Growth and Global Imbalances: The Role of Learning-by-Exporting

Transcription

1 Growh and Global Imbalances: The Role of Learning-by-Exporing B H S March 23, 2016 Absrac Rapidly growing developing economies have expored heavily and run curren accoun surpluses. This paper explores wheher learning-by-exporing helps explain his macroeconomic behavior of developing counries. I builds up a wo-counry general equilibrium growh model in which a developing economy benefis from learning-byexporing as i rades wih a developed economy. In he benchmark model, he policies are resriced o non-rade relaed ones by he World Trade Organizaion (WTO). The opimal policy for he developing counry is hen o ax non-raded goods consumpion and subsidize savings o increase expors. This policy generaes he simulaneous fas growh and curren accoun surpluses observed in he daa. However, if here were no WTO resricions, he developing counry would manipulae is erms of rade raher han is curren accoun. This policy improves he welfare of boh developing and developed counries, highlighing ha erms of rade manipulaion can be a "win-win" in he presence of learning-by-exporing. Keywords: Curren Accoun, Learning-by-Exporing, Terms of Trade JEL Classificaions: E61, F13, F32, O24 I am deeply indebed o Mark Aguiar for his suppor and encouragemen. I also hank Yan Bai, Mark Bils, Yongsung Chang, Eyal Dvir, William Hawkins, Jay Hong, Ronald Jones, Nobuhiro Kiyoaki, Ronni Pavan, Hye Mi You, and seminar paricipans a he Universiy of Rocheser, Ryerson Universiy, Ohio Sae Universiy, SUNY a Buffalo, KIPF, Ajou Universiy, KISDI, Kyung Hee Universiy, KDI, Korea Universiy, Seoul Naional Universiy, Sogang Universiy, Ewha Womans Universiy, he 2011 and 2013 Midwes Macro Meeings, he 2011 WEAI Graduae Suden Disseraion Workshop and Annual Conference, he Asian Meeing of he Economeric Sociey 2011, he 19h Inernaional Conference on Compuing in Economics and Finance, and he Yonsei Global Research Team Workshop for heir valuable commens. All errors are mine. Deparmen of Economics, The Ohio Sae Universiy, 410 Arps Hall, 1945 N. High S., Columbus, OH seok.19@osu.edu 1

2 1 Inroducion Rapidly growing developing economies such as China and oher Asian counries have expored heavily 1 and run curren accoun surpluses. The fas growh accompanied by curren accoun surpluses conradics he predicion of he open-economy neoclassical growh model ha counries wih faser produciviy growh should receive more ne capial inflows o fund invesmen and smooh consumpion. Gourinchas and Jeanne (2012) name his inconsisency he "allocaion puzzle." These fas growing counries curren accoun surpluses conribue o he worldwide curren accoun imbalances, or so-called "global imbalances." 2 Since hese economies have expored heavily, a popular view is ha expor-led growh may be behind some of hese dramaic growh phenomena. This is suppored by empirical sudies which sugges ha "learning-by-exporing" (exporers produciviy improvemen accompanied by increased expors) may be quaniaively large in developing counries. This paper akes he popular view seriously and explores wheher learning-by-exporing helps explain he key macroeconomic behavior of fas-growing developing counries. This paper also examines which policies exploi learning-by-exporing, heir implicaions for aggregaes like he curren accoun and he real exchange rae, he welfare consequences for he growing economy and he res of he world, and wheher resricing he se of policies o non-rade relaed policies maers. In order o answer hese quesions, his paper builds up a wo-counry general equilibrium growh model in which a developing economy benefis from learning-by-exporing as i rades wih a developed economy. This posiive exernaliy in he developing counry s 1 Figure 1 shows ha since 1991, China s expor has grown persisenly. The average expor o GDP raio in China from 1991 o 2007 is 25.22%. Over he same period, he average growh rae of China s real GDP per capia is 9.57%. 2 As shown in Figure 2, China has run a curren accoun surplus since he early 1990s, while he U.S. has run a persisen curren accoun defici. This curren accoun imbalance beween he U.S. and China conribue o global imbalances. 2

