THE EMPIRICAL RELEVANCE OF A BASIC STICKY- PRICE INTERTEMPORAL MODEL

Transcription

1 Deparmen of Economics THE EMPIRICAL RELEVANCE OF A BASIC STICKY- PRICE INTERTEMPORAL MODEL Massimo Giuliodori Universiy of Glasgow Glasgow, Augus 21 Absrac In his paper, we firs ouline he moneary version of he sicky price ineremporal model of Obsfeld and Rogoff (1995, 1996), in which moneary shocks unambiguously generae a permanen nominal exchange rae depreciaion and a emporary curren accoun surplus. We hen empirically invesigae hese heoreical predicions in wo srucural VAR sysems for 15 OECD counries over he period , using he long-run resricion idenificaion scheme suggesed by Clarida and Galì (1994). Our empirical findings suppor he main predicions of he basic model, as well as suggesing ha moneary shocks play an imporan role in he curren accoun flucuaions. Moreover, we find ha more open economies show greaer sensiiviy of he curren accoun o moneary shocks. Keywords: JEL Classificaion: Srucural VAR, real exchange rae and curren accoun. C32, E4, F41, F42 I am graeful o Ulrich Woiek for providing he GAUSS codes and help in implemening he economeric procedure. I also hank Jim Malley, Campbell Leih, Jordi Galì, Phillip Lane, seminar paricipans a he Universiy of Glasgow and discussans a he Scoish Docoral Programme Mehodology Conference 21, he European Economic Associaion Conference 21 (Lousanne) and he MMF Annual Conference 21 (Belfas) for useful commens and suggesions. The usual disclaimer applies.

2 1. Inroducion Over he las wo decades, many researchers have ried o provide an analyical framework, which could be a superior alernaive o he Mundell-Fleming model. To his regard, a number of sudies published in he early 198s moved owards an ineremporal approach, based on microfounded opimising models, where preferences, echnology and capial markes access are direcly included. Mos of his lieraure, however, addresses he ineremporal analysis of he curren accoun deerminaion focusing on flexible-price and non-moneary economies, in which fiscal and echnology shocks are he main sources of curren accoun flucuaions (Glick and Rogoff, 1995 and Obsfeld and Rogoff, 1996a for a survey). Only recenly, following he parial failure of his class of models in explaining he high volailiy of he curren accoun, some sudies have inroduced a role for moneary shocks as fundamenal deerminans of exernal imbalances. The Redux model by Obsfeld and Rogoff (1995, 1996) (henceforh OR) is wihou doub considered as he precursor of his lieraure, which inroduces marke imperfecions (namely price sickiness and monopolisic compeiion) in an ineremporal opimising open economy framework. 1 Their model shows ha, under floaing exchange raes, moneary shocks unambiguously affec consumpion, income and curren accoun. However, he original assumpions of he OR framework have been relaxed by many auhors, who show ha he main predicions are no robus o alernaive hypohesis on he key parameers (see Lane, 21 for an excellen survey). As a resul, we conduc some empirical evidence on his direcion and evaluae he relaive imporance of moneary shocks in explaining real exchange raes and curren accoun flucuaions in 15 OECD counries by esimaing wo srucural VARs. In order o idenify he srucural forces driving he variable dynamics over he shor--run, we make use of long-run idenificaion resricions la Blanchard and Quah (1989) and Clarida and Galì (1994). 1 For several years, he radiional aggregaive Keynesian models wih sicky prices and he flexible-price ineremporal neo-classical models have represened he main way of modelling he economy. Boh of hem, however, presen weaknesses. The former ones, alhough hey allow for nominal rigidiies, have many shorcomings in wha hey lack of microfundaions for ineremporal choice by ignoring he ineremporal budge consrains, say very lile abou curren accouns and budge deficis, and, in general, do no provide a clear descripion of he ransmission mechanisms of he moneary and fiscal policies. Moreover, hey do no allow for welfare and normaive analysis which enormously limi he possibiliy of formulaing policy prescripions. On he oher hand, he neo-classical models, while embodying many of hese cenral issues, are sill based on he unrealisic hypohesis of flexible prices and perfecly compeiive markes. 2

3 The aim of his sudy is wofold. Firsly, as poined ou by Lane (1999), his exercise can be helpful in disinguishing beween hose class of models predicing a curren accoun surplus in response o a moneary shock (Mundell, 1983 and Obsfeld and Rogoff, 1995, 1996 among ohers), and hose in which a curren accoun neuraliy or defici migh occur as in Bes and Devereux (1996, 2), Lombardo (21), Kollmann (1997) and Chari, Kehoe and McGraan (1997), Tille (21). A second aim is o invesigae he role of nominal or moneary shocks in he undersanding of he curren accoun deerminaion, which, in he previous empirical analysis, has mainly been focusing on echnology and fiscal shocks (Glick and Rogoff, 1995). The srucure of he res of he paper is organised as follows. In Secion 2, we briefly ouline he moneary version of he OR model and some of is predicions. Secion 3 summarises wo main exensions, which, by keeping he basic srucure of he Redux model, while inroducing pricing-o-marke and differen degrees of elasiciy of subsiuion beween home and foreign goods, allow for he possibiliy of differen qualiaive implicaions. Secion 4 provides an overview of some recen empirical aemps o es he role of moneary shocks on exchange rae and curren accoun deerminaion. Secion 5 oulines he economeric approach we use, while Secion 6 commens on he specificaions and provides he esimaion resuls. Secion 7 concludes. 2. A Basic Two-Counry General Equilibrium Model of Inernaional Moneary Policy Transmission (Obsfeld-Rogoff Redux Model, 1995, 1996) In wha follows we ouline a exbook version of he Redux model in which Obsfeld and Rogoff (1995, 1996) develop a perfec foresigh wo-counry general equilibrium moneary model which combines hree fundamenal blocks he firs emphasises he ineremporal decisions by individual agens where foreign rade and asse exchange open up avenues for ransferring resources over ime which are no available in a closed economy; 2. he second is based on monopolisic compeiion in he goods marke, which plays a cenral role because i rigorously defends he Keynesian assumpion ha oupu is demand-deermined in he shor-run; 2 This ouline is srongly based on he moneary version model summarised in Chaper 1 of Obsfeld and Rogoff (1996) and he original paper in Obsfeld and Rogoff (1995) 3

