Financial globalization and exchange rates

Transcription

1 Financial globalizaion and exchange raes Philip R. Lane IIIS, Triniy College Dublin and CEPR Gian Maria Milesi-Ferrei Inernaional Moneary Fund and CEPR Preliminary Commens welcome Absrac The founders of he Breon Woods Sysem sixy years ago were primarily concerned wih orderly exchange rae adjusmen in a world economy ha was characerized by widespread resricions on inernaional capial mobiliy. In conras, he rapid pace of financial globalizaion during recen years poses new challenges for he inernaional moneary sysem. In paricular, large gross cross-holdings of foreign asses and liabiliies means ha he valuaion channel of exchange rae adjusmen has grown in imporance, relaive o he radiional rade balance channel. Accordingly, his paper empirically explores some of he iner-connecions beween financial globalizaion and exchange rae adjusmen. * Prepared for he Dollars, Deb, and Deficis 60 Years Afer Breon Woods Inernaional Conference co-organized by he Banco de España and he Inernaional Moneary Fund, Madrid, June We are graeful o Marco Arena, Charles Larkin and Vahagn Galasyan for excellen research assisance. Par of his paper was wrien while Lane was a visiing scholar a he Inernaional Moneary Fund. Lane also graefully acknowledges he financial suppor of he Irish Research Council on Humaniies and Social Sciences (IRCHSS) and he HEA-PRTLI gran o he IIIS. This paper is also par of a research nework on The Analysis of Inernaional Capial Markes: Undersanding Europe s Role in he Global Economy, funded by he European Commission under he Research Training Nework Programme (Conrac No. HPRN CT ). The views in his paper do no necessarily reflec hose of he Inernaional Moneary Fund.

2 I. INTRODUCTION Financial globalizaion has been one of he mos imporan rends in he world economy in recen decades. This process has involved he accumulaion of large gross inernaional invesmen posiions, wih foreign asse and liabiliy posiions sharply rising, wheher scaled by GDP or by domesic financial variables (Lane and Milesi-Ferrei 2003a, Obsfeld and Taylor 2004). In addiion o larger gross posiions, financial globalizaion has also allowed a greaer dispersion in ne foreign asse posiions, wih a significan number of counries emerging as eiher large ne crediors or ne debors (Lane and Milesi-Ferrei 2002a). In general, financial globalizaion is one of he key rends ha has reshaped he global economy relaive o he environmen envisaged by he designers of he Breon Woods sysem in 1944 and undersanding is macroeconomic implicaions is crucial o ake a view on he appropriae fuure direcion for he inernaional moneary sysem. One consequence of financial globalizaion is ha he inernaional spillovers from asse price and currency movemens have been enhanced. In addiion o affecing he direcion and magniude of ne capial flows, asse price dynamics also generae revaluaions of exising invesmen posiions. For insance, he value of he ne liabiliy posiion of he US is quie sensiive by he relaive movemens in he US versus non-us equiy markes and by swings in he value of he dollar. Indeed, such revaluaion effecs may be as imporan as curren accoun imbalances in driving he dynamics of ne foreign asse posiions (Lane and Milesi- Ferrei 2001a, 2002a, Gourinchas and Rey 2004). Of course, asse price and currency movemens canno be viewed as exogenous influences on he value of inernaional invesmen posiions, since shifing global demands for various asses and liabiliies are an imporan driver of financial reurns and exchange raes (e.g. hrough he deerminaion of counry and currency risk premia). Moreover, here is an obvious inerplay beween he financial and rade accouns ha provides anoher link beween ne foreign asse posiions and exchange raes: a long-erm debor may require real depreciaion in order o generae he rade surpluses ha are he counerpar of susained ne invesmen income ouflows (Lane and Milesi-Ferrei 2002b, 2004). Our goal in his paper is o explore he inerconnecions beween financial globalizaion and exchange raes. To esablish he sylized facs abou financial globalizaion, he firs par of he paper examines rends in gross and ne inernaional invesmen posiions and heir componens for a large se of advanced and developing economies. This exends our previous work: in paricular, we updae our esimaes of exernal asses and liabiliies for a sample of emerging markes. 1 Given he evidence on he increased pace of financial globalizaion since he mid-1990s documened for indusrial counries in Lane and Milesi-Ferrei (2003a), i is imely o provide an overview exended o emerging markes ha encompasses he mos recen period. As noed above, a cenral aspec of our analysis is he focus on he facors 1 Lane and Milesi-Ferrei (2001a) explain he consrucion of he daa.

3 - 2 - explaining he changes in exernal posiions: no only capial flows bu also valuaion effecs, such as hose caused by asse price and exchange rae flucuaions. In he second par of he paper we firs provide an analyical framework ha is useful in undersanding he facors driving he dynamics of he ne foreign asse posiion, and hen explore he conribuion of currency movemens o he revaluaion componen of ne foreign asse dynamics. This relaionship depends on a number of facors. For insance, he impac of an exchange rae depreciaion will depend on gross foreign asse and liabiliy holdings (in addiion o he ne posiion); he currency composiion of boh sides of he inernaional balance shee; and he naure of he co-movemen beween exchange rae changes and oher financial reurns. 2 These facors will vary across counries, according o he level of developmen, counry size and oher characerisics. Along one dimension, i is well appreciaed ha a high proporion of he liabiliies of a major indusrial counry is likely o be denominaed in is own currency, whereas a ypical emerging marke economy exhibis significan liabiliy dollarizaion. Counries also differ as o he mix of shor- and long-erm deb insrumens and he levels of porfolio equiy and FDI holdings in he inernaional balance shee: he impac of currency movemens on he ne exernal posiion is undoubedly sensiive o exernal capial srucure. 3 Our analysis suggess ha heoreical work on open economy macroeconomics should srive o incorporae elemens such as persisen non-zero ne foreign asse posiions, large gross asse cross-holdings and mixed porfolios of equiy and deb insrumens and illusrae why hese feaures can make a difference o model dynamics and welfare analysis. In he las par of he paper, we also skech he implicaions of our empirical work for policy analysis. II. TRENDS IN INTERNATIONAL FINANCIAL INTEGRATION In Lane and Milesi-Ferrei (2002a, 2003a) we documened a number of sylized feaures of inernaional capial flows and exernal posiions in indusrial counries. In paricular, capial flows o and from indusrial counries increased subsanially in recen years, boh in absolue erms and as shares of GDP and domesic wealh. In his conex, he increase in foreign direc invesmen and porfolio equiy invesmen is paricularly noeworhy. In addiion, he increase in gross socks of exernal asses and liabiliies means ha valuaion effecs have become more imporan. We highligh hese feaures again below, wih an updaed daase which includes boh indusrial counries and emerging markes. 2 Tille (2003) provides an ineresing analysis for he US. 3 Lane and Milesi-Ferrei (2001b) analyze some of he deerminans of he composiion of he inernaional balance shee.

