A Note on the Estimation of the Atemporal Elasticity of. Substitution Between Tradable and Nontradable Goods

Transcription

1 A Noe on he Esimaion of he Aemporal Elasiciy of Subsiuion Beween Tradable and Nonradable Goods Özge Akinci April 28, 2017 The aemporal elasiciy of subsiuion beween radables and nonradables plays an imporan role in he analysis of he dynamic macroeconomic equilibrium models of a small open economy. Various daases and procedures were used in he lieraure o esimae his parameer, wih values lying in he inerval ranging from 0.43 o Emerging marke economies end o have a smaller aemporal elasiciy of subsiuion parameer relaive o advanced economies. This noe offers a comprehensive survey of he lieraure. Federal Reserve Bank of New York, 33 Libery Sree, NY 10045, USA. Conac a ozge.akinci@ny.frb.org. 1

2 1 Inroducion The aemporal elasiciy of subsiuion beween radables and nonradables plays an imporan role in he analysis of he dynamic macroeconomic equilibrium models of a small open economy. Exising mehodologies used o esimae his parameer can be broadly divided ino hree caegories: (1) Single Equaion Regression Models using he cross secional daase by Kravis, Heson and Summers (KHS, for shor, hereafer), e.g., Sockman and Tesar (1995) and Mendoza (1992); (2) Generalized Mehod of Momens (GMM) using annual ime series daa pooled for a group of counries, e.g., Osry and Reinhar (1992); (3) Vecor-Error-Correcion (VEC) Models using quarerly ime series daa, e.g., Rozada e al. (2004) for Argenina and Lorenzo e al. (2005) for Uruguay. In he remainder of his noe, I will summarize he underlying macroeconomic models, consrucion of he daase, he mehodology used in he esimaion, and he resuls of he leading sudies in each hese sub-groups. 2 Single Equaion Regression Models Based on he KHS Daase 2.1 Underlying Model/Preferences Sudies in his caegory, in general, consider he uiliy funcion given by he following funcional form: [ (x α f 1 α) μ + n μ] 1 γ u(x, f, n) = 1 γ ; (1) where x denoes exporable goods, f denoes imporable goods, and n denoes nonradable goods. This expression can be rewrien for convenience as u(, n) = [ μ + n μ ] 1 γ ; (2) 1 γ where denoes a composie radable good. The consan elasiciy of subsiuion (CES) beween radable and nonradable goods is given by 1 1+μ. 2

3 2.2 Daa Following Kravis and Lipsey (1988), Sockman and Tesar (1995) esimae he elasiciy of subsiuion beween radable and nonradable goods from cross-secional daa provided in he World Bank s Income Comparison Projec. The daa can be found in he book by Kravis e al. (1982), which is available online under he Research Papers secions of he Penn World Tables. 1 Sockman and Tesar (1995) calculae he elasiciy of subsiuion beween raded and nonraded goods in a sample of 30 counries (17 developing and 13 indusrialized) using daa on per capia GDP, expendiure shares on raded and nonraded goods, and price indexes for raded and nonraded goods. Mendoza (1992) uses exacly he same daa se in his analysis bu includes only indusrialized counries in he esimaion of he elasiciy parameer. The daa which can be found in Kravis e al. (1982) for all counries for he year 1975 is summarized as follows : Per capia GDP (Kravis e al. (1982), p. 12, Tables 1-2, col.5) Expendiure shares on radable and nonradable goods (Kravis e al. (1982), p. 194, Tables 6-10, cols. 8 and 9) (Raio of nonradable goods expendiure o radable goods expendiure is calculaed by he auhors) Price Index for raded and nonraded goods (Kravis e al. (1982), p. 196, Tables 6-12, cols. 8 and 9) (Raio of nonradable good prices o radable good prices, Index, U.S. =100, is calculaed by he auhors) Figure 1 in he Appendix presens he scaer plo of he raio of nonradable good prices o radable good prices versus he raio of nonradable goods expendiure o radable goods expendiure for boh developing and indusrial counries in he KHS daase. 1 Daa can be found a a hp://pw.econ.upenn.edu/papers/conens.hml. 3

