CURRENT ACCOUNT MAURICE OBSTFELD University of California, Berkeley and KENNETH ROGOFF* Princeton University Cont ent s 1. I nt r oduct i on 2. The cur r ent account : Basi c concept s and hi st ori cal over vi ew 3. I nt e r t e mpor a l appr oaches t o t he cur r ent account 3.1. Deterministic models of the current account 3.2, Stochastic models of the current account 4. Empi r i cal evi dence on t he i nt er t empor al appr oach 4.1. The relationship between national saving and domestic investment rates 4.2. Tests of intertemporal current-account models 5. How usef ul is t he t heor y? Ref er ences 1732 1733 1742 1743 1764. 1770 1776 1780 1792 1795 *We thank Geun Mee Ahn, Harald Hau, Matthew Jones, Giovanni Olivei, and Clara Wang for excellent research assistance and the National Science Foundation, the German Marshall Fund of the United States, and the Ford Foundation for financial support. Helpful suggestions were made by David Backus, Richard Clarida, Jon Faust, Kiminori Matsuyama, Cedric Tille, and participants in the March 1994 conference for the Handbook at Princeton University. Handbook of International Economics, vol. IIl, Edited by G. Grossman and K. Rogoff Elsevier Science B.V., 1995 1731 1. Introducti on The i nt ert emporal approach views the current -account bal ance as the out come of forward-l ooki ng dynami c saving and i nvest ment decisions. Int ert emporal analyses of the current account became common in the earl y 1980s as a re- sult of papers by Bui t er (1981), Obstfeld (1982), Sachs (1981), Svensson and Razin (1983), and many others, although the approach had explicit precursors in work on t rade and growth by Bardhan (1967), Bruno (1970), and Hamada (1969). 1 As usual, this new focus in open-economy macroeconomi cs resul t ed bot h from t heoret i cal advances in ot her parts of economi cs and from eco- nomi c events that existing open-economy models seemed ill equi pped to ex- amine. Lucas' s (1976) influential critique of economet ri c policy eval uat i on was one i mport ant theoretical mot i vat i on for an i nt ert emporal approach. His insistence on grounding policy analysis in the actual forward-l ooki ng decision rules of economi c agents suggested that open- economy models might yield mor e reli- able policy conclusions if demand and supply functions were deri ved from the opt i mi zat i on probl ems of households and firms rat her t han specified to mat ch reduced-form estimates based on ad hoc economet ri c specifications. Furt her impetus to devel op an i nt ert emporal approach came from events in the worl d capital market , especially the substantial current -account imbalances that fol l owed the sharp world oil-price increases of 1973-74 and 1979-80. The di vergent pat t erns of current -account adjustment by industrialized and devel- oping countries raised the i nherent l y i nt ert emporal pr obl em of characterizing the opt i mal dynami c response to external shocks. Nei t her the classical monet ar y models nor the Keynesi an models in vogue at the t i me offered reliable guidance on this question. Similarly, the explosion in recycled bank lending to devel opi ng countries aft er the first oil shock sparked fears that borrowers' ext ernal debt lev- els might become unsustainable. The need to eval uat e devel opi ng-count ry debt levels again led nat ural l y to the not i on of an i nt ert emporal l y optimal current- account deficit. This chapt er surveys the t heory and empirical work on the i nt ert emporal approach to the current account as it has devel oped since the early 1980s. 2 Re- cently, some researchers have studied dynami c stochastic i nt ernat i onal models with compl et e Ar r ow- Debr eu forward market s for uncert ai n consumption. This particular offshoot of the i nt ert emporal approach is the "compl et e-market s" 1A number of studies published in the early 1980s based exchange-rate or baLance-of-payments models on intertemporal foundations. These contributions are surveyed in the chapter by Maurice Obstfeld and Alan C. Stockman in volume 2 of this Handbook (Obsffeld and Stockman 1985). 2For a complementary survey, see Razin (t995). 1732 Ch. 34: The Intertemporal Approach to the Current Account 1733 model . Because compl et e- mar ket s model s fit mor e nat ur al l y i nt o Mar i anne Baxt er ' s chapt er in this Ha ndbook, t hey are summar i zed onl y bri efl y here. We r eser ve t he t er m " i nt er t empor al appr oach" - as well as t he bul k of our discus- sion - f or model s wi t h i nt er nat i onal bor r owi ng and l endi ng but not necessari l y wi t h compl et e i nt er nat i onal mar ket s in st at e- cont i ngent claims. The chapt er begi ns wi t h an i nt r oduct or y sect i on, Sect i on 2, t hat expl or es the concept of t he cur r ent account , its behavi or in r ecent history, and t he concep- t ual adequacy of measur es of t he cur r ent account as r e por t e d by gove r nmc m agencies. Sect i on 3 lays out basi c i nt er t empor al model s of t he cur r ent account , star1 ing wi t h t he det er mi ni st i c case and t hen expl or i ng st ochast i c model s. Sect i on 4 shows how st ochast i c model s can be used to devi se t est s of t he i nt er t empor ai appr oach, and goes on t o eval uat e t he resul t i ng evi dence. Much of t he discussion t hr ough Sect i on 4 of this chapt er focuses on posiziv~ ~ pr edi ct i ons of t he i nt er t empor al appr oach. A maj or advant age of t he appr oach, however , is its r el evance t o nor mat i v e quest i ons. Sect i on 5 t her ef or e t akes u I, t he r easons why an i nt er t er mpor al appr oach t o t he cur r ent account is essent i al f or sound pol i cy f or mul at i on. Finally, we not e t hat , gi ven t he ext ensi ve r ecent l i t er at ur e this chapt er aspires t o encompass, t her e are several i nst ances wher e space per mi t s us onl y to sket ch al gebr ai c deri vat i ons. We t r y t o al er t t he r eader whe ne ve r i nt er medi at e al gebr a has be e n pr uned especi al l y severel y, so t hat he or she will not become bogged down dur i ng a first readi ng. Space limits l i kewi se al l ow us t o pr ovi de onl y an i l l ust rat i ve r at her t han an exhaust i ve set of r ef er ences. 2. T h e c ur r e nt a c c o u n t : Ba s i c c o n c e p t s a n d hi s t o r i c a l o v e r v i e w A count r y' s cur r ent - account bal ance over any t i me per i od is t he i ncrease in r esi dent s' claims on f or ei gn i ncomes or out put s, less t he i ncrease in similar f or ei gn- owned cl ai ms on home i ncome or out put . Thus, in t heory, t he cur r ent account i ncl udes not onl y expor t s less i mpor t s ( br oadl y def i ned t o i ncl ude all t he i ncome on and payout s on cr oss- bor der assets: di vi dends, i nt er est payment s, i nsur ance pr emi a and payment s, etc.), but also net capi t al gains on existing f or ei gn assets. Fr om t he close of Wor l d War I unt i l r el at i vel y recent l y, most count r i es' hol di ngs of f or ei gn assets had been l i mi t ed bot h in quant i t y and scope, so t he l at t er consi der at i on was secondary. A focus on t he cur r ent account as the net expor t bal ance l ed some economi c t hi nker s t o vi ew r el at i ve i nt er nat i onal pri ces as its cent r al det er mi nant . Thus was bor n t he "el ast i ci t i es appr oach" to t he cur r ent account , under whi ch t he det er mi nant s of i nt er nat i onal expendi t ur e l evel s and i ncomes are hel d fixed in t he backgr ound whi l e st at i c pri ce elasticities of de ma nd and suppl y det er mi ne t he net i nt er nat i onal flow of capital. 1734 M. Obstfeld and K. Rogoff As undergraduate macroeconomics texts demonstrate, however, the current account also is national saving less domestic investment. If saving falls short of desired investment, for example, foreigners must take up the balance, acquiring as a result claims on domestic income or output. This alternative viewpoint, which led to the absorption approach, stresses how macroeconomic factors must ultimately determine international borrowing or lending patterns [Alexander (1952)]. The intertemporal approach to current-account analysis extends the absorp- tion approach through its recognition that private saving and investment deci- sions, and sometimes even government decisions, result from forward-looking calculations based on expectations of future productivity growth, government spending demands, real interest rates, and so on. The intertemporal approach achieves a synthesis of the absorption and elasticities view, however, by account- ing for the macroeconomic determinants of relative prices and by analyzing the impact of current and future prices on saving and investment. International capital flows, in the form of trade credits and commercial traffic in such assets as jewels and precious metals, were already common by biblical times. By the early fourteenth century, Italian banks spanning western Europe and the Levant had become large-scale lenders to sovereigns such as King Ed- ward III of England, whose invasion of France in 1340, aided by foreign finance, initiated the Hundred Years War. The two most powerful Florentine banking houses, those of the Bardi and the Peruzzi, were bankrupted along with many lesser banks in 1343 when Edward proved unable to meet his obligations. But as Europe recovered from this early banking crisis and from the subsequent Black Death (1348), international financial linkages grew strong once again. The Catholic church, through its usury doctrine, unwittingly promoted the in- ternationalization of banking in this period. While domestic loans for interest were prohibited, there was no definitive ban on exchanges of bills payable in different countries and currencies, even when the terms negotiated included im- plicit interest charges [de Roover (1966)]. Theological constraints thus led the largest banks to maintain extensive systems of foreign branches. The expulsion of the Jews from the Iberian peninsula at the end of the fif- teenth century, followed by widespread and continuing persecutions of Protes- tants after the Reformation, created networks of refugee communities with both the motivation and connections to move capital between countries. Dur- ing European wars of the seventeenth and eighteenth centuries, international capital markets developed further as some governments turned to large-scale debt sales to foreigners. [See Neal (1990).] By the early nineteenth century at the latest, the outlines of modern international capital markets are visible in investors' search for profit opportunities on distant shores. The era of the classical international gold standard, spanning the late nine- teenth and early twentieth centuries, is often held up as a benchmark case Ch. 34: The Intertemporat Approach to the Current Account 1735 of unf et t er ed capi t al mobi l i t y bet ween nations. Fi gur e 2.1 shows dat a on sav- ing, i nvest ment , and t hei r di f f er ence, t he cur r ent account , f or a dozen coun- tries over 1885-1913. (All dat a ar e nomi nal flows di vi ded by a nomi nal in- come or out put measur e. ) 3 The graphs i ndeed show sever al exampl es of l arge and pr ot r act ed cur r ent - account i mbal ances, i ndi cat ors of ext ensi ve t r ade across t i me. Canada r an per si st ent deficits which, by t he eve of Wor l d War I, ex- ceeded 15 per cent of gross nat i onal pr oduct ( GNP) . Thes e l arge flows wer e undoubt edl y p r o mo t e d by Canada' s close pol i t i cal and cul t ural links wi t h t he Uni t e d Ki ngdom, t he l argest l ender. But even count r i es wi t hout such close ties t o pot ent i al l ender s wer e abl e t o dr aw ext ensi vel y on i nt er nat i onal cap- ital mar ket s. J apan r an an ext er nal deficit of 10 per cent of nat i onal expen- di t ur e in financing its 1904-1905 war with Russia. Dur i ng Wor l d War I, t he count r y r an a compar abl e surplus t o hel p fi nance its fi nanci al l y bel eagur ed allies. 4 Da t a f r om t he i nt er war per i od, shown in Fi gur e 2.2, r eveal a part i al r esur gence of net i nt er nat i onal bor r owi ng and l endi ng as post war r econst r uct i on progresses, but this process comes t o a sudden hal t as rest ri ct i ons on i nt er nat i onal payment s pr ol i f er at e af t er t he onset of t he Gr e a t Depr essi on. s Post -1945 dat a disclose a r ever se evol ut i on. Initially, cur r ent - account i mbal ances wer e slight because or official rest ri ct i ons on i nt er nat i onal capi t al movement s , wi t h most i ndust ri al - count r y cur r enci es bei ng i nconver t i bl e t hr ough 1959. Af t e r t he ear l y 1970s (see Fi gur e 2.3), net i nt er nat i onal capi t al flows expanded as a resul t of pet r odol l ar recycl i ng, t he r emoval of many i ndust r i al - count r y rest ri ct i ons on i nt er nat i onal payment s fol l owi ng t he adopt i on of fl oat i ng exchange rat es, and t echnol ogi cal evol ut i on in t he fi nanci al industry. By t he 1980s, t he l argest i ndust ri al count ri es, Ger many, Japan, and t he Uni t ed States, wer e showi ng subst ant i al ext er nal im- bal ances. At t he same t i me ma ny devel opi ng count ri es, caught in a debt crisis br ought on in par t by vi gor ous bor r owi ng in t he 1970s, f ound t hemsel ves de- ni ed access to r esour ce inflows. Onl y in t he ear l y 1990s di d this st ark bor r owi ng const r ai nt begi n t o ease. Unf or t unat el y, t he saving and i nvest ment flows r epor t ed in nat i onal i ncome and pr oduct account s (NIPA) and shown in Fi gures 2.1 t hr ough 2.3 don' t always conf or m cl osel y t o t heor et i cal l y cor r ect concept s of saving and i nvest ment , par- t i cul arl y when i nt er nat i onal capi t al mobi l i t y is ext ensi ve. One especi al l y seri ous def ect is t he f ai l ur e of NI PA nat i onal i ncome measur es ful l y t o refl ect capi t al gai ns and losses on net f or ei gn assets. Dur i ng 1991, f or exampl e, t he NI PA mea- 3These data, as well as those shown in Figure 2.2 below, are from Jones and Obstfeld (1994). Finland was actually a possession of Russia during the 1885-1913 period, but it was afforded a fair amount of administrative autonomy. 4For a discussion of wars and the Japanese current account during the first part of the twentieth century, see Obstfeld and Rogoff (1996), ch. 1. 5See Eichengreen (1990) for further discussion. 1736 0.4 Australia 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . El ~0. 2 ~ : 0.1 .g -0.1 2 3 ~ ~ ...................................... ..................................................... -0,2 I l l l : l : ; l l l l [ l l ; ; : l ( l : l l l ; l l 188 1892 1899 1906 191: 0. 4 M. ObstJeld and K. Rogoff Finland 0,3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . El 0.2 ' 6 = 0,1 O u- -0.1 ..................... : : Z: ~ .................... "~-" ~ . ~ -0.2 I I I I t I I i i i i i i i i i i i , L~- +- I i i t ~ 1885 1892 1899 1906 191 0,4 Sweden 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o. 3 (~ 0. 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . = 0.1 0 -0.2 +-/-/--I I I I I ', ', ', ', ; ', ', ; ; ; ; I I t I t I I 111913 1888 1892 1899 1906 0. 4 Norway 0.3 O. Z (9 0.2 "6 c 0.1 ,0. 29 U'.o. 1 -0" 8 i 1892 1899 1906 191 0.4 Denmark 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . El ' 6 (~ 0. 2 0.1 o 0 , >. " -0.1 ............................................................................................................................. ! " 0. 2 ', ', ', : ', ', ', ', L ~ I i I I I I t I I ~ I I ++- ~- ~- I - ~- ~ 1885 1892 1899 1906 1913 0. 4 Italy 0. 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ~0.2 c 0.1 2 ~ l l : : ~ l ; : l ; ; ; ; l l l t l E I I l i t i l l ~' 1885 1892 1899 1908 1913 - I nvest ment - Savi ng - - Cur r ent Account Figure 2.1. Saving, investment, and the current account: Classical gold standard, 1885-1913. Ch. 34: The l nt ert emporal Appr oac h to the Current Ac c ount 1737 United States 0,4 ~. 0. 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Z 0. 2 ................................................................... c 0.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o u. - 0. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1885 1892 1899 1906 1913 United Kingdom 0. 4 . . . . O. 0, 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . t : 3 (=9 0. 2 = 9. 1 ~ 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . u--0.1 ...................................... 1885 1892 1899 1906 1913 0.4 Canada z m 0. 3 0 0, 2 c 0. 1 ................................................................................ . 9 0 .. . . .. ... .. ... .. ... ... .. ... .. ... ... .. ... .. ... ... .. ... .. ... ... .. ... ... . .. . .. . . .. . - 0. 1 ............ ~- ~ . . . . . . . . . . . . . . . . . . . . . ~' \ / 1885 1892 1899 1906 1913 France 0, 4 O~ 0, 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C~ 0 0. 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . c 0.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , o 0 ,~ ~ ~, ~- ~ ~. , " _0, 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0, 2 i i i i t=-i r i i i i i [ f--. +-; ; 'r ; : ; ; ; I I I 1885 1892 1899 1906 1913 0.4 Germany 0. 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . z z 0, 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . e 0. 1 0 0 2 ~ ~ " ~ ~ ' " ~ - ~ ' 2 Z " > ~ ' ~ : ~ . ~ - ~ ' . ~ - ' ~ u. . - 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0. 2 I ~ I ~ '~ : I ; : : ', : ; i i { ~ i i I -~F=t =~- i i i i L 1885 1892 1899 1906 1913 Japan 0. 4 . . . . . . . . . . . . 0, 3 ILl Z O 0, 2 c 0. 1 , o u. - 0. 1 . . . . . . . . . . . . " - ~ . . . . . . . . . . . . . . . . . . ! ~, ! . . . . . . . . . . . . . . . . . . . . . . . - 0. 2 ~- ~- i f ' , : ', : ; ; ; ; ; ; ; ; i i r ~ i i ~ [ J i 1885 1892 1899 1906 1913 - - I nvest ment - - Savi ng Fi gur e 2, 1, continued. Cur r ent Account 1 7 3 8 0. 4 Australia 0 . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o , = 0 . 1 0 0 .... . . . . . . . . . . . . . . . . . . . . . . . . . , - ' ~ ~ - " , , ; ~ . ~ ........ u . . 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . >',~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0 . 2 i r i ~ L i i i i i ~ i i i ~ i i 1 9 2 1 1 9 2 7 1 9 3 3 1 9 3 9 0, 4 M. Obs t f e l d and K. Rogof f Finland 0 . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o 0 2 c 0 . 1 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - - 0 . 1 - 0 , 2 l l q ~ l l l l l ~ l l l l l i l l 1 9 2 1 1 9 2 7 1 9 3 3 1 9 3 9 Sweden 0. 4 e. 0 . 3 . . . . . . . . . . . . . . . . . . . . . . . . . o O 0 . 2 = 0 . 1 .............................................................................. . o a. -0.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0 . 2 i i i i i i i i i i i i i r i i i - i - - + 1 9 2 1 1 9 2 7 1 9 3 3 1 9 3 9 Norway 0 . 4 0 . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 . 2 c 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o u. -0.1 ................................................................................. - 0 . 2 - - ~ - , - - , - - , - - - - ~ t - . f - - f - - ~ - - f - = . f - - t ~ f - - - i i i i 1 9 2 1 1 9 2 7 1 9 3 3 1 9 3 9 Denmark 0. 4 0 . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o . c3 O 0 , 2 . . . . . . . . . . . . . . . . . . . . . . . "6 = 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o " - 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0 . 2 i + ~ ~ ~ ~ , , ~ , ~- ~- - +- +- - ~+ , 1 9 2 1 1 9 2 7 1 9 3 3 1 9 3 9 0. 4 Italy 0 . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . a . z = 0,1 ._o u . - 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0 . 2 I I I I I I I I I I I I I I l I I I I 1 9 2 1 1 9 2 7 1 9 3 3 1 9 3 9 - - I n v e s t me n t - - S a v i n g - - Cu r r e n t Ac c o u n t F i g u r e 2 . 2 . S a v i n g , i n v e s t m e n t , a n d t h e c u r r e n t a c c o u n t : I n t e r - w a r e r a , 1 9 2 1 - 1 9 3 9 . Ch. 34: The lntertemporal Approach to the Current Account I 739 0.4 United States 0. 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . z 0. 2 . . . . . . . . . . . . . . . . . . . . . . . . c 0. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 ' "-0. 1 " . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0' 219LI j 2 + ~- ~- * ~ E i b i ~ - ~ ~ 1927 1933 1939 0. 4 0.3 Q. E3 (3 0. 2 `6 .~ o.1 a . 4) . 1 - 0. 2 1921 United Kingdom 1 J . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I J 1927 1933 ! 939 Canada 0,4 0. 3 .......................................................................................... o. z L~ 0. 2 c 0. 1 .9 " - 0, 1 .................................................................................................................. - 0, 2 ~- I p i i ] =l I I I I P q I I I I I I 1921 1927 1933 1939 Germany 0, 4 0. 3 el. z Z 0.2 `6 e o. 1 o ~ o u. -17.1 - 0' 219~1 1927 1933 1939 0.4 Japan z.u 0. 3 .................... Z (3 0. 2 = 0.1 ............................................................................................................. 0 o u. -0. 1 ....................................................................................... -0,2 ', ', = ; ', ; i i i r P i i ] I I l [ E 1921 1927 1933 1939 - I nvest ment - Savi ng Fi gur e 2. 2. continued. - Cur r ent Account 1 7 4 0 A u s t r a l i a 0. 4 0 ~ 0.20"3 ..................................................................... . ~ c 0.1 ............................................................................... 0 0 .~-- --.-.-~ ............................................................ u . - 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0 , 2 I i ~ i ) l - + - - - ) : - ~ t t - - ~ + - - ~ + - F - + 1973 1979 1985 1991 M . O b s t f e l d a n d K . R o g o ] " F i n l a n d 0. 4 O EL 0. 3 0 0. 2 = 0.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o O . . . . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . u . d ) . l ~ ' / ~ 1973 1979 1985 1991 S w e d e n 0. 4 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . n
