy of Oslo and Norges Bank May 3, 25 [PRELIMINARY AND INCOMPLETE] COMMENTS ARE WELCOME Absrac This paper analyses he effecs of moneary policy in an open economy hrough srucural VARs, paying paricular aenion o a possible inerdependence beween he moneary policy sance and exchange rae movemens. Previous sudies of he effecs of moneary policy in open economies have ypically found only small (and even puzzling) effecs on he exchange rae. However, many of hese sudies have disregarded a possible ineracion beween moneary policy and exchange rae movemens by placing zero resricion on heir conemporaneous correlaion. By insead imposing a long-run resricion on he real exchange, hereby allowing he ineres rae and he exchange rae o reac simulaneously o any news, we find he inerdependence o increase considerably. In paricular, a moneary policy shock now implies a srong and immediae appreciaion of he exchange rae. However, he qualiaive properies of a moneary policy shock found in he esablished lieraure are sill preserved. In paricular, a moneary policy shock ha increases he ineres rae emporarily lowers oupu and has a sluggish bu negaive effec on consumer price inflaion. Keywords: VAR, moneary policy, exchange rae ineracions, idenificaion. JEL-codes: E61, E52, E43. * Valuable commens and suggesions from Leif Brubakk, Sharon McCaw and Kjeil Olsen are graefully acknowledged. The views expressed are hose of he auhor and should no be inerpreed as reflecing hose of Norges Bank. 1

2 1 Inroducion Recenly, here has been a grea deal of ineres in quanifying he effecs of moneary policy on a se of macroeconomic aggregaes. These empirical sudies have been paricularly successful in exracing informaion useful o evaluae he fi of he newly developed dynamic sochasic general equilibrium (DSGE) models like hose of Chrisiano e al. (25) and Alig e al. (23) for he closed economy and Adolfson e al. (25) for he open economy. In paricular, he auhors esimae he parameers of he DSGE model by maching condiional dynamics in response o cerain srucural shocks. For insance, he impulse response o a moneary policy shock is specified and aken as given; he DSGE parameerizaion is hereafer chosen o give he same answer. According o Faus (25), Clarifying he idenificaion in his way could make more producive he effors o consrucively criique he models by boh cenral bank saff and ouside observers. The quaniaive effecs of moneary policy shocks have o a large exen been addressed in erms of vecor auoregressive (VAR) models, iniiaed by Sims (198). Through he recen applicaions of Leeper e al. (1996) and Chrisiano e al. (1999, 25), one has reached a consensus on how moneary policy affecs he closed economy (like he U.S.). In paricular, a conracionary moneary policy shock ha increases he ineres rae, gradually lowers oupu and has a sluggish bu evenually negaive effec on inflaion (ha peaks afer 1-2 years). These resuls have been found using VARs ha are idenified wih some sor of recursive idenificaion sraegy beween he macroeconomic environmen and moneary policy. However, sudies idenifying moneary policy wihou his resricion, have also found similar resuls, see for example Faus e al. (24) and he references herein 1. VAR sudies of he open economy have on he oher hand provided less of a consensus wih regard o he effecs of moneary policy, and have along cerain dimensions even provided some puzzling resuls. The exchange rae, in paricular, has been a roublesome issue. According o Dornbusch s (1976) well known exchange rae hypohesis, a conracionary moneary policy shock (ha increases he domesic ineres raes relaive o foreign ineres raes) will lead o an impac appreciaion followed by a persisen depreciaion of he domesic currency. This view underlies mos models for policy analysis, as emphasised by Walsh (23). 2 However, when confroned wih daa for he open economy, very few VAR sudies have found suppor for Dornbusch overshooing. Insead hey have found ha following a conracionary moneary policy, he real exchange rae eiher acually depreciaes, or, if i appreciaes, i does so for a prolonged period of up o hree years, hereby giving a humpshaped response ha violaes he uncovered ineres pariy (UIP) condiion. In he lieraure, he firs phenomenon has been ermed he exchange rae puzzle, whereas he second has been 1 An excepion is Uhlig (25), who using sign resricions finds moneary policy o have no clear effec on GDP, alhough prices move gradually as expeced. 2 Walsh (23, pp ) argues ha "The real exchange rae does no overshoo; i falls below he baseline and hen reurns smoohly o is iniial level. This behaviour is consisen wih he empirical evidence of Eichenbaum and Evans (1995)". However, as we will see below, his resul is acually no a all consisen wih Eichenbaum and Evans (1995), who finds a persisen overshooing of up o hree years. 2

