The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
clevelandfed.org/research/workpaper/index.cfm Working P a ~ e r9213 RELATIVE PRICE MOVEMENTS IN DYNAMIC GENERAL EQUILIBRIUM MODELS OF INTERNATIONAL TRADE by David K. Backus, Patrick J. Kehoe, and Finn E. Kydland David K. Backus is a professor of economics at the Stern School of Business, New York University; Patrick J. Kehoe is a professor of economics at the University of Pennsylvania and a consultant at the Federal Reserve Bank of Minneapolis; and Finn E. Kydland is a professor of economics at Carnegie-Mellon University, Pittsburgh, and a research associate of the Federal Reserve Bank of Cleveland. The authors thank the National Science Foundation, the Institute for Empirical Macroeconomics, and the Center for Japan-U.S. Business and Economic Studies faculty fellowship program for financial support. They also thank Ronald Jones, Maury Obstfeld, and Chris Stefanadis for helpful comments. Working papers of the Federal Reserve Bank of Cleveland are preliminary materials circulated to stimulate discussion and critical comment. The views stated herein are those of the authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System. November 1992 clevelandfed.org/research/workpaper/index.cfm ABSTRACT We examine the behavior of international relative prices from the perspective of dynamic general equilibrium theory, with particular emphasis on the variability of the terms of trade and the relation between the terms of t r a d e and net exports. We highlight aspects of the theory t h a t a r e critical in determining these properties, contrast our perspective with those associated with the Marshall-Lerner condition and the Harberger-Laursen-Metzler effect, and point out features of the data that have proved difficult t o explain within existing dynamic general equilibrium models. clevelandfed.org/research/workpaper/index.cfm 1 Introduction Relative prices are a central feature of both international trade and open-economy macroeconomics. the pure theory of Although the emphasis differs in the two branches of international economics, t o a great extent the same theory underlies theoretical and empirical work in each. of dynamic general equilibrium theory, or international Applications real business cycle theory, continue this tradition by extending, a t the aggregate level, several of the features of static trade theory t o dynamic and stochastic settings. What these applications offer, we think, is a deeper understanding of the dynamics of trade and relative prices. The Marshall-Lerner condition is, without question, the most common link between trade theory and international macroeconomics. In trade theory, this elasticity condition on import demand functions determines the direction of many comparative statics exercises and serves as a stability condition on an otherwise process static will theory, succeed macroeconomics, the telling in same us whether establishing condition is a disequilibrium equilibrium. used to In adjustment international establish a positive association between the trade balance and the terms of trade or real exchange rate. This is the level a t which the theory is presented in most textbooks and, indeed, in the popular Mundell-Fleming and Dornbusch macroeconomic models of open economies. The macroeconomic branch of international economics has also developed insights that a r e largely independent of the theory of trade. approach focuses on the and the balance of trade. theory is its suggestion accounting relation The absorption connecting saving, investment, What distinguishes this work from static trade that the trade balance, or the closely related current account, reflects the dynamic decisions by agents t o lend or borrow in international capital markets. A critical development f o r understanding clevelandfed.org/research/workpaper/index.cfm t h e relation between t r a d e and relative prices i s t h e recognition t h a t any dependence of t h e t r a d e balance on t h e terms of t r a d e implies, as a m a t t e r of accounting, a similar relation with international borrowing and lending. This connection was noted by Harberger (1950) and Laursen and Metzler (1950) and l a t e r incorporated by Obstfeld (1982) and Svensson and Razin (1983) into explicitly dynamic theories of t h e balance of trade. These l a t e r papers emphasize t h e influence of t e r m s of t r a d e movements on permanent income, and hence on saving. They argue t h a t persistent changes in t h e t e r m s of t r a d e have l a r g e r income e f f e c t s t h a n transitory changes, and t h u s give r i s e t o potentially different relations between t h e t r a d e balance and t h e terms of trade. The effect, in their analysis, is a comparison between two deterministic equilibria. We approach t r a d e and relative price dynamics f r o m a somewhat different theoretical tradition, t h a t of dynamic general equilibrium theory. Like much of s t a t i c t r a d e theory, we use competitive consumers and producers. t h a t work, however, our theory i s explicitly dynamic. Unlike And unlike t h e modern approach t o t h e Harberger-Laursen-Metzler e f f e c t , we consider not comparisons between different stochastic deterministic theoretical equilibria, economies. We but properties consider in the experiment t h a t applied economists consider in t h e data: between trade Somewhat t o and our relative surprise, price this variables approach along leads to an of equilibria theory the in same t h e correlations equilibrium substantially path. different views of t r a d e and price behavior than suggested by earlier work. The application of dynamic general equilibrium theory t o international t r a d e h a s also led, in a rapidly growing number of papers, t o attempts t o quantify t h e theory's properties and compare them t o properties of national economies. These a t t e m p t s have led t o a clearer understanding of which f e a t u r e s of t h e d a t a can be accounted f o r by t h e present s t a t e of theory and clevelandfed.org/research/workpaper/index.cfm which remain predictions anomalous. than qualitative This quantitative theory and helps approach generates sharper t o focus f u r t h e r theoretical work on clearly defined issues. We elaborate on the twin themes of theoretical development and We s t a r t , in section 2, by quantitative properties in t h e r e s t of t h e paper. documenting some of t h e salient properties of aggregate t r a d e and relative prices a number for explicit of t h e o b j e c t s of industrialized interest countries. in theoretical These properties make economies and serve as a basis of comparison with t h e theory. We develop t h e theory in a series of two-country worlds, highlighting as we go t h e roles played by d i f f e r e n t theoretical features. Section 3 i s devoted t o a n exchange economy in which each country specializes in a single traded good. Here, the variability of relative prices and between prices and t r a d e are governed by a single parameter: of substitution between foreign and domestic goods. the relation t h e elasticity Some of t h e quantitative properties of t h i s economy change when we consider preferences t h a t are not additively separable between consumption of home goods. 4. foreign and domestic goods and that favor This aspect of t h e theory i s developed in section We find, among o t h e r things, t h a t agents' risk aversion plays a r o l e in both t h e dynamics of t r a d e and prices and t h e relation between these t w o variables. In section 5, we compare t h i s theory with alternatives based on t h e Marshall-Lerner condition and t h e Harberger-Laursen-Metzler that our elasticity condition is not related to effect. the condition, which i s always satisfied in our symmetric economy. t o t h e Harberger-Laursen-Metzler effect, we find t h a t We show Marshall-Lerner With respect t h e persistence of shocks i s orthogonal t o t h e relation between relative prices and t h e balance of trade: f o r given preference parameters, t h e correlation between t r a d e and clevelandfed.org/research/workpaper/index.cfm prices is the same whether price changes last one period o r one hundred. both comparisons, dynamic general equilibrium theory provides a In different perspective on trade and price fluctuations than does earlier work. In the structure remaining that change theory's predictions. sections, we consider extensions of the theoretical some of i t s quantitative features and broaden the In section 6, we explore shocks t o government spending a s well a s t o aggregate endowments. A s one might expect, this extension has the potential t o change equilibrium comovements considerably. We find, f o r example, that the sign of the relation between the t r a d e balance and the terms of trade depends on the relative sizes of shocks t o endowments and government spending, a s well a s on the elasticity of substitution. In section 7, we embed the exchange structure of earlier sections into an economy with endogenous labor supply and capital formation. element here is capital formation. dynamics of trade now reflect, physical investment. The The critical With this modification, we find that the t o a large extent, cyclical fluctuations in most striking result is an asymmetric cross-correlation function f o r the trade balance and the terms of which the we label S-curve. This feature does not arise in trade, exchange economies, whereby construction investment is zero and the cross-correlation function is symmetric. In this sense, the dynamics of capital formation play an important role in connecting the dynamics of the trade balance and the relative price of foreign t o domestic goods. We conclude with a few remarks on the strengths and weaknesses of existing dynamic general equilibrium theories of international trade, and suggestions f o r directions the theory might take in the future. 