4 3. he hird conemplaes he presence of sicky prices. In heir model, he world is populaed by a coninuum of individual consumer-producers, indexed by z [, 1], each of whom produces a single differeniaed perishable good, also indexed by z. The home counry consiss of producers on he inerval [, n], and he remaining agens z (n, 1] reside in he foreign counry. Thus, n provides an index of he relaive size of he wo counries. Foreign variables will be denoed by a superscrip aserisk (*). Individuals everywhere in he world have preferences defined over a consumpion index, real money balances, and effor expended in producion. 3 In paricular, he ineremporal uiliy funcion of a ypical home agen j depends posiively on consumpion and real balances, and negaively on work effor, which is posiively relaed o oupu. For analyical reasons, OR ake ino accoun he specific case, in which all individuals have he same preferences defined by L = F + s H G = NM s χ M s k (2.1) U β log Cs y s ( z ) 1 ε P 2 s 1 ε I KJ where < β < 1 is he subjecive rae of discoun, ε > is a parameer inversely relaed o he elasiciy of money demand, he elasiciy of ineremporal subsiuion is equal o 1 and he elasiciy of disuiliy from oupu equal o 2. 4 Assuming ha all goods are raded and leing c j (z) be a home individual s consumpion of produc z, he variable C j can be defined as a real consumpion index, on which uiliy depends, and represened by he consan-elasiciy-of-subsiuion (CES) funcion 2 O QP, θ 1 θ 1 z θ 1 j j (2.2) C c ( z) θ dz = L N M O Q P, where he elasiciy of subsiuion beween varieies θ > 1. 5 The foreign consumpion C* is defined analogously. 3 The assumpion of ime-separable uiliy funcion is preferable for a number of reasons. In paricular, Obsfeld and Rogoff (1996) poin ou ha an ineremporal non-separable uiliy funcion would yield few concree and esable predicions, and ha empirical research has no been able o provide a superior non-separable alernaive. 4 I is easy o show ha a rise in home produciviy can be capured by a fall in k. 5 As poined ou by Obsfeld and Rogoff (1995) his assumpion is necessary because he parameer θ is also he price elasiciy of demand faced by each monopolis. They have o impose his resricion, because since 4

5 The price deflaor for nominal money balances is he consumpion-based money price index corresponding o (2.2). 6 Leing p(z) be he home-currency price of good z, i is easy o show ha he money price level in he home counry is: z 1 1 (2.3) P p( z) 1 θ 1 dz = L N M O Q P θ Being p*(z) he foreign-currency price of good z, he foreign price index P* can be wrien similarly. There are no impedimens or coss o rade beween he counries. Leing E be he nominal exchange rae, defined as he home-currency price of foreign currency, hen he law of one price holds for every good, 7 so ha (2.4) p( z) = Ep * ( z) Since boh counries residens have he same preferences, equaion (2.2) implies ha home and foreign consumer price indexes are relaed by purchasing power pariy (PPP) (2.5) P = EP* There is no capial or invesmen, bu his is no an endowmen economy because labour supply is elasic. This means ha period oupu of good z, y (z), is chosen in a manner ha depends on he marginal revenue of higher producion, he disuiliy of effor, and he marginal uiliy of consumpion, ha is, he oupu is endogenous. There is an inegraed world capial marke in which boh counries can borrow and lend. The only asse hey rade is a real bond, denominaed in he composie consumpion good. Le r be he real ineres rae earned on bonds beween and +1 and F and M he socks of bonds and domesic money held by a home residen enering dae +1. marginal revenue is negaive when elasiciy of demand is less han 1, θ > 1 ensures an inerior equilibrium wih a posiive level of oupu. However, as poined ou by Lombardo (21), he elasiciy of subsiuion beween domesic and impored goods in OR model is "unnecessary" consrained o be bigger han one. He relaxes his assumpion providing ineresing insighs. Tille (21) furher exends he basic model allowing for differen elasiciy of subsiuion across and wihin counries. See he nex Secion for a brief ouline of hese exensions. 6 The price index is defined as he nominal expendiure of domesic money needed o purchase a uni of consumpion C. 7 In his model he law of one price always holds. However, several auhors (Engel, 1998 Rogoff, 1996) have documened ha inernaional deviaions in radable prices are responsible for a large proporion of real exchange flucuaions. Following his empirical evidence, Bes and Devereux (1996, 1998, 2) have inroduced marke segmenaion ino he basic model in he form of pricing-o-marke (PTM). As above, his exension produces heoreical implicaions in he inernaional ransmission of moneary (as well fiscal) policy which differ from he OR model. Similarly, Chari e al. (1997) develop a sicky price model wih price discriminaing monopoliss, which produces deviaions from he law of one price. See he nex Secion for a brief discussion. 5

6 Individual z s period budge consrain wrien in nominal erms is: (2.6) P F + M = P ( + r ) F + M + p ( z) y ( z) PC PT where y(z) is he individual s oupu, for which agen z is he only producer, and T denoes lump-sum real axes paid o he domesic governmen (which can be negaive in he even of money ransfers). Since Ricardian equivalence holds in his model, hey assume ha he governmen runs a balanced budge each period. Therefore, all seignorage revenues are rebaed o he public in he form of ransfers: (2.7) T = M + M P 1 or The same is rue for he foreign counry. PT = M M 1 Given he consan-elasiciy-of-subsiuion consumpion index, eq. (2.2), i is easy o find he home individual j s demand for good z in period and he corresponden demand of a foreign individual. Inegraing demand for good z across all agens (ha is, aking a populaion weighed average of home and foreign demands), and making use of eqs. (2.2) and (2.4), which implies ha p(z)/p = p*(z)/p* for any good z, we can deermine he consan-elasiciy-ofsubsiuion oal world demand for good z d (2.8) y ( z) p ( z) P = L N M O Q P θ C w where world consumpion C w, which producers ake as given, is defined as w j j * (2.9) C C dj C dj nc ( 1 n) C, z z * = + = + n 1 n where hey imposed symmery on he idenical agens wihin each counry in order o simplify noaion. From he firs order condiion of he uiliy maximisaion, he following funcions are derived: (2.1) C = β( 1 r ) C (2.11) M P = L N M χc F HG 1+ i i IO KJ Q P 1/ ε, 6