4 - 3 - A. Ne flows and ne posiions, indusrial counries Figure 1 plos he ne foreign asse posiion as a raio of GDP agains he level of GDP per capia, measured in curren US dollars, for he year There is a wide dispersion in ne foreign asse posiions among indusrial counries, wih Swizerland being by far he larges credior, and New Zealand and Iceland he larges debors. The posiive relaion beween ne foreign asses and GDP per capia, which is shown in he Figure o hold in he cross-secion, holds also along he ime-series dimension as counries ge richer, relaive o rading parners heir ne foreign asse posiion ends o improve (Lane and Milesi-Ferrei, 2002a). Table 1 summarizes ne capial ouflows from indusrial counries over he period , ogeher wih changes in heir exernal posiion. In absolue erms, Japan has been he larges capial exporer, while Swizerland and Norway had he highes ne ouflows relaive o heir GDP. On he oher side, he Unied Saes had by far he larges ne inflows in absolue erms, and also as a raio of GDP. While here is clearly a posiive relaion beween ne ouflows and change in he ne exernal posiion, he Table highlighs he imporance of valuaion effecs: for example, he Unied Kingdom was a ne capial imporer during his period, bu is ne exernal posiion improved by 8.5 percen of GDP; Sweden insead was a ne capial exporer, bu is ne posiion deerioraed subsanially. 4 B. Gross flows and gross posiions, indusrial counries Figure 2 summarizes he evoluion of gross exernal asses and liabiliies in indusrial counries during he pas 20 years. The growh in inernaional financial inerdependence is sriking: during his period, aggregae asses and liabiliies ripled as a share of GDP, asses and liabiliies o GDP increased four-fold, porfolio equiy asses and liabiliies six-fold, and deb asses and liabiliies o GDP 2 ½ imes. Focusing on he mos recen period, he char also shows he effecs of he global decline in sock marke valuaions beween end-1999 and end-2002, which is he main facor behind he reducion in he sock of porfolio equiy asses and liabiliies. 5 4 The main reason for Sweden s deerioraion in ne foreign asses is he decline in value for he large holdings of Swedish direc invesmen abroad, when measured a marke value. If Sweden s FDI is measured a book value, is ne exernal posiion improved beween 1998 and Only a few counries in our sample (primarily he Unied Saes) measure FDI a marke value: hence, sock marke flucuaions have a less dramaic impac on FDI holdings compared o porfolio equiy holdings. Of course, alhough foreign porfolio equiy asses and liabiliies fell relaive o GDP, his does no imply a decline relaive o he size of domesic equiy holdings.

5 - 4 - Table 2 summarizes gross capial flows o and from indusrial counries during he period The size of gross flows is remarkable, paricularly o and from financial ceners such as Swizerland and he Unied Kingdom, bu also o and from he euro area. While here are subsanial ne flows, he daa sugges ha porfolio diversificaion, raher han ineremporal borrowing and lending, is he dominan moive for inernaional asse ransacions among indusrial counries. C. Ne flows and ne posiions, emerging markes We focus on a sample of 21 emerging markes (lised in he Appendix). Figure 3 plos he evoluion of he average curren accoun balance as a raio of GDP in our emerging markes sample. The key cycles in capial flows o emerging markes sand ou clearly from his picure: he deerioraion of curren accoun imbalances in he lae 1970s unil he deb crisis, heir sharp reversal during he remainder of he 1980s, he increase in imbalances during he early 1990s, and he new reversal following he Mexican crisis and especially he Asian crisis. The dynamics of he ne exernal posiion, expressed as a raio of GDP, is ploed in Figure 4. I shows he deerioraion caused by he deb crisis and is afermah, a subsequen sharp improvemen, he sabilizaion of he ne exernal posiion from 1990 unil he Asian crisis, he deerioraion caused by he sharp declines in GDP and real exchange depreciaion, and he subsequen improvemen associaed wih curren accoun surpluses and srenghening currencies in Asia. In he daa, here is no evidence of an increased dispersion in curren accoun balances across he counries in our sample (ha is, here is a significan common rend in he ne capial flows o his emerging marke group), while he dispersion of he underlying ne exernal posiions has increased. Figure 5 plos he ne foreign asse posiion, scaled by GDP, in relaion o GDP per capia in curren US dollars a end While here is sill a posiive relaion beween ne foreign asses and GDP per capia, his relaion is much weaker han for indusrial counries. Indeed, crediors include economies wih high GDP per capia, such as Taiwan province of China, bu also economies wih much lower GDP per capia, such as Russia and Venezuela. 6 We urn now o he evoluion of capial flows o individual counries during he mos recen period. Table 3 characerizes he size of ne capial flows among some counries in our sample, boh in absolue erms and as a raio of GDP, during he period The able shows a number of Lain American and Cenral European counries as he larges ne 6 In addiion o he sandard macroeconomic drivers of ne foreign asse posiions ha were emphasized by Lane and Milesi-Ferrei (2002a), poliical risk and naural resource endowmens are oher variables ha may be imporan, especially in reference o he counries lised here.

