"Trade Shocks and Macroeconomic Fluctuations in Africa" M. Ayhan Kose and Raymond Riezman

Transcription

1 "Trade Shocks and Macroeconomic Flucuaions in Africa" M. Ayhan Kose and Raymond Riezman CSGR Working Paper No. 43/99 Ocober 1999 Cenre for he Sudy of Globalisaion and Regionalisaion (CSGR), Universiy of Warwick, Covenry CV4 7AL, Unied-Kingdom. URL: hp://

2 Trade Shocks and Macroeconomic Flucuaions in Africa M. Ayhan Kose a and Raymond Riezman 1 Graduae School of Inernaional Economics and Finance, Brandeis Universiy; Deparmen of Economics, Universiy of Iowa. CSGR Working Paper No. 43/99 Ocober 1999 Absrac: This paper examines he role of exernal shocks in explaining macroeconomic flucuaions in African counries. We consruc a quaniaive, sochasic, dynamic, muli-secor equilibrium model of a small open economy calibraed o represen a ypical African economy. In our framework, exernal shocks consis of rade shocks, modeled as flucuaions in he prices of expored primary commodiies, impored capial goods and inermediae inpus, and a financial shock, modeled as flucuaions in he world real ineres rae. Our resuls indicae ha while rade shocks accoun for roughly 45 percen of economic flucuaions in aggregae oupu, financial shocks play only a minor role. We also find ha adverse rade shocks induce prolonged recessions. Keywords: Trade shocks, dynamic sochasic quaniaive rade model, African economies. JEL Classificaion: F41, E31, E32, D58, F11. Address for correspondence: Ray Riezman, Deparmen of Economics, Universiy of Iowa, Iowa Ciy, IA 52242, USA raymond-riezman@uiowa.edu a Graduae School of Inernaional Economics and Finance, Brandeis Universiy, Walham, MA 2254, akose@lemberg.brandeis.edu. 1 We are graeful o Simon Evene, Jan Gunning, David Richardson, and Linda Tesar for heir helpful commens. We also benefied from he suggesions of Ahmed El-Sofy, Kamil Yilmaz, and seminar paricipans a he 1998 Fall Midwes Inernaional Economics Conference a he Universiy of Michigan. The usual disclaimer applies.

3 1. Inroducion There is a large and expanding lieraure suggesing ha highly unsable domesic macroeconomic environmen is one of he primary reasons for he poor growh performance of African counries in he las hiry years. 1 The implicaion is ha o improve growh performance in Africa, we need o undersand why heir economies are so volaile. Tha is he objecive of his paper: using a dynamic, sochasic model we esablish a link beween exernal shocks and he highly volaile macroeconomic flucuaions in hese economies. We invesigae he effecs of rade and financial shocks. Surprisingly, we find ha, despie he fac ha hese counries are ypically heavily indebed, rade shocks play a much more imporan role han financial shocks. In paricular, i urns ou ha rade shocks explain almos half of he volailiy in aggregae oupu. Inernaional rade can induce macroeconomic flucuaions in a small open economy by wo channels: one channel is hrough rade in goods and services, and he oher one is by rade in financial asses. In African economies, hese wo channels have disincively imporan roles in shaping domesic economic aciviy: firs, he volume of inernaional rade on average accouns for more han half of he aggregae oupu in hese counries. Moreover, a narrow range of primary commodiies consiues a significan fracion of heir expors, and heir main impor iems are inermediae inpus and capial goods. Their expor revenues are highly unsable due o recurren and sharp flucuaions in he prices of primary commodiies. Second, mos of he African counries are heavily indebed, and a significan fracion of heir expor revenues are used o mee heir deb service obligaions. These make African counries exremely vulnerable o changes in he world ineres rae. A horough undersanding of he sources of macroeconomic flucuaions in African economies requires a good grasp of he impac of exernal shocks, namely flucuaions in he prices of expored primary commodiies, impored capial goods, and inermediae inpus, and financial shocks, namely flucuaions in he world real ineres rae, on domesic economic aciviy. In his paper, our objecive is o shed some ligh on hese issues by addressing he following quesions: firs, do rade disurbances accoun for a significan fracion of macroeconomic flucuaions? Second, how are rade shocks ransmied and propagaed hrough hese economies? We begin by documening some of he salien feaures of he economic srucure, macroeconomic flucuaions, and dynamics of rade shocks o provide empirical evidence ha here is a srong link beween inernaional rade disurbances and domesic economic aciviy in African counries. We, hen, consruc a muli-secor, dynamic, sochasic small open economy model which reflecs he srucural characerisics of a represenaive African economy. There are wo secors in he model: primary goods and nonraded final goods secors. The economy expors primary goods, and impors all of is inermediae inpus and a significan fracion of is capial goods. The households in his economy can buy and sell one-period risk free bonds in he world financial markes. We also sudy he role of domesic facors in generaing macroeconomic flucuaions as hese facors are capured by he changes in produciviy of he wo secors. 1 Hadjimichael, Ghura, Muhleisen, Nord, and Ucer (1994), Sachs and Warner (1996), Ghura and Hadjimichael (1996), and Rodrik (1998a) use a variey of growh regressions o explain he deerminans of economic performance and conclude ha macroeconomic sabiliy is an imporan facor for he long-run growh, Easerly and Levine (1997) emphasize he adverse impac of ehnic fragmenaion on growh in African counries. In a recen paper, Collier and Gunning (1998) provide a deailed survey of his lieraure. Pindyck (1991), and Aizenman and Marion (1993) provide heoreical models where volaile macroeconomic environmen 1