7 P + where we use he Fischer pariy equaion defined as ( 1+ i ) = 1 ( 1+ r ), and P θ 1 θ θ w (2.12) y ( z) ( C ) / θ = θk C which represen respecively, he sandard Euler equaion, he money demand and labour supply equaion. The same is rue for he foreign counry counerpar equaions. In order o analyse he effecs of a moneary shock, hey carry ou a log-linear approximaion around a flexible-price seady-sae equilibrium (where all he exogenous variables are * consan and where he iniial sock of ne foreign asses is, i.e. F = F = ) and solve for differences beween home and foreign log-linearized expressions. 8 In he shor-run, nominal producers prices p(h) and p*(f) are predeermined, ha is, hey are se a period in advance bu can be adjused fully afer one period. In his sense, we can inerpre he possible sources of sickiness wih menu coss of price adjusmen. Wih sicky nominal prices, oupu becomes demand deermined for small enough shocks, because he monopolis always prices above marginal cos, and i is profiable o mee unexpeced demand shocks a he pre-se price. Since prices are fixed for one period, Obsfeld and Rogoff disinguish beween he impac (firs-period) effec of a shock and is long-run (second-period) seady-sae effec. In considering he effecs of an unanicipaed permanen rise in he relaive home money supply (ha is, M ˆ Mˆ * ˆ ˆ * = M + 1 M + 1 ) he main findings and dynamics of he model can be summarised as follows. A moneary expansion in he home counry produces a nominal depreciaion and a subsequen rise in he domesic price level, followed by a decrease of he domesic relaive prices. As a consequence, and under he assumpion of monopolisic compeiion, domesic oupu emporarily expands. Wih consumpion which is based on permanen income, consumpion rises less han oupu, leading he home counry o run a curren accoun surplus. The excess of oupu over domesic consumpion is expored and, as a paymen for hese expors, he home counry receives claims agains he fuure oupu of he foreign 8 They implemen his linearizaion by expressing he model in erms of deviaions from he baseline seadysae pah. By denoing percenage changes from he baseline by has, for any variable, X $ dx / X, where X is he iniial seady-sae value. 7

8 counry. Domesic consumpion, herefore, rises alhough he increase in oupu lass for only one period. Alhough a full analyical descripion of he relevan equaions would make he analysis more horough, he main findings, which are relevan for he heoreical relaionship beween moneary policy, exchange rae and curren accoun, can be summarised in he following expressions. (2.13) $ $ ( $ $ * 1 ) ( $ $ * E = E+1 = M M C C ) ε which shows ha he exchange rae jumps immediaely o is new long-run equilibrium following a permanen relaive money shock. The assumpion behind his resul is ha if consumpion differenials and money differenials are boh expeced o be consan, hen agens mus expec a consan exchange rae as well. Anoher imporan equaion, which derives direcly from he budge consrain, is df 1 * (2.14) = ( θ 1) Ê ( Ĉ Ĉ ) C w 1 n which ells us ha he change in bond holdings (namely, in curren accoun) is a posiive funcion of he exchange rae deviaion from he seady-sae value (being θ >1) and a negaive funcion of he difference beween home and foreign per capia consumpion. The firs member of he RHS is a resul consisen wih he Marshall-Lerner condiion, bu in addiion we also have a consumpion effec which ells us ha he larger he increase of relaive consumpion, he smaller he wealh ransfer. 9 Eq. (2.14) can be rewrien as follows df * (2.15) Z = ( 1 n)( θ 1) Ê ( 1 n)( Ĉ Ĉ ) C w E where we can see ha he smaller he home counry (i.e. he smaller n), and he larger he elasiciy of subsiuion beween varieies, he larger he impac on he home curren accoun. The second expression derived from he model is given by (2.16) df 1 = C 1 n r 1 w 2θ ( ) ( * Ĉ ) Ĉ θ,, 8

9 which shows ha he consumpion change is posiively relaed o he curren accoun surplus hrough he permanen ineres income which home individuals earn from he wealh ransfer. Moreover, i is easy o see ha he increase in consumpion is less han he amoun r df (since θ >1), for he fac ha he higher wealh leads o some increase in leisure and, consequenly, in some reducion of producion. Similarly o he expression rewrie eq. (2.16) as (2.17) df C w Z C 2θ = r ( 1 n) ( θ 1) * ( Ĉ Ĉ ) We provide a represenaion of he wo lines (namely, Z E, we can Z E and Z C ) in Figure 1. We can see how heir inersecion implies ha an exchange rae depreciaion improves he curren accoun over he shor run. FIGURE 1 Effec of an unexpeced home money shock on he curren accoun. df C w Z C Z Ê * Ĉ Ĉ 9 This derives from he fac ha under he assumpion of monopolisic compeiion, domesic oupu emporary rises. Wih he permanen income hypohesis, he larger is he consumpion effec, which, however, expands less ha oupu, he smaller is he home counry curren accoun surplus. 9

10 3. Two Exensions of he Basic Model As originally Obsfeld and Rogoff poined ou, heir framework can be improved wih a number of exensions. 1 Lane (21) offers an excellen survey of he recen lieraure aemping o exend and generalise he basic Redux model by inroducing sicky wages, saggering nominal rigidiies, marke segmenaion and pricing o marke, differen household preferences and financial srucures. In wha follows, we provide a brief overview of wo of he mos recen conribuions, which, while keeping he basic srucure of he model, develop wo very ineresing exensions by inroducing PTM and allowing for differen elasiciies of subsiuion across and wihin counries. I urns ou ha hese assumpions srongly affec he inernaional ransmission of moneary shocks and provide differen heoreical predicions and welfare implicaions. Bes and Devereux (2) develop an exension of he OR model by allowing shor-run deparures by he real exchange rae from PPP, which is assumed o hold in he basic framework. In paricular, hey inroduce he hypohesis ha a fracion s of firms in each counry can price-discriminae across counries and, herefore, se differen prices in home and foreign markes. This parameer s provides a measure of pricing-o-marke (PTM) in some raded goods indusries where firms end o se prices in local currencies of sale and do no adjus prices o movemens in he exchange rae. 11 Bes and Devereux (2) show ha he nominal price sickiness and he presence of PTM increase he volailiy of he exchange rae and srongly affec he inernaional moneary ransmission mechanisms. In paricular, hey show ha high degree of PTM reduces he radiional expendiure swiching effecs of exchange rae depreciaion, which, as a resul, has lile effec on he relaive price of impored goods facing domesic consumer and on he corresponden shif in world demand. Moreover, shor-run deviaions from PPP end o generae low (wih respec o models wihou PTM) comovemens of consumpion and high posiive comovemens of oupu across counries. The degree of PTM is clearly cenral o heir resuls. In paricular, hey 1 For insance, hey sugges inroducing overlapping generaions in place of homogeneous infiniely lived agens. This assumpion no only allows us o have a more realisic model han he basic one, bu, more significanly, makes room for he possibiliy of he Ricardian Equivalence no o hold. 11 Following he lack of empirical suppor for he law of one price and PPP for raded goods, a leas a high frequency, in he las wo decades here has been a growing lieraure on he presence of PTM. Glodberg and Kneer (1997) provide a comprehensive survey of he empirical evidence on he degree of exchange rae passhrough, marke segmenaion and pricing-o-marke. Engel (1993) and Engel and Rogers (1996) provide some 1