3 referred o as delayed overshooing or forward discoun puzzle, see Cushman and Zha (1997). Mos of he recen VAR sudies of he open economy, including Eichenbaum and Evans (1995), Kim and Roubini (2) and Kim (2) for he G7 counries, Peersman and Smes (22) and Smes and Wouers (22) for he aggregae Euro area, Mojon and Peersman (23) for individual Euro area counries and Lindé (23) for Sweden, have found evidence of a leas one hese puzzles, which is in sharp conflic o convenional wisdom. On he oher hand, evidence of hese puzzles has been so persuasive ha he puzzles hemselves are now considered consensus, as is eviden from he following quoe in Adolfson e al. (25, p. 37): Sill, we find he empirical relevance of he nominal and real fricions o be such ha he impulse response funcions o a policy shock are very similar o he ones generaed in idenified VARs There is, however, one excepion and ha is he real exchange rae. Alhough he nominal fricions in he model provides some persisence in he real exchange rae following a policy shock, i is eviden ha he model does no provide us wih a humpshaped response of he real exchange rae which is a persuasive feaure of esimaed VARs. However, one major obsacle when aking he closed economy VAR o he open economy, is how o properly address a possible simulaneiy problem beween moneary policy and he exchange rae. In paricular, almos all of he sudies of open economies like hose referred o above deals wih a possible simulaneiy problem by eiher (i) resricing he exchange rae from reacing conemporaneously o a moneary policy shock or (ii) placing zero conemporaneous resricions on he response of he sysemaic ineres rae seing o an exchange rae shock. 3 The firs assumpion is equivalen o assuming ha he exchange rae mus have been fixed for he whole period under invesigaion; e.g. Mojon and Peersman (23). This assumpion is he mos implausible and leas used, in paricular in he pos Breon Woods era where mos exchange raes have been floaing. However, even in periods where he exchange rae has been parly fixed (managed floa), i is reasonable o expec a conemporaneous effec of moneary policy acions on he exchange rae. Very few (indusrialised) counries have managed o mainain a fixed exchange rae for a prolonged period, and even less so for he se of diverse counries analysed in Mojon and Peersman (23) where he exchange rae puzzle is paricularly pronounced. 4 The second se of resricions implies a reversing of he conemporaneous feedback, which is equivalen o assuming ha he moneary auhoriy ignores any surprise movemen in exchange raes ha have occurred during he ime in which decisions on he policy variables are made. As will be discussed in more deail below, o assume ha he policymakers do no pay aenion o surprise movemens in he exchange rae when seing policy is highly unlikely. Or, o cie Faus and Rogers (23, p. 146), If lierally rue, his would be quie 3 Kim and Roubini (2) provide evidence of delayed overshooing in only some individual counries (e.g. Canada and France). However, compared o mos oher sudies, hey allow for a simulaneous response beween he ineres rae and he exchange rae, and esimae insead an explici reacion funcion of he moneary auhoriies. However, he domesic ineres rae is resriced from reacing conemporaneously o a foreign ineres rae shock, which is a odds wih evidence from financial markes daa.. 4 The resricion is placed on a se of diverse counries such as Finland, France, Greece, Ireland, Ialy and Spain. 3

4 disappoining o hose of us who inform he Federal Reserve Board on a real-ime inraday basis regarding surprising movemens in financial markes. 5 Empirical sudies ha assume ha he moneary policymaker can reac o all shocks wihin a period excep hose o he exchange rae, herefore seems implausible. However, i is one hing o assume ha he policymakers are informed on a daily basis of surprise movemens in he exchange rae. I is quie anoher o assume ha he policymakers will ac accordingly o he news. This discussion involves wo disinc and separae issues. Firs, here is he issue of wheher he cenral bank should acually arge he exchange rae above wha is implied by he inflaion arge. From a heoreical poin of view, in he presence of complee exchange rae flexibiliy, he closed and he open economy version of opimal moneary policy model can be considered isomorphic o one anoher. Hence, he bes oucome is found when policymakers are inward looking, solely focusing on domesic arges (i.e. Rogoff, 2; Clardia, Gali and Gerler, 21). However, empirical evidence indicaes ha he exchange rae is no fully flexible, and ha here are subsanial deviaions from he law of one price for raded goods prices (Rogoff, 1996). Under he circumsances of incomplee pass-hrough, he analysis of moneary policy in an open economy will be subsanial differen from he one of a closed economy. In paricular, sric adherence o inward-looking policy objecives such as sabilisaion of domesic oupu and inflaion can no be opimal when firms markups are exposed o exchange rae flucuaions. Hence, he moneary auhoriies may aemp o affec he expeced fuure pah of he real exchange rae (see for example Corsei and Peseni, 21; Monacelli, 23). Secondly, even if he exchange rae is no among he arge variables, he cenral bank may sill pay aenion o where he exchange rae is heading. The fac ha he exchange rae is deermined in a forward-looking manner ha reflecs he expeced fuure reurn on he asse, may by iself provide imporan informaion abou he expeced developmen of he deerminans of he argeing variables. Thus, if he exchange rae is likely (even jus in some periods) o respond immediaely o a moneary policy shock, and moneary policy may respond quickly (a leas wihin he monh or quarer which is he usual sampling frequency in hese sudies), o an exchange rae shock, he srucural shocks canno be recovered using eiher of he recursive shor-run resricions on he parameers, a common pracice in he VAR lieraure on open economies. This poin has been forcefully made in Faus and Rogers (23), who by relaxing wha hey call dubious resricions and imposing a mos raher mild sign resricions (or shape resricions) are able o es he implicaions of he conemporaneous resricions mos commonly used in he VAR lieraure. By doing so, hey find ha he delayed overshooing resul is sensiive o dubious assumpions. By eliminaing his assumpion, heir resuls allow 5 The view ha moneary auhoriies pay aenion o financial markes is also suppored by Bjørnland and Leiemo (24), who show ha once one allow he cenral bank o reac immediaely o news in he financial marke (and sock marke in paricular), he ineracion beween moneary policy and sock marke developmens increase significanly. 4