6 - 5 - recipiens of ne capial flows, while Asian counries, ogeher wih Russia, have been on average ne capial exporers. D. Gross flows and gross posiions, emerging markes Figure 6 provides a longer-erm perspecive on he size of exernal asses and liabiliies in hese emerging markes. Boh asses and liabiliies have increased subsanially as a raio of GDP during he pas 20 years. However, here is virually no increase in average exernal liabiliies when hey are scaled by expors, raher han GDP, while he rend increase in exernal asses is sill visible. This sands in conras wih he evidence for he advanced economies, where he increase, especially since he mid-1990s, is very srong even as a share of expors. As noed earlier, an ineresing quesion is wheher he composiion of exernal asses and liabiliies has changed over ime. Table 4 provides evidence ha highlighs he increased relaive imporance of direc invesmen and porfolio equiy liabiliies. The averages hide a large degree of heerogeneiy counries such as Chile and he Cenral European economies in our sample (Czech Republic, Hungary, and Poland) have exernal equiy liabiliies above 50 percen of GDP, while he levels end o be lower in Asian economies. The Table also documens he increase in foreign exchange reserves, expressed as a share of GDP, during he pas 20 years. I should be noed, however, ha his increase has gone hand in hand wih he increase in oher exernal asses, so ha a he end of 2002 reserves acoun for he same share of oal exernal asses as in 1982 (abou one hird). Direc invesmen and porfolio equiy asses have also increased during he pas wo decades, and represened around 10 percen of GDP and 1/6 of oal exernal asses. Table 5 characerizes gross capial flows o and from some emerging economies during he period The paern reveals an ineresing dichoomy. For a number of counries, gross flows primarily reflec ineremporal borrowing or lending decisions, wih counries accumulaing ne asses or ne liabiliies eiher cumulaive inflows or cumulaive ouflows are clearly dominan (see also Table 3). Among counries ha accumulaed subsanial ne asses, a number of Eas Asian counries sand ou, paricularly Indonesia, Korea, Malaysia, and Thailand, ogeher wih oil-exporing counries such as Russia. For anoher group of counries, insead, gross inflows and gross ouflows have been large and similar in magniude, reflecing increased financial inegraion wih he world economy. Examples include Chile and China. 7 7 Of course, similar aggregae levels of gross inflows and gross ouflows can conceal significan ne imbalances wihin specific asse caegories: for insance, China is a major ne recipien of FDI flows while simulaneously accumulaing a significan volume of foreign reserves.

7 - 6 - Having provided a broad characerizaion of he growh in inernaional balance shees in recen years for boh advanced and emerging economies, we nex urn o providing a simple framework for undersanding he underlying drivers III. EXTERNAL ASSET DYNAMICS In his secion, we provide a simple accouning framework ha relaes he dynamics of ne foreign asse posiions o rade flows, growh, raes of reurn, and real exchange raes. The goal is o lay ou he various channels by which exchange raes and oher macroeconomic fundamenals can poenially affec he exernal adjusmen process. We hen provide a decomposiion of he facors underlying changes in ne foreign asses over he pas decade for a se of emerging markes. A. Accouning for exernal asse dynamics The change in he ne foreign asse posiion B can be wrien as follows: B B = CA + KA + EO + KG (1) 1 where B is he ne foreign asse posiion, CA is he curren accoun balance, KA is he capial accoun balance, EO are errors and omissions, and KG is he ne capial gain (change in sock minus underlying flow). In line wih saisical reporing pracices, all variables are expressed in US dollars. Equaion (1) can also be expressed as follows: B B = BGST + ( KA + EO ) + ( i A i L + KG ) (2) A L where BGST is he balance of rade in goods and services plus ne ransfers, A and L are exernal asses and liabiliies, respecively, and A L i, i are he yields on hese asses and liabiliies. Taking raios of GDP and indicaing such raios wih lower-case leers, we can express (2) as follows: A L i A 1 i L 1+ KG γ b b 1 = bgs + ( ka + eo) + b $ 1 (3) Y 1 + γ where γ is he growh rae of nominal GDP measured in US dollars and bgs indicaes he raio o GDP of he sum of rade balance, errors and omissions, and capial accoun. Anoher way o re-wrie he above expression is as follows: b b bgs i A i L + KG ka eo π d b g b A L = + + ( ) $ Y (1 + g)(1 + π ) 1+ g (4)

8 - 7 - where g is he economy s real growh rae, π is he rae of inflaion (measured wih he GDP deflaor), and d is he rae of nominal exchange rae depreciaion vis-à-vis he US dollar. In oher words, changes in he ne exernal posiion can be due o several facors: 1. he curren accoun; 2. he capial accoun and ne errors and omissions; 3. ne capial gains (measured in US dollars); 4. he effec of exchange rae changes on he pas ne foreign asse posiion; 8 5. he effec of real GDP growh on he pas ne foreign asse posiion. An alernaive way o express equaion (4) is in erms of overall raes of reurn on exernal A L asses and liabiliies. Define k ( k ) as he rae of capial gain on exernal asses (liabiliies), A A A L A 1+ i + k measured in US dollars, so ha k FA 1 k FL 1 = KG, and le rˆ = be he real US 1+ π rae of reurn on foreign asses, measured in US dollars, wih an analogous definiion holding for he rae of reurn on foreign liabiliies r ˆL. In his case we can re-wrie (4) as follows: rˆ g ε (1 + g ) rˆ rˆ b b bgs ka eo b a L A L 1 = + ( + ) (1 + g)(1 + ε) (1 + g)(1 + ε) (5) Equaion (5) shows several facors ha can accoun for he dynamics of ne foreign asses: he adjused rade balance, he difference beween he real ineres rae and he growh rae, adjused for he bilaeral real exchange rae vis-à-vis he US dollar, and differences in reurns beween foreign asses and liabiliies. A L If we express he real raes of reurn in domesic currency and denoe hem by r, r, equaion (5) akes he more familiar form: L A L r g r r b b 1 = bgs + ( ka + eo) + b 1+ a 1 1+ g 1+ g (6) 8 If exernal asses and liabiliies are all denominaed in domesic currency, he capial gain effec will go exacly in he opposie direcion from he exchange rae change effec. Indeed, assume for simpliciy ha asse prices in domesic currency do no change. In his case, he capial gain expressed in domesic currency is zero, bu expressed in dollars i will be given KG d by = b $ 1, which is similar (wih an opposie sign) o he erm Y (1 + π )(1 + g) π d b 1 in equaion (4). (1 + g )(1 + π )