0 0 , 2 "6 = 0. 1 ................................................................................ 0 O -- .w~,<-~-~.~. -.g~ 222 2, - ; ' ~- " ' ~ . . . . . ~ ~ ; ~ t u - 0. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0 , 2 t I - F - - I I I I I I I I ) J r - - ~ - - ~ 973 1979 1985 1991 0. 4 N o r w a y 0. 3 L 3 0 0. 2 "6 = 0.1 o o u. -0.1 - 0 . 2 ~ - 1973 ~ . . / / ......... > , - . ~:.~;/ ............................................................ i P - I I I ~ - i I I ~ i I I I I I I 1979 1985 1991 D e n m a r k 0. 4 0.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E L ?, 0.2 = 0. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . = O u . - 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 0 , 2 + I I + - - ~ I I 1 i - - f - - 4 ~ t - - + - - k - - u ~ I I 1973 1979 1985 1991 0. 4 I t a l y c~ EL 0. 3 i O 0. 2 = 0 . 1 ........................................................................... .2 0 ~ . . ~ 2 ~ - ~ - ~ . ~ g ; : ~ . z ~ - ~ - ~ - ~ - ~ - , - ~ ; .- " - 0 . 1 ....................................................................................... - 0 . 2 - t I I I . t - ~ ; i ; ; ', l ; ; ; - I I t I I 1973 1979 1985 1991 - - I n v e s t m e n t - S a v i n g . C u r r e n t A c c o u n t F i g u r e 2. 3. S a v i n g , i n v e s t me n t , a n d t h e c u r r e n t a c c o u n t : P o s t - B r e t t o n Wo o d s e r a , 1 9 7 3 - 1 9 9 1 . Ch. 34: The Intertemporal Approach to the Current Account 1741 United States 0 . 4 0 . 4 0 , 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . r~ 0 . 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . , - 0 . 1 ......................................................................................................... . 2 ~ " - 0 . 1 .................................................................................................................. 1 9 7 3 1 9 7 9 1 9 8 5 1 9 9 1 0 . 3 13, ( 9 0 . 2 c 0.1 o_ u. -0A United Kingdom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ,i 1 9 7 3 1 9 7 9 1 9 8 5 i ~ 9 ~ Canada n 4 0 . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 . 2 " 6 , - 0 . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . o - + - u . - 0. 1 ............................................................................................... 1 9 7 3 1 9 7 9 1 9 8 5 1 9 9 1 France 0 . 4 o . 0 . 3 . . . . . . . . . . . . . . . . . ( 9 0 , 2 . . . . . . . . . . . . . . . . . . - , . P ~ , - 0. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 m- 0. 1 ............................................................................................. 1 9 7 3 1 9 7 9 1 9 8 5 1 9 9 1 0 . 4 Germany 0 , 3 ............................................................................................................... 0 . 0 . 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " 6 0. 1 ................................................................................................................ O u. 4) . 1 ............................................................................................................ i 1 9 7 3 1 9 7 9 1 9 8 5 1 9 9 1 Japan o. , 0 . 3 0 . 2 ............................................................. L Z I I 7 ........................... 1 = 0. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i 0 u" " 0. 1 .......................................................................... { i 1 9 7 3 1 9 7 9 1 9 8 5 1 g)~,~ - i n v e s t m e n t - S a v i n g Fi gur e 2. 5. col~[muea. C u r r e n ~ A c c o ~ 1742 M. ObstJeld and K. Rogoff sur e of savi ng l ess i nves t ment for t he Uni t e d St at es, t he cur r ent account , was r e por t e d by t he U. S. De p a r t me n t of Co mme r c e as - $ 3 . 7 billion. The De p a r t - me n t of Co mme r c e also cal cul at ed, however , t hat on a ma r ke t - va l ue basi s, U. S. asset s hel d a br oa d at t he end of 1990 appr eci at ed in dol l ar val ue by $67.8 bi l l i on ( mor e t han t he ent i r e 1991 pur chases of f or ei gn asset s by U.S. r esi dent s, $62.2 bi l l i on). At t he s a me t i me, t he dol l ar val ue of f or ei gn cl ai ms on t he U. S. at t he end of 1990 r ose by $172.8 bi l l i on ( mor e t han 2.5 t i mes as l ar ge as t he 1991 f or ei gn pur chas es of U. S. assets, $67.0 bi l l i on). 6 I n economi c t er ms, t he t r ue dol l ar U. S. cur r ent - account deficit f or 1991 is p r o b a b l y much cl oser t o - $3. 7 bi l l i on + $67.8 bi l l i on - $172.8 = - $108. 7 bi l l i on t han t o its si mpl e NI PA me a - sure. And e ve n t hi s figure refl ect s onl y par t i al c ove r a ge of i nt er nat i onal asset hol di ngs. The numbe r s in Fi gures 2.1 t hr ough 2.3 ar e mi sl eadi ng f or a not he r r ea- son. As Sect i on 3 shows, it is changes in real, not nomi nal , asset hol di ngs t hat ma t t e r f or a count r y' s wel f ar e. Thus, NI PA me a s ur e s of t he cur r ent ac- count , e ve n if cor r ect ed t o i ncl ude nomi nal capi t al gai ns and losses, mus t be adj ust ed as wel l t o cor r ect f or t he i nf l at i onar y er os i on of f or ei gn asset s' r eal val ues. Thes e p r o b l e ms pl ague all of t he empi r i cal l i t er at ur e di scussed in Sect i on 4, al t hough it woul d be possi bl e t o r e me d y t he m s ome wha t for a f ew i ndust r i al count r i es. I n empi r i cal l y eval uat i ng t heor i es it is obvi ous l y cruci al t o achi eve t he bes t possi bl e ma t c h be t we e n t he concept ual f r a me wo r k and t he dat a br ought t o be a r in t est i ng it. 3. Intertemporal approaches to the current account A r eal i st i c e c onomi c model i ncor por at i ng all e l e me nt s r el evant t o t he t ypi cal count r y' s cur r ent account woul d be hopel essl y compl ex. I ns t ead of a t t e mpt i ng t o const r uct such a compr ehens i ve model , we t ur n in this sect i on t o a succes- si on of model s t hat i l l ust rat e what we bel i eve t o be t he ke y el ement s i nfl uenci ng s a vi ng- i nve s t me nt bal ances in t he wor l d economy. I n f ol l owi ng this st rat egy, we l ar gel y par al l el t he de ve l opme nt of t heor et i cal t hi nki ng on t he cur r ent account ove r t he pas t fi ft een year s An i mpor t a nt set of i nsi ght s can be der i ved even f r om det er mi ni s t i c model s; this is our first or der of busi ness. We t hen i nt r oduce uncer t ai nt y about t he economi c e nvi r onme nt and deduce a n u mb e r of essen- t i al modi f i cat i ons of t he pr ecedi ng nonst ochast i c f r a me wor k. The i nt r oduct i on of uncer t ai nt y is, of course, a pr er equi s i t e f or t he empi r i cal anal ysi s of t he in- t e r t e mp o r a l a ppr oa c h t hat we discuss in Sect i on 4. 6The United States capital-account surplus, equal to $67.0 billion - $62.2 billion = $4.8 billion, differs from the recorded current-account deficit by a statistical discrepancy of -$1.1 billion. Ch. 34: The Intertemporal Approach to the Current Account 3.1. Deterministic model s of the current account ! 743 The first br oad class of model s t o be descr i bed assumes t hat i ndi vi dual deci si on maker s have per f ect f or esi ght and compl et e i nf or mat i on about t hei r economi c envi r onment . Whi l e t hese model s lack real i sm al ong some di mensi ons, t hey serve at l east t wo ver y useful purposes. First, t hey el uci dat e a numbe r of que:~ t i ons f or whi ch uncer t ai nt y is of secondar y i mpor t ance, and second, t hey pr ovi de a be nc hma r k agai nst whi ch to meas ur e t he pr edi ct i ons of r i cher st ochast i c r ood els. As we shall see later, st ochast i c model s in which real bonds arc t he oni v assets count r i es t r ade i mpl y r esponses t o shocks similar t o t he i mpul se r esponses of det er mi ni st i c model s. 3.1.1. A one-good model with representative national residents Consi der a small open e c onomy t hat pr oduces and consumes a single c o mp o s i t e good and t r ades f r eel y wi t h t he rest of t he world. Fr ee t r ade i ncl udes t he i nt er nat i onal exchange of assets. We assume t hat t he onl y t r aded asset is a cons umpt i on- i ndexed bond wi t h fixed face val ue t hat pays net i nt er est at t he r at e rt bet ween per i ods t - 1 and t. I mpor t ant l y, l abor is i nt er nat i onal l y i mmo- bile. In per capi t a t erms, l et At+l denot e t he economy' s st ock of net f or ei gn claims at t he end of per i od t, Yt net domest i c pr oduct or out put in per i od r, Ct pr i vat e consumpt i on, Gt gover nment consumpt i on, and It net investmem= ~ The n t he i dent i t y l i nki ng net f or ei gn asset a c c u mu l a t i o n - t hat is, t hc curlTelli account , CA t - t o t he savi ng- i nvest ment bal ance is: CAt =A t +l - A t = rtAt + Yt - Ct - Gt - Lt (3.1) Def i ne t he ma r ke t di scount f act or f or dat e s cons umpt i on by 1 R , . , = r I = , + l ( 1 + ( 3 . 2 / ( wher e Rt,t = 1). For war d i t er at i on of eq. (3.1) l eads t o o o (1 + rt)At = Z Rt,,(C.,. + Gs + I~ - Ys) + l i m R,,sAs+l S ~ + O O S = t Because f or ei gn l ender s will not al l ow t he e c onomy to rol l over a debt indel;o i ni t el y t hr ough unl i mi t ed bor r owi ng, t he condi t i on lims-~oo Rt,sAs+l >~ 0 appl i cs above. The resul t i ng intertemporal budget constraint f or t he economy is thus: o o o o ~-'~ Rt,s(C,. + Gs +i s) < (1 + rt)At + Z Rt,sYs (3.3) s=t s=t In t he model s we sur vey no r esour ces are willingly f or gone, so (3.3) always hol ds 7We wi l l a s s u me f o r s i mp l i c i t y t h a t c a p i t a l d o e s n o t d e p r e c i a t e . 1744 M. Obstfeld and K. Rogof f wi t h equality. In t hat case, (3.3) states t hat t he pr esent val ue of the economy' s expendi t ur es must equal its initial net forei gn weal t h plus the present val ue of domest i c product i on. The i nt er t empor al budget const rai nt del i mi t s the feasi bl e choices of the econ- omy. To descri be the circumstances in which current -account i mbal ances will arise, however, one must specify how t he component s of expendi t ur e and out - put are det er mi ned. We assume t hat the r epr esent at i ve consumer maxi mi zes t he t i me- separ abl e funct i on oo ut = Z 'uICd I3.4) s =t wher e/ 3 c (0, 1), u' ( C) > O, u"( C) < O. The st rong i nt er t empor al separabi l i t y of utility assumed in (3.4) will form t he backbone of our formal analysis. Bef or e going furt her, it is wort hwhi l e to offer some j ust i fi cat i on for this decision. (1) One mi ght wish to start with a very general i nt er t empor al l y nonsepar a- ble ut i l i t y f unct i on of t he f or m Ut = U( Ct , Ct+l, ...). But this woul d yi el d few concret e and t est abl e behavi oral predictions. I nst ead we prefer to begi n wi t h a t ract abl e basic set up like eq. (3.4) wi t h st rong implications. Preferences t hen can be gener al i zed if t he basic setup seems to be l eadi ng us as t r ay/ (2) Aggr egat i on, across bot h goods and i ndi vi dual s, may cause i nt er t empo- ral dependenci es appr oxi mat el y to cancel out at t he level of t ot al per capi t a consumpt i on. (3) At t he levels of t i me aggregat i on common in macr oeconomi c model s, t he assumpt i on of i nt er t empor al separabi l i t y is not i mpl ausi bl e. In any event , while empi ri cal research on dat a at these frequenci es has rai sed i nt erest i ng quest i ons about the t i me-separabl e ut i l i t y model , it does not cl earl y poi nt to a superi or nonsepar abl e al t ernat i ve. Havi ng def ended our choice of i nt er t empor al l y addi t i ve ut i l i t y funct i ons, we now begi n to pursue t hei r implications. Let lit be t he real val ue of domest i c firms at t he end of peri od t - 1 (after peri od t - 1 di vi dends have been paid), Bt t he st ock of i nt erest -earni ng claims owned by t he domest i c pri vat e sector at t he end of t - 1, wt the real wage in peri od t, Lt t he per capita suppl y of labor, and Tt l ump- sum taxes levied by the home gover nment . Then t he i nt er t empor al budget const rai nt of t he represent at i ve consumer is 9 8pl ausi bl e non- t i me- s epar abl e al t er nat i ves t o (3.4) can be anal yzed; [see, for exampl e, De ve r e ux and Shi (1991), Obs t f el d (1982), Shi and Eps t ei n (1993), and Svens s on and Razi n (1983)]. 9Strictly speaki ng, t he way t he next const r ai nt is e xpr e s s e d as s umes per f ect f or esi ght b e t we e n dat es t - 1 and t, so t hat a r at e of r et ur n of rt act ual l y is e a r ne d on s har es ex post . Thi s as s umpt i on is ma de onl y t o avoi d cl ut t eri ng t he not at i on, and is not i mpos e d in what follows. A mor e gener al f or mul at i on si mpl y woul d r epl ace (1 + rt)Vt by t he ex pos t val ue (inclusive of di vi dends) of s har es bought on dat e t - 1. Ch. 34: The Intertemporal Approach to the Current Account 1745 O 0 0<3 ~ R,,~c~ = (1 + r, ) ( vt + Bt ) + ~ Rt,s(wsLs - "tO C3.51 s=t s=t When (3.4) is maximized subject to (3.5), consumption necessarily follows the i nt ert emporal Eul er equat i on u ' ( C t ) =/ 3( 1 + r t +l ) u ' ( Ct +l ) ~ c.', This optimality condition, stressed long ago by Irving Fisher (1930), equate~ the marginal rate of substitution of present for future consumption, whM.. ;, just / 3 u ' ( C t + l ) / u ' ( C t ) , to the price of future consumpt i on in terms of presem consumption, 1/(1 + rt +l ) . An implication is that, leaving aside discrepancies bet ween/ 3 and 1 / ( l + r t < ) , optimized consumption will follow a smoo!h, c,onst:ap.', pat h (recall that u ( C ) is strictly concave). A convenient closed-form descripli,w_ of the current account is obt ai ned by specializing further to the case in whic!~ u ( C ) takes the isoelastic form c 1-1/' ~ - 1 u(C) - 1 - 1/ o" with o- > 0 the elasticity of intertemporal substitution, In this case (3.6 9 impiic,: that optimal consumpt i on growth obeys Ct+l =/ 3~( 1 + r t <) ' ~ Ct (3.7) The consumpt i on pat h described above must satisfy the economy' s intertom poral constraint; using (3.7) to eliminate Cs ( s > t ) from (3.3) shows 1hal lhe economy' s dat e t consumpt i on will be: Ct = ( 1 + r t ) At E s =t Rt , s ( Ys - Gs - I s) ( 3 . 8 ) oo R E,=t , , , ( / 3 ~- ' I R , , ~) ~ Equat i on (3.8) leads to an illuminating general charact eri zat i onof the currcm account. Define the "permanent " level of variable X on date t , X , , by _ Es%t R,,,X~. and define ( / 3 / R ) ~ as the following weighted average of ratios of (s -- 0- per i od subjective and market discount factors raised to the power 0.: x-,oo R { s - t / R ho- ( / 3 / R ) ~ = z..,s=t t , s v~ / t,s2 ~o R E , ' , ' = t t ~ s Then, a f t e r m a n y s t e p s , eqs. (3.1) and (3.8) show the current account surplus on 1746 M . O b s t f e l d a n d K. R o g o f j : dat e t to be: C A t = ( r t - f t ) A t + ( l i t - L ) - ( G t - G t ) - ( / t - ~ ) + [1 1 ] ( At + B - - d t - i t ) ( 3 . 9 ) (As we al l uded in t he i nt r oduct i on, t her e are sever al places wher e we advi se t he r eader t o consi der hol di ng of f on r epr oduci ng a r esul t until a l at er readi ng; this is t he lirst.) Equat i on (3.9) yields a numbe r of i mpor t ant pr edi ct i ons ( each of whi ch r equi r es a cet eri s par i bus clause): 1 (1) If t he e c onomy is a net f or ei gn cl ai mant and t he worl d i nt er est r at e is above its pe r ma ne nt average, t hen t he cur r ent account will be in gr eat er surpl us as peopl e s moot h cons umpt i on in t he face of t empor ar i l y high f or ei gn i nt er est i ncome. If t he e c onomy is a net f or ei gn debt or, t empor ar i l y hi gh i nt er est r at es will have an opposi t e cur r ent - account effect . (2) Out put above its pe r ma ne nt l evel will cont r i but e t o a hi gher cur r ent - account surplus, agai n due to cons umpt i on smoot hi ng. Similarly, t he pr i vat e sect or will use f or ei gn bor r owi ng t o cushi on its cons umpt i on f r om abnor mal l y hi gh gove r nme nt cons umpt i on and i nvest ment needs. (3) The last t er m in (3.9) reflects cons umpt i on t i l t i n g due to di ver gences in t he cur r ent and f ut ur e per i ods bet ween wor l d real i nt er est rat es and t he domest i c r at e of t i me pr ef er ence, (1 - f l ) / f i . Whe n t he home count r y is on aver age mor e i mpat i ent t han t he rest of t he worl d, fl is l ower t han f ut ur e wor l d i nt er est r at es will t end t o be, resul t i ng in ( f l / R ) ~ < 1. The r e will t hen be a secul ar t endency t owar d cur r ent - account deficits, and t hus t owar d secul arl y i ncreasi ng f or ei gn debt and decl i ni ng consumpt i on. When it is f or ei gner s on t he ot her hand who ar e ( on aver age) mor e i mpat i ent , so t hat ( f i / R ) ~ > 1, consumpt i on' s t i me pat h will have an upwar d tilt. The tilting ef f ect is pr opor t i onal t o t he economy' s " pe r ma ne nt " resources. Al so, t he tilting ef f ect is st r onger t he hi gher is t he case of i nt er t empor al subst i t ut i on in consumpt i on, meas ur ed by o-. iA less general version of eq. (3.9) is presented by Sachs (1982). To derive the first entry on the right-hand side of (3.9), observe that l +r t _ r t + 2 s T _ t + l R t , s r s _ ~ s ~ t R t , s r s Ft. X2 R Oa O0 where we have made use of the fact that oo ~ ' ~ R t , s r s = 1 s - - t +l The rest is straightforward; for details, see Obstfeld and Rogoff (1996), ch. 2. Ch. 34: The Int ert emporal Appr oach to the Current Account 1747 Before exploring the short-run implications of eq. (3.9), it is useful to consider some of the model' s predictions for steady-state current-account behavior in a growing economy. As we shall see, a growing economy can run a current-account deficit indefinitely. Incorporating the investment effects of growth forccs one to model explic- itly the linkage bet ween capital accumulation and production. Assumc that the product i on function for a small economy is Cobb-Dougl as: g , = Ot K~ L 1-~ (c~ < 1), where Kt is the end of peri od t - 1 capital stock (available for production in peri od t), L is the constant l abor force (which we will normalize to 1 for convenience), and the productivity coefficient 0 grows so that Or+ 1 = (1 + g ) l - c ~ o t with g > 0. Assume also that the capital stock can adjust in a period, withoul installation costs. In the absence of unanticipated shocks, the marginal product of capital, a Ot K~ - 1 , thus must equal r, the constant world interest rate; r > g by assumption. 11 Then, in a steady-state equilibrium, investment is easily shown to be ( ~ ) 1/(1 c0 ( - ~ ) i t = Kt +l - Kt = g = Y t Out put and investment t herefore bot h grow at rate g. If government spending is zero, then one can use (3.9) to show that the optimal current account is given by C A t = A , + I - A t = - [ 1 ( l +r ) ~/ 3~] At - l + g - ( l + r ) ~ f i ~ (1 ~ ) Yt r - g Division by Y t yields a difference equation in A / Y , At+lYt+l = ( ( l + r ) ~ x j A t l 7 7 / Yt 1 + g - ( 1 + r ) ~ / 3 ~ - ~ - +gT~_- - ~ ( 1 - ~ ) Provi ded (1 + r ) " ~ < 1 +g, the st eady state is stable; it is the negative number 1 - ( a g / r ) (3.10) A / Y - - ( r - g) Notice that, because a g / r = I / Y , the long-run ratio of foreign debt to out put equals the ratio to current out put of the e nt i r e p r e s e n t v a l u e of future out put net of investment. 11For the world economy as a whole, the assumption that the growth rate does not exceed the interest rate can be justified using the standard Cass-Koopmans general equilibrium growth model. 1748 M. Obstfeld and K. Rogof f What size debt - out put rat i o does eq. (3.10) i mpl y? Suppose t he worl d real i nt erest rat e r is 8 percent per year, g is 4 percent per year, and a = 0.4. Then A / Y -- - 20, and the economy' s t rade bal ance surplus each peri od must be - ( r - g ) A / Y = 80 percent of GDP! Such l arge debt levels and debt burdens are never observed in practice: economi es t hat must borrow at mar ket i nt erest rat es rarel y have debt s as great as a single year' s GDR The anomal ous predi ct i on poi nt s to some short comi ngs of t he model . Wi t h finite lifetimes, i ndi vi dual s cur r ent l y alive woul dn' t be abl e t o bor r ow agai nst t he ent i re present val ue of t he economy' s out put . Ther e is no al l owance in t he model for soverei gn risk. Finally, we not e t hat t he worl d inter- est rat e r is det er mi ned by t he rest of the worl d' s growt h rate. In this exampl e, however, t he worl d growt h rat e must equal (1 + r)~/3 ~ [see Obst fel d and Rogof f (1996), ch. 2], which is less t han 1 + g by assumpt i on. But if a small economy wer e to grow fast er t han t he worl d for a sufficiently l ong period, it woul d cease bei ng small and t he fixed i nt erest rat e assumpt i on woul d be vi ol at ed. If, on t he ot her hand, the count ry' s growt h rat e event ual l y converges to t he worl d gr owt h rat e, the count r y' s ability and desire to bor r ow f r om worl d capital mar ket s are reduced. 3.1.2. The role o f comparative advantage The t heor y of i nt er nat i onal t rade offers a useful perspect i ve on t hese results. The r eason is t hat foreign borrowi ng and l endi ng can be vi ewed as intertemporal trade, t hat is, as t he exchange of consumpt i on avai l abl e on di fferent dates. The principle of comparat i ve advant age t hus applies. I magi ne t hat t he home count r y faces a const ant worl d i nt erest rat e equal t o l + r = 1/ f l and t hat lit > Ys for s > t. Equat i on (3.9) predicts t hat (absent ot her c ur r e nt - pe r ma ne nt di fferences) the count r y will have a current -account surplus. The principle of comparat i ve advant age makes t he same prediction. In the ab- sence of t rade, t he count ry' s aut ar ky i nt erest rates, equal to u' (Yt )/ ~' ~-t u' (Y~), lie bel ow t he correspondi ng worl d i nt erest rates, (1 + r) s~t. It t her ef or e pays for t he r epr esent at i ve resi dent to export present consumpt i on by l endi ng abroad, since pr esent consumpt i on is compar at i vel y cheap at home, and to i mpor t f ut ur e consumpt i on by receiving r epayment for t he loans at a l at er date. Similarly, t empor ar i l y hi gh i nvest ment needs can raise a count ry' s aut ar ky in- t erest rat e above t he worl d level, maki ng it an i mpor t er of present consumpt i on, t hat is, an ext ernal borrower. Thi s perspect i ve makes clear t hat it is onl y t he country-specific or i di osyncrat i c component s of shocks t hat result in current -account imbalances. The gl obal component s of shocks, in contrast, affect all count ri es similarly and t hus open up no new oppor t uni t i es for gains t hr ough i nt er t empor al t rade. Count ri es cannot all Ch. 34: The lntertemporal Approach to the Current Account 1749 smoot h consumpt i on perfect l y in the face of a shock t hat t empor ar i l y depresses worl d out put , for exampl e. Inst ead, t he world real i nt erest rat e is bid up and all count ri es tilt t hei r consumpt i on pat hs upward. 3. 1. 3. Mo d e l i n g o u t p u t f l u c t u a t i o n s a n d i n v e s t me n t Our earl i er i nvest ment exampl e assumed t hat a nat i on' s capital stock can bc adj ust ed wi t hout cost to equat e capital' s margi nal product to the worl d i nt erest rate. Empi ri cal obser vat i on makes this assumpt i on difficult to swallow: even a', t he i ndust ry level, we si mpl y do not observe large di scret e changes in stocks of capital over ver y short periods. If t here are costs to installing capital, however, capital stocks will move mor e sluggishly and pr ot r act ed devi at i ons of capital' > margi nal pr oduct from r are possible. A simple model illustrates how costly i nvest ment affects current -account dynamics. 12 As s ume t hat final out put is pr oduced according to the st andar d homogeneous and concave pr oduct i on funct i on Ot F( Kt , Lt ), where, as before, Ot capt ures shifts in t ot al fact or product i vi t y, and where Kt is given by past i nvest ment decisions. A firm maki ng decisions on dat e t maxi mi zes t he pr esent di scount ed val ue of its profits (which equal t he di vi dends sharehol ders receive): [ a I 2 ] ~ _ _ , R , , , O , F ( K , , L s ) - w , r s - 1 ~ - 5 g ] , J ( 3 . 1 1 ) S=f Above, a e 7 ( I . ; / Kt ) is the deadwei ght out put cost of installing K ~ + 1 - K , = I t ( 3 . l 2 ~ units of capital, where a > 0 is an adj ust ment -cost paramet er. Accordi ng to (3.11), t he efficiency of i nvest ment in produci ng i nst al l ed capital declines at an increasing rat e as t he rat i o of i nvest ment to i nst al l ed capital rises. Gi ven (3.11), t he economy' s net out put on dat e t is Yt = Ot F( Kt , L t ) - ~ ( I 2 / K t ) . The condi t i ons for maxi mi zi ng profits (3.11) subject to (3.12) i ncl ude I , - qt - 1 g , ( 3 . 1 3 ) a and the i nvest ment Eul er equat i on a 2 Ot+IFK(K,+I, gt+l) + ~( l , +l / Kt +l ) + q, < = (3.14) qt 1 + rt+ 1 12Matsuyama (1987) develops an open-economy model with costly investment. The effects of residential investment are studied in Matsuyama (1990). 1750 M. Obst f el d and K. Ro g o f f wher e qt (Tobin' s q) is the Lagr ange mul t i pl i er on const rai nt (3.12) and has t he usual i nt er pr et at i on as the shadow price of i nst al l ed capital. For war d i t er at i on on (3.14) shows t hat qt = ~ Rt , s [ O s F K ( K s , L s ) + 2(I s~Ks)2] s = t + l (3.15) (we have excl uded bubbles in t he price of capital). Equat i on (3.15) states t hat qt , t he e x d i v i d e n d shadow price of a uni t of capi t al at the end of peri od t, equal s the pr esent val ue of its fut ure margi nal product s plus t he present val ue of its f ut ur e cont ri but i ons to l oweri ng t he costs of i nvest ment . 