5 for an early peak in he exchange rae, which migh give a role for he convenional overshooing model. On he oher hand, he auhors find ha very lile of he exchange rae variaion is due o hese moneary policy shocks and ha he bulk of he exchange rae variaion afer a policy shock is due o large variaion form UIP. Hence, he overshooing can no be driven by he Dornbusch's mechanism. On he oher hand, using he same kind of procedure as Faus and Rogers (23), bu imposing resricions on he impulses responses for several periods, raher han on he impac effec only (as is wha Faus and Rogers do), Scholl and Uhlig (25) find more evidence of boh he exchange rae puzzle as well as he delayed overshooing puzzle. However, he peak appreciaion happens somewha faser han he hree year horizon found in Eichenbaum and Evans (1995). There are a leas wo problems relaed o his alernaive idenificaion sraegy. Firs, by dropping implausible assumpions, he auhors may no longer be able o idenify he iems of ineres. In paricular, he moneary policy shock ha is idenified and well inerpreed in he closed economy VAR applicaions, may no longer be he same shocks ha are sudied using his alernaive resricion. Second, in even analysis where acual moneary policy acions ha are perceived as a surprise on he marke are colleced and analysed, auhors such as Zeelmeyer (24) among ohers finds ha a surprise moneary policy shock ha increases he ineres rae has a subsanial appreciaing effec on he exchange rae in a small open economy. This link beween surprise moneary policy acions and iniial exchange rae responses could herefore be a feaure ha he idenified VAR models a leas on impac should be able o replicae. Sudies like Scholl and Uhlig (25) who finds ha he exchange rae iniially depreciaes afer he conracionary moneary policy shock, is herefore highly implausible This paper analyses he effecs of moneary policy in an open economy hrough srucural VARs, paying paricular aenion o a possible inerdependence beween he moneary policy sance and exchange rae movemens. In paricular, we impose a combinaion of shor-run and long-run resricion on he parameers of he VAR model, so ha wihou deviaing oo much from he exising lieraure, we can idenify moneary policy shocks wihou having o resric he conemporaneous simulaneiy beween ineres rae and exchange rae movemens. The analysis is applied o Norway. To our knowledge, here have been no similar sudies of moneary policy in Norway before. On he oher hand, i is only recenly ha Norway abandoned a regime of argeing exchange raes, and can now be appropriaely described as following some kind of Taylor rule for moneary policy. This paper herefore conribues o he lieraure on how moneary policy can be idenified and analysed in a small open economy, as well as esablishing some sylized facs on he effecs of moneary policy in Norway. We find ha a conracionary moneary policy shock has he usual effecs idenified in oher inernaional sudies: emporarily increasing he ineres rae, lowering oupu and a sluggish bu negaive effec on consumer price inflaion (and real wages). On he oher hand, 5

6 conrary o mos of he inernaional sudies, we find a subsanial effec on he exchange rae which appreciaes immediaely. In secion 2 we discuss he VAR mehodology used o idenify moneary policy shocks and secion 3 discusses he empirical resuls. Secion 4 provides robusness checks and Secion 5 concludes. 2 The idenified VAR model In his sudy we explicily accoun for he inerdependence beween moneary policy and exchange raes wihin a VAR model by imposing a combinaion of shor-run and long-run resricions. In paricular, we build on he radiional VAR lieraure in ha we idenify recursively a sandard srucure beween macroeconomic variables and moneary policy, so ha moneary policy can reac o all shocks, bu he macroeconomic variables reac wih a lag o he moneary policy shocks. However, our approach differs from he radiional mehod in ha we also allow moneary policy o respond o he conemporaneous exchange rae, which iself is allowed o reac simulaneously o all shocks. We mus have an alernaive resricion in order o idenify and orhogonalise all shocks. We herefore assume insead ha moneary policy shocks can have no long-run effecs on real exchange raes. By using only one long-run resricion, we address he simulaneiy problem wihou deviaing exensively from he esablished lieraure (i.e., Chrisiano e al., 1999 for closed economies and Lindé (23) open economies) of idenifying a moneary policy shock as an exogenous shock o an ineres rae reacion funcion (he sysemaic par of moneary policy). Once allowing for full simulaneiy beween moneary policy and he exchange rae, he VAR approach is likely o give very useful informaion abou he simulaneous ineracion beween moneary policy and exchange raes in open economies. However, given ha we essenially replace one (shor run) resricion on he conemporaneous ineracion beween ineres raes and exchange raes, wih anoher (long-run) resricion, he plausibiliy of his resricion has o be esed, hrough a series of robusness ess. The choice of variables in he VAR model are chosen o reflec he heoreical se up of a New-Keynesian small open economy model, such as ha described in Clarida, Galí and Gerler (1999) and Svensson (2). In paricular, he VAR model comprises he log of he annual changes in he domesic consumer price index 6 (CPI) (π ) referred o hereafer as inflaion, log of real GDP (y ), he hree monh (NIBOR) ineres rae (i ), he log of he foreign ineres rae (i *) and he log of he real exchange rae agains a baske of rading parners (e ) (for deails, see appendix A). 7 Below we will show ha a combinaion of shor-run and long-run resricions on he esimaed VAR model will be sufficien o allow moneary policy sance and exchange raes o reac simulaneously o he idenified shocks. 6 Adjused for energy price changes and axes (CPI-ATE). 7 However, as is demonsraed below, oher variables were also ried ou. 6

7 2.1 Idenificaion Throughou his paper, we follow wha has now become sandard pracice in VAR analysis (see for example Chrisiano e al. 1999) and idenify moneary policy shocks wih he shock in an equaion of he form MP i = f(...) + σε, (1) where i is he insrumen used by he moneary auhoriy (usually he ineres rae) and f is a linear funcion ha relaes he insrumen o he informaion se (feedback rule). The moneary MP policy shock ε is normalised o have uni variance, and σ is he sandard deviaion of he MP moneary policy shock. We assume ha ε is orhogonal o he elemens in he informaion se ( ). Having idenified he feedback rule (from he variables ha are in he informaion se) he VAR approach concenraes on deviaions from his rule. Hence, such deviaions provide researchers wih an opporuniy o deec he responses of macroeconomic variables o moneary policy shocks ha are no expeced by he marke. Obviously, we hen assume ha he marke also knows he rule. Below we firs se ou o follow sandard pracice in many recen VAR applicaions, namely o idenify he differen srucural shocks hrough a series of conemporaneous resricions on he effecs of he shocks ono he variables. In paricular, i is commonly assumed ha macroeconomic variables such as oupu and prices do no reac conemporaneously o moneary shocks, whereas here migh be a simulaneous feedback from he macro environmen o moneary variables, see for example Sims (198, 1992), and Chrisiano e al. (1999) among many ohers. Bagliano and Favero (1998) show ha when moneary policy shocks are idenified in his recursive way on a single moneary policy regime, hese shocks sugges a paern for he moneary ransmission mechanism ha is consisen wih he impulse responses of moneary policy shocks idenified insead using financial marke informaion from ouside he VAR, such as in Rudebusch (1998). This would also limi he pracical imporance of he Lucas criique, since a sable regime does no require any re-parameerizaion. However, as discussed above, a more profound problem wih his recursive idenificaion is ha once one include high frequency daa such as he exchange rae ino he VAR, i becomes difficul o assume an exclusion resricion beween he simulaneous response in he ineres rae and he exchange rae o any kind of news. To be able o solve his simulaneiy problem, we herefore use insead a long-run resricion ha does no limi he conemporaneous response of he variables. This resricion idenifies moneary policy shocks as hose shocks ha have no long-run effec on he level of he real exchange rae. This is a sandard neuraliy assumpion in he moneary policy lieraure, used among oher in Clarida and Gali (1994) in a differen applicaion. I implies ha PPP will hold in he long-run, a leas wih respec o he moneary policy shock, which is perfecly consisen wih he Dornbusch's hypohesis discussed above. 7