9 - 8 - This framework delivers several imporan insighs. Firs, he gap beween curren producion and curren absorpion (i.e. he rade balance) is only one facor in deermining he aggregae evoluion of he ne foreign asse posiion: i is vial o also keep rack of valuaion and denominaor effecs. Second, as is shown by he hird erm on he righ hand side (RHS) of equaion (5), he inheried ne foreign asse posiion exers a poenially powerful influence on is curren dynamics. Third, as capured by he las erm on he RHS in equaion (5), he gross scale of he inernaional balance shee maers in addiion o he ne posiion: even if he inheried ne foreign asse posiion is zero, he accumulaed levels of gross foreign asses and liabiliies will influence he overall dynamics o he exen ha he raes of reurn differ beween he wo sides of he inernaional balance shee. B. The evoluion of ne foreign asses in emerging markes In Table 6 we provide a simple preliminary decomposiion of changes in he raio of ne foreign asses o GDP beween end-1990 and end-2002 for a selecion of emerging markes in our sample. The breakdown follows equaion (4), so ha changes in he ne foreign asse posiion are given by he sum of he curren accoun (iself divided ino rade balance ands invesmen income), capial accoun and errors and omissions, capial gains (including he effecs of exchange rae changes on he ne exernal posiion), and he effecs of growh on ne exernal asses. A number of feaures are worh highlighing: a. despie a cumulaive curren accoun in balance or surplus, counries such as Indonesia and Thailand experienced a deerioraion in he raio of ne foreign asses o GDP. For boh counries, his occurred because of capial losses linked o he real depreciaion of heir currencies during he period. b. On he oher side, Hungary s and Mexico s exernal posiion deerioraed by much less han he large cumulaive curren accoun deficis would sugges, hanks o subsanial capial gains on heir ne exernal posiion linked o he real appreciaion of he forin and he peso beween end-1990 and end In summary, he Tables highlighs he need o focus no only on he curren accoun (which includes he yield on exernal asses and liabiliies), bu also on economic growh and he overall raes of reurn on he exernal porfolio in order o undersand he evoluion of he ne exernal posiion. IV. EXCHANGE RATES AND THE ADJUSTMENT PROCESS The framework summarized in equaion (5) highlighs he poenial conribuion of shifs in exchange raes in deermining he dynamics of exernal asse posiions. In his secion, we firs briefly review he radiional channels by which exchange raes influence he adjusmen process, before focusing on he valuaion channel (i.e. he impac of he

10 - 9 - exchange rae on he raes of reurn earned on he accumulaed holdings of foreign asses and liabiliies). A. Exchange Raes, he Trade Balance and Real Oupu The iner-connecion beween he exchange rae and he rade balance is among he mossudied quesions in inernaional economics, in boh academic and policy circles. From a long-run perspecive, he classical ransfer problem posulaes ha persisen credior naions should have more appreciaed real exchange raes. The mechanism underlying he ransfer problem hypohesis is ha he posiive inernaional invesmen reurns earned by long-run crediors have heir counerpar in rade deficis and aendan real appreciaion. Lane and Milesi-Ferrei (2002b, 2004) find considerable empirical suppor for he ransfer problem, for boh indusrial and developing counries. However, hey find he magniude of he effec differs wih counry characerisics such as openness, size and he level of developmen. In relaion o financial globalizaion, imporan findings are ha he ransfer problem is smaller in he absence of curren and capial accoun resricions and ha equiy financing reduces he size of he ransfer effec relaive o deb financing. A a shorer horizon, he inerplay beween he exchange rae and he rade balance is complex and less well undersood. In paricular, he cyclical correlaion beween he variables will depend on he naure of he shocks hiing he economy, wih nominal, fiscal and real shocks generaing differen co-movemen paerns beween he variables. However, in policy erms, here is wide consensus ha exchange rae depreciaion is ypically required if he objecive is o engineer an improvemen in he rade balance. Empirical sudies of he elasiciies of rade volumes o exchange raes and income levels provide exensive suppor for his proposiion (Hooper e al 2000). Again, he pace of financial globalizaion and real globalizaion (in erms of inegraion of produc markes) will influence hese key elasiciies. For insance, he scale of exchange rae adjusmen is eased, as foreign goods become beer subsiues for domesic goods. In erms of financial globalizaion, wider rade imbalances are more feasible, he more diversified are inernaional porfolios. In racking he dynamics of he raio of ne foreign asses o GDP, real exchange raes also operae by deermining he real value of domesic oupu in erms of inernaional price comparisons. For insance, if variables are measured in US dollars, a foreign asse ha is consan in real dollar erms will shrink relaive o he consan-dollar value of GDP if real appreciaion vis-à-vis he US dollar occurs. This denominaor effec is highlighed in equaion (8) in he previous secion and is powerful channel by which he real exchange rae may influence he dynamics of he NFA/GDP raio.