4 2.3 Mehodology The aemporal elasiciy of subsiuion beween radable and nonradable goods, 1/(1 + μ), is found by running a logarihmic regression of relaive expendiures on relaive prices and GDP per capia adjused for purchasing power. ln ( ) C 75 n,j C,j 75 = α 0 α 1 ln ( ) P 75 n,j P,j 75 + α 2 ln ( GDP Cj 75 ) ; (3) where C 75 n,j is he expendiure share on nonradable goods in 1975 for counry j, C75,j is he expendiure share on radable goods in 1975 for counry j; he expendiure shares Cn,j 75 + C75,j = GDP 75 j = 100 for each counry j. Similarly, Pn,j 75 is he price index for nonraded goods in 1975 for counry j and P 75,j is he price index for raded goods in 1975 for counry j. The raio of nonraded goods o he raded goods price index is calculaed by he auhors such ha he U.S. relaive price index=100. Finally, GDP Cj 75 is he per capia GDP daa for counry j in 1975 which is included in he regression o capure he income effec. 2.4 Esimaion Resuls wih Kravis e al. (1982) Daa Se Following Kravis and Lipsey (1988), Sockman and Tesar (1995) esimae he elasiciy of subsiuion beween raded and nonraded goods for 30 counries using KHS daa se by running he regression in equaion (3) and hey find ha here is a low degree of subsiuion in consumpion wih an elasiciy of subsiuion [1/(1 + μ)] of Mendoza (1992) esimaes he elasiciy of subsiuion beween radable and nonradable goods only for indusrial counries (13 counries) using he same daa se. This esimaion exercise produces an esimae of 1/(1 + μ) of abou 0.74 for he year

5 Table 1: Esimaion Resuls wih Kravis e al. (1982) Daa Se Sockman and Tesar (1995) Mendoza (1992) Emerging Counries (Akinci (2017)) Income Effec Income Effec Income Effec Parameers w/ w/o w/ w/o w/ w/o α (sd error) (0.498) (0.502) (1.612) (1.244) (1.008) (0.974) α (sd error) (0.225) (0.123) (0.448) (0.284) (0.314) (0.254) α (sd error) (0.087) (0.280) (0.130) Table 1 summarizes he esimaion resuls wih and wihou he income effec aken ino consideraion for all he counries in he sample (Sockman and Tesar (1995)), for only indusrial counries in he sample (Mendoza (1992)), and for emerging counries using he he KHS daase. Using he same daa se, afer replicaing hese resuls, I re-run he regression only for developing counries in he sample (17 counries) and found ha he elasiciy of subsiuion beween radable and nonradable goods is Mendoza (1995) in anoher sudy also menioned ha running he regression suggesed by Sockman and Tesar (1995) using KHS daa for developing counries yields 1/(1 + μ) = GMM Mehodology (Osry and Reinhar (1992)) 3.1 Underlying Model/Preferences Osry and Reinhar (1992) consider a small, open, endowmen economy wih an infiniely lived represenaive household whose objecive is o choose a consumpion sream ha maximizes U =[σ/(σ 1)]E 0 subjec o he series of budge consrains =0 β ( am 1 1/ɛ ) 1 1/σ + n 1 1/ɛ 1 1/ɛ, (4) p m + q n = p m + q n + x + A (1/R 1)A 1. (5) 5

6 and he ranversaliy condiion, where m (n ) denoes consumpion of imporables (nonradables), p and q denoe, respecively, he relaive price of imporables and nonradables, m, n, x show endowmens of he imporable, nonradable and exporable goods, respecively. The variable R denoes he world discoun facor and A denoes he real level of deb carried from period o period +1. In heir model, he elasiciy of subsiuion beween radable and nonradable goods is given by ɛ. The parameer β is he subjecive discoun facor and σ denoes he ineremporal elasiciy of subsiuion. The firs order condiions for an opimum (found by maximizing (4) subjec o (5)) are given by E E p R p +1 q R q +1 [ 1 1/ɛ am +1 + n 1 1/ɛ +1 am 1 1/ɛ + n 1 1/ɛ [ 1 1/ɛ am +1 + n 1 1/ɛ +1 am 1 1/ɛ + n 1 1/ɛ ] σ ɛ σ(ɛ 1) [ ] 1 m+1 ɛ m ] σ ɛ σ(ɛ 1) [ ] 1 m+1 ɛ m = 1 β, (6) = 1 β, (7) a(n /m ) 1/ɛ = p /q. (8) Since equaion (8) mus hold idenically and since equaions (6) and (7) are no independen, using raional expecaions assumpions, hey use equaion (6) o define he firs order condiion o be esimaed as he following: p u = R p +1 [ 1 1/ɛ am +1 + n 1 1/ɛ +1 am 1 1/ɛ + n 1 1/ɛ ] σ ɛ σ(ɛ 1) [ ] 1 m+1 ɛ m 1 β. (9) 3.2 Daa The parameers of he represenaive household s uiliy funcion μ =[β,ɛ,σ] are esimaed using annual pooled ime-series, cross-secion daa for 13 developing counries. The counries examined in he analysis include four African counries-egyp, Ghana, Coe d Ivoire, and Morocco; five Asian counries-sri Lanka, India, Korea, Pakisan, and he Philippines; and four Lain American counries-brazil, Colombia, Cosa Rica, and Mexico. Daa coverage for each counry begins in 1968 and ends somewhere beween 1983 and As equaion (9) highlighs, esimaion of he ineremporal and he inraemporal elasiciies 6