4 We compare he properies of he macroeconomic flucuaions generaed by his model wih hose acually observed in African counries. The resuls of his comparison sugges ha he model successfully replicaes some of he salien feaures of macroeconomic flucuaions in African economies. We quaniaively evaluae he conribuion of inernaional rade shocks o domesic macroeconomic flucuaions. Our findings indicae ha rade shocks play an imporan role in driving economic aciviy in African counries: almos 45 percen of flucuaions in aggregae oupu, is explained by world price shocks. Furher, rade shocks accoun for almos 87 percen of aggregae invesmen variaion. These resuls are consisen wih hose of Deaon and Miller (1996) who analyze he imporance of inernaional commodiy prices in driving economic flucuaions in African counries using vecor auoregression analysis. Their resuls sugges ha while a sudden 1 percen increase in commodiy prices resuls in a 6 percen increase in oupu, he price shocks mos heavily affec invesmen dynamics in African economies. Anoher ineresing oucome of our analysis is ha world ineres rae flucuaions do no have a significan impac on economic dynamics of African counries. Impulse response analysis demonsraes ha he propagaion of economic flucuaions generaed by he rade shocks is differen han ha caused by domesic produciviy shocks. Firs, produciviy shocks induce movemens in he marginal produc of facors of producion. In conras, rade flucuaions exhibi a raher indirec impac: shocks o he relaive prices of impored capial goods, for example, resul in subsiuion effecs which rigger labor supply movemens beween he wo secors. Second, while posiive produciviy disurbances resul in shor lived expansions, adverse rade shocks cause prolonged recessions. There has been a growing research program examining he link beween domesic economic aciviy and rade shocks. 2 While conribuing o his lieraure, our paper is also relaed o some recen sudies examining he sources of macroeconomic flucuaions in African counries: Deaon and Miller (1996) employ a VAR, and Hoffmaiser, Roldos, and Wickham (1998) esimae a srucural VAR model, where idenifying resricions are derived from a long-run small open economy model. While he former sudy concludes ha rade shocks play an imporan role in driving macroeconomic flucuaions in hese economies, he laer one finds rade disurbances accoun for only a small fracion of he variaion in oupu. This sudy exends he scope of his research program in several dimensions: firs, we sudy he sources of macroeconomic flucuaions in African counries in a fully specified sochasic dynamic open economy model reflecing he srucural characerisics of hese economies. 3 Since he model economy is dynamic, and involves endogenous labor-leisure choice, we are able o examine he link beween rade shocks and flucuaions in aggregae invesmen, foreign asse holdings, and labor markes. The model economy employs domesically produced capial goods, impored capial goods, and impored inermediae inpus in wo differen secors. This srucure of differeniaion in producive facors allows us o sudy he has an adverse impac on growh. Ramey and Ramey (1995), using he daa of developing and developed economies, find ha counries wih highly volaile macroeconomic environmen have relaively lower growh. 2 Balassa (1978), Moran (1983), Feder (1983), and Basu and McLeod (1992) examine he relaion beween expor insabiliy and economic growh. See also Bevan, Collier, and Gunning (1993), Easerly, Kremer, Priche, and Summers (1993), Ghosh and Osry (1994), Gavin and Hausmann (1996) and Rodrik (1998b). Bevan, Collier, and Gunning (1994) provide a compuable general equilibrium model of an open economy o examine he economic experiences of Kenya and Tanzania afer he major rade shock in See Baxer (1995) and Backus, Kehoe and Kydland (1995) for surveys on dynamic general equilibrium models of open economies. Senhadji (1995), and Mendoza (1995) analyze he imporance of erms of rade shocks, and Kose (1997a) examines he role of world price shocks in driving business cycles in developing economies using a sochasic dynamic business cycles models of small open developing economies. 2

5 impac of differen ypes of rade shocks on differen secors of he economy. Second, our sudy considers a broader definiion of rade shocks as i focuses on he price changes of he main expor and impor iems insead of erms of rade disurbances. This is moivaed by our empirical examinaion which reveals he erms of rade is no able o fully reflec he movemens in highly volaile relaive prices of main expor and impor iems of African counries. Moreover, while assessing he role of relaive price shocks, we invesigae he impac of world ineres rae flucuaions on domesic economic aciviy as well. Third, we examine he daa of a large group of African counries, and documen several sylized feaures of macroeconomic flucuaions in hese economies. The organizaion of he paper is as follows: in secion 2, we review he empirical regulariies of African counry daa. Following his, we presen he model. Model calibraion is described in secion 4. In secion 5, we firs assess he abiliy of he model in replicaing business cycle dynamics of a represenaive developing counry. Then, we quaniaively evaluae he imporance of differen ypes of shocks in generaing macroeconomic flucuaions. The model dynamics are analyzed using impulse responses. Following his, he sensiiviy of he resuls o changes in he srucural parameers of he model is briefly invesigaed. We conclude wih a brief summary and suggesions for fuure research in secion 6. An appendix, conaining informaion abou he soluion of he model, daa sources, and deailed daa ables, is also provided. 2. Analysis of he daa In his secion, we firs documen some of he salien feaures relaed o economic srucure of several African counries. Following his, we analyze he main regulariies of macroeconomic flucuaions observed in hese economies. Nex, we examine he cyclical behavior of rade shocks and heir comovemen wih aggregae oupu and he rade balance. Our analysis is based on he annual daa of weny-wo non-oil exporing African counries for he period Srucural characerisics of he African economies We begin wih an examinaion of he decomposiion of aggregae oupu o provide a beer undersanding of he srucural characerisics of he African economies. We presen informaion abou he expendiure shares of aggregae oupu and indusrial srucure in Table 1a. The G7 average of each magniude is also provided for comparison purposes. The major difference beween hese wo groups is he role played by inernaional rade in domesic economic aciviy. In African (G7) counries, expors accoun for almos 31 (2) percen of oal GDP while impors consiue more han 4 (18) percen of i. Srikingly, he volume of rade on average accouns for more han 71 percen of GDP in African counries while only 38 percen of oal GDP is aribuable o he rade volume in he G7 counries. As Table 1a indicaes, African counries have relaively large rade deficis: he average rade defici is around 1 percen of he GDP in he African economies in our sample. The difference beween he average consumpion shares of he wo groups is also noiceable: for African counries, he average share is around 7 percen ha is 1 percen more han he G7 average. 4 We examine he daa of 22 non-oil exporing African counries. 18 of hese counries are Sub-Saharan African, 4 of hem are Arabic: Burundi, Cape Verde, Egypy, Gambia, Ghana, Guinea Buissea, Kenya, Liberia, Madagascar, Malawi, Mauriania, Mauriius, Morocco, Seychelles, Sierra Lione, Sudan, Swaziland, Tanzania, Tunisia, Zaire, Zambia, and Zimbabwe. Deailed informaion abou he daa sources and definiions is given in Appendix A2. 3