2 F i r s t Look a t t h e Data Since the ultimate objective of our theory is t o account f o r empirical clevelandfed.org/research/workpaper/index.cfm regularities, we start by looking at some of t h e properties of international relative prices. denoted p, We focus here on t h r e e variables. i s t h e r a t i o of The t e r m s of t h e import price deflator t o t h e export price deflator, both taken f r o m national income and product accounts. denoted nx, i s t h e r a t i o of output in current prices. on t h e country, trade, exports minus imports, Net exports, in c u r r e n t prices, to Real output, denoted y, i s GDP o r GNP, depending in base-year (generally 1985). prices All t h r e e variables a r e constructed with d a t a taken from t h e OECD's Quarterly National Accounts. In table 4.1 we r e p o r t various properties of t h r e e countries: t h e t e r m s of trade f o r Japan, t h e United Kingdom, and t h e United States. s t a t i s t i c s r e f e r t o Hodrick-Prescott filtered variables, of t r a d e and r e a l output a r e logarithms. These and both t h e t e r m s Many of t h e same properties a r e reported in Backus, Kehoe, and Kydland (1992b) and Blackburn and Ravn (1991) for additional developed countries and in Mendoza (1992) f o r developing The sample period in table 4.1 r u n s f r o m 1955:2 t o 1989:4, which countries. enables us t o look separately at t h e periods before and a f t e r t h e collapse of Bretton Woods. We see, f o r a s t a r t , t h a t movements in t h e t e r m s of t r a d e have been both variable and persistent. The standard deviation of terms of trade f1uc;uations ranges f r o m 2.71 percent in t h e United Kingdom t o 5.97 percent in Japan. Both here and in our earlier paper, variability of t h e t e r m s of t r a d e i s considerably larger in Japan than in other countries. We are unsure, at t h i s point, how much of t h i s additional variability reflects t r u e relative price movement and how much has t o do with differences in t h e manner in which t r a d e prices a r e constructed. Both Alterman (1991) and Graboyes (1991) r a i s e questions concerning t h e quality of c u r r e n t t r a d e prices in t h e United States, and these problems may be g r e a t e r in earlier periods and other countries. Alterman (1991) estimates t h a t improved price d a t a exhibit about clevelandfed.org/research/workpaper/index.cfm 30 percent less variability t h a n those reported here. Nevertheless, we a r e probably on s a f e ground in claiming substantial variability of t h e t e r m s of trade in all countries. Persistence is evident in the autocorrelations, Table 4.1 also verifies t h a t which a r e generally in t h e neighborhood of 0.8. t h e r e h a s been much more variability of t h e t e r m s of t r a d e since t h e advent of floating exchange r a t e s than before, a f e a t u r e s t r e s s e d by Mussa (1986) f o r r e a l exchange r a t e s ( r a t i o s of consumer price indexes converted at spot exchange r a t e s ) . Standard deviations of t h e t e r m s of t r a d e a r e typically t w o t o t h r e e times l a r g e r in t h e l a t t e r period. We also include, in table 4.1, correlations of t h e t e r m s of t r a d e with net exports and r e a l output. coherence across periods, and net exports have With respect t o net exports, we find g r e a t e r but little across countries. generally been positively The t e r m s of t r a d e correlated in S t a t e s and negatively correlated in Japan and t h e United Kingdom. table 4.2, which covers t h e post-Bretton the United We s e e in Woods period f o r 10 countries, t h a t t h e United S t a t e s i s an outlier in t h i s regard: The correlation between t h e t r a d e balance and t h e t e r m s of t r a d e i s negative f o r every other country. With respect t o output, t h e r e h a s been no regularity in t h e correlation with t h e t e r m s of t r a d e , either over time o r across countries. The contemporaneous correlation between net e x p o r t s and t h e t e r m s of t r a d e f a i l s t o capture a n important regularity t h a t appears when we examine t h e complete cross-correlation function: t h e correlations, t h a t is, between pt and n ~ ~ f + o r ~various , leads and lags k. r e f e r s t o k=O. The contemporaneous correlation For positive k t h e correlations pertain t o net exports and p a s t prices, and f o r negative k t h e reverse. We find f o r Japan and t h e United Kingdom t h a t t h i s function h a s an asymmetric S shape, which we call t h e S-curve. For Japan and t h e United Kingdom, t h i s f e a t u r e appears not only in t h e postwar period as a whole, but in t h e pre- and post-Bretton Woods clevelandfed.org/research/workpaper/index.cfm subperiods as well. For t h e United States, the same p a t t e r n is evident only in the earlier period. Our earlier paper (Backus, Kehoe, and Kydland, 1992b, figure 1) documents t h i s pattern in eight of eleven countries. In t h e remainder of t h e paper, we examine these properties from the perspective of a series of successively more complex dynamic general equilibrium models, bringing additional d a t a t o bear when t h e theory suggests it. For now, we note t h a t t h e t e r m s of t r a d e has been highly variable and persistent in all t h r e e countries, t h a t i t s correlation with net exports i s generally negative, and t h a t t h e cross-correlation function f o r net exports and t h e t e r m s of t r a d e is often asymmetric (the S-curve property). 3 A Dynamic Exchange Economy One of the simplest dynamic general equilibrium models of economy has two countries t h a t t r a d e specialized endowments. from an initial date 0 to a terminal date T, possibly a world Let time t r u n infinite. The evolution of this endowment i s stochastic, given by a "Debreu tree" t h a t we describe in notation adapted from Lucas (1984). The s t a t e zL, a n element of t t h e s e t Z , denotes t h e history of the economy from d a t e 0 through t. t of these possible s t a t e s occurs with a probability n ( z 1. Each Country 1, which we call the home country, is endowed with a stream of positive quantities of the home Likewise, good, country denoted 2, the t {y (z 1) 1 foreign or, in country, shorthand is notation, endowed with simply the (yl). positive sequence (y 1 of quantities of t h e foreign good. We denote the prices of t h e 2 t t t domestic and foreign goods in s t a t e z by ql(z and q2(z 1. As in section t t t 2, we define t h e t e r m s of trade, p ( z 1 = q ( z 1/q (z 1, as t h e relative price 2 1 of imports t o exports. Each country i s represented by a single consumer, who stands in f o r a large number of like agents. The preferences of t h e consumer of country i clevelandfed.org/research/workpaper/index.cfm a r e characterized by the expected utility function with O<p<l, U(c) = cl-'/(l-'), parameter. We r e f e r t o and p O . a s the risk aversion Both agents consume composites of the foreign and domestic goods, described by the Armington aggregator functions, c (a b ) 1 1' 1 = G(al,bl), c (b , a ) 2 2 2 G(b2,a2), = with 1-a 1-a l/(l-a) [wa + b 1 , G(a,b) = where a. and b. a r e the quantities of the domestic and foreign good consumed 1 1 by the agent of country i. Thus, the agents of each country consume combinations of foreign and domestic goods, a s expressed in the function G. This theoretical device, due to Armington (19691, computable s t a t i c general equilibrium trade models. is widely used in In this economy i t is equivalent t o giving consumers preferences over foreign and domestic goods directly. The two parameters, substitution and shares of cr=l/a. a>O and w>O, govern the foreign and domestic goods. elasticity of The elasticity is With w=l, the two consumers have identical preferences, and with w>l they exhibit a preference f o r home goods: If home and foreign goods sell f o r the same price, agents consume more of the home good than the foreign good. The budget constraint of the domestic agent is The foreign agent faces an analogous constraint. A competitive equilibrium quantities ( a b. 1 and prices i' 1 in this economy consists of (qi) such that state-contingent ( i ) consumers maximize utility clevelandfed.org/research/workpaper/index.cfm given prices and budget constraints and (ii) quantities s a t i s f y t h e resource constraints, f o r each s t a t e z Negishi-Mantel t . We find i t convenient t o compute a n equilibrium using t h e algorithm in which, f o r any initial distribution of resources, a competitive allocation i s associated with a P a r e t o optimum. i s t h e solution t o a welfare ( A ,A ), 1 2 weights maximize C.A.u 1 1 problem of t h e form: choose quantities For some choice of (a a b b ) 1' 2' 1' 2 subject t o t h e resource constraints. i' Each optimum in each positive state to The supporting prices can then be identified with t h e Lagrange multipliers on t h e constraints o r derived from describes consumers' this first-order procedure in a conditions. similar context Backus and (1992, section 2 ) discusses alternative decentralization schemes. For t h e optimum problem, l e t u s denote t h e Lagrange multipliers on t h e resource constraints in s t a t e z foreign goods, respectively. t t t by ql(z 1 and q2(z ) f o r t h e domestic and The Lagrange multipliers correspond t o prices in t h e associated competitive equilibrium. price functions t t qi(z 1 = / 3 t n ( z t ) ~ i ( z1, by Qi If, f o r each i, we define spot then the optimum problem t s e p a r a t e s into a number of identical problems, one f o r each s t a t e z , of t h e form max {a ,bi} subject t o t h e state are t t t resource constraints. t /3 n ( z 1 {AIU(G[al(z ),bl(z 11) + t t A2U(G[b2(z 1,a2(z 11)) The f i r s t - o r d e r conditions f o r each clevelandfed.org/research/workpaper/index.cfm t t t t t The t e r m s of t r a d e i s p(z 1 = q ( z )/q ( z 1 = Q ( z )/Ql(z ). 