13 Mul t i pl y eq. (3.14) by Kt+l and use eq. (3.12) to express it as a 2 Ot +I FK( Kt +I , L t +I ) Kt +I q- 2 ( I ; +1 / K r + I ) q- qt+l (Kt+2 - It+~) ~l t " t +l : 1 + rt+l Because t he pr oduct i on funct i on is homogeneous of degree one and qt+l z 1 + a ( I t +l / Kt +l ) [recall eq. (3.13)], the foregoi ng rel at i onshi p, i t er at ed forward successively to el i mi nat e t erms of form q~Ks+l , i mpl i es t hat qt Kt +l = Rt , s Os F ( K, , L s ) - ws L s - l s - 2Kss ] ~ Vt+l s = t + l (3.16) Thus, t he opt i mi zi ng firm' s e x d i v i d e n d mar ket val ue at the end of peri od t, Vt+I, is t he same as the shadow val ue of its i nst al l ed capital: in Hayashi ' s (1982) t ermi nol ogy, average q equal s margi nal q.i4 To see some of the model ' s predi ct i ons for current account and i nvest ment dynami cs, let us simplify by assumi ng t hat r, 0, and L are const ant and consi der a si t uat i on in whi ch the capital stock K is i ni t i al l y bel ow its st eady st at e level. Thus t he mar gi nal product of capital, OF K( Kt , L ) exceeds r, q > 1, and bot h K and Y are expect ed to rise over time. As t he capital st ock rises t owar d its st eady- st at e level, K, wher e OFK( f f ; , L ) = r, q falls to one. is Accordi ng to (3.9) t he cur r ent account (abstracting from gover nment - spendi ng changes, and assumi ng t he economy consists of a represent at i ve firm owned by a represent at i ve agent ) 1 3 E q u a t i o n ( 3. 15) d e f i n e s t h e ex di vi dend s h a d o w p r i c e o n d a t e t b e c a u s e i t d o e s n o t i n c l u d e t h e p r o f i t e a r n e d o n t h a t d a t e a n d p a i d o u t t o s h a r e h o l d e r s as d i v i d e n d s . 14As H a y a s h i s h o ws , t h i s e q u a l i t y i s d u e t o t h e h o m o g e n e i t y o f d e g r e e o n e of t h e p r o d u c t i o n a n d i n s t a l l a t i o n - c o s t f u n c t i o n s . l S F o r d i s c u s s i o n s o f t h e s e d y n a mi c s , s e e A b e l ( 1982) a n d S u m m e r s ( 1981) . Ch. 34: The Interternporal Approach to the Current Account is gi ven by 1751 { r E O F ( K t , L ) 2Kt ] l + r s=t C A t r ~ ( 1 ~s t q s - - 1 K s ] - a 1 + r s=t - - U- Since K is rising over t i me, final out put is rising over t i me as well. Since i nves t ment also is falling, 16 dat e t i nst al l at i on costs exceed t hei r pe r ma ne nt level. Thus, l~t = O F ( K t , L ) - a : ~ ( I i / K t ) falls short of its pe r ma ne nt l evel , cont r i but i ng a negat i ve c ompone nt t o t he cur r ent account . The fact t hat i nvest ment also is above its pe r ma ne nt l evel cont r i but es a second negat i ve c ompone nt to t he cur r ent account . Suppose K i ni t i al l y lies bel ow its l ong- r un st eady- st at e val ue because t he e c onomy has j ust exper i enced an unant i ci pat ed pe r ma ne nt i ncrease in f act or product i vi t y, 0. We can see f r om t he pr ecedi ng analysis t hat t he e c onomy runs a cur r ent - account deficit. A classical exampl e of this dynami c is Nor way in t he 1970s (Fi gure 2.3), whi ch bor r owed ext ensi vel y t o bui l d up its Nor t h Sea oil pr oduct i on fol l owi ng t he first oi l -pri ce shock. Not e t hat an unant i ci pat ed pe r ma ne nt rise in 0 not onl y l eads t o a cur r ent account deficit but causes an i mmedi at e rise in out put as well. Ther ef or e, with pr oduct i vi t y shocks t he model can in pri nci pl e expl ai n t he wel l - document ed count er cyel i cal behavi or of t he cur r ent account , t7 Of course, this resul t rests on t he assumpt i on t hat t he rise in 0 is per manent . I magi ne, in cont rast , an unexpect ed rise in 0t l ast i ng onl y a single per i od. Thi s change will not affect i nves t ment pl ans f or f ut ur e dat es, and so t he onl y ef f ect is a l evel of out put t empor ar i l y above t he pe r ma ne nt level. Equat i on (3.9) shows t hat a cur r ent account surplus will arise on dat e t. Pr oduct i vi t y shocks wi t h gr eat er per si st ence cause some i nvest ment and hence smal l er initial cur r ent - account surpluses, and, if per si st ent enough, t hey can pr oduce an initial deficit, as in t he pe r ma ne nt c a s e . 16I nvest ment equal s (q - ])K/a, but wi t h q falling whi l e K is rising, it may not be obvi ous t hat 1 falls over t i me. The as s er t i on in t he t ext is al ways t r ue in t he ne i ghbor hood of t he s t eady st at e. To see why, obs er ve t hat It+i-It = Kt [ ~ q t ) + ( qt . l - a 21)(qt-1)]_ t o a f i r st - or der appr oxi mat i on. 17See t he art i cl e by Mor r i s Gol dst ei n and Mohsi n S. Khan in vol ume 2 of this Ha ndbook (Gol d- st ei n and Kha n 1985). 1752 M. Obstfeld and K. Rogof f The r eader shoul d not e t hat in t he pr es ent one- good smal l - economy model , changes in i nvest ment can affect consumpt i on, but i nvest ment itself is det er - mi ned by el ement s t hat are independent of cons umpt i on pr ef er ences. I nves t ment deci si ons ar e ma de t o maxi mi ze t he pr esent di scount ed val ue of t he count r y' s out put , eval uat ed at t he wor l d i nt er est rat e. The count r y' s saving behavi or is ir-- r el evant . Thi s i ndependence of i nvest ment f r om cons umpt i on pr ef er ences need not hol d i f t he economy uses capi t al t o pr oduce nont r aded goods and servi ces as well as t r adeabl es; it can also br eak down if t her e is a nont r aded i nput in pr oduct i on, such as labor. I nt r oduci ng nont r aded goods and servi ces l eads t o a numbe r of ot her modi fi cat i ons of our basi c one- good model ' s results, as we shall see in t he next section. 3.1.4. Nontradeables, consumption, and investment An e c onomy t hat consumes and pr oduces nont r adeabl es as wel l as t r adeabl es ma y behave qui t e di f f er ent l y f r om t he one exami ned so far. We begi n by seei ng how nont r adeabl es can affect consumpt i on decisions. In this sect i on, it will be conveni ent t o r ei nt er pr et C as a composi t e index of t he r epr es ent at i ve i ndi vi dual ' s cons umpt i on of t r adeabl es and nont r adeabl es, CT and CN, but r et ai n t he assumpt i on t hat t he per i od ut i l i t y f unct i on u(C) is i soel ast i c wi t h i nt er t empor al subst i t ut i on el ast i ci t y o-. Assume, mor eover , t hat compos i t e cons umpt i on C has t he CES form: C = [c~l/C (-1)/ + (1 - 'a)l/"C(y-l)/]/(-l) (3.17) He r e , p is t he ( i nt r at empor al ) subst i t ut i on el ast i ci t y bet ween t r adeabl es and nont r adeabl es. Take t r adeabl es as t he numer ai r e commodi t y and l et p be t he pri ce of nont r adeabl es in t er ms of t r adeabl es. Then, as one can easi l y show, t he exact consumer - pr i ce i ndex (CPI) in t r adeabl es, def i ned as t he mi ni mal cost in t r adeabl es of a uni t of subut i l i t y C, is gi ven by P = [a + (1 - cOpl-] l/(1-) (3.18) For t he speci al case O = 1 ( Cobb- Dougl as pr ef er ences) , P = pl 4. Thes e same cal cul at i ons r eveal t hat , gi ven C, t he opt i mal cons umpt i on levels f or t r adeabl es and nont r adeabl es are: CT = c~ C, CN = (1 -- c~) C (3.19) Gi ven t he i nt r at empor al al l ocat i on rul es in (3.19), t he consumer ' s intertempo- ral deci si on pr obl em can be anal yzed ent i r el y in t er ms of composi t e cons umpt i on C and t he pri ce i ndex P. The cons umer maxi mi zes (3.4) subj ect t o t he budget const r ai nt cor r espondi ng t o (3.5) but wr i t t en in t er ms of expendi t ur e on C: Ch. 34: The lntertemporal Approach to the Current Account 1753 O<3 O0 E Rt,sPsCs = (1 + rt)(Vt + Bt ) + Z n, , s( wsLs - Ts) s=t s=t ( Above, asset stocks, wages, and t axes still are expr essed in units of t r adeabl e@ The i nt er t empor al Eul e r equat i on f or t he consumpt i on i ndex C is now c , + ~ = / 3 ~ ( l + r t + ~ ) ~ ~ c~ . . . . Thi s r el at i onshi p is anal ogous to Eul e r eq. (3.7), except t hat over al l consumF, li.<,~ gr owt h depends on t he ut i l i t y-based real i nt er est f act or (1 + rf+l)(Pt/Pt+! ), and not si mpl y on t he r el at i ve i nt er t empor al pri ce of t r adeabl es 1 + rt ~1. [Dornbuscl" (1983) stresses this point. ] Combi ni ng eqs. (3.19) and (3.20) shows t hat t he Eul er equat i on f or tradeables expendi t ur e, CT , is CTt+I = / 3 ( 1 + r t +l ) ( Pt ~ -0 CTt (3.21) Pt+l J Equat i ons (3,20) and (3.21) show, in part i cul ar, t hat consumpt i on need no l onger be i nt er t empor al l y s moot hed when t he t i me- pr ef er ence r at e and worl d t r adeabl e goods i nt er est r at e coi nci de. For exampl e, if t he CPI, P, is rising over t i me, t he real i nt er est r at e will be bel ow t he own i nt er est r at e on t r adeabl es, r. So t ot al cons umpt i on expendi t ur e measured in tradeables, P C, will fall over t i me if o- > 1 and rise over t i me if o- < 1. Similarly, whi l e a rising pat h of P tilts t hat of C downwar d wi t h an el ast i ci t y or, t he pri ce changes cause CT to rise r el at i ve t o C wi t h el ast i ci t y p [see eq. (3.19)]. If p > o- and 13(1 + r) = 1, for exampl e, cons umpt i on of t r adeabl es will rise over t i me if P is rising} s Some gener al - equi l i br i um i mpl i cat i ons of t hese poi nt s can be i l l ust rat ed by a si mpl e model in whi ch t he economy' s out put of nont r adeabl es, YN, is exogenous. I f we assume ( f or si mpl i ci t y) t hat nei t her t he gover nment nor t he domest i c i nves t ment pr ocess uses up nont r adeabl es, t hen t he condi t i on of equi l i br i um in t he nont r adeabl e sect or is si mpl y CN = YN, and by (3.19) t he equi l i br i um r el at i ve pr i ce of nont r adeabl es is - - o l ) C T ] 1 / p p = [[ (1 ~ - N ] In t he Cobb- Dougl a s case p -- 1, and it is si mpl e t o wri t e down t he equilibo r i um gr owt h pr ocess f or t r adeabl es cons umpt i on by combi ni ng t he pr ecedi ng expr essi on wi t h eq. (3.21): (1 . ) ( ~ - 1 ) o- ( YNt+l "~ ,~--(~-I) CT,< = / 3 ~ ( 1 + r t < ) ~ \ YYt J CTt (3.22) 18For a panel of 13 developing countries, Ostry and Reinhart (1992) present Euler-eqnation estimates of ~r = 1.27 or 1.22 and p = 0.38 or 0.50 (depending on the instrumental variables used in estimation). These results indicate that the case o- > p may be relevant. :1754 M. Obstfeld and K. Rogof f Not i ce t hat even when nont r adeabl es out put is const ant , t he gr owt h of t r ade- abl es cons umpt i on usual l y reflects t he pr esence of nont r adeabl es in t he con- sumpt i on basket . The reason: gr owt h in CT affect s p and thus t he domest i c r eal i nt er est r at e. A f undament al i mpl i cat i on of t hese resul t s is t hat empi ri cal st udi es of t he in- t e r t e mpor a l appr oach shoul d distinguish car ef ul l y be t we e n fl uct uat i ons in t r aded and nont r aded out put s. As an exampl e, suppose t hat t he pat h of t r aded out put is flat, t hat /3(1 + r) = 1, and t hat YNt < YNt+l. Accor di ng t o eq. (3.22), con- sumpt i on of t r adeabl es will be rising if o- > 1 and fal l i ng if o- < 1. Because t he cur r ent account surplus is t he di f f er ence bet ween t he economy' s e ndowme nt of t r adeabl es and its absor pt i on of t r adeabl es, even t he sign of t he cur r ent account bal ance ma y not be r el at ed in a si mpl e way t o t he t i me pat h of t he economy' s t ot al real out put . One can add a l abor-l ei sure t r adeof f t o t he i nt er t empor al model by vi ewi ng l ei sure as a nont r adeabl e (assumi ng t her e is no i nt er nat i onal mi grat i on). In this cont ext t he r eal wage pl ays t he rol e t hat t he r el at i ve pri ce of nont r adeabl es p pl ayed above. ~9 Al so, as we suggest ed at t he end of t he pr ecedi ng section, t he pr esence of a sect or pr oduci ng nont r adeabl es per mi t s domest i c cons umpt i on pr ef er ences t o af f ect t he economy' s i nvest ment behavi or. A shift in pr ef er ences t owar d non- t r adeabl es, f or i nst ance, will r educe i nvest ment if t he nont r adeabl es sect or i s t he r el at i vel y l abor - i nt ensi ve one. 2 Al l owi ng f or nont r adeabl es is an i mpor t ant st ep t owar ds havi ng a mor e real - istic model of cur r ent account behavi or. In t he next sect i on we consi der a not he r i mpor t ant modi fi cat i on. 3.1.5. Consumer durables and the current account Ei ght een per cent of 1993 Uni t ed St at es cons umpt i on spendi ng was devot ed t o dur abl es (i ncl udi ng cl ot hi ng and shoes). But t he t he or y we have devel oped t hus f ar does not capt ur e t he possi bi l i t y t hat cons umer pur chases in one per i od ma y yi el d ut i l i t y over several peri ods. We now i l l ust rat e how t he pr esence of dur abl es can al t er cur r ent account responses. In this subsect i on, l et C st and f or t he i ndi vi dual ' s cons umpt i on of nondur abl es and l et D be t he stock of dur abl es he or she owns; all goods can be t r aded. A st ock D of dur abl es yi el ds a pr opor t i onal servi ce flow each per i od, i ncl udi ng t he per i od in whi ch it is acqui red. The cons umer in a small count r y maxi mi zes u , = / 3 s - ~ [ ~ l o g c , , + ( 1 - ~ ) l o g D , ] s t 19For an early analysis of the effects of introducing nontraded labor on current account dynamics, see Bean (1986). 2See, for example, Murphy (1986) and Engel and Kletzer (1989). Ch. 34: The l nt ert ernporal Ap p r o a c h to t he Current Ac c o u n t 1755 subj ect t o t he fi nance const r ai nt s At +l - - A t = r At + Yt - Ct - Gt - (Kt+j - Kt ) - p t [ Dt - (1 - 6) Dt - l l and a sol vency condi t i on. Her e, Dt is t he st ock of dur abl es (including newl y pur chas ed dur abl es) hel d over per i od t, Pt is t hei r pr i ce in t er ms of non- dur abl es, 6 is t hei r depr eci at i on r at e, and r, t he wor l d i nt er est r at e, is const ant . Not e t hat A, Y, C, G, and K are meas ur ed in nondur abl es. Dur abi l i t y i mpl i es 6 < 1 . In addi t i on t o t he usual i nt er t empor al Eul e r equat i on f or nondur abl es c o n sumpt i on, t he i ndi vi dual ' s fi rst -order condi t i ons i ncl ude a Dt - - P t - ~ P t < =- ~t, whi ch equat es t he mar gi nal r at e of subst i t ut i on of nondur abl es f or t he services of dur abl es t o t he us er c os t or r ent al pri ce of durabl es. As s umi ng/ 3 = 1/ ( 1 + r), cons umpt i on of nondur abl es is Ct = -~-+-Tr ( l + r ) At + ( 1 - a ) p t Dt _ , + \ ~+Tr /t (Y~ - ls - Gs ) s=t whi l e cons umpt i on of dur abl es' services is 0 - - ~ ) r 1 ( v, t,, - G) Dt - - ~( l +r ) ( l +r ) A t +( 1 - 6 ) p t Dt =l + l +r ~ " S=t Let ' s suppose f or si mpl i ci t y t hat , Pt and, hence, t he user cost, Lt, are const ant . The last equat i ons t hen i mpl y t hat t he cons umer smoot hs, not t he pat h of expen- di t ur es p [ Dt - ( 1 - 6) Dr _l ] on durabl es, but t he s e r v i c e flow f r om durabl es, which is pr opor t i onal t o Dt . Wi t h durabl es, t he cur r ent account is t he same as i mpl i ed by (3.9) (assuming, as we have her e, t hat r = (1 - / 3 ) / ~ ) , but wi t h an addi t i onal t er m t hat depends on new dur abl es pur chases ( we o mi t t he de r i v at i on) : = {(Yt - Y t ) - ( Gt - ~at) - (It - J, )} + (~ - p) AD, (3.23) C A t Not i ng t hat t he pri ce of purchasi ng a dur abl e out r i ght , p, must be gr eat er Tha n t he one- per i od user cost ~, we see t hat t he i nt r oduct i on of dur abl es neces- sari l y i ncreases t he vol at i l i t y of cur r ent - account r esponses t o unexpect ed i ncome changes. 3. 1. 6. T h e t e r ms o f t r ade a n d t he t r ans f e r p r o b l e m One of t he earl i est pr obl ems mot i vat i ng t he i nt er t empor al appr oach was t he need t o under s t and how changes in t he t er ms of t r ade - t he pr i ce of a count r y' s 1756 M. Obst f el d and K. Ro g o f f expor t s in t er ms of its i mpor t s - affect saving and t he cur r ent account . Ear l y ap- pl i cat i ons of Keynes i an model s by Ha r be r ge r (1950) and Laur s en and Met zl er (1950) had mode l e d adver se t er ms- of - t r ade shocks as real i ncome- r educt i ons t hat r educe savi ng and t he ext er nal surplus in pr opor t i on to Keynes' s mar gi nal pr opens i t y t o save. Inst ead, t he i nt er t empor al appr oach emphasi zed t he re- sponse of f or war d- l ooki ng i ndi vi dual s t o t he changes in l i f et i me cons umpt i on possi bi l i t i es t hat t er ms- of - t r ade movement s cause. The si mpl est case is of a speci al i zed e c onomy wi t h an exogenous e ndowme nt Y of its expor t good, but whi ch also consumes i mport s. As in t he model wi t h nont r adeabl es , we agai n assmne an i soel ast i c per i od ut i l i t y f unct i on def i ned over a CES i ndex C; her e it depends on t he i ndi vi dual ' s consumpt i ons C~t of i mpor t s and Cx of export s: Le t p now de not e t he pri ce of expor t s in t er ms of i mport s, whi ch is det er mi ned exogenous l y in t he wor l d mar ket . The cons umpt i on- bas ed pri ce l evel in t er ms of i mpor t s is agai n denot ed by P and gi ven by f or mul a (3.18). The nat ur al benchmar k case, assumed by Svensson and Razi n (1983), sup- poses t hat i nt er t empor al t r ade is done t hr ough bonds i ndexed t o t he consump- t i on i ndex, C. I n this case r is t he own r at e of i nt er est on t he cons umpt i on i ndex, and t he budget const r ai nt cor r espondi ng t o (3.3) in t he pr esent set up ( abst r act i ng f r om i nvest ment and gover nment ) is 2 2 Rt,sCs = (1 + rt)At + Rt,s S : t S=I The Eul e r equat i on f or t he consumpt i on i ndex is agai n (3.7), i mpl yi ng t he con- sumpt i on f unct i on cor r espondi ng t o (3.8), whi ch has I = G = 0 and p Y / P in pl ace of Y. To focus on t he t er ms of t r ade, it is hel pful t o assume t hat r = (1 - ~) / / ~ and Y are const ant , in whi ch case t he cons umpt i on f unct i on r educes to: Ct = rat + ~ Z \ l~--rr J S=t A fall in t he t er ms of t r ade l owers p (t he r el at i ve pr i ce of expor t s in t er ms of i mpor t s) r el at i ve t o P ( t he overal l CPI in t er ms of i mport s). Thus, fl uct uat i ons in t he t er ms of t r ade affect t he cons umpt i on i ndex and t he cur r ent account (whi ch he r e is meas ur ed in consumpt i on- i ndex units) exact l y like fl uct uat i ons in GDP at const ant t er ms of t rade. In part i cul ar, a t e mpor a r y t er ms - of - t r ade set back causes a cur r ent - account deficit, wher eas a pe r ma ne nt set back causes an i mmedi at e shi ft t o t he new, l ower cons umpt i on l evel consi st ent wi t h ext er nal Ch. 34. The Intertemporal Approach to the Current Account 40 C P 30 20 "8 P 10 "~ 0 O e e "6 -10 e e -20 AUS USA NOR JAP NET ITA FRA SWE B m SPA BGE R DEN & "c"" BEL -0.15 -0.1 -0.05 0 0.05 0.1 0.15 Change in Net Forei gn Asset s/ GDP Figure 3.1. Real currency appreciation versus the change ill foreign assets, 1981-1990. 1757 balance. Obstfeld (1982) and Svensson and Razin (1983) illustrate how this latter result depends on the strong intertemporal separability of utility. 21 This subsection has, thus far, focused on the response of the current account to exogenous terms-of-trade changes. When countries have some monopoly power in trade, however, shifts in their current accounts may influence terms of trade by redistributing wealth internationally. The impact of international wealth transfers on the terms of trade is a classic problem of international finance. In the 1920s, Keynes and Ohlin disagreed on the price effects of German reparations; in the 1990s, observers of the protracted United States external deficits have debated the need for a fall in the relative price of American exports. 22 For a cross-section of 15 OECD countries, Figure 3.1 plots the percent change in the trade-weighted WPI real exchange rate (a terms-of-trade proxy) against the change in the ratio of net foreign assets to output. The net foreign asset series attempt to account not only for measured current account flows, but for 21Ostry (1988) and Edwards (1989) study interactions between the terms of trade and the relative price of nontradeables. Gavin (1990) considers the terms of trade in a richer dynamic setting. For developing countries, it may be natural to assume that bonds are indexed to imports (e.g. "dollars"). In that case the intertemporal budget constraint differs from the one we have analyzed, so that the Euler equation for C is formally the same as (3.20). If the world bond rate is constant, expected terms of trade movements therefore have consumption-tilting effects through the consumption-based real interest rate. 22See, for example, the papers in Bergsten (1991). 1758 M. Obstfeld and K. Rogoff capi t al gai ns and losses. 23 The changes are cal cul at ed as t he 1986-1990 aver age less t he 1981-1985 average. The fi gure shows a distinct posi t i ve r el at i onshi p - an i ncrease in a count r y' s net f or ei gn assets appear s t o be associ at ed wi t h an i mpr ove me nt in its t er ms of t r ade. The l east - squar es r egr essi on line is A l og( pj ) = 0.039 + 1. 0422x(A/Y)j + uj; R 2 = 0.31 (0.027) (0.433) i mpl yi ng a st at i st i cal l y significant rel at i onshi p. Accor di ng t o t he regressi on, an i ncr ease of 1 per cent in t he r at i o of net f or ei gn assets t o out put is associ at ed wi t h a 1 per cent i mpr ove me nt in t he t er ms of t r ade. Obvi ousl y, this regressi on pr ovi des onl y a nonst r uct ur al cor r el at i on. To assess accur at el y t he consequences of an exogenous weal t h t r ansf er woul d r equi r e a full e c onome t r i c st ruct ural model . Gi ven t he per enni al i mpor t ance of t he t r ansf er ef f ect in pol i cy discussions, t he de ve l opme nt of empi ri cal i nt er t empor al model s t hat can expl ai n its magni t ude is a r esear ch priority. Theor et i cal l y, t he t r ansf er ef f ect can oper at e t hr ough several channel s in a gener al - equi l i br i um setting. Two pot ent i al mechani sms are home pr ef er ence f or domest i c expor t s [see, f or exampl e, Bui t er (1989)] and t he pr esence of a non- t r adeabl es sect or t hat compet es f or r esour ces wi t h t he expor t abl e sector. A speci al case of t he l at t er mechani sm is due t o t he ef f ect of a weal t h change on l abor suppl y and, hence, on t he suppl y of expor t abl es. A si mpl e t wo- count r y model exhi bi t s t he l abor - suppl y channel f or t he t r ansf er effect . As s ume t hat t he home count r y speci al i zes in pr oduci ng good X and t he f or ei gn count r y good M. Mor eover , assume t hat in bot h count r i es p, t he el ast i ci t y of subst i t ut i on in demand bet ween t he t wo pr oduced goods, equal s 1. Le t t he per i od ut i l i t y funct i on, c ommon t o t he home and f or ei gn count r i es, be an i soel ast i c f unct i on of an i ndex t hat depends on bot h cons umpt i on and l ei sure, ( H - = - ) ( H - C ~ L ) I - ~ wher e 0 < ~ < 1, H is an i ndi vi dual ' s e ndowme nt of t i me, and L his or her l abor supply. Le t home out put be pr oduced accor di ng t o t he pr oduct i on f unct i on Y = wL and f or ei gn out put accordi ng t o Y* = ~o*L*. Le t P once agai n denot e 23Real exchange r at e dat a come from Int ernat i onal Monet ar y Fund, International Financial Statis- tics. Net foreign asset positions come from OECD Economic Outlook 55 (June 1994), p. A54. The count ry sampl e is Aust ral i a (for which t he CPI was used), Bel gi um-Luxembourg, Canada, Den- mark, Finland, France, Germany, Italy, Japan, Net herl ands, Norway, Spain, Sweden, t he Uni t ed States, and t he Uni t ed Kingdom. We are skeptical t hat t he real appreci at i on of 33 percent t hat t he I MF report s for Norway is accurate. (The I MF report s a similar number based on CPIs.) OECD dat a place Norway' s real appreci at i on over 1981-1990 closer to 15 percent . However, t he observat i on on Norway is not driving t he regression results report ed below. Ch. 34: The Intertemporal Approach to the Current Account 1759 t he pr i ce of t he i ndex C in t er ms of home i mport s, and assume bonds are i ndexed t o C. The home count r y' s net ext er nal assets are A, and t he f or ei gn count r y' s, t her ef or e, ar e - A. Usi ng t he st eady- st at e cons umpt i on funct i ons to sol ve f or de ma nd and suppl y in ei t her of t he goods mar ket s, one can show ( af t er ma n y s t eps ) t hat t he equi l i br i um t er ms of t r ade satisfy t he condi t i on poea~H (1 - oe)w*H* _ (1 - (:)(1 - f l ) . A e P ~/3 Fr om this i mpl i ci t t er ms- of - t r ade equat i on one can easi l y see t he effect .,! an exogenous shift in net ext er nal assets f r om t he f or ei gn to t he home count-~y (t hat is, of a ri se in A). An absol ut e rise in p, t he r el at i ve pr i ce of home ex- port s, raises p r el at i ve t o P, and P itself rises absol ut el y. If A rises, a rise m p t her ef or e mai nt ai ns equi l i bri um. The i nt ui t i on why a weal t h t r ansf er to t he home count r y i mpr oves its t er ms of t r ade is simple. If home resi dent s r ecei ve a financial wi ndfal l , t hey spend some of it on leisure. Out put of t he home good t hus falls and its r el at i ve pri ce must rise. 3. 