6 Having noed all hese differences, i is ineresing o observe ha he shares of governmen expendiure aggregae invesmen in GDP are quie similar in boh groups. Characerisics of African economies indusrial srucures also make hem highly vulnerable o rade shocks: hey have relaively smaller indusry and service secors, and, consequenly, he share of agriculural secor is considerably larger in hese counries. To be more specific, agriculural goods on average accoun for 28 (4) percen of oal GDP while indusrial iems consiue roughly 18 (3) percen of oal domesic income in African (G7) counries. We hen sudy he decomposiion of expors and impors o deermine he relaive imporance of differen commodiy groups in he inernaional rade. Table 1b describes he srucure of expors. As his able clearly illusraes, he African economies heavily depend on primary goods for heir expor revenues. The share of primary expors on average is 77 percen. Food producs and meals consiue 47 percen and 2 percen of oal expors respecively. Ineresingly, he average share of capial good expors is less han 2 percen in oal expors. 5 If hese counries expor a variey of primary goods, hen heir dependency o primary commodiies for heir expor revenues migh no be seen as an imporan problem, since flucuaions in he prices of differen ypes of primary commodiies can offse each oher and resul in relaively sable expor earnings. In order o examine he exen of diversificaion of expors a a more disaggregaed level, we use wo differen crieria ha are presened in he las wo columns of Table 1b. Wih boh measures, he African counries in our sample seem o be much more concenraed in heir expors han he G7 counries do. Firs, we examine he number of commodiies expored by he African economies: They expor on average 54 differen goods. This number is around 213 for he G7. Second, we use he Gini-Herschman coefficien o measure he concenraion of expors. 6 A higher value of his coefficien indicaes a higher degree of expor concenraion. While he average coefficien of expor concenraion is more han 6 for he African counries, i is less han 1 for he G7. Table 1c provides informaion abou he decomposiion of impors. Two poins abou his able are noeworhy: firs, he main impor iems of hese counries are capial goods and inermediae inpus. While he impors of inermediae inpus accoun for almos half of he oal impors, he average share of capial good impors is approximaely 28 percen. Second, he average share of agriculural goods is minor in oal impors. Table 1b and 1c ogeher presen an ineresing picure: while he expor srucures of he G7 economies and African economies are quie differen, heir impor srucures are similar on some dimensions. For example, boh he G7 and African counries rely on impored manufacured goods. Movemens in he cos of servicing exernal deb also seem o be an imporan source of macroeconomic flucuaions in several African counries, paricularly highly indebed ones. 7 As Table 2 indicaes, he average raio of exernal deb o GNP is around 89 percen and deb service o expor raio, which is a widely used measure of deb burden, is around 2 percen for he African economies in our sample. Noably, he raio of shor-erm deb paymens o he expor revenues is on average 43 percen. 5 A number of African economies have recenly increased he share of he manufacured expors in heir oal expors. However, he share of he manufacured expors is sill a minor fracion of he oal expors. 6 See Appendix A2 for more informaion abou his coefficien. 7 Greene (1989) examines he deb problem of African counries and emphasizes he adverse impac of increases in he world ineres rae on his problem. Fosu (1996) finds ha he deb burden of Sub-Saharan African counries has a srong adverse impac on he growh performance of hese counries using regression esimaes. 4

7 We conclude his secion wih some broad observaions. While here migh be some differences in he economic performance of African counries, hese counries are quie similar in heir indusrial, and inernaional rade srucure. A ypical African economy ges he bulk of is expor revenues from a narrow group of primary commodiies, impors mainly capial goods and inermediae inpus, faces persisen rade deficis, and has o allocae a significan fracion of is expor revenues o mee is shor erm deb obligaions. A model designed o examine he sources of macroeconomic flucuaions in African economies should reflec hese characerisics Main regulariies of macroeconomic flucuaions We documen some of he regulariies associaed wih he macroeconomic flucuaions in African economies in Table 3. 8 We also presen deailed ables wih counry specific daa in appendix B. This invesigaion serves wo purposes: firs, we ge a beer undersanding of he magniude of economic flucuaions in hese economies. Second, we se up a group of benchmark saisics ha are going o be used in evaluaing he performance of our model economy. All properies of he daa refer o momens of Hodrick- Presco (HP(1)) filered variables. 9 We focus on he hree main feaures of macroeconomic flucuaions: volailiy, measured by sandard deviaion, comovemen, measured by correlaions, and persisence, measured by auocorrelaions. Table 3 presens he properies of economic flucuaions for main macroeconomic aggregaes: among he componens of aggregae oupu, manufacuring producion is he mos volaile one. This evidence compounded wih he observaions in he previous secion implies ha oupus of he producion secors, such as manufacuring, which heavily depend on impored capial goods and inermediae inpus, show more aggregae volailiy han he ones, such as agriculure secor, which mainly use domesically produced inpus. The variaion in consumpion is greaer han ha in aggregae oupu, since our consumpion series include durable consumpion goods. Differen counries use differen mehods o measure flucuaions in he labor marke, so he variaion of employmen across counries does no display any regular paern: i ranges from 2.7 percen o percen wih an average of 7.33 percen (see appendix B.) As expeced, invesmen exhibis high cyclical volailiy. Analysis of he disaggregaed exernal rade daa indicaes ha boh expors and impors are highly volaile. On average, expors are less volaile han impors. The rade balance is he mos volaile series. I is surprising o observe ha he African counries in our sample exhibi very similar characerisics in heir macroeconomic flucuaions: for example, in 16 ou of 19 counries for which daa is available, volailiy of manufacuring oupu is greaer han he one of service secor. We briefly examine he persisence and comovemen properies of aggregae variables. Manufacuring oupu is more persisen han he rade balance, impors, and aggregae oupu. Agriculural secor oupu is he leas persisen series wih an average auocorrelaion coefficien of.15. Excep he rade balance, all macro aggregaes are procyclical. 8 In a recen paper, Agenor, McDermo, and Prasad (1998) documen he main sylized feaures of macroeconomic flucuaions of 12 developing counries noing ha his is he firs sep o build sochasic dynamic models o see if hese models can reproduce hese sylized facs. Our sudy is he firs one documening hese ypes of sylized feaures for African economies, and consrucing a model o examine he macroeconomic flucuaions in hese counries. 5

