NATIONAL BANK OF BELGIUM

Transcription

1 NATIONAL BANK OF BELGIUM Conference 150h anniversary WORKING PAPERS - RESEARCH SERIES THE ROLE OF THE EXCHANGE RATE IN ECONOMIC GROWTH: A EURO-ZONE PERSPECTIVE Ronald MacDonald (*) (*) Deparmen of Economics, Universiy of Srahclyde, 100 Cahedral S, Glasgow, G4 OLN. NBB WORKING PAPER No.9 - MAY 2000

2 Ediorial Direcor Jan Smes, Member of he Board of Direcors of he Naional Bank of Belgium Saemen of purpose: The purpose of hese working papers is o promoe he circulaion of research resuls (Research Series) and analyical sudies (Documens Series) made wihin he Naional Bank of Belgium or presened by ouside economiss in seminars, conferences and colloquia organised by he Bank. The aim is hereby o provide a plaform for discussion. The opinions are sricly hose of he auhors and do no necessarily reflec he views of he Naional Bank of Belgium. The Working Papers are available on he websie of he Bank: hp:// Individual copies are also available on reques o: NATIONAL BANK OF BELGIUM Documenaion Service boulevard de Berlaimon 14 B Brussels Imprin: Responsibiliy according o he Belgian law: Jean Hilgers, Member of he Board of Direcors, Naional Bank of Belgium. Copyrigh Naional Bank of Belgium Reproducion for educaional and non-commercial purposes is permied provided ha he source is acknowledged. ISSN: X WORKING PAPER No.9 - MAY 2000

3 Ediorial On May 11-12, 2000 he Naional Bank of Belgium hosed a Conference on "How o promoe economic growh in he euro area?". A number of papers presened a he conference is made available o a broader audience in he Working Papers series of he Bank. This volume conains he fifh of hese papers. The oher five papers were issued as Working Paper 5-8 and 10. Absrac In his paper we consider a range of opics which connec exchange raes o he economic growh process. In paricular, we firs of all ouline he basic properies of exchange raes when hey are flexible. One key feaure of flexible exchange raes is ha hey are highly volaile and such volailiy may affec growh hrough he channels of rade and invesmen. These channels are considered in some deail in his paper. We also consider he links beween secoral and aggregae growh and he exchange rae, using he Balassa-Samuelson and Houhakker-Magee-Krugman hypoheses. The main conclusion of he paper is ha he curren exchange rae arrangemens for he euro-zone area, boh inernal and exernal, are likely o simulae economic growh. NBB WORKING PAPER No.9 - MAY 2000

4 WORKING PAPER No.9 - MAY 2000

5 TABLE OF CONTENTS: 1. INTRODUCTION SOME STYLISED FACTS ABOUT REAL AND NOMINAL EXCHANGE RATE BEHAVIOUR ESTIMATION METHODS - A DIVERSION EXCHANGE RATE MODELS AND ECONOMIC GROWTH EXCHANGE RATE REGIMES AND ECONOMIC GROWTH THE GROWTH - EXCHANGE RATE LINK Decomposing he real exchange rae: Violaions of he LOOP and he Balassa-Samuelson hypohesis The Houhakker-Magee-Krugman 45 Rule THE EXCHANGE RATE - GROWTH LINK Exchange raes and inernaional rade Theory Evidence Exchange Raes and Invesmen CONCLUSIONS REFERENCES...52 NBB WORKING PAPER No.9 - MAY 2000

6 WORKING PAPER No.9 - MAY 2000

7 1. INTRODUCTION The role of he exchange rae in he economic growh process is no immediaely apparen from a cursory glance a he growh lieraure. Indeed, he idea ha a financial price can have real effecs would a firs blush perhaps seem o be a raher odd idea. However, some clues o he likely effecs of exchange raes on growh may be gleaned from he behaviour of exchange raes when hey are flexible. Firs, in a flexible exchange rae regime here is a very close correlaion beween real and nominal exchange raes and i is widely acceped, alhough no unconroversial, ha in he presence of sicky prices i is he nominal exchange rae which drives he real exchange rae. Furhermore, once a real exchange rae change occurs ha change ends o be highly persisen or, indeed, permanen. Anoher feaure of exchange raes when hey are flexible is ha hey end o be exremely volaile and such volailiy has been argued o be excessive, in he sense ha here appears o be no corresponding volailiy in he kinds of variables driving exchange raes, such as relaive money supplies and prices. Wha is he relaionship beween such exchange rae behaviour and economic growh? In my lecure oday I am going o ake he body of economic growh heory as given and simply hink of economic growh as driven by changes in he facor proporions, along he lines of a sandard growh accouning relaionship. Wha effec does he exchange rae have on hese proporions? For he purposes of his lecure, i shall prove useful o decompose growh ino permanen, cyclical and ransiory componens as: P C y = y + y + ε, (1) where y denoes he naural logarihm of naional income,? is he firs difference operaor, and herefore? y represens he growh rae, cyclical componens of naional income and p y and c y are he permanen and ε is a ransiory erm. The permanen componen may be hough of as relaed o long-run, or seady sae, growh and he cyclical elemen is he business cycle-relaed componen. How can he exchange rae influence p y and c y? In his paper we disinguish beween wo poenial exchange rae effecs: a level and a volailiy effec. The level effec migh occur when a counry experiences, say, a susained appreciaion of is nominal and real exchange raes, due o a igh moneary policy. This could make par, or all, of he counry's raded secor WORKING PAPER No.9 - MAY