8 Assume ha Z is he (5x1) vecor of he macroeconomic variables discussed above, which can be ordered as follows: Z = ( i *, y, π, i, e ), where we for now assume ha all variables bu inflaion are firs differenced o obain saionariy. 8 The reduced form VAR can be wrien in marix form as A(L)Z =v, (2) where A(L)= j= j A L j is he marix lag operaor, A =I and v is a vecor of reduced form residuals wih covariance marix Ω. Assuming A(L) o be inverible, (2) can be wrien in erms of is moving average Z=B(L)v, (3) where B(L)=A(L) -1. The idenificaion of he relevan srucural parameers, given he esimaion of he reduced form, is a radiional problem in economerics. A srucural model is obained by assuming orhogonaliy of he srucural shocks and imposing some plausible resricions on he elemens in B(L). Following he lieraure, we assume ha he underlying orhogonal srucural disurbances (ε ) can be wrien as linear combinaions of he innovaions (v ), i.e., v=sε. (4) The VAR can hen be wrien in erms of he srucural shocks as Z=C(L)ε, (5) where B(L)S=C(L). (6) Clearly, if S is idenified, we can derive he MA represenaion in (5) since B(L) can be calculaed from he reduced form esimaion of (2). Hence, o go from he reduced form VAR o he srucural inerpreaion, one needs o apply resricions on he S marix. Only hen can one recover he relevan srucural parameers from he covariance marix of he reduced form residuals. To idenify S, we firs assume ha he ε s are normalised so hey all have uni variance. The normalisaion of cov(ε ) implies ha SS = Ω. Wih a five variable sysem, his 8 This assumpion is furher discussed and relaxed in he empirical analysis below. 8

9 imposes 15 resricions on he elemens in S. However, as he S marix conains 25 elemens, o orhogonalise he differen innovaions, en more resricions are needed. Of hese, here will be nine conemporaneous resricions direcly on he S marix. These are consisen wih a Cholesky decomposiion used on he par of he S marix ha ignores he open economy variables, and as discussed above, are sandard in he VAR lieraure on moneary policy shocks. In addiion we impose one commonly acceped neuraliy resricion on he long-run mulipliers of he C(L) marix, namely ha a moneary policy shock can have no long-run effecs on he real exchange rae. Wih a five variables VAR, we are able o idenify five srucural shocks; The firs wo are of primary ineres and can be inerpreed as moneary policy shocks (ε MP ) and real exchange rae shocks (ε ER ). We follow sandard pracice in he VAR lieraure and only loosely idenify he las hree shocks as inflaion shocks (inerpreed as cos push shocks) (ε CP ), oupu shocks (ε Y ) and foreign ineres rae shocks (ε i* ). Ordering he vecor of uncorrelaed srucural shocks as ε = (ε i*, ε Y, ε CP, ε MP, ε ER ) and following he sandard lieraure in idenifying moneary policy shocks, he recursive order beween moneary policy shocks and he macroeconomic variables implies he following resricion on he S marix i* i * S11 ε Y y S21 S22 ε CP π BL ( ) S31 S32 S33 ε = MP i S41 S42 S43 S44 S 45 ε e S S S S S ε ER (7) The sandard Cholesky resricion, namely o assume ha macroeconomic variables do no simulaneously reac o he policy variables, while he simulaneous reacion from he macroeconomic environmen o policy variables is allowed for, is aken care of by placing he macroeconomic variables above he ineres rae in he ordering, and assuming zero resricions on he relevan coefficiens in he S marix as described in (7). Foreign ineres raes are placed firs in he ordering, reflecing small counry assumpions. However, we are sill one resricion shor of idenificaion. The sandard pracice in he VAR lieraure, namely o place he exchange rae las in he ordering and assuming S 45 =, (so ha neiher macroeconomic nor moneary variables can reac simulaneously o he exchange rae shock, while he exchange rae is allowed o reac simulaneously o all oher variables), would have provided enough resricion o idenify he sysem, hereby allowing for he use of he sandard Cholesky recursive decomposiion (e.g. Lindé 23). However, if ha resricion is no valid bu is noneheless imposed, he esimaed responses o he srucural shocks will be severely biased. The sandard es in he lieraure, namely o include one variable above he oher and hen rearrange he order o es if ha makes a difference, will no produce he correc impulse responses if here is a genuine simulaneous relaionship beween he wo variables. Mos likely i will lead o he effecs of 9