2 1 2 In this preferences, section, w=l, for we which restrict the ourselves analysis can to the be done case of identical analytically. The equilibrium allocation i s then f o r countries i=1,2, with consumption s h a r e s s = hil/'/,Y i one. .A!/' J J t h a t sum t o Backus (1992) describes how t h e welfare weights, and hence t h e shares, a r e related t o t h e endowments. The properties of interest, however, do not depend on t h e choice of weights, s o we can skip t h i s additional step. The supporting prices a r e , up t o a f a c t o r of proportionality, and the equilibrium t e r m s of t r a d e i s The t r a d e balance in country 1 is, in units of t h e domestic good, In equilibrium, t h e r a t i o of t h e t r a d e balance t o output ( t h e form of t h e t r a d e variable used in table 4.1) i s Along any equilibrium path f o r this economy t h e welfare weights, hi, are 2 constant. The consumption shares, s a r e functions of t h e welfare weights i' clevelandfed.org/research/workpaper/index.cfm alone, so they a r e constant as well. Thus, fluctuations in both t h e t r a d e balance and t h e t e r m s of t r a d e a r e driven, along any equilibrium path, only by movements prices y2/y1, and in t h e endowment quantities y2/y1. ratio, a r e functions of a In t h i s sense, equilibrium single s t a t e variable, the ratio and a r e t h u s indirectly related t o each other. Although t h e theory i s f a i r l y simple, we can s t a r t t o compare some of i t s properties with those of t h e d a t a -- in particular, t h e variability and persistence of t h e t e r m s of t r a d e , p, and t h e correlation of t h e t e r m s of t r a d e with n e t exports. variability cr=l/a. of Clearly, t h e variability of t h e endowment ratio, y2/y1, p i s governed by t h e and t h e substitution parameter, A given amount of variability of t h e endowment r a t i o can produce as much price variability as we like if cr i s small enough. As a n example, consider t h e standard deviation of t h e US t e r m s of t r a d e reported in table 3 percent, f o r t h e period as a whole. 4.1: This r e f e r s t o t h e s t a n d a r d deviation of t h e Hodrick-Prescott f i l t e r e d logarithm. The standard deviation of t h e f i l t e r e d logarithm of t h e r a t i o of Japanese t o US output i s about 2.2. To generate t h e amount of price variability we see in t h e United S t a t e s , then, we need an elasticity of about cr=0.73 (since 2.2/0.73=3.01. This i s only a rough calculation, since t h e d a t a r e f e r t o a world with more t h a n one country and in which a large p a r t of t h e variability of output appears in investment, which i s obviously absent here. played by variability. the As elasticity a rule, of the But i t i s suggestive of t h e role substitution theory can in generating produce any relative amount of price price variability we like if t h e elasticity of substitution i s a f r e e parameter. With regard to persistence, the terms autocorrelation properties of t h e endowment ratio. the autocorrelation of the Japan/US output ratio of trade inherits the To continue our example, (the filtered logarithm, t h a t is) i s 0.7, s o the autocorrelation of t h e t e r m s of t r a d e , in our theory, clevelandfed.org/research/workpaper/index.cfm is also 0.7. table 4.1, This is slightly less than we see f o r t h e US t e r m s of t r a d e in but the discrepancy is not large, either economically or statistically. The final issue concerns t h e relation between t h e t r a d e balance and t h e t e r m s of trade. The contemporaneous relation is summarized by Proposition I. Let o=l in t h e Armington aggregator, G. Then t h e relation between t h e t r a d e balance, n x /y and t h e t e r m s of t r a d e , p, is governed by 1 1' c=l/a, t h e elasticity of substitution between foreign and domestic goods. c > l , t h e t w o variables a r e positively related; if If 6 1 , they a r e negatively related. More precisely, consider x1=y'/y' with x > x l . 2 1' ratio, p < p l . two states with endowment ratios Since p i s a decreasing function of t h e endowment If c > l , then nx/y < nx1/y'. Now consider the t r a d e balance. In t h i s case, t h e s t a t e with higher p also has higher nx/y variables are, in this x=y /y and 2 1 sense, positively related. If and t h e t w o c<l, the reverse is true. A similar result is implicit in Stockman and Svensson (1987, section 5.3). If c = l , t h e t r a d e balance is constant, as noted recently by Cole and (1991). Obstfeld correlation is Except either +1 o r for -1, nonlinearities unless c=l, and when nonstationarities, the trade the balance is constant and t h e correlation i s not defined. The dynamics of t h e relation between t r a d e and prices, like t h e dynamics of prices, a r e determined completely by t h e dynamics of t h e endowment ratio. Except for nonlinearities and nonstationarities, the cross-correlation function f o r t h e t r a d e balance and t h e t e r m s of t r a d e is t h e same as t h e autocorrelation function of t h e endowment ratio. t h e logarithm of Then if the endowment c > l , t h e cross-correlation By way of example, suppose r a t i o is AR(l), with autocorrelation p>O. function f o r the t r a d e balance and t h e clevelandfed.org/research/workpaper/index.cfm t e r m s of t r a d e i s tent-shaped: correlation between p t and nx The contemporaneous correlation i s +1 and t h e t+k is p l k ' . If c < l , t h e function i s V-shaped. Even with o t h e r autocorrelation patterns, t h e cross-correlation function will be symmetric since t h e autocorrelation function is. therefore, i s incapable of The exchange economy, reproducing t h e asymmetric correlation functions pictured in f i g u r e 4.1. 4 P r e f e r e n c e f o r Home Goods When consumers in t h e t w o countries have preferences t h a t f a v o r t h e i r respective home goods (w>l in t h e Armington aggregator GI, t h e behavior of t r a d e and prices changes somewhat. When preferences f o r domestic and foreign goods a r e n o t additively separable, we find t h a t t h e r i s k aversion p a r a m e t e r , 7 , a s well a s t h e elasticity of goods, c=l/a, substitution between foreign and domestic plays a role in t h e relation between t h e t r a d e balance and t h e t e r m s of t r a d e . The simplest case i s y=a: risk aversion ( 7 ) i s t h e inverse of substitution elasticity f o r foreign and domestic goods (c=l/a). restriction, preferences, between the considerably. foreign including and the domestic aggregator, goods, are additively simplifying the With t h i s separable the analysis We compute an equilibrium, as before, f r o m an optimum. The r e a d e r may verify t h a t t h e equilibrium allocations, f o r countries i=1,2, a r e a a where t h e consumption shares, s 1 = (ohl)1/~[(whl)1/'+h~'7~, s2 = 1-sa sb = 1' 1 b b 1/'/[~11/'+(w~2)1/'l, and s = 1-5 a r e constant along any equilibrium A1 2 1' path. The only difference f r o m t h e symmetric case i s t h a t t h e consumption s h a r e s now d i f f e r across goods, with l a r g e r values of w leading t o l a r g e r a s h a r e s of home good consumption, s and sb 1 2' The supporting prices are clevelandfed.org/research/workpaper/index.cfm so the equilibrium terms of trade is A s before, the relative price p is driven by the endowment ratio. The only change is the factor of proportionality, which does not affect the properties of the logarithm of p. by the variability of The variability of the terms of trade is determined the endowment ratio, y2/y1, and the elasticity of substitution, cr=l/a. The t r a d e balance f o r country 1 in this economy, expressed a s a ratio t o domestic output, is We can see that preference f o r home goods (w>l) has the effect of damping fluctuations in the balance of trade, since larger values of w imply smaller Comovements, however, do not change: values of sb 1' The sign of the effect of the endowment ratio on the balance of t r a d e once more hinges on whether cr=l/a is greater parameters. or Thus, less the than one properties and does described not in the depend on previous any other section for identical preferences (w=l) apply t o this economy a s well. It should be clear, then, that any influence of parameter, w, on price behavior between domestic and foreign goods allocation sign of solution, of goods between a-y. but straightforward. intuition -- operate nonseparabilities influenced, in this conditions do not behind