8 unprofiable. The iniial response of his exchange rae change may well be for firms exposed o rade o reduce heir labour inpus o he exising capial sock and his could have a cyclical effec on growh. If he exchange rae misalignmen was sufficienly prolonged hen pars of he radable secor could simply disappear, as occurred in he UK in he lae 1970s and early 1980s. One could also imagine such a levels effec influencing he decision o inves in new capial for he counry experiencing he misalignmen. However, perhaps he main channel by which he exchange rae is hough o influence economic growh is hrough he effec of exchange rae volailiy on he profiabiliy of inernaional rade and invesmen. Indeed, he unaracive implicaions of exchange rae volailiy for rade and invesmen has been argued o be one of he major weaknesses of floaing exchange rae regimes (see for example Group of Tweny Four (1985) and he Group of Ten (1985)) and his cerainly has been one of he driving forces for greaer fixiy of exchange raes wihin Europe, and also is behind calls for greaer fixiy of he ripolar hree exchange raes - he euro, dollar and yen. Alhough i is someimes argued ha he exisence of capial markes, and in paricular a well developed forward marke, should inernalise he unpleasan consequences of exchange rae volailiy, hedging is cosly, and someimes prohibiively so. Furhermore, such markes are ofen incomplee, paricularly a horizons of greaer han one year. We reurn o he issue of hedging below. How imporan are he exchange rae effecs discussed above likely o be for he euro-zone area? This is one of he aspecs of he growh - exchange rae relaionship I wan o address in my lecure oday. If we are prepared o hink in erms of a causaliy relaionship, hen his effec may be hough of as exchange rae movemens causing growh (posiive or negaive). There are wo dimensions o his. Firs, here is he inernal dimension - o wha exen have locking exchange raes wihin Europe squeezed ou he unpleasan consequences of exchange rae behaviour for inra-european rade and invesmen? Some insigh ino his quesion may be gleaned from an examinaion of he linkages ha exised prior o he formaion of he euro. Second, how imporan is his effec likely o be for he euro-zone area vis-à-vis is exernal rading parners? Given ha he euro-zone as an eniy is a relaively closed area in erms of inernaional rade, i may be hough ha his exernal effec is likely o be relaively small. There is, however, anoher causaliy link beween growh and he exchange rae, which is essenially he reverse of he above. Alhough here are various raionalisaions for his effec, one ha I shall discuss in his paper is relaed o he ime series properies of real 2 WORKING PAPER No.9 - MAY 2000

9 exchange raes. For example, and as I shall demonsrae below here is considerable long-run, or secular, persisence in real exchange raes. Wha explains his persisence and is he degree of persisence similar wihin and across moneary unions? Alhough here are a number of poenial candidaes o explain he persisence of real exchange raes here are wo which are paricularly perinen o he opic of his lecure. One is he socalled Balassa-Samuelson effec which posis ha a counry which has relaively high produciviy in is raded goods secor, compared o is non-raded goods secor, will have an overvalued currency relaive o is rading parner(s). Furhermore, if he produciviy growh in he home counry's radable secor is more favourable relaive o is rading parners over ime, his will impar a secular appreciaion ino is real exchange rae. Clearly, if his effec is significan i could have imporan policy implicaions for he inernal workings of a newly formed moneary union since i implies ha wih a fixed nominal exchange rae he repercussions mus be refleced in relaive prices or inflaion differenials. Are such differenials likely o be susainable? To wha exen, hen, is he Balassa-Samuelson, effec imporan for he kinds of counries paricipaing in EMU? An alernaive perspecive on he persisen naure of real exchange raes may be found in wha I will refer o as he Houhakker-Magee-Krugman (HMK) hypohesis. This hypohesis suggess ha counries wih differen long erm growh raes, relaive o heir rading parners, or counries which face differen elasiciies of impor and expor demand, may suffer secular changes in heir real exchange rae. Again, he exen o which his relaionship does, or does no hold, for euro-zone counries may have imporan policy implicaions. A final spin on he growh-exchange rae link, which has been brough ino sharp relief recenly by he sharp depreciaion of he euro, is he effec of relaive business cycle growh on an exchange rae. A number of commenaors have argued ha he euro is weak because aggregae growh in he euro-zone area is relaively slow; once growh in he euro-zone caches up wih US growh, he euro will sar o appreciae. We briefly discuss his linkage in secion 4. I am going o give my discussion of he growh - exchange rae opic an explicily European perspecive by generaing some new empirical resuls for key EU counries. For he euro-zone area we may hink of essenially wo exchange raes: he inernal and he exernal. Prior o moneary union here was some flexibiliy in he nominal and real exchange beween European counries and here was much more flexibiliy in he exernal nominal and real exchange raes vis-à-vis non-european counries, such as he US. The WORKING PAPER No.9 - MAY

10 adven of moneary union, of course, means ha inernal nominal raes are now rigidly fixed wihin Europe, alhough inernal real raes can vary, while he exernal value of boh he real and nominal euro have been flexible. Given ha he euro-zone area is relaively closed - rade and invesmen is predominanly amongs EU counries - i has been suggesed ha he exernal flexibiliy of he euro is unlikely o have paricularly large implicaions for he euro-zone area. I aemp o ge a feel for he likely effecs of exchange rae movemens on euro-zone growh by consrucing panel daa ses consising of he currencies of counries which are currenly full paricipans of EMU. These panel daa ses ry o capure he effecs of boh inernal and exernal exchange rae movemens. In sum, our approach o hinking abou he growh - exchange rae relaionship for he euro-zone area essenially involves presening a smorgasbord of opics which seem relevan o his issue. In he nex secion we se he scene by presening some sylised facs abou he behaviour of real and nominal exchange raes in a flexible rae regime. Secion 3 deails he esimaion mehods used for our empirical resuls. We hen go on in Secion 4 o look a wha a selecion of open economy macro-economic exchange rae models have o say abou exchange rae - growh linkages. In secion 5 a brief of overview of he effecs of he exchange rae regime on economic growh from an hisorical perspecive is presened. In secion 6 he relaionship beween growh and he exchange rae is considered by examining he Balassa-Samuleson and Houhakker-MageeKrugman hypoheses; some new empirical resuls are also presened in his secion. In secion 7 we focus on he poenial role of he exchange rae in creaing economic growh hrough he channels of invesmen and inernaional rade. A concluding secion gahers ogeher he various poins made hroughou he paper. 4 WORKING PAPER No.9 - MAY 2000