11 consider he wo opposie cases of s and s 1. In he former (i.e. he law of one price is mainained coninuously and PPP holds) for boh counries, a devaluaion unambiguously improves he curren accoun, consisenly wih he radiional OR model, and generaes permanen posiive effecs in he home consumpion relaive o foreign consumpion. When s 1 and full-ptm occurs, a moneary disurbances generae exchange rae devaluaion, bu no curren accoun effecs. In general, he higher he degree of PTM, a moneary shock implies a greaer exchange rae volailiy, a smaller curren accoun improvemen, and reduced permanen effecs on relaive consumpion. 12 Anoher simple, bu crucial, exension has been developed by Tille (21), who allows for he elasiciy of subsiuion across and wihin counries o differ and generalises he baseline Redux model where he wo parameers are equal and consrained o be bigger han one (i.e. θ > 1 ). Under he assumpion of law-of-one-price and PPP, hey disinguish wo possible values of he elasiciy of subsiuion beween home and foreign goods respecively greaer and less han uniy. He labels he firs case as MLR (namely Marshall-Lerner-Robinson condiion), in ha, when goods produced in differen counries are close subsiues, an exchange rae depreciaion generaes a curren accoun surplus, a permanen rise of home consumpion relaive o foreign consumpion. When he goods produced in he wo counries are poor subsiues and he elasiciy is less han uniy (i.e. NON-MLR), he curren accoun will be in defici in he shor run, deermining a permanen fall in relaive consumpion. Similarly, Lombardo (21) modifies he original OR specificaion wih differen degrees of elasiciy of subsiuion beween domesic and impored goods. He derives condiions allowing for posiive, neural and negaive curren accoun responses o a moneary expansion and currency depreciaions, arguing ha he sandard Marshall-Lerner condiion may no apply for specific inervals in he value of he relevan coefficiens. In summary, several models have been developed exending and generalising he basic Obsfeld and Rogoff s Redux model. However, he size and he sign of he effecs of moneary shocks are srongly affeced by he magniude of several key coefficiens and he srucural assumpions. As poined ou by Lane (21), he above lieraure has been mainly empirical evidence and show ha deviaions from he law-of-one-price across inernaional borders are greaer han can be due o geographical disance and ransporaion coss. 12 Similarly o Obsfeld and Rogoff (1995), Bes and Devereux (2) also focus on he implicaions of governmen spending shocks. They show ha boh emporary and permanen fiscal shocks generae shor-erm real depreciaion and fall of relaive consumpion. The curren accoun effecs crucially depend on he degree of 11

12 focusing on heoreical aspecs. However, he acual effecs of moneary shocks o exchange raes, curren accoun and oher real variables are primarily an empirical issue. While some exercises have been mainly based on calibraion mehods and ohers on he esimaion of he key parameers of he model, a number of recen papers have addressed his deficiency by esimaing impulse-response funcions generaed by VAR economeric echniques, where differen idenificaion soluions have been used. In wha follows, we provide a brief overview of he VAR empirical evidence 4. Empirical Review The wo mos influenial papers on which he following lieraure has buil are Eichenbaum and Evans (1995) (EE) and Clarida and Galì (1994) (CG). Alhough hey use differen idenificaion soluions, hey boh invesigae he effecs of shocks o moneary policy on exchange rae in a manner ha is qualiaively consisen wih he above sicky-price models. In paricular, EE use U.S. monhly daa covering he sample period 1974:1-199:5 o esimae alernaive VAR models, which share common variables (namely, indusrial producion, CPI, shor-erm ineres rae differenial and exchange raes), bu differ for he use of hree measures of shocks o U.S. moneary policy. 13 Their sandard Cholesky decomposiion idenificaion assumes ha all shocks are orhogonalised and is such ha he moneary variables are ordered prior o he indusrial producion and he consumer price level and afer he ineres rae and exchange rae variables. This corresponds o he assumpion ha he moneary auhoriy ses is policy insrumen wih curren values of he firs wo variables in mind, while hese do no respond conemporaneously o movemens of he moneary shock. They find srong evidence ha conracionary policy shocks lead o significan and persisen appreciaions in exchange rae, boh nominal and real, and conclude by poining ou ha moneary shocks conribued significanly o he overall variabiliy of U.S. exchange raes in he pos-breon Woods era. PTM, he duraion of he shock and he magniude of he elasiciy of labour supply. See Bes and Devereux (2), page for a discussion. 13 They use orhogonalised shocks o he federal funds rae, orhogonalised shocks o he raio of non-borrowed o oal reserves and changes in an index hey consruc as moneary policy proxy. 12