8 2.3. Dynamics of Prices There has been a revival of ineres o undersand he dynamics of commodiy prices, since a number of developing counries have faced a very sharp downrend in he prices of heir main expor iems over he las wo decades. 1 Since our ulimae objecive is o evaluae he effecs of rade shocks proxied by relaive price flucuaions on macroeconomic dynamics of African economies, we briefly examine he cyclical feaures of price series and provide furher empirical evidence abou he relaion beween hese series and macroeconomic flucuaions in hese counries. We sar wih a visual inspecion by presening he plos of annual change for he prices of main expor and impor iems of African counries in Figures 1a-1d (Consrucion of relaive price series of capial goods and inermediae inpus are explained in he nex paragraph.) These prices exhibi very high cyclical volailiy. The year-o-year variaions have become increasingly magnified since he early 197s. Very shor lived price booms have been generally followed by prolonged roughs (see Deaon and Laroque (1992).) Figure 1e describes he behavior of erms of rade of African counries. As hese figures clearly demonsrae, while providing some informaion abou recurren flucuaions, he erms of rade is unable o reflec he sharp peaks and deep roughs observed in he prices of major expor and impor iems (see Kouparisas (1997b).) Nex, we look a he sylized facs peraining o flucuaions in paricular relaive price series. Insead of analyzing he erms of rade dynamics only, we also examine a disaggregaed measure of he erms of rade and look a he dynamics of relaive prices of capial goods and inermediae inpus o primary goods. As we repored in Secion 2.1., hese hree groups of commodiies consiue a significan fracion of he rade volume in African counries. Table 4 documens our findings. The relaive price of capial goods o primary goods, p k, is calculaed as he raio of he U.S. producer price index of capial equipmen o he expor price index of he domesic economy. The relaive price of inermediae goods, p v, is equal o he raio of he U.S. producer price index of inermediae maerials o he expor price index of he domesic economy. 11 The erms of rade is calculaed as he raio of expor price index o impor price index of each counry. Ineresingly, he relaive prices are more volaile and more persisen han he erms of rade. The relaive prices of capial goods (inermediae inpus) o primary commodiies are 1.23 (1.11) imes more volaile han he erms of rade. The persisence of he erms of rade is.22 while he persisence of relaive price of capial (inermediae) goods is.38 (.35.) Combined wih our earlier observaions hese findings imply ha models, in which he erms of rade flucuaions are he only exernal driving shock, migh poenially underesimae he role of relaive prices in inducing macroeconomic flucuaions, since he erms of rade series alone are unable o reflec he dynamics of disaggregaed prices and he ineracions beween hese prices and aggregae oupu To find he rend pah of he ime series, his mehod solves a consrained opimizaion problem ha involves minimizing he sum of squared deviaions from he ime series subjec o he consrain ha he sum of squared second differences no be oo large. See Hodrick and Presco (1997) for more informaion abou his filer. 1 Reinhar and Wickham (1994) find ha here has been a seady and considerable increase in he volailiy of commodiy prices since he early 197s. See also Ghosh and Osry (1994), who examine he opimal policy response of a ypical developing counry o flucuaions in is expor earnings. 11 We also calculaed some oher relaive price measures using expor and impor price indices of some oher indusrialized counries. Our empirical observaions were no changed, when we used differen price series. 12 Kouparisas (1997b) provides exensive evidence ha he relaive prices are more volaile han he erms of rade. Alhough he is using a differen daa se, his resuls are very close o ours: he finds ha he relaive prices of non-fuel commodiies o manufacured 6

9 There is a vas empirical lieraure on he world ineres rae flucuaions and implicaions of hese flucuaions on macroeconomic aciviy. 13 Figure 1f plos he annual change of he world real ineres rae over weny-five years. Our world real ineres rae measure is he six-monh LIBOR (he London Inerbank Offer Rae) deflaed by changes in expor uni value index of non-fuel commodiy exporing developing counries. As his figure indicaes, here has been a subsanial increase in he world real ineres rae since he beginning of he 198s. Over he years of , he average real cos of borrowing for developing economies decreased by 16.8 percen. Following his, he cos increased by 5.9 percen beween 198 and The sudden rise in he world ineres rae in 1981 dramaically increased he burden of deb service for several highly indebed developing economies and marked he beginning of he deb crisis. 3. The Model Economy In he previous secion, we provided subsanial empirical evidence suggesing ha he economies of African counries exhibi a number of common srucural feaures. In his secion, we consruc a muli-secor dynamic sochasic small open economy model ha reflecs he main srucural feaures of a ypical African economy Preferences The economy is inhabied by a large number of infiniely lived, idenical households who do no have any conrol over he prices of is expors and impors, and he world real ineres rae. The represenaive household maximizes expeced lifeime uiliy given by U ( c, l) = E { = β [ u( c, l ) 1] } 1 σ (1) σ >, β > where he parameer β denoes he subjecive discoun facor of he household and σ is he risk aversion parameer. c is he consumpion of he non-raded final good, l represens leisure. Neiher expored nor impored goods are modeled as uiliy deriving goods in he model because of he following reasons: firs, he empirical evidence provided in he previous secion indicaes ha a significan fracion of expors is coming from he primary goods secor in African counries. These expored primary goods are generally used as inpus in producing final goods, so he conribuion of hese goods o uiliy is via final goods. Second, recen empirical sudies indicae ha consumer goods have a minor share in he oal impors of developing counries. 14 The insananeous uiliy funcion u has he form goods is 1.37 imes more volaile han he erms of rade. Kouparisas also shows ha erms of rade can be wrien as a linear funcion of relaive prices. 13 See Sachs (1989, pages 7-8) for an exensive discussion abou he role of global shocks and, in paricular, he imporance of world ineres rae flucuaions as an inducing facor of he deb crisis. Ozler and Rodrik (1992), hrough cross-counry regressions, find ha he world real ineres rae flucuaions have a significanly negaive impac on privae invesmen in developing counries. 14 Henschel (1992, page 8) analyzes he composiion of impors for several developing counries, and concludes a general characerizaion of he impors as consumpion goods is no jusified. 7