through different values of a and y. countries is The first-order the must the home preference the admit a equilibrium The case, by the simple analytic allocation is fairly A s before, the variables of interest a r e functions of the clevelandfed.org/research/workpaper/index.cfm endowment r a t i o , x=y /y 2 1' In s t a t e s with high values of x, t h e r a t i o s of foreign t o domestic good consumption, b./ai, 1 r a t i o , c /c 2 of 1' a r e also high: t h e endowment r a t i o x. and t h e aggregate consumption They a r e increasing functions, in other words, Neither property is surprising: If t h e r e is relatively more of t h e foreign good, then in equilibrium both agents consume relatively more, and aggregate consumption favors the country preferences weight t h e foreign good more (country 2, since w>l). whose In simple terms, let t h e foreign and domestic goods be bananas and apples, and suppose the foreign apples. agent prefers bananas (w>l) and the domestic agent prefers Then t h e two statements a r e , f i r s t , t h a t in s t a t e s with relatively more bananas, both agents consume relatively more bananas than apples and, second, t h a t aggregate consumption by t h e foreign agent, who h a s a s t r o n g e r preference f o r bananas, r i s e s proportionately more t h a n consumption by t h e domestic agent. Proofs of both of these statements a r e included in t h e appendix. From t h i s s t a r t i n g point, w e can deduce t h e e f f e c t s on t h e relative price of t h e foreign good and t h e balance of trade. From t h e first-order conditions, reported Consider t h e price. in t h e previous section, w e see t h a t t h e t e r m s of t r a d e satisfies the relations where f o r convenience w e have dropped explicit dependence of variables on t h e state, z t . Since b /a and b2/a2 1 1 a r e increasing functions of t h e endowment r a t i o x , t h e t e r m s of t r a d e is a decreasing function. t h e relative supply of t h e foreign good Thus, a n increase in lowers i t s relative price. With somewhat g r e a t e r e f f o r t we can characterize t h e magnitude of t h e decline. The f i r s t s t e p i s t o r e l a t e t h e endowment r a t i o t o bl/al. constraints imply The resource clevelandfed.org/research/workpaper/index.cfm b s1=bl/y2 where is the (possibly state-dependent) share of foreign good Thus, with 0 1 , the proportional consumption by the agent of country 1. change in b /a can be either greater o r less than the proportional change in 1 1 x, depending on whether sb increases o r decreases with x. 1 The second step follows from the first-order conditions f o r the foreign good: Since c /c 2 1 is an increasing function of the endowment ratio, sb 1 is increasing in x if p a , decreasing if ;r<a, and constant if ;r=a (this being the additively separable case described earlier). Thus, the (absolute value of the) slope of the relation between the logarithms of the terms of trade, p, and the endowment ratio, x, is greater than a if y>a, less than a if y<a, and equal t o a ( a s we've already seen) if y=a. risk aversion tends to increase the variability Other things equal, greater of the terms of trade relative t o t h a t of the endowment ratio. The contemporaneous relation between the trade balance and the terms of t r a d e is characterized by Proposition 2. positive number Let w>l (preference f o r home goods). cr* such that if cr>cr*, the trade Then there exists a balance, nx /y 1 1' is positively related t o the terms of trade, p, and if crccr*, the two variables a r e negatively related. Furthermore, if ;r>l, then cr*<l. Thus, the risk aversion parameter, ;r, has an influence on comovements between t h e trade balance and the terms of trade. The proof consists largely of expressing the ratio of the trade balance t o output in a convenient form. As in Proposition 1, both the trade balance and the terms of trade a r e monotonic clevelandfed.org/research/workpaper/index.cfm functions of t h e endowment r a t i o x=y /y 2 1' parameter values, decreasing in x. a With w=l, sl i s constant. The t e r m s of t r a d e is, f o r a l l As f o r t h e t r a d e balance, we let Since bl/al is, as we have shown, increasing in t h e endowment r a t i o x=y /y t h e proposition follows immediately with cr*=l. 2 1' a With w>l, t h e dependence of s on x introduces an additional element of 1 dependence on x. If ;r>l and a<;r, sa i s increasing in x. 1 t r a d e balance i s decreasing in x. parameters, Thus, f o r a s 1 t h e Since t h e solution i s continuous in t h e t h i s i s t r u e f o r values of a slightly g r e a t e r than 1 a s well. For l a r g e values of a , however, t h e t r a d e balance is increasing in x , as we show in cr*=l/a*<l t h e appendix (result [A41). Thus, there a r e numbers a*>l and t h a t divide regions of positive and negative comovement between t h e t r a d e balance and t h e t e r m s of trade. In s h o r t , t h i s economy i s much like t h a t of t h e previous section. Its similarities and differences a r e illustrated by t h e following example. Example. 1 2 1 2 Let there be two s t a t e s , with unconditional probabilities (Nonlinearities a r e irrelevant here, since t w o points c a n always be connected by a s t r a i g h t line.) Let t h e aggregate endowments (y y 1' 2 of t h e domestic and foreign goods be (l+e,l-E) in s t a t e 1, (I-e,l+e) in s t a t e 2, with e>O. Thus, t h e standard deviations of y 1 and y 2 a r e e and t h e standard is, f o r small e , deviation of t h e logarithm of t h e endowment r a t i o y /y 2 1 approximately 2e. The transition probabilities between s t a t e s i and j a r e The persistence parameter p governs t h e autocorrelation function f o r any random variable adapted t o the s t a t e , with t h e k-order autocorrelation equal to p k . clevelandfed.org/research/workpaper/index.cfm Case 1: identical preferences (w=l). The t e r m s of t r a d e i s - ~ state 1 which equals [ ( 1 + ~ ) / ( 1 - ~ ) 1 in The is, f o r small c , approximately With c=0.05 and a=5, t h e standard deviation i s 0.500. The r a t i o of s t a n d a r d deviation of a(2c). and [ ( l - c ) / ( l + c l ~ - ~in s t a t e 2. t h e logarithm of p t h e t r a d e balance t o output is Thus, f o r cr=l/a<l, t h e correlation between p and t b is -1, correlation i s +l. and f o r cr>l t h e For a = l , t h e t r a d e balance is zero in both s t a t e s and t h e correlation i s not defined. Note t h a t t h e persistence parameter, p, h a s no bearing on t h i s correlation. The cross-correlation function f o r t h e t r a d e balance and t h e t e r m s of t r a d e has t h e f o r m where cr=l/a. Case 2: As noted earlier, t h i s function i s symmetric in k. preference f o r home goods (w>l). c=0.05, w=2, and 7=2. To make t h i s concrete, l e t With a=5, t h e standard deviation of t h e logarithm of t h e t e r m s of t r a d e i s 0.497, slightly smaller than in t h e case of identical preferences. There is a critical value cr*=O.885<1 of the elasticity of substitution such t h a t f o r cr>cr*, t h e t r a d e balance and t h e t e r m s of t r a d e a r e positively correlated, and cross-correlation function i s which remains symmetric. f o r crccr* they a r e negatively correlated. The clevelandfed.org/research/workpaper/index.cfm 5 T h e Marshall-Lerner Condition a n d t h e Harberger-Laursen-Metzler E f f e c t The theory of data t h e l a s t two sections a r e relatively straightforward. and i t s relation t o time series We have found, however, that this theory gives u s a much different understanding of t h e relation between t r a d e and relative prices Marshall-Lerner than that condition Harberger-Laursen-Metzler underlying or the effect. either 1980s the widely cited on the literature In this section, we contrast our work with these other approaches. The elasticity condition of Proposition 1 i s an example of a common result in price theory, t h a t t h e sign of something depends on whether utility i s more o r less concave than t h e log (cr=l). surprising about it. difference between In t h a t sense, t h e r e i s nothing What we find interesting in t h e present setting i s t h e this condition and t h e Marshall-Lerner condition, often cited as t h e determining f a c t o r in t h e sign of t h e relation between t h e t r a d e balance and the terms of trade. The Marshall-Lerner condition is the centerpiece of textbook treatments of t h i s issue; among many other examples, see Dornbusch (1980, p. 591, Ethier (1988, pp. 402-71, and Krugman and Obstfeld (1991, p. 423). The Marshall-Lerner condition applies t o a s t a t i c economy, but we can review i t s logic in a one-period treatment follows Kemp (1987). version of our two-country world. Our We start by deriving conditions on s t a t i c import demand functions t h a t allow us t o express t h e balance of t r a d e as a n increasing function of the terms of trade. Among the conditions for equilibrium in t h e s t a t i c economy a r e t h e resource constraint f o r t h e home good, a l + a 2 = yl, and t h e f i r s t agent's budget constraint, clevelandfed.org/research/workpaper/index.cfm These conditions together imply balanced trade: Utility maximization by both agents, subject to their respective budget constraints, r e s u l t s in import demand functions, say a2(p) and bl(p). These functions define net exports as a function of t h e t e r m s of trade: This function i s