1. Z De mo g r a p h i c st ruct ure, f i s c al pol i cy, a n d t he c ur r e nt a c c o u n t The single r epr es ent at i ve agent par adi gm f ol l owed so far in this chapt er may furni sh mi sl eadi ng pr edi ct i ons about t he cur r ent account when t he economy con- sists i nst ead of het er ogeneous fami l i es bor n on di f f er ent dat es and unconnect ed by al t rui st i c links. A model based on t he over l appi ng- gener at i ons st r uct ur e in Weil (1989) i l l ust rat es some possi bl e i mpl i cat i ons of al l owi ng f or demogr aphi c compl exi t y. 24 We will i dent i f y t he economy' s t ot al popul at i on wi t h its l abor supply, Lt , whi ch grows at r at e n: Lt+ 1 = (1 + n ) L t , wi t h L0 nor mal i zed to 1. Thi s l abor f or ce, however , consists of i mmor t al unconnect ed i ndi vi dual s (or dynast i es) bor n on successive dat es. A r epr esent at i ve f r om t he gener at i on bor n on dat e v (t he gener at i on' s " vi nt age" ) maxi mi zes Uo,t = ~ . ~ t fls t u( Co, , ) subj ect t o oo OO s=t s - t ( wher e Bv,t denot es vi nt age v' s bond hol di ngs and each i ndi vi dual supplies one uni t of l abor per per i od) . Not i ce t hat cons umpt i on and weal t h are vi nt age- specific, wher eas all peopl e face t he same wage, i nt er est rat es, and (by assump t i on) l ump- sum taxes. A key assumpt i on is t hat Vv,~ = By# = 0: vi nt ages are bor n wi t h no fi nanci al weal t h, onl y with a l i f et i me e ndowme nt equal t o t he 2aFor alternative open-economy models with overlapping generations, see Buiter (1981, 1989), Blanchard (1985), Persson (1985), Frenkel and Razin (1987), and Eaton (1988). 1760 M. Obs t f el d and K. Ro g o f f pr es ent val ue of aft er-t ax l abor i ncome. ( One can t hi nk of t he model as one in whi ch pr i mogeni t ur e gover ns t he beques t of f ami l y weal t h. ) I nves t ment can be mode l e d exact l y as in Sect i on 3.1.3. The onl y di f f er ence is t hat a growi ng l abor f or ce r equi r es posi t i ve st eady- st at e i nvest ment . To ma ke t he mai n poi nt s it suffices t o wor k wi t h t he special case in whi ch t he i nt er est r at e is const ant at r and t he per i od ut i l i t y f unct i on is u ( C ) = l o g ( C) Y In this case, Cv, t --- (1 - / 3 ) (1 + r ) ( V v , t + B y , t ) + Z ( w s - T s ) (3.24) S=t is t he cons umpt i on funct i on. In i nvest i gat i ng t he economy' s aggr egat e behavi or, it is hel pful t o wor k wi t h per capi t a aggr egat e measur es of macr o- var i abl es such as cons umpt i on and weal t h. As s ume t hat t he e c onomy starts on dat e 0. Let X v , t be t he vi nt age- specific val ue of var i abl e X on dat e t. Obs er ve t hat t he size of gener at i on v = 0 is 1, t hat of gener at i on 1 is (1 + n) - 1 = n, t hat of gener at i on 2 is (1 + n) 2 - (1 + 17) = n(1 + n), and so on up t hr ough vi nt age t, whi ch is of size n(1 + n ) t a. Thus t he per capi t a aver age val ue of X on dat e t is: X t = X o , t + n X l , t + n(1 + n ) X 2 , t + . . . + n(1 + n ) t - l X t , t (1 + n ) t (3.25) To comput e aggr egat e per capi t a consumpt i on, obs er ve t hat Vt, def i ned as in t he last equat i on, is t he aver age per capi t a val ue of claims t o domest i c firms at t he end of per i od t - 1 and is still gi ven by (3.11) if t he quant i t i es in t hat equat i on ar e i nt er pr et ed as per capi t a aggregat es. Equat i ons (3.11), (3.24), and (3.25) t hus i mpl y 26 Ct = (1 - / 3 ) (1 + r ) B t + Z ( Y s - T s - I s ) (3.26) S=t Above, cons umpt i on has be e n expr essed as a f unct i on of t he t i me pat h of taxes. To under s t and t he rol e of t axat i on, however, we must f or t he first t i me in this chapt er expl i ci t l y consi der exact l y how t he domest i c gover nment coor di nat es t he t i me pat hs of publ i c expendi t ur es and taxes. Le t D t now denot e t he gover nment ' s per capi t a debt at t he end of dat e t - 1. The n if G t denot es per capi t a gover nment consumpt i on, t he i nt er t empor al publ i c- sect or budget const r ai nt on dat e t is 25This c a s e c o r r e s p o n d s t o i soel ast i e pr e f e r e nc e s wi t h o- = 1. 2670 der i ve t hi s e qua t i on, we a s s u me pe r f e c t f or es i ght b e t we e n da t e s t - 1 and t, so t ha t (1 +r ) Vt = Yt - w t - I t + Vt+l. Th e n e x t e qua t i on hol ds, however , e ve n wi t hout t he pe r f e c t f or es i ght a s s umpt i on, as c a n be s e e n by modi f yi ng c o n s u me r s ' i nt e r t e mpor a l budge t c ons t r a i nt s as e xpl a i ne d in t he f oot not e pr e c e di ng eq. (3.5) above. I n der i vi ng (3.26), we ha ve ma d e us e of t he fact t ha t t he pr oduc t i on f unct i on is h o mo g e n o u s of de gr e e one. Ch. 34." The I nt ert emporal Appr oac h to the Current Ac c ount 1761 (l+r)(l+n)tDt+ \ - ~r ] (l+n)SG* = H (1 + n)"T, S=l ,9=f This constraint equat es the present value of tax revenues to the present value of spending plus initial debt. If r < n nei t her the government ' s revenue nor its spending has a finite present value, so we assume r > n. Division of the precedi ng constraint by (1 + n) t renders it in per capita t erms as: 0o (1 +n'~ st (1 +n~ '~' (l +r)Dt +~_~\ l +rj Gs = \ l +r ] Ts S=t S=t To see what this constraint implies for private consumption, lead it one period and observe that (1 + r)Dt + Gt - Tt = (3.27) Dt+l 1 + n Solving eq. (3.27) for Tt and taking present values, we obtain \ ~--+-rr,i Ts = (1 + r)Dt - Z \ Hi nDs+ Z \ HI G,, s=t S=/ +l s=t which, aft er rearranging, can be expressed in the form: ( 1 ~s-t = ( 1 "~ s-' n \ HJ Ts HJ Gs+ ( l +r ) ( l - r ) Dt s=t s =t _ f 1 , ~s- , ( _~) -~ \ ~- ~r J n(Ds+ 1 - D, ) (3.28) S=t Expressi on (3.28) shows that in the overl appi ng-generat i ons economy, gov- er nment debt is net wealth to the private sector and higher future government deficits lower the present value of taxes for those current l y alive, holding the pat h of gover nment consumpt i on constant. Why is this so? Equat i on (3.27) shows that if Dt were a unit higher the government would have to raise taxes in all fut ure peri ods by only r - n to keep the per capita public debt constant. Thus, an extra unit of government bonds in the hands of someone alive at the start of date t raises his or her discounted stream of tax liabilities by only H (r - n) = (l + r) 1- r 5' =t How do current and future government deficits alter the tax bill of current generat i ons? Let the government cut per capita taxes by 1 + n units in peri od t and finance this tax cut by issuing enough additional debt to make Dt+l one unit higher [see eq. (3.27)]. If the government is to mai nt ai n this new higher 1762 M. Ob s t f e l d a n d K. R o g o f f per capi t a debt l evel , it mus t rai se t axes by r - n pe r capi t a f r om dat e t + 1 on. Thus, f or s o me o n e al i ve on dat e t, t he net ef f ect of a defi ci t -fi nanced t ax cut is t o l ower t he pr e s e nt val ue of t axes by 1 + n - ( r n ) = n s=t +l Def i ci t - f i nanced t ax cuts in per i ods af t er t have a cor r es pondi ng di scount ed ef f ect on t he t ax l i abi l i t i es of t hose al i ve on dat e t. The me c ha ni s m under l yi ng t hese resul t s is wel l known: t he mor e debt t he gove r nme nt issues, t he mor e t axes can be shi f t ed ont o f ut ur e gener at i ons yet unbor n. Sever i ng al t rui st i c l i nks be t we e n t hose al i ve t oday and s ome of t hose who will be al i ve t omor r ow cr eat es a si t uat i on in whi ch cur r ent gener at i ons do not ful l y i nt er nal i ze t he f ut ur e t ax l i abi l i t i es ari si ng f r om gove r nme nt debt issue. If, in cont r ast , n = 0, t her e ar e no f ut ur e ent r ant s to t he e c onomy t o be t axed and, as (3.28) shows, t he t i me profi l e of gove r nme nt debt no l onger ma t t e r s t o economi c agent s. I n this case, g o v e r n me n t debt is not net weal t h and t he Ri car di an equi val ence of deficit- and t ax- f i nanced gove r nme nt expendi - t ur e hol ds. I n t he r epr es ent at i ve- agent e c onomy cons i der ed earl i er, gove r nme nt bor r owi ng has no i nfl uence on t he cur r ent account . Thi s is not t he case here. Because t he e c onomy' s over al l net f or ei gn asset s a r e A t - - B t - Dt , subst i t ut i on of (3.28) i nt o c ons umpt i on f unct i on (3.26) shows t hat C, = (1 - f i ) D t + ~ n ( D s + l - D s ) S=t + (1 + r ) A t + \ ] ~ T J ( Y s - I s - G , . ) S~t I n per capi t a t er ms, t he dat e- t cur r ent account is (1 + n ) A t + l A t = r a t + Y t - C t - I t - G t , so, ot her t hi ngs equal , a hi gher cur r ent l evel of gove r nme nt debt or a hi gher t r aj ect or y for f ut ur e gove r nme nt deficits will r ai se cur r ent cons umpt i on and t he e c onomy' s f or ei gn bor r owi ng. De s pi t e s ome not abl e epi sodes, it has pr ove n difficult t o ver i f y a st r ong st at i s- tical cor r el at i on be t we e n budget and cur r ent - account deficits. Bel ow we pr es ent cr oss- sect i on r egr essi ons f or i ndust r i al i zed count r i es ove r f i ve- year s ubs ampl es of 1976-1990. 27 Each s ubs ampl e cr oss- sect i on obs er vat i on is f or me d by aver - 27The sampl e consists of t he OECD countries as of 1994, except for Belgium, Luxembour g and Turkey. The dat a for current -account deficits and cent ral gover nment budget deficits relative to GDP come from I nt er nat i onal Monet ar y Fund, I nt e r nat i onal Fi n a n c i a l St at i st i cs. Due to lack of data, Switzerland is omi t t ed from t he 1981-1985 regression and Canada, New Zeal and, and Switzerland are omi t t ed from t he 1986-90 regression. Regressions based on general gover nment deficits yield similar results [see Obsffeld and Rogoff (1996), ch. 3]. Ch. 34: The Intertemporal Approach to the Current Account 1763 agi ng a count r y' s annual cur r ent - account surpl us r at i o t o GDP and its cent ral gove r nme nt deficit r at i o t o GDP. 1976-1980 ( CA / Y ) j = - 0. 208 - 0.406 ( d i D/ Y) j + uj; R 2 = 0.23 (0.907) (0.171) 1981-1985 ( CA / Y ) I = 0.995 - 0.506 ( A D/ Y ) j + uj; R 2 = 0.32 (1.206) (0.173) 1986-1990 ( CA / Y ) j = - 0. 881 - 0.016 ( A D/ Y ) j + uj; R 2 = 0. 00 (0.707) (0.114) Ov e r t he first t wo s ubs ampl es t her e is a st r ong negat i ve cor r el at i on bet ween budget deficits and t he cur r ent - account surpl us, as t he over l appi ng- gener at i ons mode l woul d suggest . 2s Dur i ng t he mos t r ecent s ubs ampl e, however , t he cor r e- l at i on evapor at es . The l ast r esul t suggest s t hat f act or s ot her t han gove r nme nt budget s domi na t e d sever al count r i es' cur r ent account s ove r 1986-1990. I f one runs t he s ame r egr es s i on on a s ampl e t hat i ncl udes non- i ndust r i al i zed count r i es, t he r esul t s are weaker . Bewar e: t hese si mpl e cor r el at i ons ar e me r e l y suggest i ve and have no st r uct ur al i nt er pr et at i on. I n part i cul ar, t hey cannot tell us t he ef- fect of an exogenous i ncr ease in a gove r nme nt ' s budget deficit on t he cur r ent account . An o t h e r ser i ous issue is t o choose t he a ppr opr i a t e me a s ur e of t he in- t er gener at i onal i nci dence of budget policies. For exampl e, even if t he social secur i t y account of t he budget is bal anced, it r epr es ent s a huge t r ans f er f r om young t o old. Cl earl y, a bet t er under s t andi ng of t he "t wi n defi ci t s" will r equi r e st r uct ur al model s and car ef ul at t ent i on t o i nt er gener at i onal di st ri but i on. Model s i ncor por at i ng mor e compl ex demogr aphi cs have i mpor t a nt i mpl i - cat i ons asi de f r om t hei r pr edi ct i ons about gove r nme nt fi nanci al i mbal ances. Consi der , f or exampl e, t he i mpl i cat i ons of expect ed pr oduct i vi t y growt h. I n a r e pr e s e nt a t i ve - a ge nt f r a me wor k, hi gher pr oduct i vi t y gr owt h will t end t o we a ke n t he cur r ent account as pe opl e bor r ow t oday agai nst hi gher f ut ur e i ncome. (In- ve s t me nt effect s woul d s t r engt hen this result. ) I n a l i fe-cycl e set t i ng t he i mpl i - cat i ons of hi gher pr oduct i vi t y gr owt h ar e less clear. I f each i ndi vi dual benefi t s i dent i cal l y f r om t he change r egar dl ess of his or her age, t he resul t s are much t he s a me as in a r e pr e s e nt a t i ve - a ge nt model . But s uppos e t hat pr oduct i vi t y gr owt h rai ses t he l abor i ncomes of young wor ker s but does not af f ect t he l abor i ncomes of ol der wor ker s. Becaus e young savers will count mo r e heavi l y in aggr egat e savi ng t han ol d di ssaver s, savi ng will t end to ri se and t he cur r ent account to 28In a sample including four industrialized countries and Mexico, Bernheim (1988) finds a similar correlation between budget deficits and current-account deficits over 1976-1985. 1764 M. Obstfeld and K. Rogoff i mpr ove. De a t o n (1992) discusses how l i fe-cycl e t heor i es mi ght expl ai n t hc ob- s er ved t endency f or nat i onal savi ng and gr owt h r at es t o be posi t i vel y cor r el at ed in cross sect i ons. An i mpor t a nt pr ope r t y of over l appi ng- gener at i ons model s is t hat t hey pe r mi t a s t eady- s t at e wi t h posi t i ve cons umpt i on f or a smal l open economy, even if i ndi vi dual s have fixed t i me- pr ef er ence r at es di f f er ent f r om t he wor l d i nt er est r at e. In t he Wei l (1989) mode l j ust sket ched, f or exampl e, t he smal l e c o n o my can r each a s t eady st at e even if i ndi vi dual cohor t s ' cons umpt i on l evel s ar e ri si ng or fal l i ng ove r t i me; what is needed is t hat t he bi r t h r at e of new i ndi vi dual s be l ar ge e nough t o of f set t he di scr epancy be t we e n/ 3( 1 + r) and 1. I f / 3( 1 + r) < 1, n > 0 gua r a nt e e s t hi s out come; i f/ 3(1 + r) > 1, t he i nequal i t y/ 3( 1 + r) < 1 + n is necessary. 3.2. St ochast i c models' o f the current account A t heor y of cur r ent - account de t e r mi na t i on t hat ma k e s no al l owance f or t he uncer t ai nt y a bout t he f ut ur e under l yi ng c ons umpt i on and i nves t ment deci si ons cannot be ful l y sat i sfact ory. Unf or t unat el y, however , t he i nt r oduct i on of st ochas- t i c e l e me nt s can rai se t he t echni cal difficulty of wr i t i ng down sol ut i ons t o in- di vi dual s' maxi mi zat i on pr obl e ms by an or der of magni t ude. Bel ow we r evi ew t he pr edi ct i ons of s ome l eadi ng st ochast i c model s. A ke y t he me emer gi ng f r om t he di scussi on is t hat t he cur r ent - account r es pons e t o var i ous shocks depends on whe t he r ma r ke t s exi st f or i nsuri ng agai nst t he shocks' effect s. Thi s t h e me l eads us, at t he sect i on' s end, t o consi der a model in whi ch t he ext ent t o whi ch shocks ar e i ns ur abl e is endogenous. 3.2.1. Compl et e market s The mos t t r act abl e case of uncer t ai nt y is t hat in whi ch i nsur ance ma r ke t s exi st f or all f ut ur e cont i ngenci es, wi t h out comes ful l y ver i f i abl e and cont r act s ful l y enf or ceabl e. I n t he classic Ar r o w- De b r e u wor l d of c ompl e t e mar ket s , equi l i b- r i um r es our ce al l ocat i ons ar e effi ci ent and, f r om a f or mal poi nt of view, t he e c o n o my can be anal yzed as if per f ect cer t ai nt y appl i ed. The r e ar e si mpl y ma n y mo r e commodi t i es , commodi t i es now bei ng i ndexed by t he st at e of nat ur e in whi ch t hey ar e d e ma n d e d and suppl i ed. 29 29International models with complete markets are analyzed by Lucas (1982), Stockman (1988a), Stulz (1988), Cole and Obstfeld (1991), Stockman and Tesar (1995), Backus, Kehoe, and Kydland (1992), and Baxter and Crucini (1993a), among many others. Lucas's (1982) model does not actually assume complete markets, and in fact can contain far more states of nature than assets. However, other special assumptions made by Lucas result in an allocation the same as the one complete markets would produce. Ch. 34: The lntertemporal Approach to the Current Account 1765 Except for cert ai n special cases, mar ket compl et eness requi res t hat peopl e be abl e to t r ade as ma ny i ndependent assets as t her e are prospect i ve states of nat ure. Equi val ent l y, t hey must be able t o t rade a compl et e set of Arrow-- Debr eu securities, each of which pays off onl y in one state. The result of this t rade is t hat i ndi vi dual s everywhere in the worl d equal i ze t hei r margi nal rates of subst i t ut i on of pr esent for f ut ur e st at e-cont i ngent consumpt i on to the same Ar r ow- De br e u prices, so t hat for all count ri es i and j , and all dat es t, u ' ( q + l ) ' J u ( C t + 1 ) - ( 3 . 2 9 , u ' ( C / ) , , ' ( C / ) given shared rat es of t i me preference (that is, t he / % cancel). Condi t i on (3,29) precl udes f ur t her mut ual gains, on any dat e or in any st at e, f r om i nt er-consumer risk pooling. I nvest ment decisions are made to maxi mi ze the present val ue o~ profits eval uat ed at st at e-cont i ngent out put prices. When all peopl e t hr oughout the worl d t rade prospect i ve risks in i nsurance market s, some local economi ~ shocks effect i vel y become gl obal shocks and t hei r current account effects are di mi ni shed or even el i mi nat ed. Consider, for in- stance, a pure exchange economy t hat experi ences a t empor ar y i di osyncrat i c posi t i ve out put shock. Absent i nsurance market s, t he count r y woul d run a current -account surplus, accumul at i ng some forei gn assets so as to smoot h the benefi t s of the shock over all fut ure periods. Under compl et e market s, however, t he home economy has al r eady t r aded much of its out put risk to forei gners and purchased, in t urn, claims on t hei r risky out put processes. Thus, t he home econ- omy' s posi t i ve out put shock will cause a small synchroni zed i ncrease in every count ry' s consumpt i on under compl et e market s. But it will also cause a shift in ever y count ry' s i ncome as "di vi dend" payment s flow f r om the home count r y to its forei gn sharehol ders. Wi t h i nt er nat i onal l y i dent i cal isoelastic preferences, no current -account i mbal ances result. Indeed, in t hat case, t here are never unan- t i ci pat ed current account movement s. (Regardl ess of t he ut i l i t y funct i on, dif- ferences in t i me pr ef er ence can gener at e predi ct abl e nonzer o current accounts under compl et e market s. ) If per i od ut i l i t y is not isoelastic and identical across count ri es, worl d out put shocks will, in general , cause current -account i mbal ances. Even in this case, shocks t hat affect t he di st ri but i on of out put across count ri es, but l eave worl d out put unchanged, have no current -account implications. Ret ur ni ng to t he case of i nt er nat i onal l y i dent i cal isoelastic ut i l i t y but in- t roduci ng product i on, a positive shock to t he product i vi t y of domest i c invest- ment causes a current -account deficit, as in our earl i er perfect -foresi ght models. But t hat deficit will reflect onl y an influx of savings f r om abr oad to share in ownershi p of the i ncr ement al i nvest ment . (Indeed, the existing capital of firms will al r eady have a gl obal l y dispersed ownership. ) The deficit does not reflect 1766 M. ObstJeld and K. Rogoff cons umpt i on- s moot hi ng effects, because all count r i es' i ncome profi l es are rising in pr opor t i on. Des pi t e t he deficit - i ndeed, because of it - t he i nt er nat i onal weal t h di st r i but i on is const ant . Unde r l i t er al l y compl et e mar ket s, risks due t o changes in gover nment con- sumpt i on also woul d be per f ect l y pool ed among nat i ons. An exogenous un- expect ed rise in Uni t ed St at es spendi ng on hi ghway repai rs due, say, to bad weat her , woul d be fi nanced most l y by cont i ngent - cont r act payment s f r om for- ei gners t o t he Uni t e d States. The obvi ous adver se i ncent i ves i nt r oduced by such cont r act s i l l ust r at e why, in pract i ce, asset mar ket s ar e har dl y compl et e. If a coun- t ry' s r esi dent s have sold most of domest i c firms' f ut ur e earni ngs on f or war d mar- ket s, its gove r nme nt has ever y i ncent i ve t o rai se cor por at e t axes sharpl y af t er t he fact. Mor e general l y, under asymmet r i c i nf or mat i on, mor al hazar ds affect i ng pr i vat e as wel l as gover nment behavi or i mpede compl et e risk sharing. I nf or mal obs er vat i on and statistical evi dence bot h confi rm t hat even in a domest i c con- t ext , risks are f ar f r om bei ng pool ed as t he compl et e- mar ket s par adi gm woul d predi ct . In t he i nt er nat i onal cont ext , sover ei gn risk and di st ance, t oget her wi t h cul t ural and l egal di fferences, gr eat l y magni f y t he difficulties. 3 3.2.2. Bonds as the onl y asset Le t us l ook i nst ead at t he opposi t e ext r eme, t hat in whi ch t he onl y asset nat i ons t r ade is a one- per i od bond i ndexed t o t r adeabl e cons umpt i on goods and of f er i ng a cer t ai n one- per i od r et ur n. Thi s rest ri ct i on on t he me nu of assets is t oo sever e t o furni sh a compl et el y real i st i c model , but it does serve t o pr oduce a model qui t e compar abl e t o t he det er mi ni st i c one st udi ed earlier. La t e r on we discuss hybr i d model s in whi ch t her e are mar ket s f or some, but not all, risks. For simplicity, we agai n choose as our f r amewor k a r epr es ent at i ve- agent econ- omy in whi ch all goods are t r aded. Now t he i ndi vi dual maxi mi zes t he condi t i onal expect at i on subj ect t o a budget const r ai nt like (3.5) equat i ng t he pr esent val ue of consump- t i on t o initial financial weal t h plus t he pr esent val ue of aft er-t ax l abor i ncome. The cons umpt i on levels in (3.30) are cont i ngency pl ans for cons umpt i on t hat depend on t he i ndi vi dual ' s hi st ory t hr ough t he dat e t he pl an is i mpl ement ed. The budget const r ai nt depends on st ochast i c f ut ur e earni ngs, taxes, and i nt er est rat es. The sequence of cont i ngent cons umpt i on pl ans t he cons umer chooses on 3See Obsffeld (1995) for a survey of evidence. The limited extent of international risk sharing has prompted Shiller (1993) to propose creating international markets in GDP futures. Ch. 34: The Intertemporal Approach to the Current Account 1767 date t must satisfy the budget constraint for every prospective history of the economy. The i nt ert emporal Eul er equation derived in the certainty case now holds in expectation [cf. eq. (3.6)]: u ' ( C t) / 3( i + r t + l ) E , ( 3 . 