10 ν 1 σ u( c, l ) = ( c ψ ( 1 l ) ) ν > 1, ψ > This uiliy funcion implies he marginal rae of subsiuion beween consumpion and leisure is a funcion of he leisure only, so labor effor does no depend on he ineremporal consumpion-saving choice Technology The economy produces nonraded final goods and primary goods. Nonraded final goods producion, y f f f, uses labor, n, capial, k, and inermediae inpus, v : (2) f f f f u u y = z ( n ) [ s( k ) + ( 1 s) v ] α ( 1 α )/ u < α, s, u < 1 (3) z f represens he exogenous produciviy shock. α is he share of non-raded oupu earned by labor and s is he relaive weigh of capial. The elasiciy of subsiuion beween inermediae inpus and capial is governed by u. The empirical evidence presened in he previous secion suggess ha he cos of inermediae inpus consiues a subsanial fracion of he oal facor cos. The CES formulaion allows us o analyze he impac of degree of subsiuabiliy beween domesic capial goods and foreign inermediae inpus on he dynamics of our model. The primary goods secor produces oupu by using labor, n p, capial, k p, and land, L p which is assumed o be inelasically supplied. The producion funcion in he primary goods secor is given as p p p p p y = z ( n ) ( k ) ( L ) θ1 θ2 1 θ1 θ2 < θ, θ < 1 where z p is he echnology shock. θ 1 and θ 2 are he labor and capial income shares respecively. Capial accumulaion is modeled as 1 2 (4) Here δ is he rae of depreciaion, j j j i j k + 1 = ( 1 δ ) k + φ j ( j ) k j = f, p (5) k j i is he amoun of invesmen in secor j, and φ j (.) represens he concave adjusmen cos funcion, wih φ j (.) >, φ j (.) >, and φ j (.) >. The adjusmen cos prevens frequen movemens of capial across secors and excessive volailiy of invesmen. We assume ha primary secor capial is impored and capial used in he non-raded goods secor is domesically produced. This way of modeling capial is empirically appealing since capial goods used in primary goods secor are generally machinery and equipmen iems, and a significan fracion of capial goods used in he non-raded final goods secor can be classified as inermediae inpus, such as consrucion maerials and energy Financial Markes Each household has free access o world financial markes. However, hese markes are incomplee in he sense ha he household can rade a single financial asse, A, wih a rae of reurn, r, from period o 8

11 +1. Kose (1997b) provides empirical evidence abou he limied access of households in developing counries o inernaional financial markes. The holdings of financial asses evolve according o he formula A + 1 = nx + A ( 1 + r ) (6) where nx represens he balance of rade. The household can parially smooh is consumpion sream by borrowing and lending in he inernaional financial markes. The possibiliy of he agen playing a Ponzi game and accumulaing an infinie amoun of deb is ruled ou by imposing he condiion: 1 lim E ( ( ) ) A = 1+ r In addiion o his, we assume ha β = 1 / ( 1+ r ), r is he seady sae level of ineres rae. I is known ha when he discoun rae is smaller (greaer) han he ineres rae, he represenaive household accumulaes (decumulaes) asses in a deerminisic version of he model. In oher words, here is no seady sae equilibrium in hose cases. If he wo are equal, he economy is a a seady sae equilibrium which is compaible wih any level of foreign asse holdings Resource consrains The resource consrain for he nonraded goods secor is given by and for he primary goods secor is f f c + i = y (7) where i f and i p k p v p p i + p v + nx = y (8) are invesmen in capial goods in he nonraded and primary goods secors respecively. p k and p v denoe he relaive prices of capial goods and inermediae inpus o primary goods respecively. The price of he primary good is numeraire. The household, who has a fixed ime endowmen normalized o one, faces he following laborleisure allocaion consrain f p l + n + n = 1 (9) While labor inpu is assumed o be mobile across he secors, i is no inernaionally mobile. 15 Since models of small open economies wih ime separable preferences do no accep a sable seady sae, some researchers use models wih saionary cardinal uiliy (Mendoza (1995)), or models wih finiely lived agens (Finn (199)), or models wih an exogenously imposed supply funcion for asse holdings (Senhadji (1995)) o ge well-defined seady sae dynamics and o have a realisic mach beween he daa and he model oucome. Our formulaion is similar o Correia, Neves and Rebelo (1995), Rebelo and Vegh (1995), and Ahearne (1997). We assume ha he discoun facor is equal o he ineres rae a he seady sae of he model, since our analysis is limied o dynamics a business cycle frequencies around he seady sae of he model. Since his formulaion resuls in an equilibrium which depends on he iniial condiions imposed by he researcher, deerminaion of hose condiions migh affec he resuls. This migh be a serious problem if he researcher invesigaes he impac of a paricular policy and is implicaions associaed wih he low frequency dynamics of he model variables. However, in models such as ours, dynamics a business cycle frequencies around he seady sae of he model are analyzed. 9