7 of subsiuion requires daa on he household consumpion of radable and nonradable goods and he erms of rade. While ime series on he erms of rade are readily available, auhors consruc daa on consumpion of he radable and he nonradable goods using daa from a variey of sources: The ime series for consumpion of imporables are consruced as follows: The agriculural, mining, and indusrial secors produce raded goods; GDP originaing in hese secors hus defines domesic producion of raded goods. Privae and public services comprise he nonraded goods secor. Domesic producion of impor subsiues is calculaed as domesic producion of raded goods less expors, on he assumpion ha exporables are no consumed a home. If markes clear, all domesic producion of impor subsiues is consumed a home. Consumpion of impor subsiues plus consumer goods impors, which are oal impors less impors of inermediae and capial goods, make up he series of ineres-consumpion of imporables. Nonraded goods consumpion is residually calculaed as oal privae consumpion less consumpion of imporables. The relevan price deflaors for he consumpion of radable and nonradable goods are price indices for impors and services, respecively. All consumpion daa are convered o a per capia basis by dividing he aggregaes by he exising populaion. 3.3 Mehodology The parameer vecor μ =[β,ɛ,σ] is esimaed by fiing he firs-order condiion defined in equaion (9) o he panel daa using GMM mehodology. The esimaion proceeds under he assumpion ha he parameers ha characerize household preferences are idenical across counries and regions. Two differen ses of insrumens are employed in he esimaion. The firs vecor of insrumenal variables z1 =[consan, m 1 /m 2,n 1 /n 2,p 2 /(R 2 p 1),m 1,n 1 ], uses six insrumens. The second insrumen se replaces he levels of consumpion of imporables and nonraded wih heir raio z2 = [consan, m 1 /m 2,n 1 /n 2,p 2 /(R 2 p 1),m 1 /n 1 ]. 7

8 3.4 Esimaion Resuls Esimaion resuls of Osry and Reinhar (1992) for all he counries in he sample indicae ha inraemporal elasiciy of subsiuion, ɛ, lies in he range. Esimaion is repeaed for each group of counries in he sample separaely. Resuls show ha ɛ =1.279 for Africa, ɛ =0.655 for Asia, and ɛ =0.760 for Lain America, when insrumen se I is used in he esimaion. Table 2: Esimaion Resuls wih Osry and Reinhar (1992) Daa Se Insrumen Se I Insrumen Se II Parameers All Counries Africa Asia L. America All Counries Africa Asia L. America ɛ sd error (0.154) (0.474) (0.105) (0.172) (0.351) (0.771) (0.270) (0.383) β sd error (0.033) (0.041) - - σ sd error (0.087) (0.159) (0.201) (0.111) (0.228) (0.178) (0.235) (0.135) J-saisic p-values (0.127) (0.170) (0.140) (0.080) (0.274) (0.170) (0.298) (0.292) #ofobs The esimaion is repeaed for he same group of counries using he second se of insrumens and hey find ha ɛ = for Africa, ɛ = for Asia, and ɛ = for Lain America. Table 2 presens he parameer esimaes for each insrumen se and he minimized value of he objecive funcion, J-saisic. The J-saisics are small relaive o he degrees of freedom (for eiher insrumen se), indicaing ha he overidenifying resricions imposed by he model are no rejeced by he daa; ha is, he hree parameers esimaed do a good job of saisfying eiher he five or six orhogonaliy condiions ha depend on he insrumen se. 3.5 Esimaion Resul for Indusrial Counries Cashin and McDermo (2003), by using very similar model and slighly differen mehodology, esimae he parameer vecor μ =[β,ɛ,σ] for a group of developed counries such as Ausria, Canada, 8