11 2. SOME STYLISED FACTS ABOUT REAL AND NOMINAL EXCHANGE RATE BEHAVIOUR Some insigh ino he opic of his lecure may be gleaned by asking he quesion: how do exchange raes behave when hey are flexible? There are a number of so-called sylised facs relaing o his quesion. Firs, when exchange raes are flexible hey end o be highly volaile. This volailiy is usually gauged in a number of ways: on an hisorical basis when comparing he recen flexible rae experience wih fixed, bu adjusable, exchange rae regimes, such as he Breon Woods regime; exchange raes are volaile relaive o some measure of he expeced exchange rae, such as he forward exchange rae or he expecaion implied by survey daa. exchange raes are volaile relaive o cerain fundamenals such as relaive prices and money supplies 1.' The laer is illusraed in Table I where we presen he coefficiens of variaion for a number of exchange raes relaive o cerain key fundamenals (hese are US dollar bilaeral exchange raes for he period January 1980 hrough o December 1997). However, Table I also indicaes ha exchange rae volailiy is of a smaller order of magniude han he volailiy we observe in ineres rae yields (indeed his is also rue for oher asse yields). However, despie he laer a number of papers have demonsraed ha he volailiy of a weighed average of fundamenals is roughly he same under boh fixed and flexible exchange raes (Flood and Rose (1999)). The key disinguishing facor beween he wo regimes, as we have indicaed, is he volailiy of he exchange rae. By fixing he exchange rae his volailiy simply disappears and does no show up elsewhere in he macroeconomy. Some have concluded from his kind of evidence ha i is impossible o explain he volailiy of exchange raes in erms of sandard macroeconomic fundamenals. Table 1 - Coefficiens of Variaion for Exchange Raes and Fundamenals Counry Exchange Rae Relaive Prices Ineres Differenial France Germany Japan Swizerland UK See MacDonald (1988,2000) for an exended discussion. WORKING PAPER No.9 - MAY

12 In order o ge a feel for he relaive volailiy of currencies, as opposed o heir volailiy relaive o fundamenals, we presen in Table 2 coefficiens of variaion for a group of European currencies, including hose who have irrevocably locked heir exchange raes wihin Europe. The raes are defined wih respec o hree numeraire currencies - he DNI, he Yen and he US Dollar. These show ha he volailiy of he US and dollar raes are abou he same order of magniude, bu ha he ERM effec has aenuaed he volailiy of he DM-based currencies o around one-half of ha observed in he oher raes. Furhermore, he volailiy of all hree raes is sample-specific, wih he period of he 80's, a period when he convergence process was perhaps a is greaes in Europe, exhibiing he smalles volailiy. A second sylised fac abou exchange raes is ha here is a very close correspondence beween real and nominal exchange raes. Alhough he inerpreaion of wha causes his volailiy is conroversial, we would argue ha i is he nominal exchange rae which drives he real exchange rae. Clearly such real changes could impac on he profiabiliy of he radable secor and his could affec growh in he medium run and also, perhaps, in he longer run. The close correlaion beween real and nominal exchange raes is illusraed in Figure 1 2 and also in Table 2 where we noe ha he relaive nominal volailiy of currencies discussed above seems o ge ransferred ino roughly equivalen real volailiy. A hird sylised fac abou he behaviour of real exchange raes is ha hey are highly persisen. Evidence of such persisence may be obained from he recen lieraure on Purchasing Power Pariy (PPP) 3.' For example, single currency univariae uni roo ess sugges ha real exchange raes are effecively non-saionary, or o he exen ha hey do exhibi any mean reversion i is incredibly slow. In Table 3 we presen some univariae uni roo saisics o illusrae his persisence for he currencies examined in his paper. Three sample periods are considered: a full sample, 1980, quarer 1, o 1998, quarer 4 and wo sub-samples wihin he full sample (1980,1 o 1989,2 and 1989,3 o 1998,4). These resuls indicae an inabiliy o rejec he null for any sample period for he DM and US dollar based currencies, alhough we noe ha 5 rejecions for he yen based currencies occur in he full sample. These kind of resuls can usually be overurned by increasing he span of he daa. Here we accomplish his by sacking he hree ses of real exchange raes ino panels and consrucing Levin and Lin (1994) panel uni roo -ess and adjused -ess (adj), along wih he implied degree of quarerly adjusmen ( δ ). 2 3 The correlaion beween real and nominal exchange raes is approximaely 0.9. See, for example, MacDonald (1995,2000). 6 WORKING PAPER No.9 - MAY 2000

13 These resuls are repored in Table 4 and indicae ha he null of a uni roo can easily be rejeced in a panel conex. Table 4 - Panel - Uni - Roo Tess adj d USD DM YEN The speed of mean reversion is fases for Yen based currencies and slowes for US dollar bilaerals. A useful way of gaining exra perspecive on he sources of exchange rae volailiy is o decompose he overall real exchange rae - he exchange rae defined using CPI prices - ino he relaive price of raded goods across counries and he relaive price of goods wihin a counry, relaive o is rading parners. In naural logarihms, he real exchange rae defined for CPI prices may be defined as: q * s p p. (2) + If we assume ha he general prices enering our definiion of he real exchange rae can be decomposed ino raded and non-raded componens as: p NT T = α p + (1 α )p, (3) p * NT* T* = α p + (1 α )p, (4) where and he T p denoes he price of raded goods, NT p denoes he price of non-raded goods α 's denoe he share of non-raded goods in he overall price level (and are assumed o be he same across counries). Addiionally, assume ha a similar relaionship o (2) can be defined for raded goods as: T T T* q = s p + p (5) WORKING PAPER No.9 - MAY