11 However, in addiion o hese well-known channels, exchange raes also poenially influence he dynamics of inernaional asse holdings hrough influencing he raes of reurn on foreign asses and liabiliies. 9 We focus on his valuaion channel in he res of his secion. B. The Valuaion Channel: A Concepual Framework As oulined earlier in he paper, he dynamics of he ne foreign asse posiion depends no only on he evoluion of he rade balance bu also on he raes of reurn earned on accumulaed foreign asses and paid ou on accumulaed foreign liabiliies. In domesiccurrency real erms, he ne impac is given by RET = r A r L (7) A L 1 1 where are he inheried socks of foreign asses and liabiliies A and L are expressed in domesic currency. Since hese accumulaed socks are predeermined from a ime- perspecive, he ne valuaion impac of a change in he real exchange rae is given by RET r r RER RER RER A L = A 1 L 1 (8) I is clear from his expression ha an exchange rae movemen can have a non-zero valuaion impac even if he iniial ne foreign asse posiion is balanced, so long as he raes of reurn on foreign asses and liabiliies are differenially affeced by a shif in he exchange rae. 10 The magniude of he valuaion channel is direcly increasing in he gross scale of he inernaional balance shee: he relevance of his channel for aggregae ne foreign asse posiion dynamics is growing in line wih he specacular accumulaion of gross foreign asse and liabiliy holdings in recen years. Relaedly, he valuaion channel also depends on he composiion of he inernaional balance shee, since he sensiiviy of reurns o exchange raes will vary across invesmen caegories and will also depend on he currency composiion of foreign asses and liabiliies (and on he exen of hedging). 9 Clearly, in racking a raio, here is some discreion in erms of aribuing he impac of an exchange rae change o he numeraor or he denominaor via he choice of he reference currency. In he nex subsecion, we look a he levels of foreign asses and liabiliies in erms of real domesic currency. 10 Sricly speaking, he impac on he reurns on foreign asses and liabiliies is no he only valuaion effec of exchange rae changes. As highlighed by he debae over he Marshall- Lerner condiion and re-emphasized by he curren debae abou limied exchange rae passhrough, exchange rae movemens also exer a pure valuaion effec on he rade balance o he exen ha impor and expor volumes are unresponsive o exchange rae changes.

12 Of course, even if he exchange rae does indeed have a valuaion impac, i does no mean ha he ne foreign asse posiion will be changed one-for-one. Firs, movemens in he exchange rae also have a direc impac on he rade balance. Second, a valuaion gain represens a posiive wealh effec ha will plausibly raise consumpion and invesmen, leading o a negaive co-movemen beween he ne reurns erm and he rade balance (Lane and Milesi-Ferrei 2002a, 2002b). From anoher angle, a sufficienly-large negaive valuaion effec may lead o a sudden sop in capial flows ha forces he rade balance o move ino surplus. In addiion, equaion (8) only capures he conemporaneous impac of a change in he exchange rae. Some reurn series may respond o he exchange rae only wih a lag (for insance, he profiabiliy of FDI posiions may by dynamically affeced by curren exchange rae movemens). In addiion, curren exchange rae movemens may lead o a revision of expecaions abou fuure exchange rae changes, which in urn feed ino he ex-ane reurns required o hold paricular foreign asse and liabiliy posiions. As was discussed earlier in he paper, here are polar cases in which he impac of exchange rae movemens on raes of reurn is sraighforward. For insance, he domesic rae of reurn on an unhedged foreign asse ha offers a fixed foreign-currency reurn will fall one-for-one wih he rae of real appreciaion: a given foreign-currency reurn will be diminished by he fall in he real domesic value of foreign currency. Conversely, he domesic rae of reurn on a foreign liabiliy ha offers a fixed domesic-currency reurn will be unaffeced by a shif in he real exchange rae. However, he domesic rae of reurn on a foreign liabiliy ha offers a fixed foreign-currency reurn (e.g. foreign currency deb or domesic deb ha offers a dollarlinked rae of reurn) will also fall in proporion o he rae of real appreciaion. More generally, he ne impac of exchange rae movemens for he value of holdings ha carry a variable marke reurn depends on he naure of he comovemens beween exchange raes, asse prices and profiabiliy (in he case of non-marke asses such as FDI posiions and some bank claims). In some cases, he iner-connecions beween exchange raes and he deerminans of marke reurns can be quie suble and complex and may also depend on he underlying source of an exchange rae shock. For insance, devaluaion may be associaed wih an increase in he rae of reurn on foreign liabiliies if i is associaed wih an increase in he profiabiliy of foreign affiliaes operaing in he domesic marke or, alernaively, if i engenders an increase in he counry risk premium. On he oher hand, a devaluaion may be generaed by a negaive domesic produciviy shock ha also lowers he reurn earned by foreign invesors. Wih respec o foreign asses, domesic real depreciaion may be he resul of superior overseas economic performance ha raises he overseas rae of reurn. However, a negaive domesic produciviy shock may also reduce he overseas earnings of domesic mulinaionals, such ha devaluaion is accompanied by a decline in he overseas rae of reurn. In view of he range of possible heoreical scenarios, he srengh of he valuaion channel is

13 ulimaely an empirical issue. Accordingly, we urn o quaniaive exploraion in he nex secion. C. Regression Analysis, Indusrial Counries In his secion, we analyze he sensiiviy of raes of reurn on foreign asses and liabiliies o movemens in rade-weighed mulilaeral exchange raes. In addiion o he aggregae posiions, we also examine he raes of reurn for he separae invesmen caegories (FDI; porfolio equiy; porfolio deb; and oher (deb)), since he relaion beween exchange rae movemens and raes of reurn should depend on he specific characerisics of each invesmen class. Our specificaion is given by r = α + βdlrer + u (9) A i i i where he dependen variable is he real domesic-currency reurn on foreign asses in invesmen caegory i and he regressor is he log change in he rade-weighed real effecive exchange rae. 11 We run an analogous equaion for he rae of reurn on foreign liabiliies.) Clearly, his is a very parsimonious seup. However, in addiion o being suied o our shor daa span, capuring he simple bivariae relaion is an obvious firs sep, even if i does no rule ou he possibiliy ha any impac of he exchange rae on he rae of reurn may jus be proxying for he role played by some omied variable ha commonly influences boh he rae of reurn and he exchange rae. Moreover, for he ime being, we jus run leas squares regressions, raher han rying o conrol for any endogenous response of he exchange rae o shifs in he rae of reurn. (Indeed, we will allow for he possibiliy of such reverse causaion when discussing he resuls.) Raher, our primary ineres is jus in esablishing he direcion and magniude of he co-movemen beween he exchange rae and raes of reurn. Finally, we do no aemp o disinguish beween anicipaed and unanicipaed changes in he real exchange rae: however, real exchange raes are largely unpredicable a an annual horizon (a leas for our sample of advanced counries), such ha his may be a fairly-innocuous assumpion The rae of reurn on foreign asses in year is measured as he sum of invesmen income and capial gains earned in ha year, divided by he sock of foreign asses a he end of year In he discussion of he resuls, we do reurn o he issue of he predicabiliy of exchange rae movemens.