13 Bes and Devereux (1999), by using U.S. daa wih respec o he remaining G-7 counries, modify he EE s specificaion, and also consider a shor-run recursive idenificaion scheme. They find ha posiive innovaions o moneary policy cause a sharp and persisen depreciaions of he real and nominal exchange raes. 14 Lane (1999) exends he EE sysem o include a rade balance in order o employ daa a a monhly frequency over 1744:1-1996:12 for he U.S. wih respec o oher G-7 counries. In paricular, he esimaes a six-variable VAR sysem, imposing a se of exclusion resricions on he conemporaneous relaionship beween he variables. The rade balance is ordered las in he sysem o allow for conemporaneous effecs of all shocks on i. However, his resuls are robus o alernaive orderings and wo measures of he US moneary policy insrumen (i.e. he Federal Funds rae and he level of non-borrowed reserves). The impulse response funcions of he rade balance wih respec o a moneary expansion show a susained surplus afer a period of abou a year, alhough he maximum impac occurs jus afer monhs, according o he moneary policy insrumen used. Anoher par of he VAR empirical lieraure has focused on he use of long-run idenificaion scheme á la Blanchard and Quah (1989). 15 An influenial conribuion was made by Clarida and Galì (1994) who invesigae empirically and aemp o idenify he sources of real exchange rae movemens afer he collapse of Breon Woods for he U.S. vis-à-vis wih he U.K., Germany, Japan and Canada, respecively. They make use of a srucural VAR sysem on oupu, he real exchange rae and inflaion, whose long-run idenificaion resricions are consisen wih a sochasic version of he Obsfeld (1985) open macro model. They assume he presence of hree srucural shocks o supply, demand, and money and impose resricions in such a manner ha money shocks are expeced o have no long-run impac on eiher oupu or he real exchange rae, and he demand shock o be longrun neural for oupu. 16 They find ha for wo of he four counries (namely, Japan and Germany), he srucural VAR esimaes sugges ha nominal (money) shocks explain a subsanial amoun of he variance in dollar-dm and dollar-yen real exchange raes (41% and 35% of he variance, respecively). For he U.K and Canada, variance decomposiion 14 In Bes and Devereux (1996), hey sudy a Wold decomposiion a la Eichenbaum and Evans and look a he CPI responses o moneary policy shocks, which hey find o be quie fla and barely significan. They inerpre hese resuls in suppor of a high degree of PTM and, herefore, as if large movemens in nominal exchange raes are no refleced in impor prices. 15 Se he nex Secion for an ouline of his idenificaion scheme. 16 See he following Secion for a descripion of heir economeric approach. 13

14 resuls are less supporive of he moneary policy shocks relevance. However, consisen wih EE, for all counries nominal shocks lead o shor-run real depreciaion, a rise in U.S. oupu and a jump in U.S. inflaion. A number of recen papers have made use of he CG s VAR idenificaion scheme, by focusing on he curren accoun raher ha he real exchange rae, in line wih he qualiaive predicions of some of he sicky-price ineremporal model. Lane (1999) esimaes a hreevariable sysem where he akes he raio of he U.S. and res-of-he-world (proxied by he non-us G-7 counries) oupu, he US s curren accoun o GDP raio and he raio of he U.S. o RoW consumer price levels. He assumes ha his sysem is driven by a sequence of hree orhogonal srucural shocks, which he labels supply, absorpion and moneary shocks, respecively. His long-run idenificaion assumpions are such ha he money shock has no long-run effecs on he relaive oupu and he curren accoun, and he absorpion shock is long-erm neural on he relaive oupu. The esimaed IRFs of he curren accoun o a posiive one sandard deviaion moneary shock show a shor-erm deerioraion (which Lane inerpres as a J-curve effec), followed by a significan and persisen surplus which reaches is maximum impac afer 1 quarers. Quaniaively, he conribuion of his shock o curren accoun volailiy consiues abou half of is variaion. Similarly o Lane (1999), Cavallari (21) proposes a hree variable sysem of he raio of domesic o world oupu, he raio of curren accoun o oupu and he raio of home o foreign shor-erm ineres rae. By using boh shor-run and long-run resricions as in Galì (1992), she assumes he presence of a supply, an absorpion and a moneary shock. The laer wo are resriced o have no long-run effec on he oupu, and he moneary shock no o affec oupu conemporaneously. Esimaing his sysem over 1974:1-1997:4 for he G7 counries, she shows ha he curren accoun response o he moneary shock differs across counries. In paricular, in he case of UK, Ialy, France and Canada, he curren accoun goes ino surplus following a negaive moneary shock. For he US, Germany and Japan he reverse holds. Prasad (1999) esimaes a hree variable sysem wih he relaive oupu level, he real exchange rae and he raio of rade balance o oupu for he G-7 counries. In paricular, in accordance wih he implicaions of a modified version of Clarida and Galì (1994) heoreical model, he assumes ha moneary shocks have no long-run impac on he real exchange rae and on he relaive oupu, and ha he demand shock does no affec he 14

15 relaive oupu in he long run. His resuls show ha nominal shocks appear o have played an imporan role in he dynamics of he rade balance over he period in G-7 counries. In paricular, he shows ha posiive moneary shocks deermine significan rade surplus and generae posiive correlaions beween oupu and rade balance. Similarly o above, Lee and Chinn (1998) esimae an even more parsimonious wo-variable VAR model conaining he real exchange rae and he raio of he curren accoun o GDP for he G-7 counries, and impose a long-run neuraliy resricion of a emporary (moneary) shock on he real exchange rae. Their resuls are supporive of he real exchange rae depreciaion and curren accoun surplus following a posiive nominal shock. Many of he above sudies focus on ransmission mechanism of he moneary shocks o he nominal and/or real exchange rae and curren accoun/rade imbalances mainly in he U.S. and he remaining G7 counries. However, none of hem has been applied o a wider se of counries. In wha follows, we ry o fill his gap by sudying he qualiaive and quaniaive role of moneary shocks in he curren accoun dynamics in a number of addiional small open economies. Several sudies have shown ha for non-us counries, where US and oher foreign financial and macroeconomic condiions are sricly under observaion by he respecive cenral banks, i is necessary o modify he idenificaion allowing for feedback effecs as well as augmening he specificaion wih a number of key endogenous variables in he sysem (Kim, 1999, 2; Kim and Roubini, 1999 among ohers). To his regard, he VAR idenificaion based on conemporaneous resricions is no he mos appropriae and simple mehod o apply. On he basis of such consideraions, we apply he idenificaion scheme suggesed by Blanchard and Quah (1989) and Clarida and Galì (1994) and, as a resul, keep our specificaions as parsimonious as possible, conemporaneously esing for he main driving forces of curren accoun dynamics. Before moivaing our specificaions and providing he corresponden empirical resuls, we ouline our economeric approach. 5. The Economeric Approach The empirical sraegy we implemen builds on earlier work by Blanchard and Quah (1989) and Clarida and Galì (1994), who propose an idenificaion scheme for VARs, in which hey impose long-run resricions on he behaviour of he variables in he sysem. 15