14 By subsiuing (3), (4) and (5) in (2) he following expression may be obained: q T NT* T* NT T = q + [ α(p p ) α(p p )], (6) q T + T,NT = q q, (7) q NT,T NT* T* NT [( p p ) ( p p )] T = α (8) The firs erm in (6), T q, represens he law of one price (LOOP), or violaions of he LOOP, while he second erm, NT,T q represens he so relaive price raio and is usually associaed wih he Balassa-Samuelson effec, considered in some deail in secion 6, alhough i can also be driven by demand side influences, such as he effec of governmen expendiure or preference shifs. On he assumpion ha he LOOP holds, expression (6) predics ha if he home counry has a relaively high inernal price raio i will have an appreciaed real exchange rae defined using overall prices. This expression is useful because i allows us o hink of he volailiy, or rend, in he overall real exchange rae as being driven by he volailiy or rend in eiher T q, NT,T q or boh. How imporan is he relaive price of raded goods, T q, compared o he inernal price raio, NT,T q explaining he volailiy and persisence in he overall real exchange rae q? Engel (1993) compares he condiional variances of relaive prices wihin and across he G7 counries using disaggregaed indices of CP1s, over he period April 1973 o Sep Ou of a poenial 2400 variance comparisons, Engel finds ha in 2250 insances he variance of he relaive price wihin he counry is smaller han he variance of he relaive price across counries; ha is, V( T NT,T q ) > V( ) q and ha his difference is saisically significan. Rogers and Jenkins (1995) essenially confirm Engel's analysis using finer disaggregaions of he prices enering he CP1s of 11 OECD counries. Addiionally, however, hey also examine he relaive imporance of rends in T q and NT,T q, in explaining he sysemaic elemen of T q. They find lile evidence ha T q is an I(0) process even when a fine level of dissagregaion is used. Furhermore, hey produce very lile evidence ha T q and NT,T q are coinegraed. Taken ogeher, he empirical 8 WORKING PAPER No.9 - MAY 2000

15 evidence on he relaive imporance of he wo righ hand side elemens in would seem o favour sicky price models, such as hose of Dornbusch (1976) and Giovannini (1988). One alernaive inerpreaion is o aribue i o he pricing o marke policies of companies. However, boh Rogers and Jenkins (1995) and Wei and Parsley (1995) show ha adjusmen speeds for disaggregae relaive prices are similar o he adjusmen speeds esimaed for aggregae CPI real exchange raes, which seems inconsisen wih he pricing o marke hypohesis. Wha are he implicaions of he sylised facs noed here for growh in he euro-zone area? This is he quesion we aemp o address in some deail in he remainder of his paper. For now, hough, we presen a summary of he likely answers. Firs, he removal of nominal volailiy by locking currencies wihin Europe may have imporan implicaions for euro-zone rade, invesmen and growh. To he exen ha he persisence in real exchange raes is driven by he persisence in nominal exchange raes his may also be beneficial since, in he absence of such volailiy, inernal euro-zone real exchange raes may be beer able o reflec relaive prices wihin Europe, raher han he capricious movemens of he nominal rae and he misaligned real raes hey can imply. Clear relaive price signals are likely o improve resource allocaion wihin Europe. The fac ha he exernal value of he euro is mean-revering means ha i can adjus over ime and his may have imporan implicaions for curren accoun imbalances and growh. Addiionally, how is volailiy in he exernal value of he euro likely o affec growh wihin he euro-zone area? The above effecs all relae o he exchange rae influencing economic growh. Bu our discussion in his secion also suggess a way in which he growh process iself is likely o have an imporan influence on real exchange raes wihin he euro-zone area. Hence removing a major source of volailiy in real exchange raes, by locking nominal raes, could mean ha he so-called inernal price raio, NT,T q, is now he dominan driving force of he overall real exchange rae. As we shall see, one of he main poenial driving forces of NT,T q is produciviy differenials in he raded goods secor relaive o he non-raded secor. Does his growh effec have unpleasan consequences for inernal euro-zone exchange raes? Before closing his secion we presen a firs pass a he exchange rae - economic growh linkage by presening scaer plos of he relaive growh raes of 8 key paricipans in he euro-zone projec (Ausria, Belgium, France, Germany, Ireland, Ialy, he Neherlands, and Spain) agains he corresponding real exchange rae changes, using hree differen WORKING PAPER No.9 - MAY

16 numeraire: he German mark, he US dollar and he Japanese yen. These are presened in Figure 2 and indicae no clear-cu relaionship beween economic growh and real exchange raes. However, hese kind of figures may in fac conceal more han hey reveal. The res of he paper may be seen as an aemp o gauge how robus he resuls in Figure acually are. 10 WORKING PAPER No.9 - MAY 2000

17 3. ESTIMATION METHODS - A DIVERSION In some of he succeeding secions we presen some new esimaes of various proposiions relaing o he exchange rae - growh proposiion. These esimaes are designed o capure boh he inernal euro effecs - ha is, for he inernal real exchange rae relaionships wihin he euro area - and also for he exernal value of he euro - agains he dollar and yen. Since mos of he variables considered here are non-saionary we use single equaion coinegraion esimaors o generae our resuls. In conras, however, o he sandard wo-sep Engle-Granger coinegraion esimaors, our esimaors recognise poenial simulaneiy and serial correlaion biases. In paricular, we use he so-called single equaion dynamic ordinary leas squares (SDOLS) esimaor of Sock and Wason (1993): y + p = α + βx + Σ θ j x + j= p j + ξ, (9) where he leads and lags are included o accoun for poenial endogeneiy and serial correlaion. The second esimaor is he panel equivalen of (9), advocaed by Kao (1999) and Mark and Sul (1999). Essenially, he panel DOLS esimaor inroduces a cross secional dimension ino (9): y i + p = αi + βxi + Σ θij xi+ j= p j + ξ i, (10) where he consan now has he inerpreaion of a fixed effec. Our daa, discussed on a case-by-case basis, are exraced from he OECDs CD-ROM (1999/2) disc and he IMF's IFS CD-ROM (March 2000). WORKING PAPER No.9 - MAY