10 he shocks being underesimaed, as a recursive ordering will always eiher a) disregard he simulaneous reacion of he moneary policy sance o he exchange rae shocks, or b) exclude he simulaneous reacion of he exchange rae o he moneary policy shocks. This will effecively be demonsraed in he nex secion. Insead, we impose he resricion ha a moneary policy shock can have no long-run effecs on he real exchange rae, which as discussed above, is a plausible neuraliy assumpion securing PPP wih respec o he moneary policy shock. This can be found simply j = 54, j by seing he values of he infinie number of relevan lag coefficiens in (5), C, equal o zero. By using his long-run resricion raher han a conemporaneous resricion beween asse prices and moneary policy shocks, S 45 may be differen from zero. However, by using he long-run resricion, we have enough resricion o idenify and orhogonalise all shocks. Wriing he long-run expression of (6) as B(1)S=C(1), (8) where B(1)= B and C(1)= C indicae he (5x5) long-run marix of B(L) and C(L) j= j respecively, he long-run resricion ha C 54 (1) = implies j= j B 41(1)S 14+B 42(1)S 24+B 43(1)S 34 +B 44(1)S 44+B 45(1)S 54 =. (9) So far, oupu, price and foreign ineres rae shocks have been only loosely idenified. However, hey can be furher inerpreed by examining he firs hree columns of S. The firs column imply ha foreign ineres rae are only affeced by foreign moneary policy conemporaneously, which is a plausible small counry assumpion. The nex wo columns imply ha while price shocks can affec all variables bu oupu conemporaneously, oupu shocks can affec boh oupu and prices conemporaneously. Hence, i seems reasonable o inerpre price shocks as a cos push shock (moving prices before oupu), whereas oupu shocks will be dominaed by boh demand shocks (in he shor run) and supply shocks (in he long-run). 9 3 Empirical resuls The model is esimaed using quarerly daa from 1993Q1 o 24Q3. Using an earlier saring period will make i hard o idenify a sable moneary policy regime, as moneary policy prior o 1993 experienced imporan srucural changes and unusual operaing procedures. In paricular, prior o 1993 Norway has been argeing he exchange rae, implying ha in some periods when here was depreciaion pressure, he ineres rae was immediaely increased o offse his pressure. Hence, he ineres rae and he exchange rae have been observed o move in he same direcion in periods. Neverheless, we have also 9 We have also experienced wih alernaing he order of he firs hree variables in Z, wihou much effecs on resuls. 1

11 experimened wih an earlier saring period, Wih a few excepions, his did no change he resuls in any imporan ways (see secion 4 below). Wih he excepion of he real exchange rae which is modelled in firs differences, he variables are modelled in levels, as is sandard pracice in many of he VAR models. 1 However, Giordani (24) has argued ha following he heoreical model se up in Svensson (1997) as a daa generaing process in moneary policy sudies, raher han including oupu in levels, we should eiher include he oupu gap in he VAR, or he oupu gap along wih he rend level of oupu. However, as poined ou by Lindé (23), a pracical poin ha Giordani does no address is how o compue rend oupu (hereby also he oupu gap). We herefore insead follow Lindé (23), and include a linear rend in he VAR along wih oupu in levels. In ha way we ry o address his problem by modelling he rend implici in he VAR. Also, he use of a rend in he VAR serves as a good approximaion for ensuring ha he VAR is inverible if he variables are non-saionary, in paricular given he shor span of daa we are using. The fac ha he real exchange rae is modelled in firs differences implies ha he long-run resricion of moneary policy shocks on he level of he exchange rae is evenually zero (c.f. Blanchard and Quah 1989). However, in secion 4 we es he robusness of our resuls o oher VAR specificaions, among oher by invesigaing he ime series properies more carefully. There are no qualiaive changes o he impac of he shocks. The lag order of he VAR-model is deermined using he Schwarz and Hannan-Quinn informaion crieria and he F-forms of likelihood raio ess for model reducions. A lag reducion o wo lags could be acceped a he one percen level by all ess. Using wo lags in he VAR, here is no evidence of auocorrelaion, heeroscedasiciy or non-normaliy in he model residuals. 3.1 Even analysis Before we sar wih he VAR analysis, we perform a simple even analysis similar o ha of Zeelmeyer (24), where acual moneary policy acions ha are perceived as a surprise on he marke are colleced and analysed. We focus on he period since 1999, which is he ime a which Norges Bank announced (unofficially) ha i was going o pursue is arge of sabilising he exchange rae by argeing a sable inflaion raes vis-à-vis he rading parners. From 21 however, Norway formally adoped an inflaion argeing framework. The approach is described in more deail in appendix A. We find ha a surprise moneary policy shock ha increases he ineres rae has a subsanial appreciaing effec on he exchange rae in a small open economy. In paricular, a surprise moneary policy shock in Norway ha increase he ineres rae by one percenage poin, implies an immediae appreciaion of he exchange rae by 1-2 percenages. Hence, 1 Based he sandard Augmened Dickey Fuller (ADF) uni roo es, we can no rejec ha any of he variables excep inflaion are inegraed of firs order. However, none of he variables are coinegraed. The variables should herefore be represened in firs differences. However, due o he low power of he ADF ess o differeniae beween a uni roo and a persisen (rend-) saionary process, we can no rule ou ha he variables could equally well be represened in levels, bu wih a rend. 11

12 he impac appreciaion following a conracionary moneary policy is a feaure ha he VAR model should be able o accoun for. 3.2 Cholesky decomposiion We will sar by presening exhausive resuls using differen Cholesky orderings and varying he number of variables somewha, before we discuss in deail he resuls using he srucural model. Since our prime ineres is o undersand he ineracions beween moneary policy and he differen macroeconomic variables in all figures we focus on illusraing he impac of he moneary policy shock. The impac of he oher shocks o he model can be obained from he auhor on reques. Figure 1 gives he impulse responses of he ineres rae, exchange rae, inflaion and GDP o a moneary policy shocks (normalised so he response of he ineres rae is one percenage poin he firs quarer), for he basic Cholesky ordering discussed above (ineres rae ordered before he exchange rae). The upper and lower dashed lines ploed in each graph are one-sandard-error bands. 11 Figure 1. Response o a moneary policy shock, using he Cholesky decomposiion a) Ineres rae (percenage poin) b) Exchange rae (percenage) c) GDP (percenage) d) Inflaion (percenage poin) The effec of he moneary policy shock increases ineres raes emporarily. The exchange rae appreciaes immediaely, bu hen depreciaes back o baseline, before i appreciaes and reaches is maximum afer 7 quarers. The effec hereafer dies ou. Hence, here is evidence 11 They were generaed from 25 draws by Mone Carlo inegraions following Sims and Zha (1999). This is a Bayesian mehod based on he naural conjugae prior. 12