3 expor provides i wih an incenive o increase exporaion. The model is calibraed o mach relevan daa momens of he U.S. and China in 1991 and is simulaed for ransiion o seady sae. I consider a seup in which policies are resriced by he World Trade Organizaion (WTO) o non-rade relaed policies as he benchmark. 3 In his benchmark model, he opimal policy for he developing counry is o ax non-raded goods consumpion and subsidize savings, which shifs labor ino he radable secor and suppresses consumpion o increase expors. 4 These policies generae he simulaneous fas growh and curren accoun surpluses observed in he daa. These policies improve he welfare of he developing counry relaive o a compeiive equilibrium wihou any policies because he developing economy benefis from rapid growh due o learning-by-exporing. However, he welfare change of he developed counry beween he benchmark case and he no policy economy is quaniaively negligible. If here were no WTO resricions, he developing counry would have an incenive o manipulae is erms of rade raher han disor savings. Specifically, he developing counry subsidizes expors o reduce is consumpion of he expor good and increase consumpion of he impor good. This policy generaes a large deerioraion in he developing economy s erms of rade, couner o he sandard opimal ariff idea in which counries wan o raise heir erms of rade. The policy also reverses he predicion for is curren accoun. In 3 Since he WTO is designed o liberalize inernaional rade, i forces counries o decrease ariffs and expor subsidies. Therefore, he WTO prevens counries from manipulaing heir erms of rade. The WTO resricions in his model represen he general sae of rade rules ha preven counries from manipulaing erms of rade. For insance, a counry may no be able o manipulae is erms of rade if is rading parner can implemen rade policies in realiaion. 4 This response is consisen wih curren Chinese governmen policies. The Chinese governmen is axing he gross revenue in he service secor (Business Tax), bu in he manufacuring secor, hey are axing he difference beween a commodiy s price and is producion cos (Value Added Tax). Ping, Liang, Hao, Zhang, and Mao (2009) show ha if he ax rae of he Business Tax is convered o a Value Added Tax, i is 18.2%. This is greaer han he Value Added Tax rae of 17% ha is applied o manufacuring. Therefore, he Chinese governmen is axing he service secor heavier han he manufacuring secor, hus he policy of applying a non-raded good s consumpion ax is consisen wih China s ax regime. The opimal policy of providing a savings subsidy is also consisen wih China s policy of sockpiling a large amoun of governmen savings, which was 4.4% of GDP in 1992 and 10.8% in 2007, according o Ma and Yi (2010). 3

4 paricular, he developing economy now runs a curren accoun defici, as i no longer relies heavily on he savings disorion o promoe expors. This policy raises he welfare of boh counries relaive o he benchmark model, as i generaes faser economic growh in he developing economy and improvemen of he erms of rade in he developed economy. 5 This highlighs he fac ha erms of rade manipulaion can be a "win-win" in he presence of a posiive exernaliy in developing counries from learning-by-exporing, which conrass wih Bagwell and Saiger (1999) s view ha he WTO improves world welfare by prevening zero-sum erms of rade manipulaion. This paper also considers a "coordinaed policy" problem o obain he firs-bes oucome for he world. If boh developing and developed counries coordinae policies o achieve he world s bes allocaion, boh counries welfare rises significanly compared wih he benchmark economy. In his seup, he developing counry s erms of rade deeriorae even furher, and i runs a greaer curren accoun defici relaive o he "no-wto resricions" case. This large deerioraion of he developing counry s erms of rade reduces welfare of he developing counry and increases ha of he developed counry, compared wih he "no-wto resricions" case. However, he welfare changes of boh counries beween he "coordinaed policy" case and he "no-wto resricions" economy are quaniaively small. This resul implies ha we can obain almos he same resuls as he firs-bes oucome for he world by allowing developing counries o subsidize heir exporers. This paper is moivaed by wo disinc lines of sudy. The firs consiss of empirical micro sudies ha find large esimaes of learning-by-exporing in developing counries. 6 A possible explanaion is ha exporers in developing counries improve heir produciviy hrough imiaion and echnology spillover from developed counries. 7 The mos challenging ask of 5 This resul suggess ha if he U.S. allows China o subsidize is exporers, hen he curren accoun imbalances beween he U.S. and China will decrease and he welfare of boh counries will increase. 6 These sudies include Kraay (1999), Blalock and Gerler (2004), Aw, Robers, and Xu (2011), Park, Yang, Shi, and Jiang (2010) for Eas Asian counries, Van Biesebroeck (2005) for African counries, De Loecker (2007), De Loecker (2013) for Slovenia, and Fernandes and Isgu (2011) for Colombia. Harrison and Rodríguez-Clare (2010) provide exensive reviews of he above. 7 Empirical micro sudies poin ou ha learning-by-exporing also comes from exporers improved access 4