16 In oulining heir empirical sraegy, we consider a rivariae sysem, in which he variables are assumed o be saionary. Leing x y, z, q ' denoe he (3x1) vecor of he sysem s 3 variables and ε s, δ, v ' denoe he (3x1) vecor of he sysem s 3 srucural disurbances, 17 we assume ha x is a covariance saionary vecor process and generaed by he following srucural moving average (MA) model: (5.1) x = C( L) ε = C ε + C ε + C ε where C is he (3x3) marix of he conemporaneous srucural relaionship among y, z and q, and where we assume ha he srucural disurbances ε are muually orhogonal and have uni variance, implying ha Eε ε ' = I. The reduced-form MA represenaion for x, which is direcly esimaes, is given by (5.2) x = R( L) u = u + R u + R u where u is a (3x1) vecor reduced-form disurbance. Assuming ha here exiss a non-singular marix S such ha (5.3) u = Sε and comparing (5.1) and (5.2), we can easily see ha (5.3a) C( L) = R( L) S i.e. C = S, C = R S, C = R S This implies ha (5.3) can be also be wrien as (5.4) u = C ε. OLS can be used o obain consisen esimaes of he parameers in (5.2) as well as an esimae of he symmeric variance-covariance marix of he reduced-form disurbances u : ' (5.5) Eu u = Σ Therefore, from (5.3), (5.4) and (5.5), and he assumpion of muually orhogonal srucural shocks, ogeher wih he normalisaion condiion above, we ge ' '. (5.6) Σ = C C = SS As i is well known in he lieraure, he sysem (5.6) provides 9 equaions in only 6 unknowns, ha is, he 3 variances and he 3 covariances ha define Σ. Jus-idenificaion of 17 I is imporan o poin ou ha, in conras wih he Sims-Bernanke procedure, his idenificaion scheme does no direcly associae he hree srucural shocks wih he hree sequences {y()}, {z()} and {q()}. In paricular, we can hink of he laer ones as he endogenous variables, and he hree shocks as he exogenous variables of he sysem. 16

17 model (5.2) and, herefore, esimaion of he marix C and he srucural innovaions ε require 3 addiional resricions, which will be given consisenly wih OR Redux model. 18 In paricular, leing C( ) C + C + C +... denoe he marix of long-run coefficiens such ha (5.7) C( 1) = F H G C ( 1) C ( 1) C ( 1) C ( 1) C ( 1) C ( 1) C ( 1) C ( 1) C ( 1) I K J CG assume hree long-run neuraliy condiions such ha C( 1) is resriced o be lower riangular. This means ha he srucural shocks δ and v do no affec he variable y in he long run, implying ha (5.8) C ( 1) = C ( 1) = Similarly, he assumpion ha srucural shocks v do no influence he second endogenous variable z in he long run requires ha: (5.9) C 23 ( 1) =. In wha follows, we show ha (5.8) and (5.9) are sufficien o idenify he srucural marix C, o recover he srucural-sysem dynamics defined by C1, C2,..., as well as he srucural shocks ε From (5.3a) i easy o see ha R = I, R = C C, R = C C, and so on. Therefore, he reduced-form MA model (5.2), ha we direcly esimae, can be rewrien as: (5.1) x = R u + R u + R u where we can noe ha (5.11) R( 1) R + R + R +.. = C( 1) C Provided he above assumpions, CG define he marix: (5.12) R( 1) ΣR( 1) ' which can be compued from he esimaion of he variance-covariance marix of he reduced-form disurbances Σ, and he reduced-form long-run coefficiens R( 1 ), which we ge from (5.1). Using he definiion of R( 1 ) and equaion (5.6) o subsiue for Σ, i is possible o wrie: 18 We will discuss he acual inerpreaion of he srucural disurbances in he nex Secion. 17

18 (5.13) R( 1) Σ R( 1) = C( 1) C( 1). ' ' Leing H denoe he lower riangular Cholesky decomposiion of (5.13) and knowing ha ' ' ' his marix is unique, we have ha HH = R( 1) ΣR( 1) = C( 1) C( 1) implying: (5.14) C( 1 ) = H since we have imposed he srucural long-run coefficien C( 1 ) o be also lower riangular. Finally, given (5.14) and (5.11), we can obain (5.15) C = R( 1) H 1 which allows us o compue he srucural dynamics defined by C1, C2,... as well as he srucural shocks ε. 6. Specificaion and Esimaion Resuls The wo specificaions we use closely follow he bivariae VAR sysem of Lee and Chinn (1998) and he rivariae model in Prasad (1999), in which hey impose a long run neuraliy condiion of moneary or nominal shocks on he real exchange rae. Lane (1999, 21) argues ha heir idenificaion assumpion is no warraned in he sudy of ineremporal sicky-price models. In paricular, he poins ou ha if moneary shocks generae curren accoun imbalances, hey will also have long run effecs on he real exchange rae. 19 While his is rue in many heoreical exensions of he Redux srucure, heir neuraliy resricion is rivially consisen wih all moneary and Keynesian aggregaive models, as well as he heoreical frameworks of Secion 2 and 3, provided an appropriae definiion (or inerpreaion) of he real exchange rae. 2 An imporan advanage of our idenificaion relies on he fac ha we do no impose any long-run neuraliy resricion on he curren accoun. In he ineremporal approach, any real 19 There is a large empirical lieraure emphasising he long-erm implicaions of cumulaed curren accoun imbalances (and ne foreign asses) on he real exchange rae (see a recen work from Lane and Milesi-Ferrei, 2). In his paper, we mainly focus on he shor-run dynamics of real exchange raes and curren accoun, and, consequenly, absrac from hese long run effecs. 2 In he basic OR model, free rade implies ha he LOOP holds for individual goods. Moreover, given ha consumers have idenical preferences across goods and idenical CES uiliy funcions, PPP holds in erms of consumer price indices. As a resul, he real exchange rae defined in erms of consumer prices is always consan boh in he shor and in he long run. On he oher hand, he erms of rade (or he real exchange rae defined in erms of oupu prices) varies and represens a cenral equilibraing mechanism in he basic framework. On he basis of hese heoreical assumpions, alhough conaining a large non-radable componen, we ake he CPI-deflaed exchange rae as a proxy for he real exchange rae. 18