12 3.5. Exogenous shocks There are five shocks in he model: wo shocks o he relaive prices of impored capial and inermediae goods, a world ineres rae shock, and wo secoral produciviy shocks. The vecor of k v f p exogenous shocks is represened by Z = [ p, p, r, z, z ]. The evoluion of Z follows a firs order Markov process and is given by ln Z = Π ln Z + ε (1) k v r f p The vecor of innovaions is denoed by ε = [ ε, ε, ε, ε, ε,] where ε ~ N(, Σ ) Numerical Soluion Mehod The model economy is solved using he opimizaion problem of he represenaive household. This corresponds o a sochasic dynamic opimizaion problem which is solved by maximizing he expeced lifeime uiliy, (1), subjec o he consrains (3)-(1). Since his problem canno be solved analyically, we find an approximae soluion by using log-linear funcions of he model variables. We describe he opimizaion problem and our soluion mehod in Appendix A1. The soluion mehod we use, namely he log-linear approximaion mehod, was shown o be quie accurae in solving sochasic dynamic general equilibrium models Model Calibraion Model calibraion amouns o selecing a combinaion of parameer values ha are roughly consisen wih he long-run feaures of he economic environmen of a represenaive African economy. The parameers and business cycle saisics of his represenaive economy correspond o he averages of relevan variables of he African counries in our sample and presened in Table Preferences The risk aversion parameer, σ, is assumed o be 2.61 which is he GMM esimae from he panel sudy of a group of developing economies by Osry and Reinhar (1992). The ineremporal elasiciy of subsiuion in labor supply, 1/(ν-1), is equal o.83. Prior empirical sudies show ha he value of his parameer is beween.3 and 3.2 (see Greenwood, Hercowiz and Huffman (1988)). The value of ψ is seleced so ha he fracion of hours worked in he seady sae is consisen wih our assumpion abou he allocaion of labor hours beween he marke and non-marke aciviies. As he world real ineres rae measure, we use he LIBOR (he London Inerbank Offer Rae) deflaed by changes in expor uni value index of non-fuel commodiy exporing developing counries. The average world real ineres rae, r, is found o be 2.9 percen annually. Since he ineres rae is equal he discoun rae a he deerminisic seady sae, he discoun facor, β, is equal o See Ingram (1995) for a brief explanaion of his mehod and several oher issues relaed o he soluion and esimaion of sochasic dynamic general equilibrium models. The mehod of log-linear approximaion was popularized by King, Plosser and Rebelo (1988). 17 We are unable o provide esimaes for he parameers of he model by employing formal economeric echniques, such as generalized mehod of momens (GMM), since our sample does no have enough observaions o uilize such echniques. Insead, we base our calibraion on he esimaes provided by earlier economeric sudies. 1

13 4.2. Technology The labor share for he nonraded final goods secor, α, is assumed o be.45. We choose a value of.58 for he Allen elasiciy of subsiuion beween capial and inermediae goods. This value is consisen wih he esimaes provided by Bernd and Wood (1975). 18 The relaive weigh of capial, s, is se a.55. A he seady sae, he capial goods and inermediae inpus shares are equal o.23 and.32 respecively. We assume ha he share of labor, θ 1, is.37. By using secoral daa and he firs order condiion for primary capial, we find ha he share of land, 1-θ 1 -θ 2, is equal o.45. We selec he rae of depreciaion a.1 ha is a widely used value in he business cycle lieraure. I is enough o specify he hree parameers, which describe he cos of adjusmen funcions in he wo secors, o analyze he near seady sae dynamics of he model. Following Baxer and Crucini (1993), we assume ha φ( i / k ) = φ( i / k ) = δ and φ ( i / k ) = φ ( i / k ) = 1 a he seady sae. This f f p p f f p p specificaion implies ha here are no adjusmen coss a he seady sae of he model, and capial o is replacemen cos, Tobin s Q, is equal o 1. The elasiciy of he marginal adjusmen cos funcion, η = ( φ / φ ) / ( i / k ), for each ype of capial is fixed, so ha he volailiy of invesmen generaed by he model is equal o he one in he daa. Trade balance o aggregae oupu raio is equal o he average one in he daa (-.96.) 4.3. Exogenous shocks Produciviy Shocks We esimae he oal facor produciviy in he nonraded goods secor, z f, using he formula of he Solow residual in logarihms y f f f f log( z ) = log( y ) α log( n ) f is he oal real value added of indusry and service oupu. n is equal o he employmen index since daa on labor hours is unavailable for mos of he counries in our sample. The capial sock and inermediae inpu usage are excluded from he formula due o he following reasons: firs, i is known ha flucuaions in he capial sock are no large in he shor-run. Second, he conemporaneous correlaion beween he capial sock and oupu is negligible. Third, he daa on inermediae inpu usage is no available. We fi an univariae AR(1) process o find he parameers of he produciviy shock for each counry and hen ake an average over he whole sample of hese parameers. These averages are assumed o be he relevan parameers for he represenaive developing economy. By following he same seps and using he daa of agriculural value added, and employmen in manufacuring secor, we esimae he shock process for he primary secor oupu Exogenous Shocks We deermine he parameers of he processes of exogenous shocks by using an univariae AR(1) processes. 19 We do no have world price indices ha are specifically designed for capial goods and 18 The value of u, which governs he elasiciy of subsiuion beween capial and inermediae goods, is calculaed by using he formula u 1 ( α + σ ασ ) k, v k, v = where k, v α + σ k, v ασ k, v beween he inermediae inpus and capial goods is equal o 1/(1 + u). σ is he Allen elasiciy of subsiuion (see Sao (1967)). The elasiciy of subsiuion 11