3 l ) To obtain closed-form consumer decision rules, we approximate u ( C ) by the quadratic function h C2 u( c ) - c - With this approximation (3.31) becomes 1 1 ( /3(1+rf+1)1 ) E t Ct+ 1 = / 3 ( l + rt+l) Ct + -~ 1 Her e we see that the relation between /3 and the gross world interest rate induces a tilt in expect ed consumption growth as was also true in Subsection 3.131 . To simplify further, assume temporarily that the interest rate is constant at r, with 13(1 + r) = 1; the result is the "random walk" prediction for consumption derived by Hall (1978): E t Ct+l = Ct (3.32) In equilibrium, the economy' s intertemporal budget constraint, eq. (3.3), must hold for every possible sequence of outcomes when its elements are random. Applying the date-t expectations operat or to both sides of (3.3) and using (3.32), we have the certainty-equivalence consumption function: C, = 1-77r Et (1 + r ) A t + Z \ ] - 7 7 2 (Y~ - G , - l s ) (3.33) S = t Equat i on (3.33) yields predictions for current accounts qualitatively similar to those of the deterministic consumption function (3.8) devel oped earlier. As a result, this equation and its relatives provide the leading vehicles for empirical studies of current-account determination. The linear-quadratic formulation has, however, at least t hree conceptual drawbacks: 31In t he precedi ng equat i on t he ratio of t he gross i nt erest rat e to 1/13 ent ers bot h in the slope and in t he i nt ercept , with opposite effects on expect ed dat e@ + 1) consumption. The quadrat i c utility specification makes sense, however, only so long as C < 1/ h, ensuring t hat t he margi nal utility of consumpt i on is positive. Under this assumption, it is clear t hat t he i nt ercept effect dominates: l oweri ng/ 3, for example, lowers expect ed dat e-(t + 1) consumpt i on in t he last equation. 1768 M. Obstfeld and K. Rogo]f (1) Un d e r quadr at i c ut i l i t y u"~( C) ~- O, so t her e is no pr ecaut i onar y saving. When, i nst ead, u"~( C) > 0, margi nal ut i l i t y d ( C ) is a convex funct i on. Thi s convexi t y i mpl i es t hat an i ncrease in uncer t ai nt y over f ut ur e cons umpt i on rai ses its expect ed mar gi nal ut i l i t y and, thus, saving. 32 (2) In bot h est i mat i ng and si mul at i ng i nt er t empor al model s, it is f r equent l y conveni ent t o be able t o l i neari ze or l og-l i neari ze fi rst -order Eul e r condi t i ons. Assumi ng quadr at i c per i od ut i l i t y is r eal l y just anot her way of linearizing. The basi c pr obl em wi t h this appr oach is t hat a l i near i zed model will be ver y i naccu- r at e far away f r om t he poi nt of appr oxi mat i on. Yet if t he l i near i zat i on implies, as in (3.32), t hat consumpt i on follows a r a ndom wal k, consumpt i on will event u- ally dri ft ar bi t r ar i l y far f r om any initial level. As Obst f el d (1982) and Svensson and Razi n (1983) showed, one way t o avoi d a uni t r oot in cons umpt i on is t o assume an endogenous, r at her t han fixed, r at e of t i me pr ef er ence. If t he r at e of i mpat i ence rises wi t h t he l evel of consumpt i on, a st at i onar y cons umpt i on di st r i but i on is possible. Unde r uncert ai nt y, pr ecaut i onar y saving can mi mi c this behavi or ( even wi t h a fixed t i me- pr ef er ence r at e) because consumer s save less at hi gher weal t h levels. [See Car r ol l (1992) and De a t o n (1992).] Anot he r sol ut i on is t o assume over l appi ng cohor t s of fi ni t e-l i ved agent s who l eave no bequest s. Unf or t unat el y, no al t er nat i ve can mat ch t he l i near - quadr at i c i nfi ni t e-l i fet i me f r amewor k f or easy empi ri cal i mpl ement at i on. Ul t i mat el y t he j ust i fi cat i on f or usi ng t he l at t er set up must rest on t he pr es umpt i on t hat it yi el ds a r eas onabl e appr oxi mat i on t o behavi or away fi' om boundar i es. Mor e r esear ch pr ovi ng or di spr ovi ng this pr es umpt i on woul d be useful. 33 (3) The cons umpt i on f unct i on (3.33) does not necessari l y const r ai n consump- t i on t o be non- negat i ve in all states of t he worl d. I f negat i ve l evel s of consump- t i on are r ul ed out , consumpt i on cannot l i t eral l y f ol l ow a r andom wal k as in (3.32) .34 We t ur n next t o i nvest i gat i ng how, when bonds ar e t he onl y i nt er nat i onal l y t r aded asset, t he separ at i on bet ween domest i c i nvest ment and consumpt i on i m- pl i ed by compl et e mar ket s can br eak down. It is nat ur al t o assume agai n t hat t he domest i c firm makes its i nvest ment deci si on on dat e t t o maxi mi ze t he pr esent di scount ed val ue of profits. But wi t h uncer t ai nt y, it is no l onger obvi ous what di scount r at e shoul d be used t o val ue t he ri sky st r eam of di vi dends t he firm issues t o sharehol ders. Wi t h popul at i on nor mal i zed t o 1, t he firm' s e x d i v i d e n d val ue, Vt+l, is t he pri ce of a per capi t a 32On precautionary saving behavior, see Leland (1968). 33Clarida (1990) develops an exact general-equilibrium model of international borrowing and lending under endowment uncertainty, heterogenous fixed discount rates, and lower limits on in- dividual assets. He finds that when there is no aggregate output uncertainty, there is a stationary distribution of wealth and consumption levels in which some households are borrowing-constrained. 34The problem of nonnegativity constraints afflicts the commonly used consumption functions derived in continuous-time models by stochastic dynamic programming. See Cox and Huang (1989). Ch. 34: The l nt ert emporal Approach to the Current Account 1769 s har e i n t he fi rm i n d a t e t asset t r adi ng. Us i n g d y n a mi c p r o g r a mmi n g ( t h e d e t a i l s a r e o m i t t e d h e r e ) , o n e c a n s how t hat Vt +l f ol l ows a s t oc ha s t i c pr oces s such t hat t he f ol l owi ng Eu l e r e q u a t i o n hol ds: V t + l U' ( C t ) = f l E t { [ O t + l V ( K t + l , Lt +l ) - Wt +l Lt +l - I t + l - a I { +l / Z Kt +l + Vt +2 ] u ' ( Ct +l ) } I t e r a t e d f o r wa r d s ubs t i t ut i ons f or V yi el d a s t ochas t i c ve r s i on of (3. 16), O s F ( K s , L s ) - w~. Ls - Is - 2-~sJ (3.34) Gi v e n t ha t t he f i r m is o wn e d e nt i r e l y by d o me s t i c r es i dent s , t he pr e s e nt val ue of a cl ai m t o its f ut ur e di vi dends i n a ny pa r t i c ul a r s t at e of n a t u r e d e p e n d s on d o m e s t i c c o n s u me r ' s ma r g i n a l ut i l i t y i n t hat st at e of n a t u r e r el at i ve t o c ur r e nt ma r g i n a l utility. We c a n d e c o mp o s e (3. 34) f ur t he r as 3s s = t + l b l t ( C t ) E t O s F ( K s , L s ) - w s L s - I s - 2 K , + f i s - t C o v t u~- 7~t ) , O s X ( K s , Ls ) - w s L s - Is,. - 2Ks.--- S=t+] De f i n e R L t o be t he ma r k e t di s c ount f a c t or b e t we e n pe r i ods t and s , t hat i s , t,s t he pr i ce of s ur e ( t hat is, n o n c o n t i n g e n t ) da t e s c o n s u mp t i o n i n t er ms of da t e t c o n s u mp t i o n . Of cour s e, R L is s i mpl y t he i nver s e o f t he gr oss s hor t r at e, t,t+l 1 + rt+1.36 Th e (s - t ) - p e r i o d a na l og o f eq. (3. 31) is s - t L . ' ( c , ) = / R t , ) E t so t he p r e v i o u s e q u a t i o n can be wr i t t e n as o~ { a / 2 . ~ Vt +l = ~ R L , . E t Os F ( K ~ , L s ) - w s L s - I s - 2 K~ J s~t+] + f l , _ t C o v t ( , u ' ( C s ) O s F ( K s , L s ) - w s C s - I s - 2 K s s : , + l I , 35The result below uses the fact that if X and X ~ are two random variables, E(XX/) E(X)E(X ~) + C o v ( X , X t ) . 36In a deterministic model, R~s = Rt,s as defined in eq. (3.2). The equality breaks down here because future short rates are stochastic, whereas R~s is known on date t. 1770 M. Obstjeld and K. Rogoff Equat i on (3.35) shows t hat t he firm' s val ue is t he convent i onal present val ue of di vi dends plus a risk premi um: t he firm is val ued mor e hi ghl y if it pays out unexpect edl y hi gh di vi dends when the margi nal ut i l i t y of owners' consumpt i on is unexpect edl y high. The presence of a risk pr emi um i nt roduces an addi t i onal channel t hr ough which shifts in consumpt i on pr ef er ences can influence invest- ment behavior. The firm' s i nvest ment behavi or can be charact eri zed by maxi mi zi ng t he sum of current profits and (3.34) subject to (3.12). As in t he det ermi ni st i c case cur- r ent i nvest ment is governed by (3.13), where qt = V~+I/Kt+I and Vt+l is given by (3.34). The resul t is a ri cher q model of i nvest ment with current -account predi ct i ons similar to t hose of Section 3.1. 37 3.2.3. Partially complete markets In real i t y count ri es t rade not onl y bonds but a rich menu of assets, includ- ing equi t y shares, cur r ency- denomi nat ed i nst rument s, and ot her securities wi t h st at e- cont i ngent payoffs. Thi s t rade ensures t hat some, if not all, consumpt i on risks can be pool ed and t hat the current -account effect s arising in t he last set of model s will be mut ed. Equat i on (3.31) implies t hat whenever t here is free t rade in noncont i ngent bonds bet ween t wo count ri es i and j, t he expect ed growt h rates of t hei r resi- dent s' mar gi nal utilities f r om consumpt i on are equal , assumi ng a common rat e of t i me preference. Thus, { } { } , u ( c , + 1 ) = E , ( 3 . 3 6 ) E, u'(C[) u'(C{) Unde r compl et e mar ket s [see eq. (3.29)], this equat i on holds ex post, not j ust in expect at i on. When onl y (3.36) holds, however, di fferences in margi nal rat es of i nt er t empor al subst i t ut i on can occur aft er t he fact. When some, but not all, risks can be t r aded bet ween countries, consump- t i on behavi or will be i nt er medi at e bet ween the predi ct i ons of (3.29) and (3.36). Specifically, (3.29) predicts t hat margi nal rat e of i nt er t empor al subst i t ut i on dif- ferences do not arise, while (3.36) predicts t hese di fferences can arise unexpect - edly. In t he i nt er medi at e case, the ex post di fference t i u ' ( C / ) . ' ( c j ) will be condi t i onal l y uncor r el at ed wi t h any da t e - ( t +l ) r andom vari abl e on which 37See, for example, Baxter and Crucini (1993b) and Glick and Rogoff (1995). Ch. 34: The Intertemporal Approach to the Current Account 1771 contingent contractual payments can be conditioned. Thus, if Xt+l is a random variable on which contracts can be written prior to date t + 1, u ( c , + 1 ) , x , + l = 0 (3. . 37) C o v , u ' ( C / ) u ' ( C l ) For example, if people in different countries can effectively pool the idiosyn- cratic consumption risks due to nominal exchange-rate fluctuations through for- eign exchange market deals, then realized exchange rate fluctuations will be sta- tistically unrelated to international differences in the growth of u/(C).3~ Because it is only partial, however, such insurance dearly leaves scope for unexpected current-account movements. Svensson (1988) develops a two-period model in which period 1 asset trading serves to pool consumption risks for period 2. Svensson shows that the usual logic of trade theory can be extended to analyze not only the current account under uncertainty, but also the asset composition of gross capital flows between countries. He develops a two-period pure exchange model of international trade in a possibly incomplete set of risky assets. In that model, a multi-commodity comparative advantage principle applies [see Dixit and Norman (1980), p. 95]: the inner product of a country's net asset import vector with the vector of home mi n u s foreign autarky asset prices is positive. When countries trade equity shares as well as noncontingent bonds, the sep- aration between domestic investment and consumption may still obtain even though asset markets are incomplete. As (3.37) shows, domestic and foreign residents must attach the same values to the state-contingent profits of a firm that they trade. A sufficient condition for the separation property to hold is that investment decisions themselves do not change these common valuations [see Ekern and Wilson (1974)]. 3.2.4. E n d o g e n o u s ma r k e t i nc ompl e t e ne s s We have seen that some of the model's predictions concerning current account behavior depend critically on the structure of asset markets and in particular the degree to which complete markets prevail. We have argued that the complete markets model is inadequate empirically, but if so it would be helpful to have a deeper understanding of the frictions that impinge on market completeness, rather than just assuming market limitations exogenously. In this section we present an example, drawn from Gertler and Rogoff (1990), of how international capital flows behave in the presence of moral hazard de- riving from asymmetric information at the microeconomic level. A key insight 38For a formal derivation, see Obstfeld (1994). 1772 M. Obstfeld and K. Rogof f f r om the model , which carries over t o ot her settings, is t hat i nt er nat i onal asset mar ket s can bri ng the global al l ocat i on of resources part of t he way, but onl y par t of t he way, t oward t he ful l -i nformat i on opt i mum. Thus, t he i nt er t empor al approach, as sket ched above, may well get t he d i r e c t i o n s of net i nt er nat i onal capital flows ri ght while overst at i ng magni t udes. We now adopt a t wo-peri od, single good set up in whi ch each of t he numer ous at omi st i c resi dent s of a small count ry maxi mi zes u ( c l , c 2 ) = G ( 3 . 3 8 ) gi ven exogenous endowment s E1 of t he consumpt i on good in per i od i and E2 in peri od 2. The utility funct i on (3.38) is obvi ousl y ver y special (clearly C1 = 0 is opt i mal ) but it allows us to simplify the analysis whi l e still maki ng our mai n points. The focus, i nst ead, is on i nvest ment . Each resi dent has two ways to t r ansf or m E1 i nt o fut ure consumpt i on. He or she may l end in t he worl d capi t al mar ket and ear n a riskless net rat e of r et ur n r. Al t ernat i vel y, current resources can be i nvest ed in a risky domest i c project. If a r esi dent invests K in per i od 1, t hen t he proj ect ' s stochastic payof f in per i od 2 is Y, where 4' > 0 wi t h probabi l i t y or(K) Y = 0 wi t h probabi l i t y 1 - ~r(K) Above, ~-(K) is increasing, strictly concave, and twice cont i nuousl y di fferen- tiable, wi t h ~r(0) = 0, 1r(oo) = 1, and 1 + r < 4'~-~(0) < ec. These assumpt i ons ensure t hat hi gher i nvest ment increases t he l i kel i hood of a successful out come (but at a di mi ni shi ng rat e) and t hat the prospect of success woul d j ust i fy nonzer o i nvest ment under symmet r i c i nformat i on. Ever y i ndi vi dual has one pot ent i al in- vest ment proj ect and di f f er ent i ndi vi dual s' i nvest ment out comes are statistically i ndependent . In this model , t he opt i mal ful l -i nformat i on i nvest ment level is K*, defi ned by ~-'(K*)4' = 1 + r (3.39) At this poi nt , t he margi nal ret urn to investing equal s the r et ur n t hat can be ear ned t hr ough l endi ng abroad. Let us assume t hat Ez K* E J + l + r < Thi s i nequal i t y implies t hat each resi dent needs to borrow a positive amount f r om forei gners to be able to invest optimally. Fur t her mor e, t he condi t i on is equi val ent to t he i nequal i t y E2 < (1 + r ) ( K * - E l ) , which states t hat it is in- feasi bl e for l enders to finance the i nvest ment level K* t hrough a risk-free loan. If an i nvest ment project fails, even t he borrower' s ent i re peri od 2 endowment is insufficient to repay l enders (1 + r ) ( K * - E l ) . Thus, l oan cont ract s t ake t he Ch. 34: The lntertemporal Approach to the Current Account 1773 f ol l owi ng, s t a t e - c ont i nge nt , f or m: i n r e t ur n f or a l oan o f si ze L in p e r i o d 1, t hc b o r r o we r p r o mi s e s t o r e p a y Zg i n pe r i od 2 i n t he e v e n t his o r her i n v e s t me n t is successf ul a n d Z b < z g i n t he e ve nt it fails. Le n d e r s ar e c o mp e t i t i v e and do not of f e r a l oa n c o n t r a c t unl ess its e xpe c t e d gr os s r e t ur n equal s 1 + r : ~ r ( K ) Z g + [1 - ~r ( K) ] Z/ ' = (1 + r ) L (3. 40) Th e pr ovi s i ons o f l oa n c ont r a c t s ar e a s s u me d t o be f ul l y e nf or c e a bl e . Th e mo r a l h a z a r d p r o b l e m unde r l yi ng t he mo d e l ar i ses f r o m its i n f o r ma t i o n s t r uct ur e. Le n d e r s o b s e r v e a b o r r o we r ' s e n d o wme n t s a nd t he o u t p u t of t he i n v e s t me n t pr oj e c t ( t hat is, wh e t h e r t he pr oj e c t is successf ul o r not ) ; of cour s e, t h e y al so k n o w t he si ze o f t he l oan. Th e y c annot , howe ve r , ve r i f y t he l evel of i nve s t me nt , K. Thus , b o r r o we r s c a nnot c o mmi t t he ms e l ve s t o a ny speci fi c l evel o f i n v e s t me n t b y wr i t i ng pr omi s e s a b o u t K i nt o t he l oa n cont r act . Fo r e xa mpl e , t he r e is n o wa y l e nde r s can p r e v e n t b o r r o we r s f r o m i nves t i ng n o t h i n g at all and i ns t e a d s ecr et l y pl a c i ng all of t hei r r es our ces , b o r r o we d as wel l as n o n b o r r o we d , i n f or e i gn asset s ( an a c t i on r e mi ni s c e nt o f " c a pi t a l f l i ght ") . But i f b o r r o we r s do so, t he r e is no c ha nc e t hei r pr oj ect s wi l l s ucceed. Th e i n f o r ma t i o n a l a s y mme t r y a l l owi ng t hi s ki nd o f b e h a v i o r l eads t o i nef f i ci ent i n v e s t me n t and b o r r o wi n g l evel s, as we now show. 39 Gi v e n t he avai l abl e l oa n c ont r a c t , a t ypi cal b o r r o we r ma xi mi z e s e x p e c t e d s e c ond p e r i o d c o n s u mp t i o n E{ C2} = ~( K) ( q5 - Z ~') - [1 - ~r ( K) ] Z b + (1 + r ) ( E l + L - K ) + E2 (3. 41) s ubj ect t o t he c ons t r a i nt E1 + L _> K (3. 42) Co n s t r a i n t (3. 42) doe s n o t bi nd wh e n t he b o r r o we r is s ecr et l y i nves t i ng re- s our ces a b r o a d r a t h e r t ha n at h o me . Th e ne c e s s a r y Ku h n - T u c k e r c ondi t i ons f or a ma x i mu m ar e: zr/(K)[q5 - ( Z g - ZI ' ) ] = 1 + r + h (3. 43) A > 0, A(E1 + L - K) = 0 (3. 44) Co n d i t i o n s (3. 43) and (3. 44) i mpl y t ha t e ve n a b o r r o we r wi t h access t o a l oan L l ar ge e n o u g h t o p e r mi t t he des i r ed i n v e s t me n t l evel ( so t ha t h = 0) pi cks a K t ha t sat i sfi es 39If borrowers could commit to invest K*, lenders would break even by lending them L = K* - El and setting Z l~ = E2, z g = (1 + r) [K* - E1 - (1 - ~r(K*))E2J/Tr(K*). [Because ~-(K) is strictly concave and ~r(0) = 0, Zg < ~b + E2.] In this model, there is an implicit assumption that foreign direct investment cannot substitute perfectly for lending; that is, foreigners cannot circumvent the agency problem by purchasing investment opportunities in the borrowing country and exploiting them optimally themselves. This assumption could be justified by a threat of nationalization, or simply by a comparative informational advantage of local residents in finding suppliers, monitoring workers, greasing the collective palm of local officialdom, etc. 1774 M. Obstfeld and K. Rogoff ~v'(K)[4, - ( Z g - zb) ] = 1 + r (3.45) and, thus, is st ri ct l y below t he f ul l - i nf or mat i on opt i mum l evel descr i bed by eq. (3.39). The r eason is t hat t he change in payof f t o t he bor r ower when a pr oj ect succeeds is 4 , - z ~ - ( - z I , ) = 4 , - ( z ~ ' - z ~ ) < 4 , Al t hough l ender s cannot obser ve K directly, t hey have r at i onal expect at i ons and t hus can fi gure out t he l evel of K bor r ower s will choose f or a gi ven l oan con- tract. The y t her ef or e of f er a l oan cont r act such t hat t he K t he bor r ower chooses, gi ven L, z g, and Z b, satisfies t he r equi r ed r at e of r et ur n condi t i on (3.40). Com- pet i t i on among l ender s ensur es t hat this cont r act is opt i mal f or t he bor r ower , subj ect t o t he const rai nt s ment i oned. Formal l y, t he opt i mal i ncent i ve- compat i bl e cont r act (L, z g, Z ~) maxi mi zes E{C2} subj ect t o (3.40), (3.42), (3.45), and t he i nequal i t y Z b < E2. The opt i mal cont r act satisfies Z b = E2 (so as t o mi ni mi ze t he gap ;~ - Z g- Z b and, thus, t he gap bet ween K and K*). 4 Fur t he r mor e , t he cont r act satisfies (3.42) wi t h equal i t y: i ncreasi ng L above K - E1 woul d force a rise in z g and, wi t h it, a wor seni ng in i nvest ment i ncent i ves. Combi ni ng t hese facts wi t h t he i ncent i ve- c0mpat i bi l i t y const r ai nt (3.45) and t he l ender s' zer o- pr of i t condi t i on (3.40), we see t hat t he opt i mal cont r act is t he sol ut i on to t he t hr ee- equat i on syst em ~ ' ( K ) ( 4 , - 2 ) = ~ + r 2 = (1 + r ) K - E,, - T T ; r / ~ ( K) L =K- E1. Thi s sol ut i on has a numbe r of i mpor t ant i mpl i cat i ons: (1) The i ncent i ve- compat i bl e i nvest ment l evel is bel ow t he f ul l - i nf or mat i on opt i mum of (3.39). Accordi ngl y, per i od 1 capi t al inflows are bel ow t he l evel a f ul l - i nf or mat i on model woul d predi ct . (2) An i ncr ease in t he pr oduct i vi t y par ame t e r 4, i ncreases t he per i od 1 capi t al inflow and i nvest ment level, but it also wi dens t he spr ead Z, bet ween t he good- and bad- out come l oan payment s. Thus, ( under mi l d rest ri ct i ons on ~-(K)) t he resul t i ng capi t al inflow is less t han in t he f ul l - i nf or mat i on case. (3) The expect ed margi nal pr oduct of capital, ~r~(K)4,, exceeds t he wor l d ( and domest i c) riskless i nt er est rat e, 1 + r. (4) An i ncr ease in ei t her endowment , E1 or E2, raises i nvest ment by l ower i ng Z. In t he first case this ef f ect is due t o a l esser r el i ance on ext er nal funds, in 4See Gertler and Rogoff (1990) for details. Ch. 34: The Intertemporal Approach to the Current Account 1775 t he second to t he possi bi l i t y of a l ar ger l oan pa yme nt in t he bad- out come state. The i nvar i ance of i nves t ment wi t h respect t o i nt er t empor al pr ef er ences t hat char act er i zed t he si ngl e-sect or model s descr i bed ear l i er need not hol d when i nves t ment is subj ect t o mor al hazard. 41 In a t wo- count r y gener al equi l i br i um ver si on of t he model , one can show t hat endogenous di f f er ences in capi t al mar ket i mper f ect i ons can dr amat i cal l y r educe t he flow of capi t al f r om rich t o poor count ri es, and even r ever se it [see Ger t l cr and Rogof f (1990)]. The pr ecedi ng model has abst r act ed bot h fron~t t he cons umpt i on smoot hi ng mot i ve behi nd t he cur r ent account and f r om t he bor r ower s ' desi re to engagc in asset t r ades t hat r educe t he uncer t ai nt y of second- per i od consumpt i on. It is st r ai ght f or war d t o add second- per i od risk aversi on t o t he model by assumi ng t hat (3.38) is r epl aced by h 2 u ( C1 , C~) = C2 - ~ C 2 In or der t o focus on t he new issues t hat arise, it is conveni ent t o suppose l hat E1 _> K*. In this case, domest i c resi dent s' e ndowme nt is l arge enough so t hat t hey have no ne e d t o bor r ow t o fi nance i nvest ment , and t hei r onl y mot i ve f or t appi ng t he i nt er nat i onal capi t al mar ket is t o i nsure per i od 2 consumpt i on. He r e agai n we have a mor al hazar d pr obl em because nonver i f i abl e i nvest ment deci- sions affect t he pr obabi l i t y di st ri but i on of second- per i od out put . It can be shown t hat under t he opt i mal i ncent i ve- compat i bl e cont ract , cons umpt i on i nsurance will be par t i al and K will be bel ow K*. 