18 4. EXCHANGE RATE MODELS AND ECONOMIC GROWTH Wha is he relaionship beween income, growh and he exchange rae in macroeconomic exchange rae models? The flexible price moneary model has become somehing of a workhorse in open economy macroeconomics, being he long-run soluion o he celebraed Dornsbusch (1976) overshooing model and a model in is own righ (see, for example, Frenkel (1976) and Mussa (1979)). Alhough he moneary model is usually moivaed as an asse marke model, i is in fac a simple exension of PPP which fleshes ou he deerminaion of prices in each counry by imposing coninuous money marke clearing. In paricular, assume ha he demand for money in he home and foreign counry is given by a (log-linearised) Cagan money demand funcion and ha he supply of money is coninuously equal o he demand a some exogenous level, m m > p = α y α i, α, α 0, (11) * * * * m > p = α y α i, α, α 0, (12) On rearranging (11) and (12) for he home and foreign counry price levels, respecively, and subsiuing hese ino an absolue PPP condiion we obain he so-called flex-price moneary reduced form: * * s = p p = (m m ) α (y y ) + α (i i ) (13) 0 * 1 * which simply saes ha he nominal exchange rae is driven by relaive excess money supplies. Income, and herefore by implicaion growh, affecs he exchange rae in his model indirecly hrough he demand for money. Oher hings equal a counry enjoying posiive income growh will enjoy an appreciaing currency: posiive income growh raises he real demand for money which, for an exogenously deermined supply of money can only be saisfied by a fall in he price level and an exchange rae appreciaion. Recen empirical research suggess ha his kind of model has some validiy boh as a long- run and also a shor-run relaionship (see, for example, La Cour and MacDonald (2000)). MacDonald and Swagel (2000) survey he poin esimaes of no repored in a number of papers and find ha he vas majoriy of esimaes are significanly negaive as prediced. Of course his effec could also arise in sicky price varians of he moneary model 12 WORKING PAPER No.9 - MAY 2000

19 (Dornbusch (1976) and Rankel (1979)), and in such models a rise in income can have a reinforced effec on income o he exen ha i pulls up nominal and real ineres raes in he process. The paern of a srong exchange rae and srong economic growh (weak exchange rae and weak economic growh) is usually hough of as he business cycle growh - exchange rae relaionship and is usually driven by ineres raes. Indeed, o he exen ha ineres raes conain informaion abou fuure growh, he exchange rae can appreciae in anicipaion of srong economic growh. The above growh - exchange rae relaionship may have implicaions for inra-euro-zone inflaion differenials which are he opposie of hose implied by he Balassa-Samuelson effec, discussed in secion 6. For example, wih a common euro-zone wide moneary policy deermined in effec by he average income and inflaion growh acros he euro-zone, a counry wih above (below) average growh will, ceeris paribus, have negaive (posiive) inflaion. How imporan is his effec likely o be? The α 0 coefficien is normally esimaed a beween -0.5 and -1. If we ake an average number of hen his suggess ha a counry which has an annualised growh of 1 % above he average of is euro-zone parners will find is inflaion rae falling by 0.75 per cen per annum. Relaive o he inflaion numbers menioned below his effec is raher small bu could, neverheless, help o offse he implicaions of increased produciviy growh in radable secors for inflaion. The real business cycle, or supply side, models of Sockman (1980) and Lucas (1982) essenially append a supply side o equaion (13). A ypical reduced form from his class of model would be: s = z + [ α + ρ ], (14) 0 where z is a vecor comprising he variables on he lef hand side of (13), is a relaive ase shock and ρ is a relaive echnology shock. The laer variables are seen as driving a counry's real exchange rae and his, in urn, is seen as providing an explanaion for he close correlaion beween real and nominal exchange raes noed in secion 2 (see Sockman (1987)). The model has no role for exchange rae movemens causing economic growh. However, for causaliy o run from he real o he nominal exchange rae he volailiy of fundamenals should have increased during he recen floaing period. The WORKING PAPER No.9 - MAY

20 fac ha hey have no is perhaps he mos convincing piece of evidence agains his class of model. A model which does poenially have an explici role for he exchange rae in he growh process is he sandard exbook Mundell-Fleming model. In he mos basic form of his model here are no supply side consrains and expecaions are formed saically. A cenral relaionship in his model is he aggregae demand funcion and a cenral elemen in his is he real exchange rae: y d = η(s p ) σr. (15) An expansionary moneary policy in his model can, for example, generae a permanen rise in oupu. We reurn o his kind of relaionship in a couple of places laer in he paper. However, i hardly needs saying ha his model is no longer a paricularly fashionable vehicle for hinking abou exchange rae issues. The so-called Mundell- Fleming-Dornbusch model appends sluggish shor-run price adjusmen, and long-run price flexibiliy, along wih forward looking expecaions o (15) o produce a differen seady sae predicion beween he exchange rae and income. The addiional relaionships in his model are: p = (1 θ) Ε p + θp, (16) 1 m p = y λi, (17) i = Ε s + +, (18) 1 µ r ( i Ε ( ρ ) p ). = + (19) 1 Equaion (15) is an IS-relaion, relaing aggregae demand ( y d ) o he real exchange rae ( q s p ) and he expeced real ineres rae r. Equaion (16) is jus he price adjusmen equaion where he bar denoes he permanen componen of he price level ( p ). A money marke equilibrium condiion is given by equaion (17), while (18) is an uncovered ineres pariy condiion augmened by ( µ ) ha could be inerpreed as a risk premium. 14 WORKING PAPER No.9 - MAY 2000

21 The supply side of he model is specified by wo random walks: y s s 1 + = y z, (19) m = m + ν, (20) 1 where z and ν denoe supply and money shocks, respecively. The seady sae of his model is given by: s y = y, (21) q 1 = [ y + σr ], (22) η p = m y λi. (23) This model predics ha oupu shocks can affec he long-run exchange rae. Alhough he exchange rae can affec oupu in he radiional way over he business cycle, i has no long-run effec (see Hoffman and MacDonald (2000)) for a furher discussion of he empirical implicaions of his model. WORKING PAPER No.9 - MAY