14 inermediae inpus. However, his ype of daa is available a he counry level. We conjecure ha world prices of hose goods closely follow he prices of he same goods produced in he U.S. So, he U.S. producer price indices of capial equipmen and inermediae goods are used o represen he prices of impored capial and inermediae goods respecively. The price series of primary commodiies correspond o he expor uni values of each counry. This assumpion is easily jusified because a significan porion of African counry expors come from primary commodiies. The relaive price of capial goods (inermediae inpus) o primary commodiies is he raio of he U.S. producer price index of capial equipmen (inermediae inpus) o he expor uni value index for each economy. In order o esimae he world real ineres rae, we use he sixmonh LIBOR (he London Inerbank Offer Rae) deflaed by changes in expor uni value index of non-fuel commodiy exporing developing counries. 2 We find he variance-covariance marix of innovaions by using he covariances beween he residual erms of esimaed processes for each counry. Then, we ake he average of hese marices over he sample. The resuling specificaion for exogenous processes is given by.44 Π = Σ =.11 (.4) (.1) (.4) (.2) (.2) (.1) The bold values in Σ represen he correlaions beween he innovaions. 5. Resuls of he simulaions This secion sars wih an evaluaion of our model s abiliy in erms of capuring main regulariies associaed wih macroeconomic flucuaions in a ypical African economy. Then, we examine he imporance of differen ypes of shocks in generaing macroeconomic flucuaions employing variance decomposiions. Nex, he propagaion mechanisms of exogenous shocks in he model economy are analyzed using impulse responses. Following his, we provide a brief discussion abou he sensiiviy of our resuls o changes in he deep parameers of he model. 19 See Deaon and Miller (1996) for a similar AR(1) modeling of price series. 2 The LIBOR is used as a benchmark ineres rae measure by inernaional organizaions and commercial banks when hey give loans o developing counries. See World Economic Oulook (1993, p. 83) for he use of his measure as a proxy for real cos of borrowing for developing economies. 12

15 5.1. How successful is he model? While i is no our primary objecive o examine he abiliy of our model in erms of maching he main characerisics of macroeconomic flucuaions in African economies, we sill hink ha his is an useful exercise since our model economy is he firs one in is class designed o sudy economic dynamics of African counries. The heoreical model is simulaed wih he specificaion described in he previous secion. The momens of acual daa refer o hose momens calculaed for he represenaive African counry and are presened in Table 6. Our focus is on he predicions of he model relaed o volailiies of main macroeconomic aggregaes and heir conemporaneous correlaions wih aggregae oupu. Table 7 summarizes our findings abou he sample second order momens of he model variables. Each saisic is he sample average of across 1 simulaions of he same lengh as he daa (23 years). The simulaed daa is also derended wih he HP(1) filer o make he resuls comparable o he daa analysis in secion 2. In erms of maching volailiy properies of macro aggregaes, he model is quie successful. Qualiaively, i replicaes mos of he feaures of acual daa: boh rade balance and invesmen are more volaile han aggregae oupu. The model also capures he volailiy ordering of oupus of producion secors: he primary secor oupu has he larges variabiliy, and aggregae oupu is he leas volaile series. From a quaniaive perspecive, he model is able o reproduce some of he sylized facs. For example, i is able o mimic volailiies of secoral oupus and aggregae oupu wih a small margin. The prediced sandard deviaion of he rade balance is slighly higher han he acual one. We se he relevan elasiciies of adjusmen cos so he model can exacly replicae he volailiy of invesmen. The volailiies of consumpion and employmen relaive o oupu are seemingly low in he model economy. This is no a surprising resul and should no be inerpreed as a weakness of he model: firs, he only available daa on consumpion in African counries, which we have access o, includes boh non-durable and durable consumpion expendiures. Unlike he daa, our model does no ake ino accoun durabiliy. Hence, a direc comparison of he model generaed consumpion daa wih he acual one migh resul in an inaccurae assessmen of he model. I is known ha he volailiy of durable goods consumpion is wo o four imes higher han ha of non-durable consumpion. 21 Second, he labor supply variaion in he model is capured only along he inensive margin. Conversely, we have employmen daa which measures he labor supply flucuaions only along he exensive margin. Earlier empirical sudies indicae ha he volailiy of employmen is wo o hree imes higher han ha of labor hours. Ineresingly, he predicion of he model concerning employmen flucuaions is also consisen wih his empirical regulariy: he volailiy of labor hours in he model is approximaely hree imes as low as he one of employmen series in he daa. We nex evaluae he performance of he model in replicaing comovemen properies of he daa. While quie closely maching he correlaion beween he primary secor oupu and aggregae oupu, he model overpredics he aggregae oupu-final secor oupu correlaion. The correlaions beween consumpion and oupu, and beween invesmen and oupu in he model are higher han hose in he daa. 22 The oupu-labor hours correlaion in he model is higher han he oupu-employmen correlaion in he daa. 21 See Backus, Kehoe and Kydland (1995) for a similar argumen abou he volailiy of durable goods consumpion. Baxer (1996) provides an exensive discussion abou he durables vs. nondurables goods differeniaion and is macroeconomic implicaions. 22 In our model, he represenaive household canno inraemporally subsiue consumpion goods in response o he relaive price flucuaions, since she derives uiliy by consuming only nonraded goods. I is possible o remedy near perfec procyclical behavior of consumpion by allowing he household o consume a variey of goods, such as exporable and imporable goods. This resuls in 13