42 The gener al concl usi on is t hat asymmet r i c i nf or mat i on need not cause finan- cial mar ket s to br eak down entirely. Inst ead, financial mar ket s may do onl y par t i al l y t he j ob t hey coul d do in a wor l d of full i nf or mat i on. Not e also t hat in t hi nki ng about t he i ncompl et eness of mar ket s, it ma y be mi sl eadi ng t o t hi nk of risks as bei ng ei t her i nsur abl e or noni nsur abl e. In ma ny cases, t hr ough what is basi cal l y a coi nsur ance mechani sm, s o m e gains f r om t r ade across states of nat ur e will be r eal i zed even under mor al hazard. 43 The same poi nt appl i es to t r ade across t i me. We have f ocused on capi t al - mar ket i mper f ect i ons arising f r om asymmet r i c i nf or mat i on at t he mi cr o level. Anot he r ver y i mpor t ant cause of i nt er nat i onal capi t al - mar ket i mper f ect i on is sover ei gn def aul t risk at t he macr o level. Sovcr- ei gn risk need not be r el at ed t o asymmet r i c i nf or mat i on, but can have qual i t a- 41It is agai n easy to see t hat Zg < q~ + E 2. 42See Obst fel d and Rogoff (1996), ch. 6. 43The implications of adverse selection probl ems can be quite different, al t hough we do not consider t hem here. At keson and Lucas (1992) study a different moral hazard probl em, one m which peopl e (or countries) wish to insure against preference shocks. They find t hat under optimal i ncent i ve-compat i bl e ar r angement s, t he degree of consumpt i on inequality in t he world increases continually. 1776 M. Obstfeld and K. Rogoff t i vel y si mi l ar i mpl i cat i ons. 44 J ona t ha n Ea t on and Ra que l Fer nandez pr es ent a det ai l ed anal ysi s of t he effect s of s over ei gn ri sk in t hei r chapt er in this Ha n d - book. A c o mmo n t heme in sover ei gn- r i sk model s is t hat t he cons umpt i on- s moot hi ng and r i sk- shar i ng rol es of i nt er nat i onal capi t al ma r ke t s still oper at e, but ar e t e mp e r e d by def aul t risk. 4. Empirical evi dence oll t he i ntertemporal approach The i nt e r t e mpor a l appr oach t o t he cur r ent account has been subj ect ed to ext en- si ve f or ma l t est i ng; much of t he me t hodol ogy used gr ows out of Hal l ' s (1978) s emi nal wor k on t he i mpl i cat i ons of t he r at i onal - expect at i ons as s umpt i on f or f or war d l ooki ng cons umpt i on t heori es. A less f or ma l empi r i cal met hodol ogy, pi one e r e d by Fel dst ei n and Ho r i o k a (1980), has be e n used t o ar gue t hat t he cl ose r el at i ons hi p be t we e n nat i onal savi ng and i nve s t me nt r at es in pos t - Wor l d Wa r I I dat a f ur ni shes a pri ma facie case agai nst t he pr act i cal r el evance of t he i nt e r t e mpor a l appr oach. The s e t wo r es ear ch avenues ar e cl osel y i nt er t wi ned, as any a t t e mpt t o r econci l e t he Fe l ds t e i n- t t or i oka fi ndi ngs wi t h t he i nt e r t e mpor a l a ppr oa c h rest s on t he val i di t y of model s such as t hose s ur veyed in Sect i on 3. 45 4.1. The relationship between national savi ng and domest i c i nvest ment rates I n a cl osed economy, nat i onal savi ng equal s domes t i c i nves t ment and t he cur- r ent account is al ways zero. Fur t he r mor e , any obs e r ve d i ncr ease in nat i onal savi ng will aut omat i cal l y be accompani ed by an equal ri se in domes t i c i nvest - ment . A basi c pr e mi s e of t he i nt e r t e mpor a l a ppr oa c h is t hat capi t al is to s ome de gr e e i nt er nat i onal l y mobi l e, so t hat cur r ent account i mbal ances ar e a pos- sibility. Gi v e n t hi s pr emi se, t he i nt e r t e mpor a l a ppr oa c h pr edi ct s a n u mb e r of si t uat i ons in whi ch di ver gences be t we e n savi ng and i nves t ment will arise. An empi r i cal fi ndi ng t hat nat i onal savi ng r at es af f ect domes t i c i nves t ment r at es wi t h uni t coeffi ci ent s woul d t her ef or e a ppe a r t o be st r ong evi dence agai nst t he appl i cabi l i t y of t he i nt e r t e mpor a l appr oach. Fel dst ei n and Ho r i o k a (1980), in t he first of a seri es of r el at ed paper s by Fel dst ei n and coaut hor s, ar gued t hat capi t al mobi l i t y is suffi ci ent l y l i mi t ed, at 44Atkeson (1991) presents an interesting analysis incorporating both asymmetric information and sovereign risk. 45We do not survey the methodology of calibrating open-economy models with incomplete mar- kets so as to match moments of actual aggregate data. Interesting recent work along this line, exemplified by Baxter and Crucini (1993b) and Kollmann (1992), is discussed in Marianne Baxter's chapter in this Handbook. We also refrain from more than a brief and highly selective account of the copious literature on measuring international capital mobility. See Frankel (1993) and Obstfeld (1995) for recent surveys. Ch. 34: The l nt ert emporal Approach to the Current Account 0.33 0.30 0.27 0.24 0.21 0.18 0.15 0.12 0.12 0.15 J A P F I N P O R N O R A U S N Z A U T I R E = S P A I C E I T A G E R Q . E " c A . " " F . . " N E T D E N - S W E U S A B E L U K %, 0. 18 0.21 0. 24 0. 27 0. 30 0. 33 S/Y Figure 4. l . Saving and i nvest ment averages, t982-1991 (as a fraction of GDP). 1777 least over long horizons, that changes in national saving rates ultimately feed through fully to domestic investment rates. As evidence, they reported cross- sectional regressions of gross domestic investment rate averages ( I / Y ) on gross national saving rate averages ( S / Y ) . These ratios of nominal flows, of course, suffer from all the conceptual deficiencies discussed at the end of Section 2. For a sample of 16 OECD countries over 1960-74, Feldstein and Horioka found that: ( l / Y ) j ~- 0 . 0 3 5 + O. 8 8 7 ( S / Y ) j + uj ; R 2 = 0.91 (0.018) (0.074) Feldstein and Bacchetta (1991) report similar results for a 1974-86 sample of 23 OECD members. Figure 4.1 shows the cross-sectional saving-investment as- sociation in the OECD sample over the decade 1982-91, with Luxembourg, which is an outlier, omitted. The estimation result for this sample, leaving out developing Turkey, is: ( I / Y ) j = 0.088 + 0 . 6 2 2 ( S / Y ) y + uy; R 2 -~- 0.69 (0.020) (0.094) This equation shows a weakening, but still very significant, positive association. Feldstein and his collaborators argue that if capital indeed were highly mobile among countries, slope coefficients like the one above should be much smaller than 1, as a country's savings would then be free to seek out the most produc- tive investment opportunities worldwide. Although the intertemporal approach is consistent with a world in which changes in saving behavior impinge on do- 1778 M. Obstfeld and K. Rogof f mest i c i nvest ment , it cer t ai nl y does not suppor t t he pol i cy concl usi on, pr ef er r ed by Fel dst ei n, t hat gover nment measur es t o rai se a count r y' s saving r at e will aut omat i cal l y cause a l ong-run pari passu i ncr ease in its i nvest ment rat e. I f capi t al is t r ul y i mmobi l e and t he i nt er t empor al appr oach i r r el evant , how- ever, t hen t he t i me- ser i es r el at i onshi p bet ween i ndi vi dual count r i es' saving and i nves t ment rat es, l i ke t he cross-sect i onal r el at i onshi p among l ong savi ng and i nves t ment aver ages, also shoul d be dose. We can t hi nk of t he t i me-seri es rel a- t i onshi p among cont empor aneous det r ended saving and i nvest ment r at es as cap- t ur i ng t he coher ence of hi gh- f r equency changes, whi l e t he Fe l ds t e i n- Hor i oka re- gressi on capt ur es t he associ at i on bet ween l ow- f r equency or "sust ai ned" changes. I ndeed, it is har d t o see how capi t al coul d be t r ul y i mmobi l e in t he l ong run but not in t he shor t run, since t he long r un is just a successi on of shor t runs. An d even if i nt er nat i onal t r ade in l ong- t er m i nst r ument s or l ong-l i ved assets wer e hi ghl y l i mi t ed - a hypot hesi s t hat t he dat a do not suppor t - shor t - t er m i nst r ument s can be r ol l ed over. Th e t i me- ser i es and cross-sect i on aspects of t he s avi ng- i nves t ment r el at i on- ship are qui t e distinct: t he t i me-seri es r el at i onshi p coul d be close and t he cross- sect i on r el at i onshi p not , or vice versa. An exampl e is pr ovi ded by t he Uni t ed Ki ngdom dur i ng t he pr e- Wor l d War I gol d st andar d ( Fi gur e 2.1). It is appar ent t hat t he shor t - r un savi ng- i nvest ment cor r el at i on is ver y close. 46 Nonet hel ess, t he U. K. r an cur r ent - account surpluses appr oachi ng 10 per cent of GDP in this same per i od. It is a "st yl i zed fact ", somewhat sensi t i ve t o t he det r endi ng me t hod adopt ed, t hat t he t i me- ser i es savi ng- i nvest ment cor r el at i on is fai rl y st rong in r ecent data. ( See Fi gur e 2.3 f or some i ndust r i al - count r y dat a, but t her e are except i ons, t he most gl ari ng of whi ch is Nor way wi t h its hi ghl y negat i vel y cor r el at ed saving and i nves t ment rat es. ) For t he OECD count r i es i ncl udi ng Luxe mbour g but ex- cl udi ng Tur key, t he aver age cor r el at i on bet ween saving and i nvest ment rat es is 0.495 over 1974-90 af t er l i near t i me det r endi ng. The cor r el at i on is 0.512 when t he dat a are f i r st - di f f er enced to r emove t rend. Eve n in a wor l d of compl et e capi t al mobi l i t y t hough, such cor r el at i ons are not necessar i l y surpri si ng and can easi l y be expl ai ned on t he basis of t he i nt er t em- por al appr oach. For exampl e, t he discussion of gr adual capi t al - st ock adj ust ment in Sect i on 3 i mpl i es t hat a shock t o t ot al f act or product i vi t y, if short -l i ved but not compl et el y t ransi t ory, raises saving as well as i nvest ment . Baxt er and Cruci ni (1993a,b) and Mendoza (1991) (t he l at t er in a smal l - count r y setting, t he for- me r in a gl obal - economy setting) have shown t hat i nt er t empor al mobi l e- capi t al model s based on i nvest ment - adj ust ment costs can easi l y pr oduce t i me- ser i es savi ng- i nvest ment cor r el at i ons at least as l arge as t hose in t he dat a. Par t of t he mechani s m under l yi ng t he Baxt er - Cr uci ni findings is al l owance f or gl obal 46See Obstfeld (1986) for a more detailed statistical analysis of the U.K. data. Ch. 34: The Interternporal Approach to the Current Account 1779 economi c shocks, whi ch obvi ousl y will i nduce positive savi ng-i nvest ment corre- lations. We have been focusi ng on time-series evi dence, but Fel dst ei n and Hor i oka' s cross-sectional findings can also be rat i onal i zed by the presence of common factors t hat mi ght si mul t aneousl y influence count ri es' saving and i nvest ment rates. It seems l i kel y t hat of t he many pot ent i al expl anat i ons of t he Fel dst ei n- Hor i oka results, no single one fully explains t he behavi or of all countries. Taken t oget her, however, and combi ned wi t h ot her evi dence i ndi cat i ng subst ant i al in- t er nat i onal mobi l i t y of capital, t he ar gument s bel ow suggest t hat t he Fel dst ei n- Hor i oka finding provi des no basis at all for dismissing the basic premi ses of the i nt er t empor al approach: (1) Even post-1973, gover nment s have somet i mes adj ust ed fiscal or monet ar y policies to avoi d l arge and pr ot r act ed current -account i mbal ances. The evi dence on this current -account t arget i ng hypot hesi s is most l y anecdot al , however, and t her e are of course pr omi nent instances (like t he Uni t ed St at es in t he 1980s) in whi ch macr oeconomi c policies have i nst i gat ed maj or ext ernal i mbal ances. 47 (2) OECD count ri es may be sufficiently well endowed wi t h capital to have r eached stochastic st eady states for t hei r ext ernal debt or asset levels. In this si t uat i on t he i nt er t empor al budget const rai nt of t he economy woul d i mpl y t hat l ong averages of savi ng-i nvest ment di fferences are small. Devel opi ng count ri es, which pr esumabl y coul d real i ze great er gains from i nt er t empor al t r ade t hr ough borrowi ng for i nvest ment purposes, are l i kel y to be mor e di st ant f r om a sta- t i onar y di st ri but i on of forei gn debt. This i nt er pr et at i on seems bor ne out by t he cross-sectional results for devel opi ng count ri es pri or to t he onset of t hei r debt crisis in 1982. For this sampl e, the cross-sectional savi ng-i nvest ment association is much l ooser t han for t he OECD sample. 48 (3) The Ger t l er - Rogof f i nvest ment model discussed in Sect i on 3 shows why i nvest ment may respond posi t i vel y to hi gher r et ai ned earnings, t hat is, to hi gher cor por at e saving. Thus, it seems pl ausi bl e t hat in count ri es wi t h hi gher saving rates, t he cost of capital will be l ower and i nvest ment higher. A mai n predi ct i on of t he mor al - hazar d model is t hat risk-free i nt erest rates are equal i zed among count ri es - as i ndeed t hey most l y are in t he i ndust ri al i zed worl d (Obst fel d 1995) - whereas t he margi nal pr oduct of capital is high in count ri es wi t h low corpo- rat e weal t h. Al t hough t her e is some evi dence in favor of this hypot hesi s, an account of how cor por at e saving and i nvest ment are r el at ed need not have st rong i mpl i cat i ons for t he rel at i onshi p bet ween t ot al saving and i nvest ment . For exampl e, pri vat e domest i c owners of firms may pierce t he cor por at e veil 47The fragility of the econometric evidence is illustrated by Feldstein and Bacchetta's (1991) reinterpretation of the regressions Summers (1988) offers as evidence of current-account targeting. 48See Fieleke (1982), Dooley, Frankel, and Mathieson (1987), and Summers (1988). 1780 M. Obstfeld and K. Rogof f and offset corporate saving decisions through their own consumption. To the extent that the investing firms are owned by foreigners, their decision to retain earnings increases foreign rather than domestic saving, other things equal. (4) In a Heckscher-Ohlin framework, a nonspecialized economy experiencing a rise in saving can absorb more capital without a fall in capital's domestic rate of return. Alternative explanations of the saving-investment association based on this observation are suggested by Fukao and Hamada (1994) and Obstfeld (1995). (5) In the life-cycle theory of consumption, sustained demographic or pro- ductivity changes that increase a country's long-term investment rate also may increase its saving rate. Section 3.1.7's discussion of overlapping-generations models provides a leading example: higher productivity growth that affects most strongly the incomes of young workers will cause saving as well as investment to rise. Feldstein and Bacchetta (1991) and Summers (1988) have dismissed this line of explanation, notwithstanding some supportive evidence offered by Tesar (1991). In a more recent contribution, however, Taylor (1994) revisits the Feldstein-Horioka equation, controlling for (a) measures of domestic relative prices, (b) the age-structure of the population, and (c) the interaction of the age structure with the growth rate of domestic output. He finds that for a number of country samples the cross-sectional saving-investment association disappears. Far from showing the irrelevance of the intertemporal approach, the large literature spawned by Feldstein and Horioka' s (1980) nonstructural exploration suggests to us that models like those reviewed in Section 3 capture key elements behind the cross-sectional and time-series regularities governing saving and in- vestment rates. The further empirical challenge for the intertemporal approach is to show that structural forward-looking models of the current account are not grossly incompatible with actual experience. We turn next to tests of such models. 4 . 2 . Tes t s " o f i n t e r t e m p o r a l c u r r e n t - a c c o u n t m o d e l s Most structural time-series studies of the intertemporal approach to the current account essentially test versions of eq. (3.9), according to which the current account depends on deviations of interest rates, output, government spending, and investment from "permanent" levels. Indeed, most (but not, as we shall see, all) of these studies focus on the special case of (3.9) with a constant real interest rate r = (1 -/3)//3 C A t = ( Y t - Y t ) - ( G t - G t ) - ( I t - I t ) (4.46) Though less general than eq. (3.9), eq. (4.46) embodies many central ele- ments of the intertemporal approach. It therefore is reasonable to ask whether Ch. 34: The Intertemporal Approach to the Current Account 1781 t her e is any evi dence in f avor of (4.46) bef or e t urni ng t o mor e compl ex model s. For exampl e, do t e mpor a r y rises in gover nment spendi ng cause cur r ent - account deficits? Quest i ons like this one seem si mpl e enough, but a numbe r of empi ri cal subt l et i es arise in answeri ng t hem. 4. 2. l . M e a s u r i n g p e r m a n e n t v al ue s ' : A d i g r e s s i o n Eve n bef or e t ur ni ng t o t he economet r i c studies, it is useful to address what is per haps t he most pr obl emat i c issue of all, t he const r uct i on of t he expect ed pe r ma ne nt val ues Y, G, and I. A first difficulty is t hat it is not obvi ous what real i nt er est r at e to use to di scount expect ed f ut ur e out put flows. 49 Most of t he st udi es sur veyed bel ow use fai rl y low di scount rat es, in t he r ange of 2 to 4 per cent per year. Thes e number s cor r es pond r oughl y t o aver age ex post real r et ur ns on U.S. Tr easur y bills post - Wor l d War II. But is a ( nomi nal l y) riskless r at e t he appr opr i at e onc f or di scount i ng ver y ri sky f ut ur e out put flows? For t he Uni t ed States, t he mean r at e of r et ur n on ri sky assets has hi st ori cal l y been much hi gher t han t hat on bonds [Mehra and Pr escot t (1985)], at l east since t he l at e ni net eent h century. To t he ext ent dat a are avai l abl e, a si mi l ar resul t seems t o hol d for a numbe r of ot her count ri es. Be r na nke (1985) argues t hat an annual real i nt er est r at e as high as 14 per cent is ne e de d t o r at i onal i ze U.S. c ons umpt i on- i nc ome rel at i onshi ps in a r el at ed cl os ed- economy setting. A second di ffi cul t y concer ns t he sensi t i vi t y of empi ri cal measur es of Y, G, and i t o appar ent l y beni gn di f f er ences in t he t i me seri es process gener at i ng t he under l yi ng val ues of Y, I, and G. Thi s pr obl em i nt er act s wi t h t he pr evi ous one, because t he sensi t i vi t y t o t he dat a- gener at i ng process is especi al l y acut e under a l ow i nt er est rat e. 5 Consi der t he fol l owi ng exampl e, in whi ch out put ( expr essed as a devi at i on f r om me a n out put ) is gener at ed by t he pr ocess Y t = P Y t - 1 + v t (4.47) wi t h 0 < p _ 1, v t a whi t e- noi se error, and t i me meas ur ed in years. If a l ow real i nt er est r at e is used t o const r uct pe r ma ne nt out put r E Y+I + _ _ + . . (4.48) #, - - - - - l + r 1+---7- ( l + r ? di st ant f ut ur e i ncomes will have r el at i vel y l arge weights. Whe n r -- 0.03, Et Yt+20, f or exampl e, t hough di scount ed, still has a wei ght mor e t han hal f t hat of cur r ent out put , Y r . Thi s means t hat when r is low, est i mat es of Yt ma y be ver y sensi t i ve t o t he ser i al - cor r el at i on pa r a me t e r p, especi al l y in t he nei ghbor hood of p = 1. Not i ng t hat under eq. (4.47) Et Yt+k = p k Y t , one sees t hat eq. (4.48) implies: 49Sometimes this issue can be finessed [see t he discussion of Glick and Rogoff (1995) below.] 5This discussion draws on Glick and Rogoff (1995). 1782 M. ObstJeld and K. Rogoff Yt -- r Yt - txYt (4.49) l +r - p When p = 1, /~ = 1 and so lTt = Yt, regardless of the value of r. But when r = 0.03 and p --= 0.97 (a value differing from 1 by an amount generally t oo small to detect empirically), /x drops to only 0.5: permanent out put is half of current output. As p becomes small this hypersensitivity abates. In practice, unfortunately, p tends to be quite close to 1, so that the presence of a unit root is difficult to reject. It should be clear that estimates of Yt can be similarly sensitive to estimates of time trends. 51 The present-value calculations are less sensitive to p when the real interest rate is higher. How high must real interest rates be'? With r = 0.14 (the value ment i oned by Bernanke) and P -- 0.97 in eq. (4.49), /~ = 0.824. With r - 0.5, /x -- 0.943. Real interest rates high enough to make /x insensitive to p in the vicinity of a unit root appear implausible. With these cautions in mind we proceed to look at the literature. 4.2.2. Earl y tests' Earl y economet ri c tests of the intertemporal approach as represent ed by eq. (4.46) include Ahmed (1986, 1987), Hercowitz (1986), and Johnson (1986). Hercowitz, who looks at Israeli data over 1950-1981, presents some support for an i nt ert emporal model but also finds that the model exaggerates the cur- rent account' s response to output fluctuations. Johnson focuses on Canada over 1952-1976. He rejects Ricardian equivalence, but concludes that Canada' s pri- vate sector can plausibly be model ed in line with a version of the i nt ert emporal approach that allows for some liquidity-constrained consumers. 52 Ahmed' s papers are distinctive in their use of long historical data series on government expenditures from the Uni t ed Kingdom. In the 1986 paper, Ahmed looks at annual 1908-80 data to gauge the impact of U.K. government spending on the current account (actually, the trade balance, TB). The 1987 paper ana- lyzes a pre-Worl d War I sample on public military spending and trade balances running from 1732 to 1913. Ahmed argues that the expenditures accompany- ing Britain' s wars were largely exogenous and were almost certainly viewed as t emporary by the public. Thus, on the basis of the i nt ert emporal approach, one 51The same problems arise in the macroeconomic literature on estimating consumption functions. Deaton (1987), for example, argues that if income is stationary in growth rates (a hypothesis that is difficult statistically to reject given the limited post-WWII time series), then consumption should move more than one-for-one with income innovations. There is, on the other hand, no "Deaton's paradox" if income is highly serially correlated but still stationary. 52Roubini (1988) combines the intertemporal approach with the tax-smoothing theory of govern- ment deficits, finding mixed results for a sample of OECD countries. Ch. 34: The IntertemporaI Approach to the Current Account 1854 Pounds (Mi l l i ons) 20 1783 Military Spending 15 10 A C u r r e n t A c c o u n t 0 1701 1722 1743 1764 t 7 8 5 1806 1827 1848 t 8 6 9 1890 1911 1932 Figure 4.2. United Kingdom: Military spending and current account, 1701-1938 (detrended annual data). might expect Britain to have been running external deficits during wars. 53 In the twentieth century, swings in British government spending have been dominated by the two world wars, both of which were accompanied by large current- account deficits. Figure 4.2 uses data from Ahmed' s papers to graph military spending against the current account over 1701-1938. 