9 New Zealand, Unied Kingdom and Unied Saes for he period Their resuls indicae ha esimae of ɛ lies in he range depending on he counry examined. 4 Coinegraion/VEC Mehod for Argenina and Uruguay 4.1 The Case of Argenina (Rozada e al. (2004)) Underlying Model/Preferences Rozada e al. (2004) consider a small, open economy wih radable, T and non-radable goods, N. Households maximize where E β U ( C T (s ),C N (s ) ), =0 U(C T,C N )= ( ω(c T ) η +(1 ω)(e u )(C N ) η+θ) 1/η for n θ, U(C T,CN )=( ω(c T ) η +(1 ω)(e u )lnc N ) 1/η for n = θ, subjec o he budge consrain A 0 + =0 p(s )(Y d (s ) p N (s )C N (s ) C T (s ) (K +1 (s ) (1 δ)k (s ))) 0, (10) where C T, C N denoe he consumpion of radable and nonradable goods respecively, u is normally disribued preference shock, p(s ) is he price of a uni good T consumpion in sae s in erms of uni goods T a ime 0, Y d is household s disposable income measured in radable goods and p N is he relaive price of non-radable goods and K +1 (1 δ)k denoes he accumulaion of capial and K 0 is given. If he parameer θ in he uiliy funcion differs from zero, preferences are no homoheic and here will be income effecs in he relaive demand for radable and non-radable goods. The firs order condiion for his problem imply he following equilibrium condiion: ( C N ln C T ) = 1 1+η ln1 ω ω 1 1+η lnpn + θ 1+η ln ( C N ) + ɛ. (11) The parameer of ineres, he elasiciy of subsiuion in he demand for nonradable goods relaive o radable goods, is given by v =1/(1 + η). 9

10 4.1.2 Daa The esimaion of he elasiciy of subsiuion in he demand for nonradable goods relaive o radable goods in Argenina is performed wih he daa obained from official sources. Rozada e al. (2004) creaed wo daa ses for he period The firs one, which is creaed based on he Naional Accouns Procedure described below, conains daa on he consumpion of radable and nonradable goods consruced from he naional income and produc accouns following he flow of goods approach (see equaion (12)). The corresponding prices are he implici prices in he naional income accouns. The second daa se, which is creaed based on Simplified Naional Accouns Procedure, uses daa on he consumpion of non-durable goods (radable goods) and services (non-radable goods) and price indices for hese caegories of goods compued from he consumer price index daase. In he remainder of he his sub-secion, I will briefly describe hese wo mehodologies used in he consrucion of he daa: 2 Naional Accouns Procedure: According o his procedure, oal GDP is assumed o be he sum of oal value added in he nine secors of he economy (secors are (i) Agriculure, Huning and Fishing; (ii) Mining; (iii) Manufacuring; (iv) Public Uiliies: Elecriciy, Gas and Waer; (v) Consrucion; (vi) Wholesale and Reail Trade, Resaurans and Hoels; (vii) Transporaion, Communicaion and Sorage; (viii) Finance, Insurance, and Real Esae; (ix) Personal, Social and Communiy Services). As briefly menioned before, Rozada e al. (2004) are able o ge he final privae consumpion expendiure daa for each of he nine secors from he official sources. However, in he paper, hey also consruc heir own daa on he consumpion of radable and nonradable goods as a way of checking he daa from he official sources. Rozada e al. (2004) have ime series for he nominal and he real GDP in erms of he nine secors 2 Rozada e al. (2004) consruc heir own series for he consumpion of radable and nonradable goods using he Naional Accouns Procedure, and obain ime series for consumpion of radable and non-radable goods from he official sources which uses he same mehodology in he consrucion of consumpion and price daa. In he esimaion exercise, however, he use he laer because hey argue ha he levels of consumpion look more plausible. An ineresed reader can refer o Appendix 2 of he working paper version his aricle (see Neumeyer and Rozada (2003)) for furher discussion abou he consrucion of he daa. 10