5 hese sudies is o conrol for he effecs of he unobserved differences in firm characerisics beween exporers and non-exporers. In order o conrol for his selecion bias, Van Biesebroeck (2005) uses ehniciy of he owner and sae ownership as insrumens, De Loecker (2007) uses mached sampling echniques based on an underlying model of self-selecion ino expor markes, and Park, Yang, Shi, and Jiang (2010) use exogenous firm-specific exchange rae shocks as insrumens. These sudies find significan evidence of learning-by-exporing afer conrolling for his selecion bias. Anoher issue regarding learning-by-exporing is o deermine if i can be disinguished from learning-by-doing. A wide range of sudies (for insance, Blalock and Gerler (2004), Van Biesebroeck (2005), De Loecker (2007), and De Loecker (2013)) show ha here is a jump in firms produciviy accompanied by he iniiaion of exporing, which canno be explained by learning-by-doing. One migh also hink ha learning-by-imporing is as imporan as learning-by-exporing. According o Keller (2004), however, here has been lile empirical suppor for learning-by-imporing. The second lieraure focuses on eiher global imbalances or he allocaion puzzle and aemps o develop mechanisms o explain he observed daa paerns. Suggesed explanaions regarding global imbalances include he lack of financial asses in developing counries (Caballero, Farhi, and Gourinchas (2008); Mendoza, Quadrini, and Rios-Rull (2009)) and weakened precauionary savings moive in he U.S. caused by he "grea moderaion" (Fogli and Perri (2006)). Sudies on he allocaion puzzle provided consumpion habis (Carroll, Overland, and Weil (2000)), financial fricions faced by privae firms in China (Song, Soresleen, and Ziliboi (2011)), limied commimen and poliical economy disorions (Aguiar and Amador (2011)), and a flaer cross-secional age-income profile accompanied by fas growh (Guo and Perri (2013); Song and Yang (2010)) as poenial explanaions. However, his lieraure focuses on ne rade (ypically in one-good models) and is silen on comparaive advanage. My paper no only provides a new explanaion regarding hese puzzling daa paerns bu also analyzes he connecion beween growh, curren accoun o advanced producion echnologies, echnical assisance from foreign buyers, compeiion wih foreign firms and higher qualiy sandards in inernaional markes relaive o domesic markes. 5