increasing in p if + & &* > 1, where = c -(abl/ap)[p/bl) and = e* a r e the domestic and foreign (aa2/ap)(p/a2) import elasticities. The inequality is the well-known Marshall-Lerner condition. We now apply t h i s analysis t o our economy. The relevant import demand functions a r e bl(p) where, as before, cr=l/a domestic goods. = yl/(p+pD) i s t h e elasticity of substitution between foreign and The import elasticities a r e therefore E: = cr-1 (l+crp )/(p+pC), clevelandfed.org/research/workpaper/index.cfm Adding these two expressions, we find t h a t f o r all positive p and cr, E: In other words, + t h e Marshall-Lerner economy, regardless of condition i s always satisfied t h e value of foreign and domestic goods. when consumers t h e elasticity of in this substitution between This example illustrates a more general result, The Marshall-Lerner cited by Ethier (1988, section A.3): satisfied > 1. E* in the two countries condition is always have identical no connection homothetic preferences. Clearly, the Marshall-Lerner elasticity condition of condition has Proposition 1 and therefore with the has no bearing on the correlation between t h e t r a d e balance and the terms of t r a d e f o r time series d a t a generated by economies like ours. The difference in results stems, we think, from t h e difference between dynamic modeling and t h e s t a t i c analysis t h a t underlies the Marshall-Lerner condition. Despite t h e intuitive appeal of t h e l a t t e r , we find t h a t when the dynamics a r e worked out explicitly, we get a different interpretation of this property of the data. The theoretical s t a t e of t h e art regarding t h e relation between the t r a d e balance and t h e t e r m s of trade, however, i s not t h e Marshall-Lerner condition but the 1980s revival of the Harberger-Laursen-Metzler initiated by Obstfeld (1982) and Svensson and Razin (1983). effect These papers, and others t h a t followed, start with the central insight of the absorption approach: that trade imbalances reflect differences between investment. The theoretical economies of these two papers a r e deterministic but share with ours t h e f e a t u r e t h a t dynamics a r e explicit. saving and They come, however, t o much different conclusions regarding t h e f a c t o r s t h a t lead t o a positive association between t h e t r a d e balance and t h e t e r m s of trade. These papers suggest t h a t two f a c t o r s a r e critical in determining t h e pattern of clevelandfed.org/research/workpaper/index.cfm comovements: the persistence of the shock and the form of dependence of the discount factor, o r r a t e of time preference, on future utility. Transitory shocks typically lead, in their analysis, t o movements in the terms of trade and the trade balance of opposite sign. We find, in contrast, that the relation between the trade balance and the terms of trade is independent of the dynamics of prices. The aforementioned papers also find that the effect of permanent shocks depends on the behavior of the discount factor. p. 100) put it: "A permanent A s Svensson and Razin (1983, terms-of-trade deterioration ... causes a deterioration or improvement in the real trade balance, depending on whether ... the the r a t e level of of time preference decreases o r welfare." Obstfeld increases, (1982) assumes that respectively, the rate of with time preference is increasing in utility and therefore predicts a decline in the trade balance. production must In his words (1982, p. experience a fall in 2511, "...an aggregate economy specialized in spending and a current [account] surplus a s a result of an unanticipated, permanent worsening in i t s terms of trade." In both papers there is no effect of a permanent change in price on the t r a d e balance if the r a t e of time preference is constant. In our economies, the r a t e of time preference is constant, fixed by the discount factor p. The conclusion should then be t h a t permanent movements in the terms of trade have no effect on the trade balance. We find, instead, t h a t the relation between the trade balance and the terms of trade is determined by the elasticity of substitution, regardless of the persistence of shocks. As in our analysis of the between our approach and that of stems, in part, Marshall-Lerner the difference the Harberger-Laursen-Metzler literature from our definition of the condition, issue. In our analysis, the relation between the trade balance and the terms of trade pertains t o the correlation between these two variables f o r a single time series realization, clevelandfed.org/research/workpaper/index.cfm like t h e q u a r t e r l y series f o r States described in table Japan, 4.1. the United This Kingdom, corresponds, in and t h e our United theoretical economies, t o t h e correlation between t h e t w o variables along a n equilibrium path. In t h e analysis of however, the Obstfeld (1982) and Svensson and Razin Harberger-Laursen-Metzler effect pertains between t w o different deterministic equilibria: to a (19831, comparison one in which t h e t e r m s of t r a d e i s high, and one in which i t i s low. Apparently these t w o thought experiments emphasize much different f e a t u r e s of t h e theory. We would a r g u e t h a t our thought experiment i s closer t o what we have in mind when we compare theory and data. A closer look also suggests, as brought out in Backus (1992) and Stockman and Svensson (19871, t h a t t h e theory requires explicit treatment of the stochastic structure of the economy, something that deterministic analysis obviously cannot provide. One way in which these t w o points of view might be reconciled is t o consider economies in which agents have more limited ability t o hedge risk than they do in t h e complete market economies of sections 3 and 4. e f f e c t , income e f f e c t s play a 1980s analysis of t h e Harberger-Laursen-Metzler central role. In t h e In our economies, however, t h e r e a r e no income e f f e c t s along a n equilibrium path. budget constraint. With complete markets, each agent has a single, date-0 As a result, each has a state-invariant marginal utility of income, reflected in t h e constant welfare weights of our optimum problem. Backus (19921, Kehoe and Richardson (19851, and Mendoza (1992) suggest t h a t some of t h e flavor of t h e Harberger-Laursen-Metzler literature may c a r r y over t o dynamic stochastic settings with some types of market incompleteness. In short, we find t h a t explicit dynamic stochastic analysis of t r a d e and relative prices their leads t o much comovements. Even different views the role of of that Marshall-Lerner condition, may need t o be reconsidered. t h e f a c t o r s determining textbook standard, the clevelandfed.org/research/workpaper/index.cfm Government Spending 6 The theory t h u s f a r h a s focused on fluctuations in t r a d e and prices arising f r o m movements in endowments. Here we consider a n extension t o t h e exchange economy of section 3 in which we have, in addition, exogenous shocks t o government (1992), Buiter spending. (19891, Related Hodrick Reynolds (1991), and Yi (1991). analyses have been (19881, Macklem provided (19901, by Baxter Obstfeld (1989), We f i n d t h i s extension both interesting in i t s own right and a useful s t e p toward introducing a wide range of impulses into dynamic general equilibrium models of trade: shocks, f o r example, t o t a x e s , t a r i f f s , and possibly even monetary policies. To keep t h e analysis simple, we r e s t r i c t ourselves t o t h e case of symmetric preferences (w=l in t h e Armington aggregator GI. In t h i s new economy, t h e government i s a n additional consumer of goods. Let us s a y t h a t in s t a t e zt t h e government of country i consumes t h e quantity t g.(z ) of i t s home good. This spending i s financed with lump-sum t a x e s , say 1 t ri(z ). An equilibrium then consists of quantities (a and b i ) , prices (qi), i a n d government policies ( g . , ~ . ) such t h a t 1 prices and budget 1 constraints, ( i ) agents maximize utility given (ii) quantities satisfy the resource constraints, a n d (iii) policies satisfy governments' budget constraints. With this structure, the endowments" yi-gi, r a t h e r than y section 3 with little change. economy i' is equivalent to one with "net and we can apply most of t h e r e s u l t s of The equilibrium allocation includes clevelandfed.org/research/workpaper/index.cfm for i=1,2, with consumption w e l f a r e weights A.. 1 shares s i x!/'1 /z.x!/'J = J f o r some choice of The equilibrium t e r m s of t r a d e i s The variability of t h e t e r m s of t r a d e is governed, then, by t h e variability of t h e net endowment r a t i o and t h e elasticity of practice, t h e addition of substitution, cr=l/a. In government purchases has l i t t l e influence on t h e variance of p, since g is only a f r a c t i o n of output and i s generally less variable. The same reasoning applies t o persistence: purchases of Introducing government goods and services does l i t t l e t o change our prediction t h a t relative prices retain t h e persistence of output ratios. The most interesting consequences of government purchases concern trade. If w.=y.