22 5. EXCHANGE RATE REGIMES AND ECONOMIC GROWTH This brings us ino he issue of wha has been he exchange rae regime mos closely associaed wih economic growh. A number of papers have sough o address his issue. For example, Bordo and Schwarz (1998) provide a comprehensive comparison of he growh of real per capial income over a number of key regimes of he inernaional moneary sysem, spanning he period The regimes covered are: he classical gold sandard, , he iner-war period 'mixed regime', , he Breon Woods period, and he recen floaing rae period, 1973 o presen. The Breon Woods period is furher subdivided ino he preconverible phase, , and he converible phase Also he recen floaing period is subdivided ino an inflaion period, , and a disinflaion period, In summary, Bordo and Schwarz find he following: he Breon Woods period, and paricularly he converible period, exhibied he mos rapid average oupu growh of any moneary regime and he iner-war 'mixed regime' period produced he lowes. However, ineresingly, aking he enirey of he Breon Woods period, heir is a higher variabiliy of growh han in he recen floaing rae period. In conras, however, Ghosh e al (1996) find ha here is lile correlaion beween an adherence o fixed exchange raes and economic growh, once accoun is aken of he 1960s period. Indeed, Bordo and Schwarz concede ha he link beween he kind of fixed exchange raes provided by Breon Woods and high economic growh seems less compelling han for oher aspecs of economic performance, such as inflaion, and hey aribue his o a number of facors. Firs, hey argue ha here is an apparen absence of a link beween exchange rae volailiy and eiher invesmen or rade flows and economic growh. Thus, alhough Ghosh e al (1996) find evidence linking real growh o he growh of invesmen and rade for pegged counries, hey also find oal facor produciviy growh o be an imporan channel of growh for floaers. Furhermore, insiuions ouwih he Breon Woods regime may have been imporan for growh, such as OEEC, EPU, European Coal and Seel Communiy (ECSC). Third, Bordo and Schwarz argue ha he Breon Woods sysem may have conribued o growh by providing an overall framework of rules which allowed Wesern European naions o solve a hierarchy of co-ordinaion problems, which allowed hem o encourage invesmen in growh-generaing expor secors. Fourh, he Breon Woods regime may have conribued o pos-war growh by being par of an overall package generaing poliical and economic sabiliy - he so-called 16 WORKING PAPER No.9 - MAY 2000

23 Pax Americana. In heir view, herefore, Bordo and Schwarz argue ha i is difficul o disenangle he effecs of he exchange rae regime per se from he insiuional facors associaed wih ha regime. WORKING PAPER No.9 - MAY

24 6. THE GROWTH - EXCHANGE RATE LINK We now urn o wo poenial avenues hrough which growh can affec he real exchange rae and, in paricular, generae he eviden persisence in real exchange raes. The firs of hese, he Balassa-Samuelson effec, focuses on he inernal price raio in (6), and argues ha unbalanced growh in a counry's raded secor relaive o is non-raded secor, can impar a secular rend ino he real exchange rae. This sory can have poenially imporan implicaions for he inernal relaive inflaion raes of he euro-zone counries and, also, for real ineres differenials wihin Europe. Alhough hese relaive effecs are ofen seen in he conex of a cach-up hypohesis, and herefore deemed o be only ransiory, i is possible ha here may be more permanen implicaions of hese kinds of effecs. The BS hypohesis is also likely o have implicaions for he exernal value of he euro. In paricular, wha are he implicaions for he sance of euro-zone moneary policy and, relaedly, he implicaions for he kind of exchange rae regime he euro should paricipae in? The BS effec is also likely o have imporan implicaions for counries, such as he cenral european counries, seeking o ener he euro-zone, since secorally unbalanced growh can produce exchange rae and inflaion combinaions which are inconsisen wih he convergence crieria. The Balassa-Samuelson hypohesis is a supply side effec relaing o he longer run rend in he real exchange rae. The second srand in he growh-exchange rae link considered in his secion is more closely associaed wih he relaive price of raded goods across counries and is relaed more o he medium run rend in he real exchange rae. We label his effec he Houhakker-Magee-Krugman hypohesis, as i was firs noed by Houakker and Magee (1969) and formalised by Krugman (1989) ino he so-called 45 rule. This hypohesis represens a parial equilibrium approach o inerpreing secular rends in real exchange raes. In paricular, he hypohesis suggess ha if a paricular lock does no hold beween a counry's relaive growh rae and is relaive expor and impor income elasiciies, his could have imporan consequences for he secular drif in is exchange rae. 18 WORKING PAPER No.9 - MAY 2000

25 6.1 Decomposing he real exchange rae: Violaions of he LOOP and he Balassa-Samuelson hypohesis Perhaps he bes known explanaion for secular rends in he real exchange rae is he Balassa-Samuelson (BS) biased produciviy growh hypohesis. The BS hypohesis focuses on he role ha he so-called inernal price raio - he raio of non-raded o raded goods prices - can play in inroducing sysemaic rends ino real exchange raes. In paricular, he proposiion is ha a counry wih relaively high produciviy in is raded goods secor will have an appreciaed real exchange rae, defined using overall price levels. Furhermore, if ha counry exhibis relaively high produciviy growh in is radables secor over ime i will have a secular appreciaion of is real exchange rae. The BS hypohesis focuses on he implicaions of rends in produciviy for long-run real exchange raes, ignoring shor-run adjusmens. The long-run naure of he model means ha relaive prices are driven by supply side facors, wih demand side facors being ignored. The BS hypohesis may be explained in he following way. Assume ha producion echnology for he home counry is given by a simple Cobb-Douglas specificaion (a similar se of relaionships are assumed o hold in he foreign counry): T (1 λ) T = η(k ) (L ) T λ (24) NT(1 δ) NT NT = υk (L ) δ (25) where T and NT denoe producion of raded and non-raded goods, respecively, η and ν represen shocks o oal facor produciviies. In he home and foreign counry capial and labour are assumed o be fully employed in he producion of raded and non-raded goods: L T NT + L = L (26) K T NT + K = K (27) Assuming ha compeiion ensures ha labour is paid he value of is marginal produc and ha nominal wages, W, are equalised across secors, hen: WORKING PAPER No.9 - MAY