6 surplus, and WTO resricions in he comparaive advanage framework. Furhermore, i explores he opimal polices in he presence of learning-by-exporing and welfare consequences for developing and developed counries, boh of which have no been addressed by previous sudies. The remainder of he paper is organized as follows. Secion 2 presens he model; secion 3 describes he calibraion of he model; secion 4 discusses he resuls; secion 5 explains a "coordinaed policy" problem and is resuls; secion 6 provides he welfare analysis; secion 7 does a sensiiviy analysis on he degree of learning-by-exporing; and secion 8 concludes my findings. 2 Model I presen a wo-counry general equilibrium growh model. Time () is discree and runs from 0 o infiniy. The Norh counry, denoed by N, corresponds o a developed economy and owns he mos developed echnology. This model absracs from physical capial o focus on savings side. 8 The Norh s human capial sock H N is assumed o be consan, reflecing ha he Norh has fully exhaused produciviy gains from learning-by-exporing. The Souh counry, denoed by S, represening a developing economy, has an inferior echnology H S [ H S 0, H N). Given he assumpion ha H N is consan, only he Souh counry grows hrough learning-by-exporing as i rades wih he Norh counry. Each counry produces one non-raded commodiy and boh counries share wo raded goods (z {1, 2}). There is also an inernaional financial marke ha buys and sells risk-free bonds b i wih a reurn denoed by 1 + r. Each economy is populaed by firms ha produce goods and workers 8 Using aggregae daa, Gourinchas and Jeanne (2012) show ha he savings wedge is he key in explaining developing counries fas growh accompanied by curren accoun surpluses. Moivaed by heir resul, his paper builds up he mos parsimonious model for savings side. Rapidly growing developing counries, which run curren accou surpluses, end o undergo a lo of invesmen bu save even more. The saving side is paricularly puzzling given he fas growh. We have models of why here may be capial wedges (enforcemen, for example), bu few regarding savings wedges. Including capial in he model will affec he quaniaive resuls, bu he key mechanism sill sands ha absen he abiliy o manipulae he erms of rade, learning-by-exporing calls for a savings wedge. 6

7 who provide domesic firms wih labor. Lasly, he Souh counry has a governmen ha implemens policies o ake advanage of learning-by-exporing. 2.1 Firms Counry i {N, S} firms in he rade goods secor use labor n i (z) o produce oupu y i (z) according o a consan reurns o scale producion funcion y i (z) = A i (z) n i (z), i {N, S}, z {1, 2}, where A i (1) H, i A i (2) ( H) i 1+φ, φ > 0. Labor produciviy in he rade goods secor depends on each counry s human capial. Since φ is greaer han zero, he second raded commodiy s producion is more human capial inensive han he firs raded commodiy s. The Souh counry has a comparaive advanage in he firs raded commodiy s producion because i has less human capial H S < H N : A S (2) A S (1) = ( H S ) 1+φ H S < AN (2) A N (1) = ( H N ) 1+φ H N. Therefore, he Souh counry expors he firs raded commodiy. Labor is hired by he firms in a compeiive domesic labor marke ha clears a an equilibrium wage w i. Firms in he raded goods secor maximize heir profi p (z) A i (z) n i (z) w i n i (z). Since I assume a perfec compeiion in he raded goods secor, he law of one price holds. Thus, he world price of raded commodiy z is p (z) wi A i (z). Since he Souh counry produces he firs raded commodiy and he Norh produces he second raded commodiy, he Souh s domesic wage is ( ) ( ) w S p (1) A S = (1). p (2) A N (2) 7

8 A firm in he non-raded goods secor uses labor n i o produce oupu y i according o a consan reurns o scale producion funcion y i = n i, i {N, S}. I assume ha labor produciviy in he non-raded goods secor is equal o one in boh counries because he focus is on produciviy improvemen in he raded goods secor. 9 The non-raded goods secor firm maximizes is profi p i n i w i n i. Therefore, each counry s non-raded commodiy price is p N = w N = 1 and p S = w S. The Norh s non-raded commodiy/ labor is he numeraire. 2.2 Domesic Workers A represenaive worker supplies labor N i inelasically for domesic firms in boh non-raded and raded goods secors, and can rade a risk-free bond b i from he inernaional financial marke. The worker enjoys uiliy flows from consumpion of he non-raded commodiy c i and wo raded goods c i (z), z {1, 2}. The worker discouns he fuure uiliy wih a discoun facor β (0, 1) and has preferences: where β u(c), i =0 u(c) i (Ci ) 1 γ 1 ; 1 γ C i ( c i ) 1 ψ c i (1) ηψ c i (2) (1 η)ψ ; η (0, 1), ψ (0, 1). 9 Even if I allow he produciviy in he non-raded commodiy secor o differ across hese wo counries, all he qualiaive resuls carry hrough. If he Norh labor produciviy in non-raded goods secor is greaer han ha of he Souh, he Norh produces and consumes more non-raded goods han before, and all oher allocaions do no change. 8