14 We begin by examining he raes of reurn on foreign asses in Table The resuls for oal foreign asses are given in column (1). In all cases, he esimaed coefficien is negaive: real appreciaion is associaed wih a fall in he domesic-currency rae of reurn earned on foreign asses, wih he fixed-effecs panel poin esimae being For a number of counries, he esimaed coefficien is in fac very close o -1 (UK, Germany, Ialy, Swizerland and Finland). For hese counries, his one-o-one mapping is consisen wih a process by which he foreign-currency real reurn on foreign asses is orhogonally deermined and he exchange rae jus acs o conver he foreign-currency reurn ino domesic erms. The smalles esimaed coefficien (in absolue value) in he sample is for he US a This admis a number of inerpreaions. Firs, some proporion of US foreign asses is denominaed in US dollars and hence heir value is no direcly affeced by exchange rae movemens. Second, dollar appreciaion could be associaed wih an increase in he rae of reurn on foreign-currency foreign asses. One example would be a US posiive produciviy shock ha boh appreciaes he dollar and raises reurns in US financial markes. If foreign financial markes posiively co-move wih he US, foreign-currency asse reurns would also rise a he same ime. Wih respec o FDI, a posiive US produciviy shock could also raise he profis earned overseas by US mulinaionals, such ha foreign-currency reurn rises in ha case as well. In some cases, we observe a coefficien above uniy, which means ha real appreciaion is associaed wih a fall in he foreign-currency rae of reurn on foreign asses: he domesic invesor loses wice by suffering boh a low foreign-currency reurn and an unfavorable conversion rae back ino domesic real erms. Such a paern could be generaed, for insance, if he domesic business cycle is asymmeric wih respec o he inernaional business cycle: he domesic currency appreciaes when inernaional parners are doing badly (as proxied by poor foreign-currency raes of reurn). This, of course, is a riskleveraging paern of co-movemens wih foreign-currency reurns and he domesic real exchange rae. 13 In erms of counry selecion for he regressions, we only include hose wih a leas hireen years of daa on raes of reurn. In addiion, we rule ou observaions ha may be conaminaed by facors such as revisions in mehodology and oher correcions.

15 The resuls for FDI asses are given in column (2). 14 For mos counries, FDI posiions are sill measured a book value, raher han marke value, and herefore he valuaion channel is ypically undersaed. However, currency movemens should sill maer, since hese would affec facors such as he curren replacemen cos of capial goods and fixures, boh domesically and overseas. In column (2), he fixed-effecs panel esimae of he impac of real appreciaion on he real reurn on FDI foreign asses is -0.76: since he pass hrough is less han proporional, real appreciaion episodes on average coincide wih periods of improved foreign-currency real reurns. However, here are some cases in which he coefficien is subsanially above uniy: for hese counries, real appreciaion ends o be associaed wih low foreign real reurns on FDI asses. We nex urn o he reurns on porfolio equiy asses in column (3). The fixed-effecs panel esimae is very close o -1, which indicaes a profile in which he conribuions of exchange rae movemens and foreign-currency raes of reurn o he domesic-currency real rae of reurn are orhogonal. Two excepions o his rule are provided by Germany and Swizerland ha display significan coefficiens of (-2.49,-2.43) respecively: for hese counries, real appreciaions have coincided on average wih periods of srongly negaive world sock marke reurns, and hence disappoining foreign-currency reurns on heir equiy porfolios. Column (4) displays he resuls for foreign asses in he porfolio deb caegory. The explanaory power of he exchange rae in explaining domesic-currency reurns is ypically quie good and he fixed-effecs panel coefficien esimae of is fairly represenaive. This paern is consisen wih he foreign-currency reurns on foreign porfolio deb asses exogenously deermined wih respec o he domesic real exchange rae. However, here are excepions: for insance, he Unied Saes coefficien is only -0.65, consisen wih he fac ha a considerable proporion of is foreign bond holdings are denominaed in US dollars. Finally, we urn o he oher invesmen caegory in column (5). This caegory largely comprises bank lending. Since banks do no mark o marke all asses and liabiliies bu raher carry a high proporion a book value, he raes of reurns in his caegory will be dominaed by he yield componen, wih capial gains and losses undersaed. However, on he asses side, he broad picure is quie similar o ha for porfolio deb. An imporan excepion is he US, where he coefficien esimae is an insignifican -0.11: again, a good candidae explanaion is ha a high proporion of is foreign lending is in US dollars. 14 In he sample represened in his able, only he Neherlands and Ausralia record FDI a marke value. The US repors posiions measured a boh book and marke value. For comparabiliy wih he counries ha only repor book values, he US esimaes in hese ables refer only o he book value measure of FDI. However, for he US, if we use he rae of reurn based on FDI a marke value hen he exchange rae coefficien in he FDI asse equaion is (-sa 1.38) and i is 0.15 (-sa 0.37) in he FDI liabiliy equaion.