19 and nominal shock ypically are assumed o have no long run effec on his variable. As a resul, we allow he sysem o freely deermine is dynamics, which, herefore, can be considered over-idenifying resricions useful o inerpre and disenangle he srucural shocks. Moreover, he uni roo ess srongly rejec he non-saionariy hypohesis of he curren accoun o GDP raio. This resul, which is consisen wih ineremporal models, clearly sugges ha making he assumpion ha real shocks are he source of a uni roo in his variable and ha he nominal shock has a long-run neuraliy effec on he curren accoun migh no be advisable. In wha follows, we show he resuls of wo esimaed SVARs based on he long-run resricion idenificaion scheme oulined in he above Secion. Daa descripion and uni roo ess are in Appendix 1. Esimaion of he bivariae VAR The firs specificaion is an exension of Lee and Chinn (1998) bivariae VAR x = REER, CA / Y o 15 OECD counries, 21 which we use as a benchmark specificaion for esing he robusness of he resuls in he rivariae sysem of he nex secion. In his model, we assume he exisence of wo ypes of srucural shocks generaing curren accoun and real exchange rae dynamics: a permanen shock and a emporary shock. In his framework, we can hink of permanen shocks as produciviy disurbances, and emporary shocks as moneary shocks. 22 On he basis of his inerpreaion, we consrain he laer disurbance so ha i has no long-run effec on he real exchange rae variable. In doing so, we force he PPP o hold in he long run, alhough we allow for shor-erm deviaions as in he PTM exension. 23 The advanage of such a model is on is simpliciy, as well as on he fac ha i is capable of idenifying he impac of he wo shocks on he curren accoun and exchange raes, wihou consraining he shor-run dynamics or he conemporaneous exclusion of any srucural shocks from any equaion. 21 Namely, Ausria, Belgium, Canada, Denmark, Finland, France, Germany, Ialy, Japan, Neherlands, Porugal, Spain, Sweden, he U.K. and he U.S. 22 This specificaion does no allow us o disinguish beween money demand and money supply shocks. In order o do his, we would have needed o include moneary variables and imposed addiional long-run (and/or shor-run) resricions. 23 This resricion is consisen wih he PPP assumpion of he basic OR model and he generalisaion by Tillie (21) and Lombardo (21), as well as wih Bes and Devereux (2) where shor-run flucuaions of he real exchange rae are allowed in he shor run. In doing so, we can observe he dynamics as well as he saisical significance of shor-run deviaions from PPP. 19

20 This sysem is esimaed independenly for each of he 15 OECD counries, by aking he firs difference of he REER series and he level of he raio CA/Y, in accordance wih saionariy properies of he series. 24 Similarly o Lee and Chinn (1998), we use wo lags, which is a fair balance beween he lag lenghs chosen by Schwarz informaion crierion (SIC) and Akaike informaion crierion (AIC). In paricular, while he SIC ends o prefer he mos parsimonious specificaion (i.e. 1-2 lags), he AIC selecs 2-3 or, in mos of he cases, more lags. However, wih he excepion of Denmark where we use 4 lags, our resuls are qualiaively robus o alernaive lag lenghs. To save space, and given ha resuls are very similar across counries, Figures 2-5 show he impulse response funcions (IRFs) of he level of he real exchange rae and he raio of he curren accoun o oupu in response o he wo differen ypes of shocks for four counries: Finland, France, Sweden, he U.S. 25 The doed lines are 95% confidence inerval bands obained wih he boosrap-afer-boosrap procedure described in Kilian (1998), which accouns for he bias of small-sample disribuion of he impulse response parameers. The resuls, which for he G-7 are obviously similar o Lee and Chinn (1998), show ha a posiive one sandard deviaion emporary shock generaes a saisically significan shor-run improvemen of he curren accoun. The level of he real exchange rae immediaely depreciaes in he shor erm, hen gradually dies ou in he long run, deerioraing he curren accoun. Differenly from Lee and Chinn (1998), however, hese poin esimaes are no saisically significan. This resul migh derive from he differen mehod used in esimaing he confidence bands. The U.K. and Belgium provide some anomalous resuls in ha, despie he presence of an exernal balance surplus o a moneary shock over he shor-run, he level of he exchange rae appreciaes, alhough no saiically significan. Alhough permanen shocks are no he objec of his analysis, i is worh poining ou ha, in all counries, hese shocks improve he curren accoun (wih he excepion of Belgium and he U.K.). However, hese resuls migh be due o he parsimony of he model, which does no include oher key variables of he ransmission mechanism or disinguish beween 24 Oher sudies find ha he curren accoun o oupu raio has a uni roo (Lane, 1999; Cavallari, 21). Alhough in our sample period we canno rejec is saionariy (see Appendix), we have re-esimaed he sysem wih boh variables in firs-difference. Resuls are qualiaively similar, bu generae permanen effecs of he idenified shocks on he curren accoun. 2

21 oher real shocks. As poined ou by Lee and Chinn (1998), however, he fac ha he wo idenified shocks generae opposie correlaions beween he curren accoun and he exchange rae suggess he cenraliy of he naure of he shock in disenangling his relaionship. The esimaed IRFs, however, are in line wih he empirical sudies oulined in Secion 4 and provide some qualiaive suppor on he inerpreaion of emporary shocks as moneary ones. Finally, Table 4 shows he variance decomposiions o illusrae he quaniaive conribuion of he wo shocks o exchange rae and curren accoun volailiy for he four represenaive counries. 26 According o our esimaes, for mos of he 15 OECD counries he nominal shock plays a greaer role in explaining he variaion of he curren accoun in all he 24 period horizons (i.e. over 7-8%). The only excepions are France and, in paricular, he U.S., where he role of moneary disurbances for exernal imbalances is minimal (around 2% on average). Conemporaneously, real shocks dominae he volailiy of he real exchange raes in all counries. Esimaion of he rivariae VAR sysem In order o verify he robusness of he above resuls, we augmen he above specificaion wih a key endogenous variable (i.e. he relaive oupu) and esimae a hree-variable sysem * x = ( Y / Y ), REER, CA / Y for he 15-OECD counries. We chose his specificaion for a number of reasons. Firsly, he simple bivariae sysem migh be subjec o problems of misaggregaion of shocks. This migh be suggesed by he fac ha he idenified moneary shock accouns for mos of he variance of he curren accoun. Our specificaion allows us o idenify, and possibly disenangle, an addiional real shock which can have only emporary effecs on he relaive oupu. As in Clarida and Galì (1994), Prasad (1999) and Lane (1999), we can inerpre his innovaion as a non-moneary demand or absorpion shock. In doing so, we disenangle a furher leading force affecing he endogenous variables and improve he correc idenificaion of he moneary shock, which is assumed o have no long run effec on he relaive oupu and he real exchange rae. 25 The above four counries are a mix of small and large economies, as well as a combinaion of counries which over he sample period operaed in a flexible exchange rae and semi-fixed exchange rae regime. The remaining counries IRFs and resuls are available from he auhor upon reques. 26 Resuls are available from he auhor upon reques. 21