9 Noe ha Cobb-Douglas preferences feaure a uni elasiciy of subsiuion across he nonraded commodiy and wo raded goods. Wih his form of uiliy funcion, he expendiure share on raded goods is ψ. Wihin raded goods, he proporions of he firs and second raded commodiy are η and 1 η, respecively. The Souh counry s real exchange rae is where e S P S P N ( p S = p N ) 1 ψ = ( w S {( ) ) 1 ψ p (1) = p (2) ( )} A S 1 ψ (1), A N (2) ( ) p P i i 1 ψ ( ) ηψ ( ) (1 η)ψ p (1) p (2), i {N, S}. 1 ψ ηψ (1 η) ψ Since he law of one price holds in he raded goods secor, he Souh counry s real exchange rae is defined by he raio of each counry s non-raded commodiy price. I assume ha he Norh does no levy axes, so he represenaive worker in he Norh counry maximizes uiliy subjec o a budge consrain: c N + p (1)c N (1) + p (2)c N (2) + b N +1 = N N + (1 + r ) b N. However, he represenaive worker in he Souh counry maximizes uiliy subjec o a budge consrain: ( 1 + τ NT ) p S c S + ( ) 1 + τ EX p (1)c S (1)+p (2)c S (2)+b S +1 +T = w S N S +{1 + (1 + τ r ) r } b S. The Souh governmen can ax or subsidize non-raded commodiy consumpion ( ) τ NT, exporing commodiy consumpion ( ) τ EX, and/ or domesic savings (τ r ). In addiion, he governmen can use a lump-sum ax or ransfer (T ). 10 Wihou loss of generaliy, I normalize axes on impors o zero. 11 For he benchmark, I consider a seup in which he policies are resriced by he WTO o non-rade relaed policies. Thus, in he benchmark model, I assume τ EX = 0. This 10 I am primarily ineresed in he long run rend of he pas 20 years. Since moneary policy is neural in he long run, I focus on fiscal policies. 11 Wihou loss of generaliy, he Souh governmen axes only consumers. The governmen can use he full se of axes, and his is no he only way o decenralize he sysem. Assume ha he Souh workers do no have access o he inernaional financial marke and is governmen rades a risk-free bond b i on behalf of workers. The model implicaions for key macroeconomic variables do no change. 9

10 ( p(1) p (2) means ha he Souh governmen canno direcly subsidize expors or manipulae is erms ) of rade. Then, I will compare he resuls of he benchmark model o hose of a "nopolicy" compeiive equilibrium ( τ NT = τ EX = τ r = T = 0 ) and he "no-wto resricions" case in which he Souh governmen can ax or subsidize exporing commodiy consumpion ( τ EX 0 ). 2.3 Law of Moion for Souh Human Capial I assume he Norh has exhaused learning-by-exporing, so only he Souh counry grows hrough learning-by-exporing as i rades wih he Norh counry. Two common findings in empirical micro sudies on learning-by-exporing are ha an exporer s produciviy improves as heir value of expors grows, and his expor-produciviy relaionship becomes sronger as firms expor o relaively more developed counries. On he basis of hese evidences, I model he degree of learning-by-exporing as an increasing funcion of boh he Souh value of expors and he difference in human capial socks beween Norh and Souh. Thus, he law of moion for Souh human capial is H+1 S = H S + ( ) { ( )} κh N H S 1 exp EXS, (1) α }{{} "Learning-by-Exporing" where EX S max {( y S (1) c S (1) ), 0 }. The Souh s human capial can grow up o κh N, where κ (0, 1) 12, hrough learning-byexporing. 13 The difference beween Norh and Souh human capial socks is hen represened by ( κh N H S ). 14 The value of Souh expors ( EX S ) is he firs raded commodiy 12 If κ is equal o one, he comparaive advanage disappears a he seady sae ( H S = κh N). Therefore, he seady sae is indeerminan. 13 The funcional form ha I use for he law of moion for Souh human capial does no allow he Souh human capial ( ) H S o converge o κh N in finie periods. Therefore, I consider ha his model economy arrives a he seady sae when he Souh human capial ( ) H S reaches 99% of κh N. 14 Noe ha he Souh s learning-by-exporing depends on he difference in human capial socks beween Norh and Souh. If he Norh human capial grows over ime, he Souh s produciviy gains from learning- 10