16 We urn o he raes of reurn on foreign liabiliies in Table 8. In erms of he resuls for oal foreign liabiliies in column (1), we see quie a mixed paern in erms of he esimaed exchange rae coefficiens across counries. For he Unied Saes, he rae of reurn paid ou on foreign liabiliies is oally unaffeced by movemens in he real exchange rae consisen wih he fac ha foreign liabiliies are almos enirely dollar-denominaed and offer reurns ha no linked o exchange rae flucuaions (e.g. bank deposis or fixed-ineres deb insrumens). A he oher exreme, he esimaed coefficien for Finland is -1.8 (albei significan only a he 10 percen level): a 10 percen real appreciaion is associaed wih a 18 percenage poin decrease in he real reurn paid ou on foreign liabiliies, as measured in domesic real erms. Alhough a paern of high real reurns plus exchange rae depreciaion is also eviden during he early 1990s, he mos sriking period for Finland is he pos-emu period: 1999 saw a very large rae of reurn paid ou on is foreign liabiliies (driven by gains in Nokia s share price during he equiy marke boom), while is real exchange rae depreciaed (on accoun of he fall in he exernal value of he euro). In conras, he sock marke reversals of were accompanied by an appreciaing real exchange rae (wih he recovery in he exernal value of he euro). This case is a vivid illusraion of he imporance of undersanding he naure of co-movemens beween exchange raes and asse prices. Moreover, i also underlines he fac ha exchange raes need no always move in a risksharing manner, which applies a foriori for members of a currency union ha have lile influence on he exernal value of he currency. For he oher counries in he sample, he esimaed coefficiens generally are in he (-1,0) range, wih he fixed-effecs panel esimae a In a more aenuaed form han in he Finnish case, his generally indicaes ha exchange rae appreciaion is associaed wih some deerioraion in domesic real reurns on foreign liabiliies. For a couple of counries, he coefficien is quie close o uniy: in approximae erms, i seems for hese counries ha foreign liabiliies can be viewed as offering a arge reurn in foreign currency, wih exchange rae movemens jus proporionaely shifing he domesic currency counerpar. In no case is he esimaed coefficien significanly posiive. This is quie surprising, since some of he mechanisms discussed earlier in order o explain a negaive relaion beween exchange rae appreciaion and he rae of reurn on foreign asses should symmerically imply a posiive associaion beween exchange rae appreciaion and he rae of reurn paid ou on foreign liabiliies. For insance, a posiive domesic produciviy shock migh be expeced o raise he profiabiliy of foreign affiliaes operaing in he domesic marke and generally boos domesic asse prices, while a he same ime generaing real appreciaion. Wih respec o FDI liabiliies, column (2) of Table 8 suggess ha, wih he marginal excepion of he Neherlands, exchange rae flucuaions appear o exer no influence on he real domesic reurns. In par, his may be aribued o he fac ha FDI posiions are mosly measured a book value bu he insignificance of he exchange rae also suggess ha he earnings of foreign affiliaes in he domesic marke are no (conemporaneously) affeced by

17 exchange rae swings. This paern is worh exploring furher bu would require he availabiliy of higher-qualiy daa. Similar o he case for FDI liabiliies, mos of he esimaed counry coefficiens for porfolio equiy liabiliies are insignificanly differen from zero: he domesic-currency real reurn offered by porfolio equiy liabiliies is no sysemaically affeced by he exchange rae. Again, his is somewha surprising o he exen ha we migh expec domesic sock marke booms o be associaed wih real appreciaion. Indeed, here is only one significan counry coefficien (Germany), bu i is negaive and large (-2.9). This means ha declines in he German sock marke are ypically associaed wih real appreciaion. Since he poin coefficien is fairly similar for boh asses and liabiliies, he correc inerpreaion in fac is ha German real appreciaion ends o occur during phases of disappoining global sock reurns, since he reurns on German overseas asses fall in addiion o he reurns paid ou on domesic socks owned by foreign invesors. On he liabiliy side, only Canada and Ausralia show a significan connecion beween exchange rae movemens and he rae of reurn paid ou on foreign bond liabiliies. For he ohers, he resuls suppor he caricaure of bond liabiliies ha offer a domesic rae of reurn ha is invarian o exchange rae flucuaions. Even for Canada and Ausralia, he paern of co movemen is negaive: real appreciaion is associaed wih low domesic raes of reurn on heir foreign bond liabiliies. In par, his may sugges ha exchange rae movemens for hese counries have a subsanial predicable componen, since foreign invesors would be prepared o accep a low domesic-currency reurn if real appreciaion were anicipaed. The predicabiliy hypohesis receives some suppor from he empirical work of Chen and Rogoff (2003), who show ha commodiy currencies (such as he Ausralian and Canadian dollars) are more predicable han oher currencies. Of course, anoher poenial conribuory facor is he exen o which hese counries issue bond liabiliies in foreign currency. On he liabiliy side, a number of counries display significanly negaive coefficiens, wih he esimaes far above uniy for Ausralia and Spain. For his pair, he paern is akin o ha experienced by emerging markes: real depreciaion raises he real reurn paid o foreign invesors more han proporionaely, wih he foreign-currency reurn on ne increasing. In summary, he regression analysis in Tables 7-8 delivers a number of ineresing lessons. Firs, especially on he foreign asses side, exchange rae movemens are an imporan deerminan of raes of reurn: he valuaion channel is powerful. Second, real appreciaion is ypically associaed no only wih lower real reurns on foreign asses, bu also lower real reurns on foreign liabiliies: a leas for small ne posiions, his implies ha he ne valuaion impac of exchange rae movemens on he ne foreign asse posiion has been limied. Third, he sensiiviy of reurns o exchange raes does vary across invesmen caegories: he composiion of he inernaional balance shee is an imporan deerminan of he aggregae valuaion effec. Fourh, he Unied Saes behaves quie differenly o oher counries in ha he raes of reurn on is liabiliies (in all invesmen caegories) are unaffeced by currency movemens. Since dollar depreciaion raises he reurn on is foreign asses, his means ha he valuaion channel in he US case may indeed be a powerful adjusmen mechanism in