54 This 238-year sample provides a more demanding testing ground than the twentieth century alone, as the period is punctuated by many wars. 55 A negative correlation between surges in government spending and the cur- S3The result that temporary war-time increases in public spending should lead to an external deficit does not necessarily hold if the whole world is at war. In an all-encompassing global war, higher government spending everywhere would push up world interest rates with current account implications that would depend mostly on countries' net external asset holdings. This point illustrates the distinction between global and country-specitic shocks emphasized in Section 3 above. 54Both series have had a 2 percent annual growth trend removed, as in Ahmed (1987). The current account has been constructed from Ahmed' s trade balance data using a 3 percent per annum sterling rate of return on foreign assets starting from an assumed zero net foreign asset position in 1701. We thank Shaghil Ahmed for providing us with the data from his papers. 55The possible gain or loss of colonial territories and privileges, which was the motive for much warfare before World War I, naturally could be expected to exert an additional wealth effect on the current account Ideally, this effect should be controlled for in estimation. 1784 M. Obstfeld and K. Rogoff r ent account is ful l y consi st ent wi t h t heor i es ot her t han t he i nt e r t e mpor a l ap- pr oach, f or exampl e, a Keynes i an mul t i pl i er model . To r educe t he set of al t er- nat i ve t heor i es consi st ent wi t h t he dat a, Ah me d expl or es f or mal e c onome t r i c tests. A r epr es ent at i ve resul t f r om Ah me d (1986) is TBt = - 0. 21 (Gr - Gt ) - 0.44 Gt + et; R 2 = 0.28, D. W. = 2.32 (0.05) (0.88) whi ch is e s t i ma t e d over 1908-80. 56 Thi s r egr essi on shows t hat t he t e mp o r a r y c o mp o n e n t of gove r nme nt spendi ng has a si gni fi cant negat i ve i nfl uence on t he cur r ent account , wher eas t he pe r ma ne nt c ompone nt i t sel f does not , consi st ent wi t h (4.46). Unf or t unat el y, t he r egr essi on' s speci fi cat i on l eaves ope n t he possi - Nl i t y t hat Gt is t he onl y significant de t e r mi na nt of t he cur r ent account and t hat Gt pl ays no r ol e at all - as an a t e mpor a l Keynes i an model woul d predi ct . Us i ng our es t i mat ed cur r ent account in pl ace of t he t r ade bal ance, we have r un a si mi l ar r e gr e s s i onon t he 1701-1938 dat a, but whi l e keepi ng Gt - Gt in t he r egr essi on we r epl ace Gt by Gt so as t o e nc ompa s s t r ans par ent l y t he i nt er t em- por al and Ke yne s i a n al t er nat i ves wi t hi n a single test. 87 The r esul t i ng r egr essi on, r un usi ng a Co c h r a n e - Or c u t t cor r ect i on f or f i r st - or der seri al cor r el at i on is 58 CAt ---- - 0 . 0 1 6 ( Gt - Gt ) - 0.028 Gt + et; p = 0.907 (0.013) (0.093) (0.027) I n this speci fi cat i on, nei t her cur r ent nor p e r ma n e n t gove r nme nt spendi ng is i ndi vi dual l y significant. As usual , it is uncl ear whe t he r t he i nt e r t e mpor a l ap- pr oa c h is s i mpl y fal se, or whe t he r t he ma n y e xt r a ne ous si mpl i fi cat i ons and mai n- t ai ned hypot he s e s i mpos ed by t he economet r i ci an ar e to bl ame. I t is t he r e f or e useful t o t ur n t o a newer empi r i cal a ppr oa c h bas ed on a less r est r i ct i ve f r a me - wor k. 4.2.3. Pr es ent - val ue mode l s o f t he current account Gh o s h (1995), Ot t o (1992), and Sheffri n and Woo (1990) appl y an al t er nat i ve me t h o d o l o g y t hat ma ke s use of t he i nf or mat i on e mb o d i e d in past cur r ent ac- count s t o ma k e mor e accur at e pr edi ct i ons of I 7", G, and [. Thes e st udi es bui l d on t he me t h o d o l o g y devel oped by Ca mpbe l l (1987) and by Ca mpbe l l and Shi l l er (1987). 59 56See Table 1, regression I (with p = 0.02) on p. 211. 57To form G we used an autoregressive forecasting model of detrended government spending. 58Similar results are obtained when we use the trade balance in place of the current account, as Ahmed (1987) does. 59Campbell's (1987) study of private U.S. saving is especially relevant to the current-account studies we are about to discuss. Comparison with Campbell's results is difficult, however, because he examines a different question, the accumulation of private wealth of all kinds in response to fluctuations in after-tax labor income. Ch. 34: The Intertemporal Approach to the Current Account 1785 Defi ne Q~ Y - G- I The st art i ng poi nt for t he present -val ue met hodol ogy is again eq. (4.46), ex- pressed in the f or m CAt = Qt - Qt (4.50) The new vari abl e Q can be t hought of as the net pri vat e noni nt er est cash flow. Al t hough the mai n i nnovat i on of the Campbel l - Shi l l er approach does not real l y requi re it, Ghosh, Sheffri n-Woo, and Ot t o all fol l ow Campbel l and Shiller in rewri t i ng eq. (4.50) as CAt = --Et ~ ~ AQs (4.5:l) s=t+l where &Qt = Qt - Qt-1 is the di f f erence of t he cash flow variable. Equat i on (4.51) says t hat t he current account bal ance t ends to be negat i ve when net cash flow is expect ed t o rise, and positive when net cash flow is expect ed to fall. Condi t i onal expect at i ons are assumed t o be l i near proj ect i ons on available i nf or mat i on. What advant age is t here to est i mat i ng eq. (4.51), wher e Qt ent ers in dif- f er enced form, r at her t han eq. (4.50), where its level ent ers? The di fferenced versi on is appr opr i at e if one is concerned t hat t here is a uni t root in Qt, so t hat removal of a t i me t r end is not sufficient for stationarity. 6 If Qt is i ndeed I(1) (has a uni t root ), t hen, as Campbel l and Shiller have emphasi zed, eq. (4.51) allows one to use the st at i onar y vari abl e zXQt as a regressor wi t hout havi ng to di fference bot h CA and Q, which is inefficient in the l i kel y event t hat CA is stationary. 61 The f undament al di fference bet ween the Ghosh, Shef f r i n- Woo, and Ot t o ap- pr oach and earl i er studies concerns how one proxi es for pri vat e agent s' expec- t at i ons of fut ure val ues of Q. The basic insight of t he Campbel l - Shi l l er met hod- ol ogy is t hat as l ong t he i nf or mat i on set used by the economet r i ci an does not cont ai n all the i nf or mat i on available to pri vat e agents, t hen past val ues of CA cont ai n i nf or mat i on useful in const ruct i ng est i mat es of agent s' expect at i ons of f ut ur e val ues of Q. Obviously, i ncorporat i ng this insight doesn' t act ual l y requi re using first di fferences, as in eq. ( 4 . 5 1 ) , r at her t han levels, as in eq. (4.50). Suppose, for exampl e, t hat one forms expect at i ons of f ut ur e val ues of d~Qt by first est i mat i ng a first-order VAR (the general i zat i on to hi gher-order VARs is st rai ght forward) 6Ghosh (1995) cannot reject the hypothesis of a unit root in Ot for his sample. Sheffrin and Woo (1990) and Ot t o (1992) report similar results. 61Trehan and Walsh (199l) discuss conditions under which stationarity of CA is necessary for a country' s intertemporal budget balance. 1786 M. Obstfeld and K. Rogof f [ C A t - i ] LeztJ and t hen makes use of i t s i mp l i c a t i o n t hat E t k C A , + k ] = 4,3 4 , 4 ] LCA,] ( 4 . 5 3 ) to form an est i mat ed current account, CAt . 62 If i is the 2 x 2 identity matrix and ~ the matrix of the 4,s, then eqs. (4.51) and (4.53) imply that C- A t : - [ 1 0 ] [ ( l +r ) - - l ~] [ I _ ( 1 + r ) _ 1 ~ ] - 1LCAt ][A Q t [ A e , l -~ [CbaQ q)CA] [ CAt ] (4.54) If the version of the intertemporal approach embodi ed in (4.51) is true, then the theoretically predicted value of [eaQ @cA] in (4. 54)is simply [0 1]! The reason this restriction emerges is obvious when the VAR captures all informa- tion peopl e use to forecast future cash flow. The same restriction also holds true, however, when the VAR captures only a subset of that information. The reason is that - CA t captures the representative consumer' s best estimate of the present value of future cash-flow changes, regardless of what ot her information he or she has. Ap p l y i n g the above approach, Sheffrin and Woo find that the restriction CAt = CAt is rejected for Canada, Denmark, and the U.K. in their 1955- 85 sample, although it is not rejected for Belgium. Ghosh, whose sample peri od is 1960-88, finds that the restriction is not rejected for the U.S., but that it fails for Canada, Germany, Japan, and the Uni t ed Kingdom. Equat i on (4.54) leads to a stringent test of the model, but a number of more general tests less sensitive to maintained hypothesis could be applied. One basic implication of the model is that CAt should Granger-cause AQt. Ghosh finds that in his full sample, even this weaker test still is passed only by the Uni t ed States data. Sheffrin and Woo arrive at more positive results. Anot her approach is adopt ed by Ot t o (1992), who tests the restriction that Et-1 { CA, - AQ, - ( 1+ r ) CAt _t } = 0 which follows straightforwardly from eq. (4.51). Ot t o rejects the present-value model for Canada and the U.S. after finding that lagged variables help in pre- dicting CAt - AQt (1 + r) CAt _l . 62Once agai n, t hi s r e qui r e s a choi ce of t he c ons t a nt r eal r i s k- f r ee i nt er es t r at e r. Ch. 34: The Intertemporal Approach to the Current Account /787 Whi l e the f or mal evi dence t her ef or e is very mi xed, 63 Ghosh, Sheffrin and Woo, and Ot t o all stress t hat the i nformal evi dence obt ai ned by simply lining up act ual current account s wi t h t he model ' s predi ct i ons can be quite impressive. Thi s perspect i ve is useful, because no empi ri cal model is l i kel y to be literally true. In Fi gure 4.3, we graph two illustrative cases, Sweden and the Uni t ed Ki ngdom, using post war data; bot h figures are based on a first-order VAR with AQ and CA as discussed above. The model performs ver y well for Sweden, but poor l y for the Uni t ed Ki ngdom. One pr obl em mi ght be t hat t he model does not explicitly i ncor por at e the effects of oil prices changes, which have been i mpor t ant for Bri t ai n in recent years. I ndeed, if one ext ends the dat a on Gr eat Bri t ai n over a l onger historical peri od, t he model ' s per f or mance looks much better. For annual British dat a over t he per i od 1870-1991, a first-order VAR for dxQ and CA yields 64 AQt] = [ 0. 24 -0.14] [AQt-11 CAt] - 0. 11 0.84 ] [CA,_I] Fi gure 4.4 is const r uct ed using t he above est i mat es and assumi ng a real in- t erest rat e of 4 per cent per annum. Ext endi ng t he dat a set yields a dramat i cal l y bet t er fit t han when one est i mat es t he model over post -Worl d War II dat a alone. Though t he visual evi dence is fai rl y striking, t he model still fails a formal test of t he rest ri ct i on embodi ed in eq. (4.54). Fr om t he above VAR est i mat es, one o b t a i n s [~)zxO ~CA] = [--0.26 0.54], which differs significantly from t he null hypot hesi s val ue of [0 1 I. A common t heme in the graphical evi dence pr esent ed by Ghosh, Shef f r i n- Woo, and Ot t o is t hat t he actual current account is of t en far mor e vol at i l e t han t he pr edi ct ed current account. This seems to cont radi ct t he Fel ds t ei n- Hor i oka concl usi on t hat current account movement s are rel at i vel y small compar ed to what one woul d expect in theory. Ghosh f or mal l y compares t he vari ances of the pr edi ct ed and act ual current account series and finds t hat , except in the U.S. case (where he cannot rej ect equal i t y of the variances), t he vari ance of t he actual series is higher. Ot t o si mi l arl y finds t hat Canada' s current account is six t i mes as vol at i l e as t he pr edi ct ed series. 65 Ghosh i nt erpret s his finding as evi dence of "t oo 63Ghosh and Ost ry (1995) apply the present -val ue approach to devel opi ng countries and argue t hat , if anything, it performs bet t er t han for industrialized countries. They find t hat across a large sampl e of developing countries, t he level and volatility of net capital movement s predi ct ed by t hei r consumpt i on-smoot hi ng model closely parallels those in t he data. This finding is puzzling - devel opi ng count ri es' capital mar ket s t end to be less open t han those of t he industrialized countries - but one possible expl anat i on relies on t he distinction bet ween global and country-specific shocks t hat we make below. Plausibly, developing countries are relatively mor e susceptible to country- specific as opposed t o global shocks, so t hat t he present -val ue model, which assumes a given and const ant real i nt erest rate, does somewhat better. 64Historical dat a are from Fei nst ei n (1972) and Maddi son (1991). 6SSheffrin and Woo' s dat a (which t hey generously supplied to us) yield similar results. 1788 M. Obstfeld and K. Rogoff 1985 Kronor per Capita 4000 3000 2000 1000 0 -1000 -2000 -3000 Actual 'kt Predicted I'l 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1985 Pounds per Capita 200 100 -100 -200 -300 V V Predicted J 19501954195819621966197019741978 19821986 1990 Figure 4.3. Sweden: Actual and predicted current account balance (annual data). Ch. 34: 7"he Intertempora! Approach to the Current Account 1789 1914 Pounds per Capi t a 6 2 0 . . . . . . . . . . . . . . . -2 -4 -6.8 Predicted A -10 -12 1871 1882 1893 1904 1915 1926 1937 1948 1959 1970 1981 1991 Figure 4.4. United Kingdom: Actual and predicted current account balance (annual data). much" capi t al mobi l i t y, in cont r ast t o t he Fe l ds t e i n- Hor i oka cl ai m of t oo little. One possi bl e expl anat i on of t he Gh o s h - Ot t o findings is t o vi ew t hem as r el at ed t o Deat on' s par adox of excessi ve consumpt i on smoot hness. The De a t on par adox can be r esol ved by st i pul at i ng t hat i ncome, t hough hi ghl y seri al l y cor r el at ed, does not l i t eral l y cont ai n a uni t root . ( When i ncome is st at i onar y in growt h rat es r at her t han in levels, a small change in cur r ent i ncome can i mpl y a ver y l arge change in pe r ma ne nt i ncome and, hence, in pr edi ct ed consumpt i on. ) Just as t he assumpt i on of a unit r oot in i ncome can l ead t o t he concl usi on t hat cons umpt i on is t oo smoot h, it can also pr oduce t he resul t t hat saving or t he cur r ent account is t oo vol at i l e. Thi s may hel p expl ai n t he Gh o s h - Ot t o vol at i l i t y results, t hough f ur t her i nvest i gat i on is r equi r ed. 4.2.4. Gl obal versus country-specific shocks' and the current account One shor t comi ng of t he studies consi der ed so far is t hei r i mpl i ci t assumpt i on t hat all shocks t o cash flow ar e pur el y i di osyncrat i c. 66 In reality, even a smalI count r y' s out put gr owt h or i nvest ment may be hi ghl y cor r el at ed wi t h t hat in t he rest of t he worl d. Out put shocks whi ch i dent i cal l y i mpact all count r i es should, however , expr ess t hemsel ves pr i mar i l y t hr ough t he gl obal i nt er est rat e, and not 66This criticism, of course, applies with equal force to much of the macroeconometric literature on consumption. 1790 M. Obstfeld and K. Rogoff in i ndi vi dual count ri es' current accounts. Gl i ck and Rogof f (1995) argue t hat this issue is empi ri cal l y i mpor t ant and try to conf r ont it. Thei r model also at t empt s t o trace shocks t o out put and i nvest ment back to changes in fact or productivity, al l owi ng for mor e st ruct ure t han the empi ri cal studies j ust discussed. 67 A si mpl er versi on of t he Gl i ck- Rogof f f r amewor k suffices t o i l l ust rat e t hei r mai n points. Let ' s label t he country-specific component of cash flow Q~: and the gl obal component QW, so t hat A Q = A Q c + A Q w The gl obal component is the part of Q t hat is perfect l y correl at ed with average worl d Q. Then, assumi ng t hat QC is 1(1), and t hat initial net forei gn asset po- sitions are zero, one can ( a f t e r m a n y s t e p s ) show t hat eq. (4.51) is repl aced by t he appr oxi mat e current -account equat i on 68 C A t = - E t -1-+-~r A Q,~ (4.55) s=t+l wher e t he i nt erest rat e used is t hat prevai l i ng al ong an initial st eady-st at e pat h. Accor di ng t o (4.55), onl y country-specific shocks affect current accounts. 69 To separ at e QC from QW, Glick and Rogof f consi der annual dat a for the G-7 count ri es over 1960-90, t reat i ng these count ri es as t he worl d (which, in t erms of economi c size, isn' t a bad appr oxi mat i on for most of t hei r sampl e period). They consi der t wo al t ernat i ve met hods of separat i ng shocks i nt o local and gl obal component s. The simplest is to form QC as Q - QW, where QW is t aken to be a mean- GNP wei ght ed average for t he ent i re group. The mor e sophi st i cat ed approach is to regress each count ry' s AQ on an i ndex of the r emai ni ng coun- t ri es' cash flows, defining QC as the regressi on residual. Glick and Rogof f find t hat the t wo approaches yi el d similar results for t hei r ul t i mat e current -account and i nvest ment equat i ons. Overall, gl obal shocks appear to account for a ver y significant por t i on of t ot al product i vi t y shocks in t he G-7 count ri es, r oughl y 50 percent . 7 Gl i ck and Rogof f find t hat t he gl obal versus country-specific dis- t i nct i on great l y i mproves t he ability of the i nt er t empor al approach to expl ai n 67Leiderman and Razin (1991) develop a model similar to Glick and Rogoff's, although they do not distinguish between global and country-specific shocks. 68The derivation of equation (4.55) requires that the variances of the underlying productivity shocks be constant. The global component of the shock, QW, affects world interest rates but not the current account. 69If initial net foreign asset positions are not zero, the interest-rate effects of global shocks can redistribute income from debtors to creditors in a way that alters current accounts. [Recall eq. (3.9) above.[ Gliek and Rogoff show that this effect is empiricaUy small. 7Costello (1993) and Stockman (1988) have found, for slightly different country samples and industry-level data, that global productivity or output shocks seem to be less correlated between similar manufacturing industries in different countries than between different industries in the same country. That evidence apparently points to a greater role for country-specific shocks, and it remains to reconcile it with the results discussed in the text. Ch. 34." The Int ert emporat Appr oach to the Current Account 1791 act ual current accounts: the coefficients on t he global shocks are i nvari abl y much smal l er t han t hose on the country-specific shocks, and are usual l y insignificant. 4 . 2 . 5 . E x t e n s i o n s The empi ri cal consumpt i on- smoot hi ng model s discussed so far all i gnore thc pr ecaut i onar y mot i ve for saving, as was r emar ked earlier. Cabal l ero (1990) has shown, however, t hat under specified assumpt i ons one can obt ai n cl osed-form consumpt i on funct i ons based on t he peri od ut i l i t y funct i on u ( C ) - - e x p ( - u C), wher e v is t he coefficient of absol ut e risk aversion. Thus, t here is no need to rel y exclusively on a l i near-quadrat i c f or mul at i on for cl osed-form solutions. Ghosh and Ost r y (1992, 1994) appl y Cabal l ero' s results to add a pr ecaut i onar y saving effect t o the present -val ue model of the current account. 71 The key new par amet er appeari ng in t hei r ext ended f r amewor k is o-g = Var(~'), where ~t = 0 t -- g r - I 0 t is t he dat e-t i nnovat i on to expect ed per manent pri vat e cash flow. Ghosh and Ost r y (1992) l ook at quar t er l y 1955-90 time-series dat a for Canada, Japan, the Uni t ed Ki ngdom, and the Uni t ed States. Because t hey are l ooki ng at time- series r at her t han cross-section dat a, t hey must negot i at e t he difficult issue of t i me var i at i on in ~ , which is a condi t i onal vari ance in a dynami c setting. Long enough i nt erval s must be al l owed for accurat e measur es of o-~, but intervals shoul d not be so l ong as to precl ude enough dat a poi nt s for meani ngf ul time- series regressions. 72 Usi ng two- to five-year intervals to measur e o- i , Ghosh and Ost r y find t hat t hei r pr ecaut i onar y vari abl e usual l y ent ers significantly and wi t h t he correct sign in present -val ue current account regressions. Ghosh and Ost r y (1994) find similarly positive results for devel opi ng countries. Thei r poi nt est i mat es suggest t hat pr ecaut i onar y savings are of t he or der of magni t ude of 5 per cent of i mport s for the Af r i can region, 4 percent for commodi t y export ers, and 14 per cent for fuel exporters. None of t he empi ri cal studies discussed thus far distinguishes bet ween dura- bles and nondur abl es or bet ween t radeabl es and nont r adeabl es. As we empha- sized in our t heor et i cal discussion, bot h distinctions can be i mpor t ant for the current account. Bur da and Ger l ach (1993) argue t hat , in theory, durabl e-goods i mport s are much mor e sensitive to expect ed movement s in the real exchange rat e (because of t he resul t i ng expect at i ons of capital gain or loss) t han arc nondur abl es i mport s. Est i mat i ng t hei r model poses a number of difficulties: in particular, the t heor y calls for a measur e of t he stock of consumer durables, but this is difficult to obt ai n in practice. Usi ng a vect or error-correct i on time- 71An alternative theoretical treatment is Rodriguez (1993). 72An alternative approach would be to adopt an explicit parameterization of Var~ ,(~:t). 1792 M. Obstfeld and K. Rogoff series specification including quarterly data on the current account, expected permanent net income, the relative price of durables in terms of nondurables, and a variable capturing expected changes in that price, Burda and Gerlach find that the expected price changes have a significant correlation with the U.S. current account over 1970-88. This finding, they argue, provides support for the empirical role of the durables versus nondurables distinction. It is difficult to compare these results with those of the empirical models discussed earlier because the Burda-Gerlach setup, with its very general lag structure, imposes much less theoretical structure. It would be interesting to pursue an alternative approach based on eq. (3.23). 73 Rogoff (1992) incorporates nontraded goods into an empirical intertemporal model, although his primary focus is on explaining the well-documented near random-walk behavior of real exchange rates. His main result is that intertem- poral consumption smoothing in traded goods might account for the persistence of innovations in real exchange rates. In the simplest case, assume exogenous output of tradeables and nontradeables, a Cobb-Douglas period utility function [p = 1 in eq. (3.17)], and an intertemporal substitution elasticity, or, of 1. In this case, as we saw earlier, people smooth their consumption of tradeables inde- pendently of the evolution of their nontradeables consumption. As above, the real exchange rate, the relative price of nontradeables on date t, depends on Grt, YNt, and consumers' expenditure shares. A permanent shock to tradeables output raises Ca- permanently and thereby permanently raises the relative price of nontradeables. But even a t emporary rise in traded-goods output raises Gr permanently because of consumption smoothing. The model thus can explain why the persistence in real exchange rate movements might be much greater than that of the underlying exogenous shocks. A country's ability to borrow and lend in international markets is the key to this result: it would not obtain absent international capital mobility. Rogoff (1992) applies his model to data for Japan and the United States, finding some support, though further testing is required. 