-g. 1 1 1 i s t h e endowment net of government purchases, net exports in country 1 a r e t nxl(z 1 = t (1-sl)wl(z 1 - t t -a t [w2(z )/wl(z 11 slw2(z The r a t i o of net exports t o output i s t t t t 1-a t t nxl(z l/yl(z 1 = ((1-sll - sl[w2(z )/wl(z 11 } wl(z )/yl(z 1. In our earlier analysis, g 1 w a s z e r o and t h e last t e r m was, therefore, one. As a result, t h e e f f e c t of t h e endowment r a t i o on t h e t r a d e balance, and t h e and t h e t e r m s of association between movements in t h e t r a d e balance, nx /y 1 1' t r a d e , p, depended only on t h e value of cr=l/a. positive; f o r cr<l t h e reverse. of t h e net endowment, w 1' For cr<l t h e association was Here we find a n additional f a c t o r , t h e r a t i o t o t o t a l output, y 1' clevelandfed.org/research/workpaper/index.cfm We can g e t some idea of t h e contributions of output and government spending shocks on t r a d e and price fluctuations by considering special cases. Consider, f i r s t , t h e case in which g 1 i s proportional t o y 1' Then wl/yl is constant, and t h e relation between t h e t r a d e balance and t h e t e r m s of t r a d e i s determined by cr, as i t w a s in Proposition 1. With a>l ( o r u<l), t h e t r a d e balance and t h e t e r m s of t r a d e a r e positively related along a n equilibrium path; with cr<l, they a r e negatively related. y. a r e constant and g 1 1 i s t h e only shock. Alternatively, suppose outputs Then t h e t r a d e balance and the t e r m s of t r a d e a r e positively associated, regardless of t h e value of cr. This example i s like many others in economics in which t h e comovement between t w o endogenous variables depends on t h e source of t h e i r fluctuations. This analysis suggests a second look at t r a d e and price special attention paid t o government purchases. t h e r e h a s been l i t t l e regularity of government purchases a c r o s s countries relative to that correlation between these two variables. of data, with As we see in table 4.2, in either t h e real output, variability or in the The same statement applies t o t h e correlations of government purchases with t h e t r a d e balance and t h e t e r m s of trade. That i s not t o say t h a t government purchases have not played a role in t r a d e and price behavior, but t h a t t h i s role i s not simple enough t o show u p in summary s t a t i s t i c s of t h i s form. Froot and Rogoff (1991) document somewhat stronger indications of a relation between government spending and r e a l exchange r a t e s using different methods. 7 T r a d e a n d C a p i t a l Formation In t h e exchange economies of sections 3 and 4, we compared properties of t h e d a t a with analogous properties of theoretical economies. closer look. t r a d e and relative prices in simple This analysis brought up two questions t h a t deserve a We found, f o r one thing, t h a t t h e variability of t h e t e r m s of clevelandfed.org/research/workpaper/index.cfm t r a d e i s governed by the variability of the output ratio and the elasticity of substitution between foreign and domestic goods. By choosing a small enough elasticity, the theory can generate literally any amount of price variability. in the theory elasticity. not, for relative The question, in this case, i s whether price variability and the data are close for reasonable values of this In another respect, we found t h a t the exchange economy could any choice of parameter values, mimic the data: cross- the correlation function f o r the t r a d e balance and the terms of trade. In the d a t a this function i s generally asymmetric, a feature we document in figure 4.1 and label the S-curve. In the exchange economy, however, the function i s symmetric by construction, since both the t r a d e balance and the terms of t r a d e a r e functions of the same s t a t e variable. The question here i s whether this property changes when we introduce physical capital formation. The introduction of capital formation brings us closer t o the theme of the absorption approach t o the balance of payments: t h a t fluctuations in t r a d e reflect differences between saving and investment. A s a matter of pure accounting, this connection i s undeniable, but i t also shifts one's away from within-period save and invest. relative prices t o the intertemporal attention decisions t o Thus, Sachs (1981) argues that trade deficits often reflect investment booms and Stockman and Svensson (19871 tie both t r a d e and relative prices t o fluctuations in, among other things, fixed capital formation. We continue this tradition by introducing capital formation t o an economy t h a t i s otherwise like our earlier ones. Kehoe, and Kydland (1992b). The structure i s adapted from Backus, The emphasis, a s in earlier sections, i s on the dynamics of the t r a d e balance and the terms of trade. The theoretical economy has the following elements. before, two countries that specialize in different goods. There are, as Preferences of the representative consumer in each country i a r e characterized by a n expected clevelandfed.org/research/workpaper/index.cfm utility function of t h e f o r m where c it and n it a r e consumption and hours worked in country i, U(c,l-n) = [cp(l-n)'-'1'-~/(1-7), and 720. The primary difference in preferences f r o m section 3 i s t h e appearance of t h e economy of leisure in agents' utility functions. Goods in t h e t w o countries, labeled country 2, are produced using a f o r country 1 and k, and capital, labor, n, b with 8 1-8 homogeneous production functions of t h e same form, F(k,n) = k n for linear . This gives r i s e t o t h e date-t resource constraints, in countries 1 and 2, respectively. The quantity y i, measured in units of t h e local good, and a t w o goods in country i. exchange setting of it it denotes GDP in country and bit denote uses of t h e If k and n a r e constant, t h i s reduces t o t h e pure section 3, with proportionate output fluctuations. productivity shocks giving The vector zt = (zlt,z2t) rise to i s a stochastic shock t o productivity whose properties will be described shortly. Consumption, investment, and government spending in each country a r e composites of t h e foreign and domestic goods, with 1-a 1-all/(l-a) where, as before, G(a,b) = [wa +b both positive, and the domestic goods i s cr=l/a. elasticity of . The parameters a and w a r e substitution between foreign and As noted earlier, t h i s s t r u c t u r e i s widely used in clevelandfed.org/research/workpaper/index.cfm s t a t i c general equilibrium models of trade. Capital stocks evolve according to where 6 i s t h e depreciation rate. the exchange economy of Two differences between t h i s economy and sections 3-5 are the introduction of capital formation and t h e assumption here t h a t government spending may have some foreign content. Finally, t h e underlying shocks t o our economy a r e independent bivariate autoregressions. where E Z The technology shocks follow i s distributed normally and independently over time with variance The correlation between t h e technology shocks, z vz' by t h e off-diagonal elements of A and VZ. 1 and z i s determined 2' Similarly, shocks t o government spending a r e governed by where g t = ( g , g 1 and It 2t cg i s distributed normally with variance V g' Technology shocks, z, and government spending shocks, g, a r e independent. With these elements, and t h e parameter values listed in table 4.3, we can approach t h e behavior of t h e t e r m s of trade. The critical parameters, f o r our purposes, a r e t h e elasticity of substitution, cr, which we s e t equal t o 1.5, and t h e steady-state r a t i o of imports t o GDP, which w e s e t equal t o 0.15 by choosing w appropriately. In t h i s benchmark version of t h e economy, foreign and domestic goods a r e b e t t e r substitutes t h a n they would be with Cobb-Douglas preferences (cr=l) and imports a r e , on average, 85 percent of GDP. The choice of elasticity i s in t h e range of estimates f r o m a large number of studies, a s documented by Whalley (1985, ch. 4). Estimates of t h e clevelandfed.org/research/workpaper/index.cfm elasticity a r e generally close t o one, and often slightly larger. The import s h a r e i s slightly l a r g e r than we s e e in t h e United S t a t e s , Japan, o r a n aggregate of European countries (with intra-European t r a d e netted out), but smaller than we s e e f o r most countries individually. A number of properties of t h e theoretical parameter settings a r e reported in table 4.4. in t h e t e r m s of trade. economy with alternative Consider, f i r s t , fluctuations The standard deviation of t h e t e r m s of t r a d e with our benchmark parameter values i s 0.48 percent, which i s a f a c t o r of s i x less than we see f o r t h e United S t a t e s in table 4.1. With smaller values of cr, t h e theoretical economy generates g r e a t e r variability of t h e t e r m s of trade. A s illustrated in f i g u r e 4.2, t h e standard deviation of p g e t s larger as we decrease c , and f o r cr less than 0.03 t h e standard deviation exceeds 2. Thus, i t appears t h a t while t h e theory can produce as much variability in t h e t e r m s of t r a d e as we s e e in t h e d a t a , i t requires a n elasticity of substitution much smaller t h a n most existing estimates. The value of cr required t o match t h e variability of t h e t e r m s of t r a d e in US d a t a i s substantially smaller in t h i s model (less than 0.03) than in our calculation in section 3 f o r t h e United S t a t e s and Japan ( f o r which we estimated t h a t w=0.73 would be