18 correcing is large exernal liabiliy posiion. We reurn o he feasibiliy of his opion laer in his paper. D. Exchange raes and raes of reurn, emerging markes In general, we would expec he relaion beween domesic-currency raes of reurn and changes in he real exchange rae o be even sronger for emerging markes, which in general have less scope for borrowing or lending in domesic currency. Careful empirical work has o face he severe difficulies in measuring such raes of reurn: among hese, he lack of precise hisorical daa on inernaional invesmen posiions; sock-flow discrepancies; deb reducion and deb forgiveness agreemens, and defaul episodes. While aware of hese limiaions, we have consruced rough esimaes of raes of reurn on exernal asses and liabiliies for our emerging-marke sample. The mehodology is based on esimaing he sock of exernal asses and liabiliies (making use of flow daa wih valuaion adjusmens, as in Lane and Milesi-Ferrei, 2001), and using daa on ineres paymens and capial flows o back ou raes of reurn. A simple panel regression wih fixed effecs of real domesic-currency raes of reurn on exernal liabiliies on changes in he real effecive exchange rae gives a coefficien of wih a -saisic of As Figure 7 shows, his relaion holds no only along he ime-series dimension, bu also in he cross-secion. The Figure shows a srong negaive relaion (ploed for he year 1997, a year of large exchange rae depreciaions in he Asian counries of our sample) beween he domesic currency rae of reurn on exernal liabiliies and he real effecive exchange rae. E. Case Sudies: he US and Ausralia I is possible o gain exra insigh ino he quaniaive imporance of he valuaion impac of exchange rae counries for hose counries ha repor he accouning decomposiion abou he relaive imporance of capial flows, marke value capial gains and exchange rae capial gains in deermining he dynamics of foreign asse and liabiliy posiions. This is possible o for wo counries in our sample (he US and Ausralia). Of course, an accouning decomposiion does no reveal he complee conribuion of he exchange rae valuaion channel, since i does no ake ino accoun he poenial indirec impac of he exchange rae on marke values or on he revaluaion of invesmen income flows. Tables 8 and 9 presen he decomposiions for he US and Ausralia respecively. The ables show he average annual relaive conribuions of each componen in proporion o he inheried socks of foreign asses (liabiliies): 15 A similar regression for asses gives a coefficien of -1, consisen wih he fac ha foreign asses are enirely foreign-currency denominaed.

19 CON _ FLOW CON _ MVFA CON _ ER FA FA CAPFLOW = FA KG _ MV = FA KG _ ER = FA FA FA FA In addiion, he ables display he sandard deviaions for hese componens. For boh he Unied Saes and Ausralia, he ranking of he hree componens in erms of heir relaive imporance is clear for boh asse and liabiliy dynamics: capial flows are he primary driving facor, followed by he conribuion of marke value gains, wih he capial gains offered by exchange rae movemens leas imporan. For he Unied Saes, he average conribuion of he exchange rae channel is close o zero over However, on he foreign asse side, i grew in imporance during relaive o During he more recen sub-period, he dollar was in an appreciaion phase, such ha he exchange rae aced o reduce he value of accumulaed foreign asses. Of course, should record a marked posiive conribuion of he exchange rae channel o he value of US foreign asses, in line wih he recen dollar depreciaion. In accordance wih our earlier resuls, he impac of he exchange rae valuaion channel on US foreign liabiliies remains close o zero hroughou. Anoher noeworhy feaure of he US daa is ha marke-value capial gains are a subsanial facor, reaching nearly 50 percen of he conribuion of capial flows o foreign asse accumulaion. Marke value gains are also imporan on he liabiliy side, albei a smaller fracion of he capial flow conribuion. 16 In erms of he relaive sabiliy of he various componens, capial flows have been much less volaile han eiher of he capial gain erms. In paricular, he sandard deviaion has been more han wice as large as he (absolue) value of he average for he marke-value capial gain erm. Turning now o he Ausralian evidence, Table 9 shows ha he exchange rae valuaion componen has been much more imporan han in he US case. Especially during he period, he exchange rae erm has been nearly imporan as he marke value erm in driving accumulaion dynamics: moreover, on he asses side, he combined capial gain has exceeded he conribuion of capial ouflows. During his period: he exchange rae has aced asymmerically, wih he capial gain on foreign asses much larger han on foreign liabiliies. 16 The difference in he raio of marke value gains o capial flows beween he asse and liabiliy sides is he counerpar o he large excess of capial inflows over capial ouflows, which has acceleraed since he mid-1990s.

20 This reflecs he subsanial real depreciaion of he Ausralian dollar during his period. In relaion o his period, many commenaors have commened on he success of Ausralia in avoiding he wors impac of he Asian crisis: he daa show ha he depreciaion sraegy operaed no jus hrough is impac on he rade balance and capial flows bu also hrough a favorable valuaion effec. Finally, we noe ha he exchange rae componen for Ausralia has been much more volaile han for he US. These case sudies have provided suggesive evidence abou he imporance of he valuaion channel in driving flucuaions in inernaional asse holdings. However, he daa only permi us o examine a relaively shor ime period: anoher elemen o his quesion is o invesigae he sources of persisen shifs in ne foreign asse posiions. For insance, he revaluaion effecs induced by currency and asse price movemens may generae much volailiy in he value of exernal asses and liabiliies bu i is an open quesion as o how imporan are revaluaion effecs versus curren accoun imbalances in driving he lower-frequency componen of ne foreign asse dynamics. On he agenda for fuure research is a beer characerizaion of he relaion beween exchange rae movemens and he ne foreign asse posiions over ime: for insance, he impac effec may primarily operae hrough he revaluaion effec, bu wih also a longer-erm conribuion via gradual adjusmen in he curren accoun o a susained real exchange rae movemen. F. The Valuaion Channel in Inernaional Macroeconomic Models The sandard approach in boh he radiional Mundell-Fleming approach and conemporary new open economy macroeconomics is o consider scenarios in which he iniial ne foreign asse posiion is zero. This obviously rules ou any consideraion of he valuaion channel in erms of macroeconomic behavior and he analysis of alernaive policies. However, wo imporan recen excepions are provided by Benigno (2001) and Tille (2004). In a wo-counry model, Benigno (2001) shows ha moneary shocks have much larger real effecs if he iniial global seady sae is characerized by ne foreign asse posiion imbalances, since exchange rae movemens generae a ne valuaion effec ha has obviously asymmeric effecs on he home and foreign counries. One implicaion is ha counries will disagree abou he opimal policy, since he valuaion channel acs o ransfer wealh beween home and foreign ciizens. Tille (2004) considers he case of iniially-balanced ne foreign posiions bu allows for differen levels of scale in erms of gross holdings of foreign asses and liabiliies. Maching he US daa, he shows ha in he case ha he foreign-currency share of foreign asses is larger han he foreign-currency share of foreign liabiliies, an increase in gross crossholdings of domesic-currency and foreign-currency bonds means ha he welfare impac of a surprise moneary expansion is grealy magnified. Indeed, his calibraion suggess ha he welfare impac of he valuaion channel is 350 percen more powerful han he radiional channel, since devaluaion confers a sizeable capial gain on he home counry in his seup.