11 of he economic aciviy for he period Final privae consumpion for each of he nine secors is available only for four poins in ime (1960, 1963, 1973 and 1997) in which Inpu-Oupu Marix was esimaed by Naional Insiue of Saisics and Census (INDEC). The daa is reliable bu scarce and would no allow hem o run he regressions. In order o obain ime series daa for he final privae consumpion for each of he nine secors in GDP, hey use he following relaion: C i, = Y i, j IC ij, (X i, M i, ) I i,. (12) where C i, Y i, X i, M i, I i denoe consumpion, gross producion, expors, impors and invesmen in secor i, andic ij denoes he inermediae consumpion of good i in secor j. The corresponding prices are he implici prices in he naional income accouns. Final privae consumpion in secor i is GDP ne of he consumpion of inermediae inpus, invesmen and ne expors. As hey discuss in he paper, he firs problem for implemening his equaion is ha here is no invesmen daa for each secor. In Argenina gross fixed invesmen is divided ino wo secors: non-residenial consrucion and manufacuring. They assume ha invesmen in every oher secor is zero, so heir series for consumpion will include he change in invenories. I = I m + C nrc where I m and C nrc denoe Invesmen in Manufacuring and Consumpion in Non-Residenial Consrucion. The second problem is ha hey know he inpu-oupu srucure of he economy only for four poins in ime. Assuming a consan inpu-oupu marix for a long period of ime in which relaive prices suffer a large variaions would be unsaisfacory. To solve his problem, hey assume ha final consumpion and he demand for inermediae inpus behave similarly and wrie (12) as (13) assuming (14): C i, + j IC ij, = Y i, (X i, M i, ) I i,, (13) C i, j IC ij, = a i. (14) They obain a i as an average of final privae consumpion for each secor ou of inermediae consumpion for each secor, from he Inpu-Oupu Marix. Once he consumpion daa for he nine secors are obained, hey assume he firs four secors are radable and he las five are non-radable, and consruc he series for radable and nonradable consumpion. The correlaion of nominal (real) nonradable o 11

12 radable consumpion raio, C N /C T, beween he daa calculaed by he auhors and he one obained from he official sources is 0.64 (0.7). The auhors argue ha official sources daa have a beer esimaion of he consumpion of secors (2) and (5) due o he fac ha hey use more disaggregae daa. Therefore, hey use he official source daa in he esimaion exercise. Figure 2 in he Appendix presens he ime series daa obained from he official sources. Simplified Naional Accoun Procedure: The second daa se employed in his sudy is consruced as daase for non-durable goods (radable goods) and services (non-radable goods) and he price indices for hese caegories of goods compued from he consumer price index daase. They did no give deailed explanaion abou he consrucion of his daa se in he paper Mehodology Afer geing he daa series, hey esimae he equilibrium relaionship given in (11) o find he inraemporal elasiciy of subsiuion, v. The firs sep in heir esimaion procedure is o ransform he daa by compuing he relevan variables in logs; ha is, hey calculae he log of he real radable and nonradable consumpion, he log of he implici prices of non radable and radable goods. The ime rend in he raio of he consumpion of nonradable o radable goods moivae he flexible funcional form for preferences ha allow for he non-homoheic case. Since he daase comprises variables varying wih ime, second sep is o check he order of inegraion by implemening he sandard augmened Dickey-Fuller es on each variable. Augmened Dickey-Fuller es on he variables shows he presence of a uni roo in each of he series. Then hey check he presence of he coinegraing vecor by using Johansen s coinegraion es among he raio beween consumpion of non radable and radable goods, he raio beween he prices of he non radable and radable goods and he consumpion of non radable goods in logarihms and expressed in real erms. They find an evidence of coinegraion among he variables as prediced by he heoreical model. Finally, hey esimae a VEC model for he hree variables. The specificaion of he error correcion model corresponds o he lags, six, and i includes hree dummy variables o accoun for he seasonal cycles. 12

13 4.1.4 Esimaion Resuls They presen he resuls of VEC model for each of he ime series hey used: VECM using daa from NIPAS ln ( C N C T ) = lnp N (0.20) ln ( C N (0.05) ) (15) All esimaed coefficiens are saisically significan a he usual significance levels. Equaion (15) indicaes ha he long-run elasiciy of subsiuion beween radable and non radable goods is abou 0.40, which is closer o he esimae of 0.44 obained by Sockman and Tesar (1995) The coefficien on C N is ighly esimaed and indicaes srong income effecs. VECM using daa from simplified NIPAS ln ( C N C T ) = lnp N (0.366) ln ( C N (0.005) ) (16) In order o have an alernaive esimaion of he elasiciy of subsiuion, hey repea he procedure described above bu use he consumpion of nondurable goods as a proxy for he consumpion of radable goods and he consumpion of services as a proxy for he consumpion of nonradable goods. The implici prices of nondurable and services goods are used o consruc he raio beween he prices of nonradable and radable goods. Equaion (16) shows ha he long-run elasiciy of subsiuion beween radable and nonradable goods is abou Robusness: Differen ime span The working paper version of his sudy by Neumeyer and Rozada (2003) also presens he esimaion resuls based on he daa covering he period 1980 firs quarer o 1988 fourh quarer for consumpion of nondurable goods and services. Using he same approach, he consumpion of nondurable goods is used as a proxy for he consumpion of nonradable goods, while he consumpion of services is 13