11 expors. If here is produciviy improvemen from he second raded commodiy expors, he Souh counry expors boh raded goods in he seady sae. In order o exclude his exreme case, he degree of learning-by-exporing from he second raded commodiy expors is se a zero. The parameer α > 0 governs he degree of learning-by-exporing, which is a decreasing funcion of α. 2.4 Compeiive Equilibrium A compeiive equilibrium consiss of a se of quaniies { c i, c i (z), b i, H S }, a se of prices {p i, p (z), w, i r }, and a se of axes { } τ NT, τ EX, τ r, T, where i {N, S} and z {1, 2}, such ha: 1. given prices and axes, workers maximize uiliies 2. given prices, firms maximize profis 3. he Souh human capial evolves according o he law of moion saed in equaion (1) 4. he Souh governmen budge consrain is saisfied: τ NT p S c S + τ EX p (1)c S (1) + T = τ r r b S goods markes clear: c i = y i, i {N, S} ; c N (z) + c S (z) = y N (z) + y S (z), z {1, 2} 6. labor markes clear: n i + n i (1) + n i (2) = N i, i {N, S} 7. bond marke clears: b S + b N = 0. by-exporing are no exhaused. The Souh canno converge o seady sae. 15 Wihou loss of generaliy, I assume ha he governmen runs a balanced budge using a lump-sum ax or ransfer (T ). 11

12 2.5 Ramsey Problem The Souh governmen recognizes he law of moion for is human capial and implemens policies in order o ake advanage of learning-by-exporing. 16 The Souh governmen s problems in he benchmark model and he "no-wto resricions" case are he Ramsey problem of choosing a compeiive equilibrium o maximize he Souh worker s uiliy, given H S 0 and b S 0. Following he primal approach o he Ramsey problem (Jones, Manuelli, and Rossi (1997)), I formulae he Souh governmen s problems as if he governmen chooses an allocaion, subjec o consrains, ha ensure he exisence of prices and axes such ha he seleced allocaion is consisen wih he opimizing behavior of workers and firms Benchmark The allocaion seleced by he Souh governmen has o saisfy he law of moion for Souh human capial, boh counries domesic labor markes clearing condiions, and all goods markes clearing condiions. In addiion o hese sandard consrains, he allocaion should also saisfy: (i) he Norh worker s opimaliy condiions and presen-value budge consrain, (ii) he opimaliy condiions of he Norh firms, and (iii) WTO resricions. The Norh represenaive worker solves: max β u(c N ), subjec o ( =0 i= r i ) =0 (c N + p (1) c N (1) + p (2) c N (2) N N) = b N 0. Therefore, he Norh worker s opimaliy condiions are: β +1 1 = ; 1 + r +1 (z) = p (z), z {1, 2}, 16 Noe ha firms do no inernalize learning-by-exporing in his model. If firms recognize learning-byexporing, he only hing firms can do o ake advanage of learning-by-exporing is dumping. However, firms canno use dumping because of he WTO resricions. Therefore, a governmen s macroeconomic policy is essenial o ge he full benefi from learning-by-exporing. 12