16 This is also an expeced resul since he form of he uiliy funcion implies ha he number of hours worked a ime is deermined by aggregae oupu. One of he imporan feaures of he model economy is is abiliy o generae counercyclical behavior of he rade balance series. Ineresingly, compared wih he acual daa, here is a relaively high negaive correlaion beween he rade balance and aggregae oupu in he model. This resul migh be due o he presence of rade shocks which are negaively correlaed wih secoral produciviy shocks in he model. We furher invesigae his possibiliy in secion How imporan are he rade shocks? Our main objecive is o deermine he relaive conribuion of he rade shocks o macroeconomic flucuaions in African economies. We apply he variance decomposiion mehod on he soluion of he model o deermine he relaive imporance of shocks in explaining economic variaions. In oher words, we decompose he variances of macroeconomic variables ino fracions explained by exogenous shocks. This mehod requires us o impose a cerain informaion ordering on he shocks because he relaive conribuion of each disurbance o macroeconomic flucuaions is sensiive o is place in he shock specificaion. Since our model represens a small open economy, here is a naural ordering of shocks. By consrucion, he small open economy does no have any conrol over he exernal shocks i faces in he world markes. This implies ha domesic shocks do no have any impac on he exernal shocks, i.e. he exernal shocks precede secoral produciviy shocks in our specificaion. 23 The resuls of he variance decomposiions, which are obained by using he informaion ordering in (1), are repored in Table 8. The firs hree columns presen he fracion of variance due o rade shocks. Srikingly, a significan fracion of macroeconomic flucuaions is explained by rade shocks. They accoun for more han 44 percen of he variaion in aggregae oupu. Our resuls indicae ha shocks o he relaive price of capial goods o primary goods play a more imporan role han shocks o he relaive price of inermediae inpus. While almos 25 percen of variabiliy in aggregae oupu is due o he changes in relaive prices of capial goods, less han 2 percen of he flucuaions is due o he disurbances o relaive prices of inermediae inpus. The domesic produciviy disurbances play an imporan role in driving economic aciviy: roughly 55 percen of he oupu variaion is due o produciviy disurbances. Ineresingly, mos of he variaion explained by he produciviy shocks is due o he domesic produciviy movemens in he final goods secor. In our model, rade shocks have a direc impac on oupu flucuaions, since boh secors of he economy use impored goods as facors of producion. A significan fracion of he macroeconomic volailiy in he final goods producing secor, ha heavily relies on impored inermediae inpus and domesic capial goods, is explained by he rade shocks. Roughly 46 percen of he oupu variaion in he nonraded final goods producing secor is due o he rade shocks. Ineresingly, rade disurbances play a more imporan role in explaining consumpion flucuaions han hey do in oupu variaion: almos 8 percen of he variaion in consumpion is due o he rade shocks. Our resuls also show ha rade shocks have a large impac on macroeconomic flucuaions in producion facors: more han 86 percen of he volailiy of aggregae invesmen is explained by rade disurbances. In paricular, shocks o he relaive prices of primary capial goods accoun for more han 98 pronounced consumpion subsiuion effecs in response o he relaive price flucuaions and decreases he correlaion beween aggregae oupu and consumpion. 14

17 percen of he variaion in primary invesmen. This resul can be explained by he fac ha all invesmen goods in he primary good producing secor are impored capial goods. More han 42 percen of he variaion in inermediae inpus is explained by he disurbances o he relaive prices of inermediae goods. Shocks o he prices of capial goods and inermediae inpus also play an imporan role in inducing flucuaions in he labor marke. Movemens in he relaive prices of capial goods (inermediae inpus) accoun for more han 42 (37) percen of he variaion in he oal labor hours. Domesic produciviy shocks explain a larger fracion of he variaion in he primary goods secor han he final goods secor. Trade balance dynamics and foreign asse holdings are also heavily affeced by he price flucuaions in he world markes. This is an inuiively appealing resul: in secion 2 we repor ha expors, impors, and rade balance series are exremely volaile in African economies. Then, we consruced a model o represen a ypical African counry where he households expor all he primary goods hey produce while imporing inermediae inpus and capial goods. Now, we esablish he connecion beween highly volaile price shocks and rade balance dynamics: almos 74 percen of he flucuaions in he rade balance is accouned for by he rade shocks. Shocks o he relaive prices of capial goods explain more han 45 percen of he volailiy in he foreign asse holdings. These resuls are consisen wih hose of Deaon and Miller (1996) who analyze he imporance of inernaional commodiy prices in driving economic flucuaions in African counries using vecor auoregression analysis. Their resuls sugges ha while a sudden 1 percen increase in commodiy prices resuls in a 6 percen increase in oupu, he price shocks mos heavily affec invesmen dynamics in African economies. Ineresingly, we find ha world real ineres rae shocks do no play a significan role in driving domesic economic aciviy in our model. For example, hey accoun for less han 1 percen of he oupu volailiy. As expeced, hese shocks have a relaively more imporan role in driving he dynamics of asse holdings, bu heir influence is sill very small compared o he role of oher shocks: less han 6 percen of he variaion in foreign asse holdings is explained by ineres rae disurbances. Why does he world ineres rae play such a minor role in driving economic flucuaions in African counries? Firs, world ineres rae shocks do no have a direc impac on oupu flucuaions in our model: a rise in he world ineres rae, for example, resuls in a shif of domesic savings o foreign asses and decreases invesmen in producion. The change in invesmen affecs he dynamics of oupu. On he oher hand, relaive price shocks have a relaively more direc effec on flucuaions in producive inpus and oupu, and play a relaively more imporan role in inducing macroeconomic flucuaions. Secondly, when we look a he daa, we observe only a few large ineres rae changes (see Figure 1f). Capuring he effecs of hose large and shor lived ineres rae flucuaions on economic aciviy requires he use of differen echniques which allow a researcher o examine he relaionship beween macroeconomic flucuaions and world ineres rae shocks over shor ime periods The dynamic effecs of shocks We sudy he dynamic effecs of rade and produciviy shocks by using impulse response analysis. This invesigaion provides informaion abou he differences beween he propagaion mechanisms of produciviy and rade disurbances. We analyze he impulse responses of model variables o a 1 percen 23 We also analyzed he sensiiviy of hese resuls o differen orderings of he shocks. This invesigaion indicaes differen informaion orderings do no cause significan changes in he resuls repored here. 15