sufficient). of t h i s difference. Three f a c t o r s account f o r most The f i r s t i s t h a t t h e theoretical economy has, in t h e benchmark case, about 25 percent less variability in t h e output r a t i o t h a n we calculated f o r Japan and t h e United States. Modifications of t h e theory t h a t bring t h e magnitude of business cycles closer t o t h e d a t a will also bring t h e theory and data closer together second f a c t o r i s capital formation. by setting 8=0 in the with respect price variability. The If we eliminate capital (which w e can do production function), considerably g r e a t e r price variability, despite of outputs. to the economy less variability The final f a c t o r i s t h e import share. generates in t h e r a t i o If t h e import s h a r e i s clevelandfed.org/research/workpaper/index.cfm raised from 0.15 t o value of cr. 0.25, the variability increases substantially a t every For cr=1.5, the benchmark value, the standard deviation of the relative price rises from 0.48 t o 0.58. second A property of the model is the between net exports and the terms of trade. contemporaneous correlation In the data, this correlation has been positive f o r the United States and negative f o r Japan and the United Kingdom (see benchmark expect table 4.1). In the theoretical economy we find, Propositions elasticity of increases with 1 and substitution. cr, and is We 2, this see in positive for correlation figure 4.3 elasticities is the A s we might parameter values, that the correlation is -0.41. from for sensitive that the greater the correlation than This feature, too, is strongly influenced by capital formation. to cr*=2.76. In the model without capital (8=0), the economy is much like that described in Proposition 2, with a critical elasticity cr*=0.94. For cr>cr*, the trade balance and the terms of trade a r e positively correlated; f o r crccr*, the reverse. A third property of interest is the impact of government spending on the correlation between the trade balance and the terms of trade. We see in table correlation 4.4 that with only shocks t o government spending, the between net exports and the terms of trade shifts from negative t o positive. This mirrors productivity, a z, similar and result government in section spending, 6. g, With we find shocks that to the both former dominate, in the sense that the economy's properties a r e similar t o those with shocks t o productivity alone. Finally, we look at the exports and the terms of trade. complete cross-correlation function for net A s pictured in figure 4.4, this correlation has the same asymmetric shape we documented f o r the data in figure 4.1. Some intuition f o r this behavior is provided by figure 4.5, in which we graph the dynamic responses t o a one-standard-deviation shock t o domestic productivity. clevelandfed.org/research/workpaper/index.cfm Following t h i s shock, increase. output and t h e relative price of foreign goods both Consumption also rises, but by less than output. Investment grows initially by much more than consumption, as resources a r e t r a n s f e r r e d t o t h e home country to exploit its expected f u t u r e productivity capital accumulates, t h i s resource t r a n s f e r diminishes. advantage. As The t r a d e balance, which i s t h e difference, at market prices, between output and t h e sum of consumption and investment, by surplus. exhibits an initial period of deficit, followed These dynamic responses give r i s e t o t h e asymmetric cross- correlation function of figure 4.4. In short, much of t h e intuition f o r t h i s dynamic general equilibrium t r a d e model i s available f r o m t h e exchange economy of sections 3-5. What t h e exchange economy misses completely a r e the dynamics of t h e relation between t h e t r a d e balance and t h e t e r m s of trade: t h e asymmetric cross-correlation function t h a t we documented in t h e d a t a and cross-correlation labeled t h e S-curve. function between these two variables exchange economy f o r all parameter values. t h a t t h e dynamics of The is symmetric in the In t h i s section, we have seen capital formation provide a plausible basis f o r a n asymmetric pattern. 8 F i n a l Thoughts We have argued t h a t a dynamic general equilibrium approach t o aggregate t r a d e theory provides both a new level of understanding of t h e interrelations between t r a d e and price movements and a framework in which these relations can be quantified. With regard t o t h e former, we have seen t h a t t h e relation between t r a d e and price variables is much different f r o m t h a t suggested by t h e Marshall-Lerner condition cited in most textbooks. With r e g a r d t o t h e l a t t e r , we suggest t h a t t h e dynamic relation between t r a d e and t h e relative price of foreign goods can be understood as a consequence of t h e dynamics of clevelandfed.org/research/workpaper/index.cfm capital formation. Thus, the dynamics of trade variables a r e inseparable from the dynamics of the r e s t of the economy. Future work will undoubtedly focus, however, not on these contributions, but on dimensions in which the theory, in i t s current incarnation, provides a relatively poor approximation t o the dynamics of actual economies. obvious example i s the variability of the terms of trade. The most In the economy of section 7, and in Stockman and Tesar (1991) a s well, the standard deviation of the terms of t r a d e is substantially smaller than we estimate in the data. This discrepancy between theory and data helps t o motivate theories in which monetary policy influences relative prices (Grilli and Roubini 1992, Schlagenhauf and Wrase 19921 and in which international market segmentation, possibly in conjunction with imperfect competition, also plays a part (Dumas 1992, Giovannini 1988, Lapham 1990). Ongoing research will likely tell us how important each of these features is, and how they modify the lessons of the theory outlined above. clevelandfed.org/research/workpaper/index.cfm REFERENCES Alterman, W. 1991. "Price trends in US trade: New data, new insights." In P. Hooper and J.D. Richardson (eds. 1, International Economic Transactions: Issues in Measurement and Empirical Research. Chicago: University of Chicago Press. Armington, P. 1969. "A theory of demand f o r products distinguished by place of production." IMF S t a f f Papers 27, 488-526. Backus, D. 1992. "Interpreting comovements in t h e t r a d e balance and t h e t e r m s of trade." Journal o f International Economics, forthcoming. Backus, D., Kehoe, P., and Kydland, F. 1992a. "International r e a l business cycles." Journal o f Political Economy 100 (August), 745-775. Backus, D., Kehoe, P., and Kydland, F. 1992b. "Dynamics of t h e t r a d e balance and t h e t e r m s of trade: The S-curve." Federal Reserve Bank of Cleveland, Working Paper 9211, October. Baxter, M. 1992. "Fiscal policy, specialization, and t r a d e in t h e two-sector model: The r e t u r n of Ricardo?" Journal o f Political Economy 100 (August), 713-744. Blackburn, K., and Ravn, M. 1991. "Contemporary macroeconomic fluctuations: An international perspective. " Unpublished manuscript, January. Buiter, W. 1989. Budgetary Policy, International and Intertemporal Trade in the Global Economy. Amsterdam: North Holland. Cole, H., and Obstfeld, M. 1991. "Commodity t r a d e and international risk-sharing." Journal o f Monetary Economics 28 (August), 3-24. Dornbusch, R. 1980. Open Economy Macroeconomics. New York: Basic Books. Dumas, B. 1992. "Dynamic equilibrium and the r e a l exchange r a t e in a spatially separated world. " Review o f Financia l Studies 5, 153-180. Ethier, W. J. 1988. Modern International Economics, Second Edition. New York: Norton. clevelandfed.org/research/workpaper/index.cfm Froot, K., and Rogoff, K. 1991. "Government spending and the real exchange r a t e in the Bretton Woods era." Unpublished manuscript. Giovannini, A. 1988. "Exchange r a t e s and traded goods prices." Journal of International Economics 24, 45-68. Graboyes, R. 1991. "International trade and payments data." Federal Reserve Bank of Richmond, Quarterly Review 77 (September/October), 20-31. Grilli, V., and Roubini, N. 1992. "Liquidity and exchange rates. " Journal of International Economics 3 2 (May), 339-352. Harberger, A. 1950. "Currency depreciation, income, and the balance of trade." Journal of Political Economy 58, 47-60. Hodrick, R. 1988. "US international capital flows: perspectives from rational maximizing models. " Carnegie-Rochester Conference Series on Public Policy 3 0 (Spring), 231-288. Kehoe, P., and Richardson, P. 1985. "Dynamics of the current account: Theoretical and empirical analysis." Federal Reserve Bank of Minneapolis, Working Paper. Kemp, M. 1987. "Marshall-Lerner condition." The New Palgrave: A Dictionary of Economics. London: Macmillan. Krugman, P., and Obstfeld, M. 1991. International Economics: Theory and Policy, Second Edition. New York: HarperCollins. Lapham, B. 1990. "A dynamic, general equilibrium analysis of deviations from the laws of one price." Unpublished manuscript, Queen's University, September. Laursen, S., and Metzler, L. 1950. "Flexible exchange r a t e s and the theory of employment. " Review o f Economics and Statistics 32, 281-299. Lucas, R. 1984. "Money in a theory of finance. " Carnegie-Rochester Conference Series on Public Policy 21, 9-45. Macklem, R.T. 1990. "Terms-of-trade disturbances and fiscal policy in a small clevelandfed.org/research/workpaper/index.cfm open economy." Bank of Canada, Working Paper 90-7, December. Mendoza, E. 1992. "The terms of t r a d e and economic fluctuations." Unpublished manuscript, International Monetary Fund, February. Mussa, M. 1986. "Nominal exchange r a t e regimes and the behavior of real exchange rates." In K. Brunner and A. Meltzer (eds.), Real Business Cycles, Real Exchange Rates, and Actual Policies. Carnegie-Rochester Conference Series. Obstfeld, M. 1982. "Aggregate spending and the terms of trade: Is there a Laursen-Metzler effect?" Quarterly Journal o f Economics 97 (May), 251-270. Obstfeld, M. 1989. "Fiscal deficits and relative prices in a growing economy." Journal o f Monetary Economics 23 (May), 461-484. Reynolds, P. 1991. "Capital formation, government spending and international economic fluctuations. " Unpublished manuscript, Northwestern University, January. Sachs, J. 1981. "The current account and macroeconomic adjustment in the 1970s. " Brookings Papers on Economic Activity (I), 201-268. Schlagenhauf, D., and Wrase, J. 1992. "A monetary, open-economy model with capital mobility." Unpublished manuscript, Arizona S t a t e University. Stockman, A., and Svensson, L. E. 0. 