26 δt / δl T δnt / δl NT = W /P W /P T NT P = P NT T (28) where NT T P /P is he inernal price raio referred o in secion 2 and a similar expression is assumed o hold in he foreign counry. Given his se up, a shock o oal facor produciviy in he raded secor will increase he marginal produc of labour in ha secor, raise wages and he relaive price of nonraded goods. Reurning o equaion (6) we see ha if, as he BS hypohesis assumes, he LOOP is coninually saisfied (and herefore T q is always zero or consan) and produciviy is unchanged in he foreign counry, his produciviy shock will appreciae he overall real exchange rae. If, furhermore, he home counry has relaively rapid growh in is raded goods secor over ime he predicion is ha i will have a secular appreciaion of is real exchange rae. Usually his effec is a is mos dramaic when comparing a developed o a developing counry. The findings, discussed in secion 2, indicaed ha he dominan source of volailiy in real exchange raes comes from he relaive price of raded goods, T q. This, of course, does no necessarily imply ha ha he Balassa-Samuelson effec is in iself unimporan or insignifican, i is jus ha he above evidence suggess ha wih flexible exchange raes he dominan componen of real exchange rae behaviour is nominal exchange rae volailiy. So how imporan is he Balassa-Samuelson effec? A number of papers have examined he effec for boh developing and developed counries (see Chinn and Johnson (1999) for a survey). Recen ess of he BS hypohesis (see for example Canzoneri, Cumby and Diba (1999), Chinn and Johnson (1999) and MacDonald and Ricci (2000)) use eiher oal facor produciviy or average labour produciviy differences as he produciviy measures are based on he following regression equaion: T NT x = α + β(pr pr ) + ν + ε (29) where x is eiher he real exchange rae, q, or he inernal price raio, ipr, and ν is a vecor of oher condiioning variables In sum, his srand or research finds significan and correcly signed effecs of produciviy differences on he inernal price raio, he relaive produciviy erm ends o overpredic he ipr and a saisically significan relaionship 20 WORKING PAPER No.9 - MAY 2000

27 beween relaive produciviy differences and he overall real exchange rae is found, especially if a panel esimaor is used (and he relaive produciviy erm ends o underpredic he real exchange rae). Canzoneri, Diba and Eudey (1996) es he BS hypohesis for a group of eleven European counries (Ausria, Belgium, Denmark, Finland, France, Germany, Ialy, Porugal, Spain and Sweden) using annual daa on average labour produciviy in he raded and non-raded secors for he period 1970 o Using some simple saisical ess and single equaion coinegraion ess Canzoneri e al show compelling evidence o sugges ha rends in he produciviy raio are good predicors of long-run rends in overall real exchange raes. Kohler (1999) uses an unbalanced panel daa se of 28 counries for he period o examine how imporan secoral produciviy is in explaining pas price movemens. Using a sandard fixed effecs panel esimaor she finds slope coefficiens which are significanly above zero bu also significanly below uniy (a range of approximaely 0.5 and 0.7) and inerpres his as a reflecion of a failure of wage equalisaion across counries (some suppor for his is o be found in Aleberola and Tyrvainen (1998) who show ha condiioning on his differenial produces a coefficien on he relaive produciviy erm of uniy). Addiionally, using he panel coinegraion esimaor of Pedroni, which allows for he esimaion of he individual BS coefficiens for each counry, Kohler finds a fairly wide dispersion of poin esimaes ranging from -I for Ialy, Belgium and Finland o -0.6 for Germany. We inerpre he above evidence as suggesing ha he Balassa-Samuelson hypohesis is in he daa for euro-zone counries. Wha, if any, are he likely consequences for his for he fuure of EMU? In paricular, wha are he implicaions for he behaviour of real exchange raes and inflaion wihin he euro-zone and also for he euro-zone relaive o is exernal rading parners? Assume, as before, ha (3) and (4) hold. Then we may define he inflaion rae, π, for he home counry as: T π P /P (1 α) p + α p = p + α ipr (30) NT T where ipr is he inernal price raio. On using (5), (30) and (27) we may calculae he inflaion differenial beween he home and foreign (numeraire) counry as: * π π = s + α( ipr ipr ) (31) * WORKING PAPER No.9 - MAY

28 T NT T* NT* [ ( mpl mpl ) ( mpl mpl )]. = s + α (32) where mpl denoes he marginal produc of labour. If he nominal exchange rae is flexible s 0 hen as we have seen BS explains he evoluion of he real exchange rae across counries: i is he role of moneary policy o decide how hese exernal effecs are spli beween nominal exchange rae changes and changes in he inflaion differenial. If, alernaively, he exchange rae is rigidly fixed, as in he euro-zone area, BS explains he inflaion differenial across paricipaing counries, where he inflaion differenial is simply he change in he inernal real exchange rae. How imporan are such inflaion differenials likely o be wihin he euro-zone area? Are hey likely o undermine he credibiliy of he fixed pariies? Before answering hese kinds of quesions we noe anoher implicaion of biased produciviy growh. Expression (31) may be used o say somehing abou he behaviour of real ineres raes across counries paricipaing in he euro-zone. For example, defining he change in he real exchange rae as q = ( π π + s), wih * flexible exchange raes, or q = ( π π ), wih fixed raes, and assuming real ineres pariy holds: * q = (r r ) (33) * hen (31) implies ha relaive produciviy differences will drive real ineres raes over ime: * T NT T* NT* [ ( mpl mpl ) ( mpl mpl )]. (r r ) = α (34) In paricular, if, say, home produciviy is growing faser han foreign produciviy, he home real ineres rae has o be lower han he foreign real rae. This, in urn, could have implicaions for relaive growh raes across euro-zone counries o he exen ha hese real ineres differenials influence gross capial formaion. In sum, he biased produciviy growh amongs paricipans of a moneary union may cause boh inernal and exernal real exchange rae changes and here is herefore an issue of how susainable hese are likely o be. In passing, i is worh menioning an alernaive explanaion for produciviy differences, refered o as he labour absorpion hypohesis (Canzoneri e al (1996)). In his effec, 22 WORKING PAPER No.9 - MAY 2000