5. How useful is the theory? Even in its most rudimentary forms, the intertemporal approach to the cur- rent account has proved valuable for analyzing a host of important problems. Without an intertemporal approach, it would be hard to analyze or evaluate the current-account patterns that followed the two oil shocks of the 1970s. The dynamic budget constraints emphasized throughout this chapter are also essen- tial in analyzing episodes of capital-market disruption, such as the developing- country debt crisis of the 1980s. True, the standard intertemporal models must 73For some preliminary empirical results, see Obstfeld and Rogoff (1996), ch. 2. Ch. 34: The I nt er t empor al Ap p r o a c h to t he Current Ac c o u n t ! 793 be ext ended t o t ake account of defaul t risk but, as we have seen, the mai n qual i t at i ve insights do not change. More generally, model s t hat fail to i nt egrat e i nvest ment , saving, and growt h make it vi rt ual l y i mpossi bl e to under st and why some count ri es have persi st ent current account i mbal ances. Why, for exam- ple, are Canada' s and Aust ral i a' s current accounts per enni al l y in deficit, and Japan' s in surplus, despi t e wi de swings in t hei r currenci es' real exchange rates'? Over l appi ng- gener at i ons vari ant s of the i nt er t empor al model are i ndi spensabl e for t hi nki ng about how, say, the aging of Japan' s popul at i on coul d event ual l y l ead to a fall in Japan' s persi st ent t rade surpluses. As positive descri pt i ons of t he current account, t he simple i nt er t empor al t heori es are not wi t hout t hei r limitations. As we saw above, simple time-series model s based on consumpt i on smoot hi ng seem to work fai rl y well for some count ri es (for exampl e, Sweden) but, in ot her cases, cl earl y miss much of the action. Fur t her research allowing for t i me-varyi ng i nt erest rates, mul t i pl e goods, durabl es, nomi nal price rigidities, and some l i qui di t y-const rai ned consumers may l ead to bet t er descriptive power. If the si mpl est i nfi ni t el y-l i ved represent a- tive agent model s are t o be believed, t hen it is a puzzle t hat ratios of forei gn debt t o out put sel dom exceed 1:1 when plausible par amet er est i mat es suggest t hat rat i os of 5:1 or 10:1 could easily be sust ai nabl e and even optimal. Obser ved de bt - GDP ratios are easi er to rat i onal i ze in economi es with finite- lived dynast i es, but such model s, while capabl e of embr aci ng a wi der set of empi ri cal phenomena, also pose empirical puzzles. A fai rl y robust i mpl i cat i on is t hat gover nment deficits l ead to current -account deficits, but the empirical evi dence support i ng this predi ct i on, while suggestive, is har dl y a basis for st rong conclusions. The striking i ndust ri al -count ry correl at i on observed over 1976-1985 is not cl earl y evi dent l at er on. Promi si ng directions for f ut ur e i nvest i gat i on in- cl ude model s wi t h mor e det ai l ed i nt er gener at i onal st ruct ures and a mor e com- prehensi ve account i ng of the i nt er gener at i onal t ransfers i mpl i ed by fiscal and social i nsurance policies. The model s we have discussed in this chapt er provi de onl y a st art i ng point. Obviously, the t ask of bui l di ng and empi ri cal l y appl yi ng ri cher and mor e realistic i nt er t empor al model s will not be an easy one. But t here is no avoi di ng this chal l enge, since t he t wo l eadi ng al t ernat i ves to the i nt er t empor al model are seri ousl y flawed. One al t ernat i ve t hat has been expl ored ext ensi vel y in recent research is the compl et e mar ket s model , in which country-specific shocks of all types - to hu- man as well as financial weal t h, to personal as well as cor por at e taxes - can be i nsured i nt ernat i onal l y. If this approach is correct, of course, t hen t he current account is little mor e t han an account i ng convent i on wi t hout maj or significance even for a count ry' s rel at i ve weal t h position. (See t he discussion in Section 3.2.1.) We have argued t hat real -worl d i nt er nat i onal capital mar ket s are very far f r om the frictionless, ful l -i nformat i on, compl et e- mar ket s ideal. Fact ors ino 1794 M. Obstfeld and K. Rogoff hibiting complete domestic capital markets include moral hazard problems in lending at the microeconomic level, finite lifetimes, and difficulties in insur- ing labor income. In international markets, these problems are compounded by sovereign default risk, difficulties in insuring national government spending shocks, and cultural and institutional differences. Of course, it would be vastly preferable to model explicitly these capital-market imperfections rather than simply to assume limited asset trade, especially for understanding the impact of government policy. We have discussed some work along these lines and pre- sented a simple example. Until these models have been more fully developed, however, the intertemporal model seems to provide a much closer description of reality than does the complete markets model. Complete-market models represent an extreme alternative to the intertem- poral approach. At the opposite pole are variants of the open-economy IS-LM model due to Mundell (1968) and Fleming (1962). This approach, which ignores intertemporal choice and even intertemporal budget constraints, remains over- whelmingly dominant in policy circles. But as a framework for addressing funda- mentally dynamic phenomena such as the current account and government debt, the Mundell-Fleming paradigm, even when jerry-rigged with dynamic add-ons, is fatally handicapped. The Mundell-Fleming approach offers no valid benchmark for evaluating ex- ternal balance. In practice, policymakers often strive to avoid a negative current account. Just as efficient international trade generally requires unbalanced trade across commodity groups, however, efficient trade across time often calls for an unbalanced current account. The intertemporal approach identifies circum- stances, for example, a transitory fall in output or a rise in domestic investment productivity, that justify a current account deficit. On these issues, the Mundell- Fleming approach has nothing to say. Evaluating the real exchange rate consistent with full employment and exter- nal balance is another prime concern of policy: intervention or realignment dc- cisions may hinge on the determination that a currency's value is "misaligned". Since the Mundell-Fleming approach has nothing to say about external balance, it is, afortiori, unable to address the possibility of misalignment. Because it lacks microfoundations or even the most basic intertemporal bud- get constraints, the Mundell-Fleming approach provides no grounds for nor- mative judgments on current accounts or international macroeconomic policies. No economist would take seriously an assessment of tax, trade, or regulatory policy based on a model with these shortcomings. The intertemporal approach to the current account offers a viable framework for assessing macroeconomic policy, one that must supplant the Mundell-Fleming framework for normative questions. It is hard to portray Mundell-Fleming as a successful positive current-account theory, either. Without denying the theory's empirical appeal in capturing short- Ch. 34." The lntertemporal Approach to the Current Account 1795 run ma c r o e c o n o mi c d e v e l o p me n t s ove r s o me e pi s o de s , t he c or e mo d e l has no cl ear, mu c h l es s t e s t abl e , pr e di c t i ons a bo ut c ur r e nt - a c c o unt dynami c s . Ag a i n , as i nt e r t e mpo r a l mo d e l s b e c o me mo r e t ract abl e and e n j o y wi de r empi ri cal t es t i ng, it s e e ms t o us t hat t h e y mus t ul t i ma t e l y c o me t o s uppl a nt mo di f i e d Mu n d e l l - Fl e mi n g mo d e l s f or po s i t i v e as we l l as n o r ma t i v e que s t i o ns . Re f e r e n c e s Abel, A.B. (1982), "Dynami c effects of permanent and temporary tax policies in a q model ol investment", Journal of Monetary Economics 9:353-373. Ahmed, S. (1986), "Temporary and permanent government spending in an open cconomy: Some evidence for the Uni t ed Kingdom", Journal of Monetary Economics 17:197-224. Ahmed, S. (1987), "Gover nment spending, the balance of trade and the terms of trade in British history", Journal of Monetary Economics 20, 195-220. Alexander, S.S. (1952), "Effects of a devaluation on a trade balance", International Monetary Fund Staff Papers 2:263-278. Atkeson, A. (1991), "International lending with moral hazard and risk of repudiation", Economet- rica 59:1069-1089. Atkeson, A. and R.E. Lucas, Jr. (1992), "On efficient distribution with private information", Review of Economic Studies 59:427-453. Backus, D.K., EJ. Kehoe, and F.E. Kydland (1992), "International real business cycles", Journal of Political Economy 100:745-775. Bardhan, PK. (1967), "Opt i mum foreign borrowing", in: K. Shell, ed., Essays ill the theory of optimal economic growth (MIT Press, Cambridge, MA). Baxter, M. and M.J. Crucini (1993a), "Explaining saving-investment correlations", American Eco- nomic Revi ew 83:416-436. Baxter, M. and M.J. Crucini (1993b), "Business cycles and the asset structure of foreign trade", Working Paper, University of Rochester. Bean, C.R. (1986), "The terms of trade, labour supply and the current account", Economic Journal 96:38-46. Bergsten, C.E, ed. (1991), International adjustment and financing: The lessons of 1985-1991 (Insti- tute for International Economics, Washington, DC). Bernanke, B. (1985), "Adjustment costs, durables, and aggregate consumption", Journal of Mone- tary Economics 15:41-68. Bernheim, B.D. (1988), "Budget deficits and the balance of trade", in: L. Summers, cd., Tax policy and the economy, vol. 2 (MIT Press, Cambridge, MA). Blanchard, O.J. (1985), "Debt , deficits, and finite horizons", Journal of Political Economy 93:223- 247. Bruno, M. (1970), "Trade, growth and capital", Working Paper, MIT Press, Cambridge, MA. Buiter, W.H, (1981), "Time preference and international lending and borrowing in an overlapping- generations model", Journal of Political Economy 89:769-797. Buiter, W.H. (1989), Budgetary policy, international and intertemporal trade in the global economy (North-Holland, Amsterdam). Burda, M. and S. Gerlach (1992), "Intertemporal prices and the U.S. trade balance", American Economic Review 82:1235-1253. Caballero, R. (1990), Consumption puzzles and precautionary savings, Journal of Monetary Eco- nomics 25:113-136. Campbell, J.Y. (1987), "Does saving anticipate declining labor income? An alternative test of the permanent income hypothesis", Econometrica 55:1249-1274. Campbell, J.Y. and R.J. Shiller (1987), "Cointegration and tests of present value models", Journal of Political Economy 95:1062-1088. 1796 M. Obstfeld and K. Rogoff Carroll, C.D. (1992), "The buffer-stock theory of saving: Some macroeconomic evidence", Brookings Papers on Economi c Activity 2:61-135. Clarida, R.H. (1990), "International lending and borrowing in a stochastic stationary equilibrium", International Economic Review 31:543-558. Cole, H.L. and M. Obstfeld (1991), "Commodity trade and international risk sharing: How much do financial markets matter?", Journal of Monetary Economics 28:3-24. Costello, D,M. (1993), " A cross-country, cross-industry comparison of productivity growth", Journal of Political Economy 101:207-222. Cox, J.C. and C. Huang (1989), "Optimal consumption and portfolio policies when asset prices follow a diffusion process", Journal of Economic Theory 49:33-83. Deaton, A. (1987), "Life cycle models of consumption: Is the evidence consistent with the theory?", in: T.E Bewley, ed., Advances in econometrics: Fifth world congress, vol. 2 (Cambridge University Press, Cambridge, UK). Deaton, A. (1992), Understanding consumption (Clarendon Press, Oxford, UK). de Roover, R. 0966), The rise and decline of the Medici bank, 1397-1494 (W.W. Norton, New York). Devereux, M.B. and S. Shi (1991), "Capital accumulation and the current account in a two-country model", Journal of International Economics 30:1-25. Dixit, A.K. and V. Norman (1980), Theory of international trade: A dual, general equilibrium approach (James Nisbet & Co. Ltd. and Cambridge University Press, Welwyn and Cambridge, U K ) . Dooley, M., J. Frankel, and D.J. Mathieson (1987), "International capital mobility: What do savi ng- investment correlations tell us?", International Monetary Fund Staff Papers 34:503-530. Dornbusch, R. (1983), "Real interest rates, home goods, and optimal external borrowing", Journal of Political Economy 91:141-153. Eaton, J. (1988), "Foreign-owned land", American Economic Revi ew 78:76-88. Edwards, S. (1989), Real exchange rates, devaluation, and adjustment (MIT Press, Cambridge, MA). Eichengreen, B. (1990), "Trends and cycles in foreign lending", in: H. Siebert, ed., Capital flows in the world economy (J.C.B. Mohr, Tubingen). Ekern, S. and R. Wilson (1974), "On the theory of the firm in an economy with incomplete markets", Bell Journal of Economics and Management Science 5:171-180. Engel, C. and K. Kletzer (1989), "Saving and investment in an open economy with non-traded goods", International Economic Review 30:735-752. Feinstein, C.H. (1972), National income, expenditure, and output of the United Kingdom, 1855-1965 (Cambridge University Press, Cambridge, UK). Feldstein, M. and E Bacchetta (1991), "National saving and international investment", in: B.D. Bern- heim and J.B. Shoven, eds., National saving and economic performance (University of Chicago Press, Chicago, IL). Feldstein, M. and C. Horioka (1980), "Domest i c saving and international capital flows", Economic Journal 90:314-329. Fieleke, N.S. (1982), "National saving and international investment", in: Saving and government policy (Federal Reserve Bank of Boston, Boston, MA). Fisher, I. (1930), The theory of interest (Macmillan, New York). Fleming, J.M. (1962), "Domest i c financial policies under fixed and under floating exchange rates", International Monetary Fund Staff Papers 9:369-379. Frankel, J.A. (1993), "Quantifying international capital mobility in the 1980s", in: On exchange rates (MIT Press, Cambridge, MA). Frenkel, J.A. and A. Razin (1987), Fiscal policies and the world economy: An intertemporal ap- proach (M1T Press, Cambridge, MA). Fukao, K. and K. Hamada (1994), "International trade and investment under different rates of time preference", Journal of the Japanese and International Economies 8:22-52. Gavin, M. (1990), "Structural adjustment to a terms of trade disturbance: The role of relative prices", Journal of International Economics 28:217-243. Gertler, M. and K. Rogoff (1990), "North-South lending and endogenous domestic capital market inefficiencies", Journal of Monetary Economics 26:245-266. C7~. 34: The lntertemporal Approach to the Current Account 1797 Ghosh, A. (1995), "Capital mobility amongst the major industrialized countries: Too little or too much?", Economic Journal 105:107-128. Ghosh, A. and J. Ostry (1992), "Macroeconomic uncertainty, precautionary savings, and the current account", Working Paper, International Monetary Fund. Gfiosh, A. and J. Ostry (1994), "Export instability and the external balance in developing countries". International Monetary Fund Staff Papers 41:214-235. Ghosh, A. and J. Ostry (1995). "The current account in developing countries: A perspective from the consumption-smoothing approach", World Bank Economic Revi ew 9. Glick, R. and K. Rogoff (1995), "Global versus country-specific productivity shocks and the current account", Journal of Monetary Economics 35:159-192. Goldstein, M. and M.S. Khan (1985), "Income and price effects in foreign trade", in R. g( Jonc~ and PB. Kenen, eds., Handbook of international economics, vol. 2 (North-Holland, Amsterdam). Hall, R.E. (i978), "Stochastic implications of the life cycle-permanent income hypothesis: Theory and evidence", Journal of Political Economy 86:971-987. Hamada, K. (1969), "Opt i mal capital accumulation by an economy facing an international capital market", Journal of Political Economy 77:684-697. Harberger, A.C. (1950), "Currency depreciation, income and the balance of trade", Journal of Political Economy 58:47-60. Hayashi, E (i982), "Tobin' s marginal q and average q: A neoclassical interpretation", Econometrica 50:213-224. Hercowitz, Z. (1986), "On the determination of the external debt: The case of Israel", Journal of International Money and Finance 5:315-334. Johnson, D. (1986), "Consumption, permanent income, and financial wealth in Canada: Empirical evidence on the intertemporal approach to the current account", Canadian Journal of Economics 19:189-206. Jones, M.T. and M. Obstfeld (1994), "Saving and investment under the gold standard", Working Paper, University of California at Berkeley. Kollmann, R. (1992), "Incomplete asset markets and international real business cycles", Working Paper, University of Montreal. Laursen, S. and L.A. Metzler (1950), "Flexible exchange rates and the theory of employment", Revi ew of Economics and Statistics 32:281-299. Leiderman, L. and A. Razin (1991), "Determinants of external imbalances: The role of taxes, government spending, and productivity", Journal of the Japanese and International Economies 5:421-450. Leland, H.E. (1968), "Savings and uncertainty: The precautionary demand for savings", Quarterly Journal of Economics 82:465-473. Lucas, R.E., Jr. (1976), "Economet ri c policy evaluation: A critique", Carnegi e-Rochest er Confer- ence Series on Public Policy 1:119-46. Lucas, R.E., Jr. (1982), "Interest rates and currency prices in a two-country world", ,i(annal <~ Monetary Economics 10:335-359. Maddison, A. (1991), Dynamic forces in capitalist development (Oxford University Press, Oxl or d) Matsuyama, K. (1987), "Current account dynamics in a finite horizon model", Journal ot ~nteH~a tional Economics 23:299-313. Matsuyama, K. (1990), "Residential investment and the current account", Journal ()f In~c~ nati~))~e6 Economics 28:137-153. Mehra, R. and E.C. Prescott (1985), "The equity premium: A puzzle", Journal of Monetary Eco- nomics 15:145-161. Mendoza, E.G. (1991), "Real business cycles in a small open economy", American Econcmlic Review, 81:797-818. Mundell, R.A. (1968), International economics (Macmillan, New York). Murphy, R.G. (1986), "Productivity shocks, non-traded goods and optimal capital accumulation", European Economic Revi ew 30:1081-1095. Neal, L. (1990), The rise of financial capitalism: International capital markets in the age of reason (Cambridge University Press, Cambridge, UK). 1798 M. Obstfeld and K. Rogoff Obstfeld, M. (1982), "Aggregat e spending and the terms of trade: Is there a Laursen-Met zl er effect?", Quarterly Journal of Economics 97:251-270. Obstfeld, M. (1986), Capital mobility in the world economy: Theory and measurement", Carnegie Rochest er Conference Series on Public Policy 24:55-103. Obstfeld, M. (1994), "Ar e industrial-country consumption risks globally diversified?", in: L. Leider- man and A. Razin, eds., Capital mobility (Cambridge University Press, Cambridge, UK). Obstfeld, M. (1995), "International capital mobility in the 1990s", in: RB. Kenen, ed., Understand- ing interdependence: The macroeeonomics of the open economy (Princeton University Press, Princeton, N J). Obstfeld, M. and K. Rogoff (1996), Foundations of international macroeconomics, (MIT Press, Cambridge, MA). Obstfeld, M. and A.C. Stockman (1985), "Exchange-rate dynamics", in: R.W. Jones and RB. Kenen, eds., Handbook of international economics, vol. 2 (North-Holland, Amsterdam). Ostry, J.D. (1988), "The balance of trade, terms of trade, and real exchange rate", International Monetary Fund Staff Papers 35:541-573. Ostry, J.D. and C.M. Reinhart (1992), "Private saving and terms of trade shocks", International Monetary Fund Staff Papers 39:495-517. Otto, G. (1992), "Testing a present-value model of the current account: Evidence from US and Canadian time series", Journal of International Money and Finance 11:414-430. Persson, T. (1985), "Deficits and intergenerational welfare in open economies", Journal of Interna- tional Economics 19:1-19. Razin, A. (1995), "The dynamic optimizing approach to the current account: Theory and evidence", in: RB. Kenen, ed., Understanding interdependence: The macroeconomics of the open economy (Princeton University Press, Princeton, N J). Rodriguez, A. (1993), "Precautionary saving and the Laursen-Met zl er effect", Journal of Interna- tional Money and Finance 12:332-343. Rogoff, K. (1992), "Traded goods consumption smoothing and the random walk behavior of the real exchange rat e", Bank of Japan Monetary and Economic Studies 10:1-29. Roubini, N. (1988), "Current accounts and budget deficits in an intertemporal model of consumption and taxation smoothing", National Bureau of Economic Research Working Paper No. 2773. Sachs, J.D. (1981), "The current account and macroeconomic adjustment in the 1970s", Brookings Papers on Economic Activity 1:201-268. Sachs, J.D. (1982), "The current account in the macroeconomic adjustment process", Scandinavian Journal of Economics 84:147-159. Sheffrin, S.M. and W.T. Woo (1990), "Present value tests of an intertemporal model of the current account", Journal of International Economics 29:237-253. Shi, S. and L.G. Epstein (1993), "Habits and time preference", International Economic Review 34:61-84 Shiller, R.J. (1993), Macro markets: Creating institutions for managing society's largest economic risks (Clarendon Press, Oxford, UK). Stockman, A.C. (1988a), "Fiscal policies and international financial markets", in: J.A. Frenkel, ed., International aspects of fiscal policies (University of Chicago Press, Chicago, IL). Stockman, A.C. (1988b), "Sectoral and national aggregate disturbances to industrial output in seven European countries", Journal of Monetary Economics 21:387-409. Stockman, A.C. and L.L. Tesar (1995), "Tastes and technology in a two-country model of the business cycle: Explaining international comovements", Ameri can Economic Revi ew 85:168-185. Stulz, R. (1988), "Capital mobility and the current account", Journal of International Money and Finance 7:167-180. Summers, L.H. (1981), "Taxation and corporate investment: A q-theory approach", Brookings Pa- pers on Economic Activity, 67-127. Summers, L.H. (1988), "Tax policy and international competitiveness", in: J.A. Frenkel, ed., Inter- national aspects of fiscal policies (University of Chicago Press, Chicago, IL). Svensson, L.E.O. (1988), "Trade in risky assets", American Economic Review 78:375-394. Svensson, L.E.O. and A. Razin (1983), "The terms of trade and the current account: The Harbergm= Laursen-Met zl er effect", Journal of Political Economy 91:97-125. Ch. 34." The lntertemporal Approach to the Current Account 1799 Taylor, A.M. (1994), "Domest i c saving and i nt ernat i onal capital llows reconsi dered", Nat i onal Bu- r eau of Economi c Resear ch Working Paper No. 4892. Tesar, L.L. (1991), "Savings, i nvest ment , and i nt ernat i onal capital flows", Journal of Int ernat i onal Economi cs 31:55-78. Trehan, B. and C,E. Walsh (1991), "Testing i nt ert emporal budget constraints: Theory and applica- tions to U.S. federal budget and current account deficits", Journal of Money, Credit and Banki ng 23:206-223. Weil, R (1989), "Over l appi ng families of infinitely-lived agent s", Journal of Public Economics 38:183-198.