1987. "Capital flows, investment, and exchange rates," Journal o f Monetary Economics 19, 171-201. Stockman, A., and Tesar, L. 1991. "Tastes and technology in a two-country model of the business cycle: explaining international comovements." Federal Reserve Bank of Cleveland, Working Paper 9019, April. Svensson, L., and Razin, A. 1983. "The terms of trade and the current account: The Harberger-Laursen-Metzler effect." Journal o f Political Economy 91 (February), 97-125. clevelandfed.org/research/workpaper/index.cfm Whalley, J. 1985. Trade Liberalization Among Major Trading Areas. Cambridge, MA: MIT Press. Yi, K.-M. 1991. "Can government purchases explain recent US net e x p o r t deficits?" Unpublished manuscript, presented at t h e NBER Universities Conference, May. clevelandfed.org/research/workpaper/index.cfm APPENDIX: Proof o f Proposition 2 The algebra behind Proposition 2 i s s t r a i g h t f o r w a r d but f a i r l y tedious. We start by reducing t h e economy t o t w o equations in t w o unknowns. the consumption shares of the first agent, sa=a /y 1 1 1 unknowns are b s =b /y 1 1 2' The f i r s t - o r d e r conditions and resource constraints then imply The and and b a where m=b /a -xs /s >O and v=2/a>0. With these substitutions, t h e t w o 1 1 - 1 1 a equations determine s and sb as functions of t h e endowment r a t i o x=y /y 1 1 2 1' Note t h a t if either w = l o r a=;r, t h e r i g h t side of (A21 i s one and sa i s 1 constant. Preliminaries: a Thus, s and s: 1 (a1 If we differentiate t h e f i r s t equation we g e t a r e positively related and we can use t h i s relation b s u b s t i t u t e out any d s 's we get. 1 b a ( b ) Differentiate t h e r a t i o (s /s 1: 1 1 b a d ( s /s 1 1 1 = b 1 (S /S a b a v a a 1 [(s1/S 1 10 -11 dsl/sl. 1 (c) Differentiate m: dm/m (dl Inequalities. b a v a a = dx/x + I ( s l / s l l w -11 dsl/sl. From (A11 and w>l: b a v ( s l / s l ) w -1 = a b (sl-sll/(l-s:) > 0. to clevelandfed.org/research/workpaper/index.cfm We now prove t h e claims in t h e t e x t . a We show t h a t sl is increasing in x if 7>a, and decreasing if 7<a. 1. We differentiate (A2) and find, a f t e r rearrangement: If we substitute t h e expression f o r dm/m [ ( c ) above1 we g e t a n equation of the form so that a sl i s increasing in x if B/A>O, and decreasing otherwise. The coefficient of dx/x i s Since w>l, B has t h e same sign as 7-a. a a The coefficient of d s /s i s 1 1 a Clearly if 7<a, A i s positive and d s /dx i s negative. 1 A/7 = 1-a l+v 1-a 2+v) 1-a (w+m ) (l+w m ) + 8 (1-w m , a b where 8=(s -s )(7-a)/;r<l. 1 1 in t h i s case, too. 2. If p a , then Combining terms makes i t clear t h a t A i s positive Thus, the sign of dsa/dx i s t h e same as t h e sign of 7-a. 1 An immediate corollary i s t h a t c /c i s increasing in x: 2 1 The f i r s t - order conditions imply The behavior of c 2 /C 1' sy with respect t o x implies t h a t t h e consumption ratio, i s increasing in x , as claimed in t h e text. 3. We now show t h a t b /a =m i s increasing in x. 1 1 From (A31 we have clevelandfed.org/research/workpaper/index.cfm From the definition of B, Since A>O, m is increasing in x. 4. We turn t o the dependence of the trade balance on x. A s in the text, Differentiating, we find which is positive if From (A4) and (AS) we can show that this inequality holds f o r large enough a. clevelandfed.org/research/workpaper/index.cfm Table 4.1 Properties of the Trade Balance and the Terms of Trade Standard Dev. (%) Country Japan United Kingdom United States Autocorrelation CrossCorrelation Period P nx P 1955-89 1955-70 1971-89 5.97 2.17 7.76 .97 .98 .94 .87 .73 .87 nx .77 .66 .83 1955-89 1955-70 1971-89 2.71 1.51 3.38 1.08 .78 1.21 .76 .38 .79 .66 .54 .65 - .54 1955-89 1955-70 1971-89 2.99 1.31 3.84 .45 .30 .55 .82 .65 .84 .80 .79 .80 .30 .28 .30 (P,nx> (P,Y) - .46 - .09 - .55 - .51 -. 15 -.60 .41 - .27 .20 .56 .10 - .09 .47 - .23 Variables are p, the terms of trade, logarithm; nx, the ratio of net exports to output; y, real output, logarithm. Data are quarterly from the OECD's Quarterly National Accounts. Statistics refer to Hodrick-Prescott filtered variables. SOURCE: Authors' calculations. clevelandfed.org/research/workpaper/index.cfm Table 4.2 The Trade Balance, the Terms of Trade, and Government Purchases Standard Deviation Country Y g Australia 1.47 Austria Cross-Correlations @,nx> (g,nx> 1.90 -.I1 -.I5 1.27 .45 - .25 Canada 1.49 1.16 -.06 France .91 .66 - .50 Germany 1.47 1.22 -.09 Italy 1.70 .69 Japan 1.48 Switzerland .I1 -.I5 (g,~) (g,~) .15 .28 -.02 .17 -.23 -.22 -.45 .24 -.I1 -.I6 .23 -.66 .ll -.42 1.54 -.51 .19 -.35 .02 1.94 1.01 -.61 -.I5 -.29 .28 United Kingdom 1.60 1.07 -.60 -.06 -.01 .06 United States 1.93 1.47 .31 .ll -.28 .13 -.01 .12 The sample period is 1971:l to 1989:4. Variables are p, the terms of trade, logarithm; nx, the ratio of net exports to output; y, real output, logarithm; g, real government purchases of goods and services, logarithm. Data are quarterly from the OECDYsQuarterly National Accounts. Statistics refer to Hodrick-Prescott filtered variables. SOURCE: Authors' calculations. clevelandfed.org/research/workpaper/index.cfm Table 4.3 Benchmark Parameter Values Preferences Technology 8 = 0.36, 6 = 0.025, a = l l a = 1.5, import share = 0.15 Forcing Processes var ~f = var E: = 0.008522, C O ~ T ( E=~0.258 ,E~ gt = 0 SOURCE: Authors' calculations. clevelandfed.org/research/workpaper/index.cfm Table 4.4 Properties of Theoretical Economies with Capital Formation Standard Deviation (%) Autocorrelation Correlation nx Y P nx Y P .30 (.02) 1.38 (. 18) .48 .63 (. 10) .83 (.05) -.64 (.M) .61 (.07) Large Elasticity .33 (.03) 1.41 (.18) .35 (.05) .63 (.07) .64 (. 18) Small Elasticity .37 (.03) 1.33 (. 18) .76 (.07) .61 (.07) Big Share .63 (.04) 1.37 (. 18) .58 (.08) .59 (.07) Small Share .08 (.01) 1.38 (.18) .43 (.06) Two Shocks .33 (.03) 1.33 (. 15) Govt. Shocks .16 (.03) .17 (.02) Economy Benchmark ( W Y ) (WP) @,P) (.07) -.41 (.08) .49 (.14) .88 (.03) -.57 (.08) -.05 (.09) .43 (.14) .63 (. 10) .77 (.05) -.66 (.07) -.80 (.09) .51 (.16) .64 (. 10) .83 (.04) -.61 (.07) -.41 (.07) (. 13) .62 (.07 .63 (. 10) .81 (.05) -.65 (.07) -.41 (.08) .48 (.14) .57 (.07) .62 (.08) .65 (.08) .78 (.06) -.57 (.15) -.05 (.17) .39 (.17) .30 (.05) .67 (.ll) .67 (.08) .67 (. 11) -.55 (.13) 1.00 (.00) -.55 (.13) .52 Variables are defined in Table 4.1. Statistics refer to Hodrick-Prescott filtered variables. Entries are averages over 20 simulations of 100 quarters each; numbers in parentheses are standard deviations. Parameters as in Table 4.3, except large elasticity, a = 2.5; small elasticity, a = 0.5; big share, import share = 0.25; small share, import share = 0.05; two shocks, mean of g = diag(0.2,0.2), B = diag(0.95,0.95), and V, = diag(0.0042,0.0042);government shocks, as in two shocks plus z, = 1, all t. SOURCE: Authors' calc~.llations. clevelandfed.org/research/workpaper/index.cfm Fig. 4.1 S-CURVES IN THE DATA JAPAN 0.75 ALL U -0.50 -- \ / I ' \ \ -0.75 I I I I I I I 1 d-/ / \ > I I I I I I I PRE-72 POST-72 / I I I I I 1 1 1 -- , , , I I I I I 1 UNITED KINGDOM 0.75 7 UNITED STATES 0.75 7 -- \ / / / \ --0.75 ! \ / ALL / PRE-72 POST-72 - - I _ I I -8 I I I I I I I I -6 -4 I I I I I I I I I I I I -2 0 2 Lag k of p behind nx SOURCE: Authors' calculations. I I I I 4 I I I I 6 ---I I 8 clevelandfed.org/research/workpaper/index.cfm Fig. 4.2 VARIABILmY OF THE TERMS OF TRADE Elasticity of Substitution, o SOURCE: Authors' calculations. clevelandfed.org/research/workpaper/index.cfm Fig. 4.3 CORRELATION OF THE TRADE BALANCE AND TERMS OF TRADE 0.8 - 0.6 -0.4 -- 3 2 4 5 Elasticity of Substitution, o SOURCE: Authors' calculations. 6 clevelandfed.org/research/workpaper/index.cfm Fig. 4.4 S-CURVE FOR THE BENCHMARK ECONOMY Lag k of p over nx SOURCE: Authors' calculations. clevelandfed.org/research/workpaper/index.cfm Fig. 4.5 DYNAMIC RESPONSES TO DOMESTIC PRODUCTIVITY SHOCK ... .. 1.00 -- 120 8 *, s C) 8y 0.80 -- OUTPUT .. -.. 3 X 3m 0.60-- ---.- - - - - - - - - _- - - - - - _- - - - - - - - - _ _ PRODUCI'M~ b o u SHOCK 0.40 -- 8 - - 2 - 020-- / I I - - TERMS OF TRADE 4 4 - 4 - 4 - 4 - 4 4 4 - 2 . - 0 2 0 l ~ I 1 1 I I I 1 I I I I I I I I 1 I I I I I I I 0 3 6 9 12 15 18 21 Time in Quarters SOURCE: Authors' calculatio~is.