29 increased inegraion in Europe has forced he raded goods secor o become more compeiive and should shed excess labour. This surplus labour has been absorbed by governmen employmen, hereby reducing average produciviy of he nonraded secor and, since his secor is shelered from compeiion, increasing he price. However, Canzoneri e al argue ha for his effec o be a valid explanaion of real exchange rae movemens wihin Europe would require he real exchange raes o overpredic produciviy rends (which are proxies for marginal coss); however, in heir work he opposie appears o be he case. How imporan is he BS effec likely o be for he euro-zone? A number of sudies have examined he kind of relaive price movemens which seem o be consisen wih he operaion of exising moneary unions. For example, De Grauwe (1992) examines he relaive price behaviour of five German Lander and finds inflaion differenials beween 0.2 and 1.2 per cen. Poloz (1990), Bayoumi and Thomas (1995) and Bui and Sapir (1998) examine inflaion differenials wihin Canada and he Unied Saes, respecively, and find inflaion differenials of beween 0.5 and 2 per cen. Canzoneri e al (1996) use heir esimaed produciviy equaions discussed above o calculae he inflaion differenials of heir group of European counries relaive o Germany implied by he rends in relaive labour produciviy. The counries can be divided ino hree groups: Belgium, Ialy and Spain form a group in which produciviy rends imply ha hey should have inflaion raes which are abou 2 % higher han German raes, while he relaive produciviy growhs of Porugal, Denmark, Ausria, France, UK and Sweden imply hey should have inflaion raes on average 1 % higher han German raes and Finland should have an inflaion rae abou he same as he German average. These kind of inflaion differenials are no inconsisen wih he Maasrich crierion, nor do hey seem o be inconsisen wih he size of differenials found wihin exising moneary unions, referred o above. Based on her esimaes of produciviy differenials, discussed above, Kholer (1999) esimaes implied inflaion (CP1) differenials for EMU counries. She finds ha he upper band for his is in he range 1-3% wih he higher figure represening he growh experience over he las 30 years, and he smaller number being derived from he growh experience over he las 15 years. Alhough he inflaion differenials implied by relaive produciviy growh raes do no seem inconsisen wih inflaion differenials in exising moneary unions, Canzoneri e al use he daa se of Bayoumi and Eichengreen (1993) o demonsrae ha here is much less regional variaion in produciviy wihin he US and indeed ha he implied differenials in WORKING PAPER No.9 - MAY

30 regional inflaion are only abou one fifh he size of Europe. Do counries wih differing produciviy rends belong in he same moneary union? Will full economic inegraion cause produciviy rends o converge in Europe? We would argue ha indeed his is wha has caused he homogeneiy in exising moneary unions. We presen some new empirical evidence on he imporance of he Balassa-Samuelson hypohesis for he euro~zone area by running he following regression using a single equaion DOLS esimaor: q NT,T = α + βq + ε. (35) In order for his equaion o represen a es of he BS hypohesis we mus assume ha he LOOP holds up o a consan and ha he inernal price raio is picking up produciviy differences and no oher demand side facors. However, even if i is no a pure es of BS i may neverheless be insrucive in indicaing he imporance of he inernal price raio in driving inernal and exernal real exchange raes for he euro-zone. If he BS is valid, he β coefficien is expeced o be significanly negaive. As is sandard in he exchange rae lieraure, we proxy he price of raded goods wih he producer price index and he consumer price index is our proxy for non-raded goods. These daa were exraced from he Inernaional Moneary Fund's Inernaional Financial Saisics CD-ROM (March 2000). We presen ses of esimaes for hree sample periods. The full sample period, 1980q1-1998q4, and wo sub-periods, 1980q1-1989q4, and 1990q1-1998q4. Esimaing his relaionship for he wo sub-samples should give some indicaion of he sabiliy of he relaionship and, in paricular, if he convergence process has affeced i. The counries chosen for hese ess are lised in Table 4 and mos are members of he euro-zone area. Three numeraire currencies have been chosen for hese ess: he US dollar, he Japanese yen and he German mark. The former wo raes pick up he exernal euro real exchange rae, while he laer picks up he behaviour of inernal real exchange raes. For he US dollar and yen based sysems we noe ha for he full sample period, and he wo sub-samples, he majoriy of coefficiens are correcly signed and saisically significan. We noe also ha in he majoriy of cases he coefficien for hese wo exernal sysems suggess ha a one per cen increase in he inernal price raio has a more han proporionae effec on he real exchange rae and his effec seems o be sable across he sub-samples. For he sysem based on he mark, we noe ha he majoriy of coefficiens are correcly signed and significan bu ha for he full sample he absolue 24 WORKING PAPER No.9 - MAY 2000

31 magniude of he coefficiens is less han uniy indicaing ha a one percen rise in NT,T q has a less han proporionae effec on he overall real exchange rae. The resuls for he firs sub-sample are similar o he full sample bu in he second sample we noe a number of coefficiens are above uniy and hese end o be for counries mos likely o be involved in a cach-up process -Spain Ialy and Ireland. Panel DOLS esimaes are consruced solely for he paricipans of he euro-zone area (ha is excluding boh Denmark and he UK) and are repored in Table 5. These resuls generally confirm he poins made regarding he single equaion esimaes. In sum, he resuls based on equaion (35) sugges ha he here is a significan and correcly signed Balassa-Samuelson effec for he inernal real exchange raes of euro-zone counries and ha he magniude of his effec does no appear o be inconsisen wih hese counries paricipaion in a moneary union. Furhermore, here also seems o be a correcly signed and significan Balassa-Samuleson effec for euro-zone counries relaive o he wo key exernal currencies. The larger magniude of he exernal effec would perhaps sugges ha he exernal nominal value of he euro should be flexible. We conclude his secion by arguing ha he exisence of produciviy differenials wihin Europe is unlikely o generae movemens of inernal real exchange raes which would pu a srain on EMU. There will, however, ineviably be imporan and, perhaps significan, differences in he shor-run as counries cach-up wih heir moneary union parners (and Ireland is a classic example of his a he momen), bu once such counries have caugh up he differenials would no be expeced o be any larger han hose observed for exising moneary unions. If agens do indeed recognise ha hese inflaion differenials are ransiory i would seem unlikely ha he implied real ineres differenials will have significan implicaions for differenial capial formaion. To hey exen ha hey do, his could acually moderae he inernal real exchange rae movemens o he exen ha hey increase produciviy in he service secors. A he end of he day he imporance of he Balassa-Samuelson effec in he euro-zone conex boils down o wheher i is seen as a good or a bad in he European conex. I would seem ha one of he key raionales for EMU is o allow counries which were originally relaively poor o cach up. Fundamenally, wha EMU does is o allow counries o rade-off real exchange rae variaion due o nominal variabiliy from T NT,T q for variabiliy due o q. As we shall argue subsequenly, WORKING PAPER No.9 - MAY