14 used as a proxy for he consumpion of nonradable goods. The prices for non durable goods and services are consruced using a geomeric price index from he CPI. ln ( C N C T ) = lnp N (0.21) (17) The long run elasiciy of subsiuion beween radable and nonradable goods is abou 0.39, which is close o he esimae of he long-run elasiciy of subsiuion for he nineies and i is also close o he 0.44 obained by Sockman and Tesar (1995). 4.2 The Case of Uruguay Lorenzo e al. (2005) employ he same mehodology o esimae he aemporal elasiciy of subsiuion. Auhors creae quarerly ime series of consumpion and prices of radable and nonradable goods for Uruguay for he period In he creaion of he daabase, similar procedures described for he case of Argenina are followed; ha is, he consumpion of radables and nonradables and he corresponding prices are creaed using daa from official resources according o naional accouns procedure and simplified naional accouns procedure. The economeric esimaion of he parameer of ineres is performed wih VEC models. The esimaion exercise gives a long-run elasiciy of subsiuion of Parameric sabiliy ess are performed on he principal model, and he predicive abiliy of he model is also esed. They conclude ha no only is he parameer of ineres sable over ime, bu he model also has good predicive properies. 5 Conclusion The exising sudies for he esimaion of he aemporal elasiciy of subsiuion show ha using he KHS cross secion daa, esimaes lie in [0.43, 0.74] range according o he specific subgroup of counries included in he esimaion when he income effec is aken ino accoun. The esimaed value of he elasiciy using annual daa for 13 developing counries by Osry and Reinhar (1992) varied in [0.75, 1.50] range according o he specific subgroup of counries sudied and he insrumen se used. The research projec by Iner American Developmen Bank for he esimaion of he inraemporal elasiciy 14

15 of subsiuion for he Lain American counries (Argenina and Uruguay) indicaes ha he elasiciy lies in [0.4, 0.5] range. 15

16 6 Appendix 6.1 KHS Daase Figure 1: Scaer Plo of he Raio of Nonradable Good Prices o Tradable Good Prices versus he Raio of Nonradable Good Expendiure o Tradable Good Expendiure for Developing Counries (upper panel) and Indusrial Counries (lower panel) in he KHS Daase. 16

17 References Cashin, Paul and C. John McDermo, Ineremporal Subsiuion and Terms-of-Trade Shocks, Review of Inernaional Economics, , 11 (4), Kravis, Irving B. and Rober E. Lipsey, The Assessmen of Naional Price Levels, Working Paper 1912, Naional Bureau of Economic Research Sepember Kravis, Irving; Heson Alan and Rober. Summers, World produc and income: Inernaional comparisons and real GDP, Balimore, MD: Johns Hopkins Universiy Press, Lorenzo, Fernando, Rosa Osimani, and Paricio Valenzuela, The Elasiciy of Subsiuion in Demand for Non-Tradable Goods in Uruguay, RES Working Papers 3171, Iner-American Developmen Bank, Research Deparmen April Mendoza, Enrique G., The Effecs of Macroeconomic Shocks in a Basic Equilibrium Framework, Saff Papers - Inernaional Moneary Fund, 1992, 39 (4), Mendoza, Enrique G, The Terms of Trade, he Real Exchange Rae, and Economic Flucuaions, Inernaional Economic Review, February 1995, 36 (1), Neumeyer, Pablo Andres and Marin Gonzalez Rozada, The elasiciy of Subsiuion in demand for Non radable Goods in Lain America. Case Sudy: Argenina, Deparmen of Economics Working Papers 027, Universidad Torcuao Di Tella November Osry, Jonahan D. and Carmen M. Reinhar, Privae Saving and Terms of Trade Shocks: Evidence from Developing Counries, Saff Papers - Inernaional Moneary Fund, 1992, 39 (3), pp Rozada, Marin Gonzalez, Pablo Andres Neumeyer, Alejandra Clemene, Diego Luciano Sasson, and Nicholas Tracher, The Elasiciy of Subsiuion in Demand for Non Tradable Goods in Lain America: The Case of Argenina, RES Working Papers 3179, Iner-American Developmen Bank, Research Deparmen Augus

18 Sockman, Alan C and Linda L Tesar, Tases and Technology in a Two-Counry Model of he Business Cycle: Explaining Inernaional Comovemens, American Economic Review, March 1995, 85 (1),