13 of delayed overshooing and if we consider he iniial response as small and insignifican, also of an exchange rae puzzle. Despie he puzzle in he exchange rae, and consisen wih oher sudies of small open economies, oupu falls gradually and reaches is minimum afer a year. The effec hereafer quickly dies ou. Inflaion falls by lile iniially, bu hen reaches a minimum afer approximaely wo years. Ineresingly, here is no evidence of any price puzzle (where prices acually increase iniially) which is commonly found in he lieraure. The small and sluggish effec of he moneary policy shock on inflaion has also been found in oher sudies of open economies such as Lindé (23) for Sweden, bu also in radiional VARs for he US economy such as Chrisiano e al. (1999) and recenly Faus e al. (24), who idenify moneary policy shocks based on high-frequency fuures daa. Figure 2. Response o a moneary policy shock, using wo differen Cholesky orderings 1 a) Ineres rae b) Exchange rae Ineres rae Ii-exc) Ineres rae (exc-i).25.2 Exchange rae (i-exc) Exchange rae (exc-i) ) The solid line corresponds o he Cholesky decomposiion where he ineres rae is ordered before he exchange rae in he VAR. In he alernaive ordering (exc-i), he ineres rae and he exchange rae swap places. If here is srong simulaneiy beween shocks o moneary policy and exchange rae, we would no expec ha a Cholesky decomposiion of he effecs on shocks would pick up his simulaneiy, since one of he shocks is assumed o have no immediae effec on one of he variables. This is invesigaed in Figure 2, which shows he impulse responses for he ineres rae and he exchange rae from a moneary policy shock, using wo differen Cholesky decomposiions. The solid line correspond o he assumpion ha an exchange rae shock has no immediae effec on he ineres rae (he same assumpion underlying Figure 1 and used among ohers in Chrisiano and Eichenbaum, 1995, and Lindé, 23), whereas he doed line corresponds o an ordering where he ineres rae and he exchange rae swap places as he ulimae and penulimae variables, so ha a moneary policy shock has no immediae effec on he real exchange rae (as was assumed in Mojon and Peersman, 23). Hence, whereas in he firs ordering he ineres rae reacs o an exchange rae shock wih a lag (one quarer), in he second ordering he ineres rae can respond immediaely o an exchange rae disurbance, bu a he cos of leing he exchange rae reac only wih a lag o moneary policy shocks. Figure 2 illusraes ha under he resricion ha eiher he moneary policy shock has no immediae effec of exchange raes or he exchange rae shock has no immediae effec on ineres raes, dose no imply a lo of difference, as he exchange rae does no appear o be 13

14 very responsive o moneary policy shocks. On he oher hand, assuming ha boh he exchange rae and moneary policy reac imporanly o shocks in he oher secor, and ineracion is imporan, he resricion imposed by eiher Cholesky ordering will disor he esimaes of he wo shocks in such a way ha he degree of ineracion will seem unimporan. Hence, Figure 2 may no provide us wih he rue responses. Figure 3. Response o a moneary policy shock, 8-VAR; Cholesky decomposiion a) Ineres rae b) Exchange rae Exchange rae (5 VAR) 1.8 Ineres rae (5 VAR) Ineres rae (8 VAR).8.6 Exchange rae (8 VAR) c) GDP, consumpion and invesmen d) Inflaion and nominal wage GDP (8-VAR) Consumpion (8-VAR) Invesmen (8 VAR) -.2 Inflaion (5 VAR) Inflaion (8 VAR) Nominal wage (8 VAR) Leeper e al. (1996) and Faus (1998) have criicised he VAR approach for lack of robusness when addiional variables are added o he model. Before we proceed using he srucural model, we herefore invesigae o wha exen adding imporan variables will change he overall conclusions, in paricular wih respec o he exchange rae. We expand he model o an eigh variables VAR (8-VAR), where we now also include consumpion, invesmen and he nominal wage ino he model. The model resembles ha of Chrisiano e al. (25) for he closed economy, (alhough we of course sill include he open economy variables; he real exchange rae and he foreign ineres rae), and he variables are ordered equivalenly in he Cholesky decomposiion. 12 Figure 3 gives he responses in he differen variables o a moneary policy shock. Where relevan, we compare he impulses wih he same impulses from he 5-VAR. In Frame A we plo he response in he ineres rae, whereas in frame B we graph he response in he real exchange rae (wih he original sandard error bands). Frame C 12 The model is no idenical o ha of Chrisiano e al (25), as we use nominal insead of real wage in he VAR. However, we are in paricular ineresed in he nominal wage, as i is idenified by he cenral bank in Norway as an imporan indicaor for fuure inflaion pressure. 14