22 Similarly o he bivariae VAR, we choose a lag lengh which represens a good balance beween he AIC and he SIC. 27 In order o allow for a comparison wih previuous empirical sudies on he curren accou deerminaion, Figures 6-12 repor he impulse responses of he curren accoun o GDP raio in response o a posiive one sandard deviaion of he hree srucural shocks in he G-7 counries. 28. The doed lines are 95% confidence inerval bands obained wih he boosrap-afer-boosrap procedure described in Kilian (1998), which accouns for he bias of small-sample disribuion of he impulse response parameers. Given he focus of our analysis, we do no presen he impulse responses for he relaive oupu and he real exchange rae, which, however, are quie reasonable. 29 In paricular, we find ha in all G7 counries he curren accoun is posiively and emporarily affeced by he posiive moneary shock. This improvemen is saisically significan in all counries and confirms he resuls obained in he bivariae sysem. I also suppors he basic predicion of he basic Redux model and he corresponden exensions characerised by low degree of PTM behaviour and high elasiciy of subsiuion beween domesic and impored goods. The responses of he curren accoun o he supply and demand shocks vary across counries, alhough hey appear o be saisically insignifican in mos of he cases. In paricular, posiive supply shocks generae a decreasing improvemen of he curren accoun raio (excep for Canada, Ialy and he U.K.). Similarly, demand shocks improve he curren accoun in he U.S., Canada, Japan and he U.K. and decrease i in Germany, France and Ialy, alhough he effec is small and no saisically significan Accordingly, we chose a lag lengh of four for Ausria, Ialy, Japan, Spain, he U.K and he U.S., six for Canada, wo for Belgium, France, Germany, Porugal and Sweden, and one for Denmark, Finland and he Neherlands. Our esimaes, however, are qualiaively similar regardless of he lag lengh used. 28 The remaining OECD counries show similar resuls, which are available from he auhor upon reques. 29 The idenified permanen shock, which could be inerpreed as a relaive supply or echnology shocks, permanenly and significanly increases oupu. I also generaes a persisen and significan (excep for he U.K., Ialy and Germany) appreciaion of he real exchange rae. This resul is also found in our bivariae VAR and oher empirical sudies, which use similar mehods (Prasad, 1999 and Clarida and Galì, 1994) or conemporaneous resricion idenificaion mehods (Muscaelli and Spinelli, 1999). The non-moneary demand shock causes a permanen and saisically significan appreciaion of he exchange rae and end o emporary, bu no significanly, increase he relaive oupu in mos counries. Finally, again wih he only excepion of he U.K., he nominal shocks lead o emporary and saisically insignifican depreciaion of he real exchange rae. The only puzzling resul is given by he emporary decrease of he relaive oupu following he moneary shock for some of he counries under invesigaion. The corresponden IRFs, however, are no saisically significan. 3 These resuls do no help us o provide a clear-cu inerpreaion of his shock. However, in he Redux model, Obsfeld and Rogoff (1996) analyse he effecs of permanen posiive home governmen spending shocks which are qualiaively consisen wih he esimaed IRFs. 22

23 Table 5 shows he variance decomposiion of he curren accoun. We can easily noice he new sysem confirms he quaniaive imporance of he moneary shock. In fac, apar from France, Germany and he U.S. where he supply shock is dominan, 31 nominal disurbances sill accoun for more han 5 per cen in Canada, Ialy, Japan and he U.K. over he 24 period horizon. The resuls (no shown) for he remaining 15 OECD economies are even more supporive, in ha he moneary shock dominae he variance decomposiion of he curren accoun (4-8 %). Differenly from he bivariae specificaion, he demand shocks dominae he forecas error variance decomposiion of he real exchange rae. The only excepions are Japan and he U.S., where supply and nominal shock, respecively, provide he major conribuion. No surprisingly, hese final resuls sugges ha he relaive imporance of hese shocks migh be sensiive o differen specificaions. However, boh sysems predic a significan role of nominal shocks in he curren accoun flucuaions for a number of OECD counries. Moreover, he esimaed IRFs show ha in almos all counries a posiive moneary shock generaes a emporary and saisically significan curren accoun surplus. Boh resuls suppor he empirical relevance of he basic moneary inererporal Redux model. Degree of openness and curren accoun flucuaions In his Secion, we focus on an addiional predicion of he OR model which refers o relaionship beween counry size and curren accoun effecs of moneary shocks. In paricular, by focusing on he wo final equaions (2.15) and (2.17) of he basic framework, i can be seen ha he larger (smaller) he home counry ha is, he greaer (he smaller) n he less (he more) he posiive impac of a home money increase on is curren accoun (Obsfeld and Rogoff, 1995, page 643). As a resul, we can expec ha smaller and more open counries will show relaively greaer effecs of moneary shocks on he curren accoun. Table 1 provides wo ofen used indicaors of openness o inernaional rade: he 31 In he U.S. case, Lane (1999) provides similar resuls and shows ha he idenified moneary shock accouns for abou 5% of he variance of he curren accoun. His specificaion, however, does no provide similar resuls for he oher counries. In paricular, we have esimaed he same rivariae VAR sysem (i.e. relaive oupu, he raio of curren accoun o oupu and relaive prices) and imposed he same long-run resricions for he G7 over our sample period. Resuls are much less clear-cu in ha, while he IRFs are qualiaively similar o ours, hey are no saisically significan. Moreover, wih he only excepion of he U.S., he size of he conribuion of moneary shocks in curren accoun flucuaions is less imporan (i.e. beween 5-1 per cen 23