15 graphs he response in GDP, consumpion and invesmen, whereas in frame D we plo he response in inflaion and nominal wage. Following a similar conracionary moneary policy shock as in he baseline scenario, we find ha he nominal wage respond hree imes more han inflaion, which implies a more persisen response in he real wage han in inflaion (which is confirmed if he real wage is included in he VAR insead of he nominal wage). Consumpion falls by less han oupu whereas invesmen falls by 3-5 imes more. As before, he peak response in he real variables is found afer a year, whereas inflaion (and wages) peaks afer wo years. Compared o he baseline five variables VAR, we see ha he responses remain as in he baseline VAR, alhough he maximum impac is slighly magnified. The real exchange rae is no excepion, alhough i now iniially depreciaes, before i follows he same paern as previously (inside he sandard error bands). Hence, adding addiional variables did no solve he exchange rae puzzle. 3.3 Srucural idenificaion scheme The alernaive o he simple Cholesky decomposiion was oulined in Secion 2. Figure 4 shows he impulse responses of a moneary policy shock on he ineres rae, exchange rae, GDP and inflaion o a moneary policy shock (normalised so he response of he ineres rae is 1 pp. he firs quarer). As for he Cholesky decomposiion, he moneary policy shock increases ineres raes emporarily. There is a degree of ineres-rae ineria in he model, as a moneary policy shock is only offse by a gradual lowering of he ineres rae. The nominal ineres rae reurns o is seady-sae value jus afer a year and hen goes below is seady-sae value. Boh he ineres-rae ineria and he reversal of he ineres rae sance are consisen wih wha has become considered known o be good moneary policy conduc. As Woodford (23a) shows, ineres-rae ineria is known o le he policymaker smooh he effecs of policy over ime by affecing privae secor expecaions. Moreover, he reversal of he ineres rae sance, hough arriving lae, is consisen wih he policymaker rying o offse he adverse effecs of he iniial policy deviaion from he sysemaic par of policy. Conrary o wha has been found in oher open economy sudies, here is no evidence of any exchange rae puzzle as he moneary policy shock has a srong and immediae impac on he exchange rae, which appreciaes (falls) by around.8 percen for each 1 percenage poin increase in he ineres rae. The exchange rae remains appreciaed for wo quarers, before i gradually depreciaes back o baseline. Hence, by allowing he ineres rae and he exchange rae o reac conemporaneously o all news, he ineracion increases considerable. These resuls are also consisen wih wha we found using he even sudy discussed above. Alhough he exchange rae remains appreciaed for a few quarers, here is however, no evidence of he u-shaped response ha has been found in oher sudies, hence he UIP condiion does no seem o be violaed. Consisen wih he srong impac on he exchange rae, boh oupu and inflaion respond by more han in he Cholesky decomposiion, and he peak respond is delayed by 1-2 quarers. Hence, GDP peaks by approximaely.25 percen afer 5-6 quarers and inflaion by.1 percenage poins afer 11 quarers. 15

16 Figure 4. Response o a moneary policy shock, using he srucural VAR a) Ineres rae (percenage poin) b) Exchange rae (pc.) c) GDP (percenage) d) Inflaion (percenage poin) Robusness of resuls [To be compleed] The robusness of he resuls repored above deserve furher discussion on a leas hree issues: i) Addiional variables and sample sabiliy, ii) he ime series properies of daa and implicaions for he specificaion of he VAR, iii) he imporance of using some alernaive, bu equally plausible idenificaion schemes. This is examined nex. Below we compare he response in he real exchange rae using he srucural VAR, bu now allowing for some addiional variables ino he model. Figure 5 graphs and compare he baseline scenario wih a VAR ha has added he nominal wage. Clearly, he effec remains virually unchanged, wih an iniial appreciaion ha remains for up o hree quarers before i depreciaes back o equilibrium. 16

17 Figure 5. Response o a moneary policy shock, using baseline srucural VAR plus model augmened wih nominal wage.6.4 Baseline Baseline + nominal wage s Concluding remarks The quaniaive effecs of moneary policy shocks have o a large exen been addressed in erms of vecor auoregressive (VAR) models, iniiaed by Sims (198). Through he applicaions of Leeper e al. (1996) and Chrisiano e al. (1999, 25), one has reached a consensus on how moneary policy affecs he closed economy (like he U.S.). However, VAR sudies of he open economy have provided less of a consensus wih regard o he effecs of moneary policy, and have along cerain dimensions even provided some puzzling resuls. In paricular, many VAR sudies have found ha following a conracionary moneary policy, he real exchange rae eiher acually depreciaes, or, if i appreciaes, i does so for a prolonged period of up o hree years, hereby giving a hump-shaped response ha violaes he uncovered ineres pariy condiion. Similar resuls have been found in a series of papers recenly. The resuls have been so persuasive ha he puzzles hemselves are insead considered consensus (or sylized facs) of which many newly developed DSGE models may seek o replicae. However, here is one major obsacle when aking he closed economy VAR o he open economy. Tha is, how o properly address a possible simulaneiy beween moneary policy and he exchange rae. In paricular, mos of he sudies of open economies are placing zero conemporaneous resricions on he response of he sysemaic ineres rae seing o an exchange rae shock. However, recenly Faus and Rogers (23) have shown ha he delayed overshooing feaure of he open economy VAR is very sensiive o his kind of resricion. VAR models of he open economy should herefore seek o idenify moneary policy wihou resricing he conemporaneous response. 17

18 This paper herefore analyses he effecs of moneary policy in an open economy hrough srucural VARs, paying paricular aenion o a possible inerdependence beween he moneary policy sance and exchange rae movemens. We explicily accoun for he inerdependence beween moneary policy and exchange raes by imposing a combinaion of shor-run and long-run resricions. In paricular, we build on he radiional VAR lieraure in ha we idenify recursively a sandard srucure beween macroeconomic variables and moneary policy, so ha moneary policy can reac o all shocks, bu he macroeconomic variables reac wih a lag o he moneary policy shocks. However, our approach differs from he radiional mehod in ha we also allow moneary policy o respond o he conemporaneous exchange rae, which iself is allowed o reac simulaneously o all shocks. We mus have an alernaive resricion in order o idenify and orhogonalise all shocks. We herefore assume insead ha moneary policy shocks can have no long-run effecs on real exchange raes. By using only one long-run resricion, we address he simulaneiy problem wihou deviaing exensively from he esablished lieraure of idenifying a moneary policy shock as an exogenous shock o an ineres rae reacion funcion (he sysemaic par of moneary policy). Once allowing for full simulaneiy beween moneary policy and he exchange rae, we find ha a moneary policy shock now implies a srong and immediae appreciaion of he exchange rae. However, he qualiaive properies of a moneary policy shock found in he esablished lieraure are sill preserved. In paricular, a moneary policy shock ha increases he ineres rae emporarily lowers oupu and has a sluggish bu negaive effec on consumer price inflaion. References Adolfson, M., Laséen, S., Lindé J., and M. Villani (25), Bayesian Esimaion of an Open Economy DSGE Model wih Incomplee Pass-Through Sveriges Riksbank Working Paper Series No Alig, D., Chrisiano, L., Eichenbaum, M. and J. Lindé (23), The Role of Moneary Policy in he Propagaion of Technology Shocks, manuscrip, Norhwesern Universiy. Bagliano, Fabio C. and Carlo A. Favero (1998), Measuring moneary policy wih VAR models: An evaluaion, European Economic Review, 42, Bjørnland, Hilde C. and Kai Leiemo (24), Idenifying he Inerdependence beween US Moneary Policy and he Sock Marke, unpublished manuscrip. Blanchard, Olivier and Danny Quah (1989), The Dynamic Effecs of Aggregae Demand and Supply Disurbances, American Economic Review, 79,

19 Chrisiano, Laurence J., Marin Eichenbaum and Charles L. Evans (1999), Moneary Policy Shocks: Wha Have we Learned and o Wha End?, in John B. Taylor and Michael Woodford, eds., Handbook of Macroeconomics. Volume 1A. New York: Elsevier Science, 1999, pp Chrisiano, L., Eichenbaum, M. and C.L. Evans (25), Nominal Rigidiies and he Dynamic Effecs of a Shock o Moneary Policy, Journal of Poliical Economy 113, Clarida, R. J. Galí and M. Gerler (1999) The Science of Moneary Policy, Journal of Economic Lieraure, 37, pp Clarida R. and J. Gali (1994) Sources of real exchange rae flucuaions: how imporan are nominal shocks? Carnegie-Rocheser Conference Series on Public Policy, 41, Clarida, R. J. Galí and M. Gerler (21) Opimal Moneary Policy in Open Versus Closed Economies: An Inegraed Approach, American Economic Review Papers and Proceeding, 91, pp Corsei, G. and P. Peseni (21) Inernaional dimension of moneary policy, Mimeo Universiy of Rome III and Federal Reserve Bank of New York. Dornbusch, R. (1976), Expecaions and Exchange Rae Dynamics, Journal of Poliical Economy, 84, Eichenbaum, M. and C. Evans (1995), Some empirical evidence on he effecs of shocks o moneary policy on exchange raes, Quarerly Journal of Economics, 11, Faus, Jon (1998), The robusness of idenified VAR Conclusions Abou Money, Carnegie- Rocheser Conference Series in Public Policy, 49, Faus, Jon (25), Is Applied Moneary Policy Analysis Hard? Manuscrip. Faus, Jon and Jon H. Rogers (23), Moneary policy's role in exchange rae behaviour, Journal of Moneary Economics, 5, Faus, Jon, Eric T. Swanson and Jonahan H. Wrigh (24), Idenifying VARS based on high frequency fuures daa, Journal of Moneary Economics, 51, Giordani, Paolo (24), An alernaive explanaion of he price puzzle, Journal of Moneary Economics, 51, pp Kim (21) JME 19

20 Kim, S. and N. Roubini (2), Exchange rae anomalies in he indusrial counries: A soluion wih a srucural VAR approach, Journal of Moneary Economics, 45, Leeper, Eric M., Sims, Chrisopher A. and Tao Zha, (1996), Wha does moneary policy do? Brookings Papers on Economic Aciviy, 2, Lindé, Jesper (23), Moneary Policy Shocks and Business Cycle Flucuaions in a Small Open Economy: Sweden , Sveriges Riksbank Working Paper Series No.153. Mojon B. and G. Peersman (23), "A VAR descripion of he effecs of moneary policy in he individual counries of he Euro area", In Moneary policy ransmission in he Euro area, I. Angeloni, A. Kashyap and B. Mojon (eds), Cambridge Universiy Press, chaper 3. Monacelli, T. (23), Moneary Policy in a Low Pass-Through Environmen, European Cenral Bank Working Paper No Rudebusch, Glenn D. (1998), Do Measures of Moneary Policy in a VAR Make Sense?, Inernaional Economic Review, 39, pp Sims, C. A. (198), Macroeconomics and Realiy, Economerica, 48, pp Sims, Chrisopher A. (1992), Inerpreing he Macroeconomic Time Series Facs: The Effecs of Moneary Policy, European Economic Review, 36, Sims, Chrisopher A. and T. Zha, (1999), Error bands for impulse responses, Economerica, 67, Smes, Frank and Raf Wouers (22), Openess, imperfec exchange rae pass-hrough and moneary policy, Journal of Moneary policy, 49, pp Svensson, Lars E.O. (1997), Inflaion Forecas Targeing: Implemening and Monioring Inflaion Targes, European Economic Review 41 (6), pp Svensson, Lars E.O. (2), Open-Economy Inflaion Targeing, Journal of Inernaional Economics, 5, pp Walsh, Carl, (23), Moneary Theory and Policy, 2nd. ed., The MIT Press. Zeelmeyer, Jeromin (24), The impac of moneary policy on he exchange rae: evidence from hree small open economies, Journal of Moneary Economics, 51, pp