13 where (z) is he Norh worker s marginal uiliy of consumpion for he raded commodiy z {1, 2}, and is he Norh worker s marginal uiliy of consumpion for is non-raded commodiy. Noe ha he Norh worker s opimaliy condiions and presen-value budge consrain are summarized as he following implemenabiliy condiion: =0 β ( c N + (1) c N (1) + (2) c N (2) N N ) = 0 b N This implies ha any compeiive equilibrium mus saisfy he Norh implemenabiliy condiion, and any allocaion ha saisfies his condiion and goods marke clearing condiions can be decenralized as a compeiive equilibrium. The opimaliy condiions of he Norh firms are summarized as follows: if p (z) = (z) < if p (z) = (z) = 1 A N (z), nn (z) = 0; 1 A N (z), nn (z) > 0, z {1, 2}. The firms in he Norh s raded goods secor do no produce he raded commodiy z if is world price p (z) is less han he firms uni labor cos 1 A N (z). The WTO resricions are represened by u c S (1) u c S (2) = (1). (2) Since he Souh governmen canno direcly subsidize expors or manipulae erms of rade in he benchmark model, he Souh s domesic relaive price of he expor good o impor ) good is equal o he world price. ( p(1) p (2) Therefore, he Souh governmen s problem in he benchmark model 18 is formulaed as follows: he Souh governmen solves max β u(c S ), 17 See he appendix for he derivaion of he implemenabiliy condiion. 18 See he appendix for he compuaion algorihm used o solve he benchmark model. =0 13

14 subjec o H S +1 = H S + ( κh N H S ) { ( 1 exp EXS α N i = n i + n i (1) + n i (2); c i = n i, i {N, S} ; )} ; =0 c N (z) + c S (z) = A N (z) n N (z) + A S (z) n S (z); (z) 1 A N (z) ; nn (z) 0; n S (z) 0, z {1, 2} ; β ( c N + (1) c N (1) + (2) c N (2) N N ) = 0 b N 0 ; No WTO Resricions u c S (1) u c S (2) = (1). (2) If here were no WTO resricions, he allocaion chosen by he Souh governmen has o saisfy all consrains of he benchmark model above excep he las equaion ). ( uc S (1) u c S (2) = (1) (2) Therefore, he "no-wto resricions" problem drops he las consrain which implies he Souh governmen can manipulae erms of rade. Consequenially, he Souh s domesic ) relaive price of he expor good o impor good can be differen from he world price. ( p(1) p (2) 3 Calibraion This secion explains how I se parameer values of he benchmark model economy. I inerpre he Norh counry as he U.S. and he Souh counry as China. A se of parameers are adoped from relaed lieraure and he U.S. daa. The model period is one year. The discoun facor β is se a 0.96, which implies 4% real ineres rae per annum a he seady sae, and he preference parameer γ, which deermines he ineremporal elasiciy of subsiuion, is se a 2. The expendiure share on raded goods ψ is , which is he average 14

15 U.S. GDP share of raded goods secor 19 from 1991 o and he parameer η is se a so ha he expendiure share on he firs raded commodiy (ηψ) maches he average U.S. impors o GDP raio from 1991 o 2007 (0.1345) 22. The parameer κ, which deermines he Souh s human capial a he seady sae, is 0.99 in order o preven muliple soluions. 23 Boh he Norh ( N N) and he Souh s ( N S) labor is normalized o 1, and he Souh s iniial deb ( b S 0 ) is se a 0. The remaining parameers are chosen so ha he model can replicae relevan daa momens of he U.S. and China. The Souh s iniial human capial H S 0, he Norh s human capial H N, and he parameer φ, which governs he labor produciviy in he second raded commodiy s producion, are seleced so ha he model maches hree arges: (i) China s labor produciviy of is manufacuring indusry relaive o is service indusry in 1991 (0.5269) 24 ; (ii) he U.S. labor produciviy of is manufacuring indusry relaive o China s in 1991 ( ) 25 ; and (iii) he U.S. relaive labor produciviy of is exporers in 1992 (1.169) calculaed by Bernard and Jensen (1999). The parameer α, which governs he degree of learning-by-exporing, is chosen so ha he model maches he average growh rae of China s real GDP per capia relaive o he U.S. from 1991 o 2007 (0.0752) 26. The degree of learning-by-exporing under his calibraion implies ha if he Souh counry s expors increase by 10%, is produciviy rises by 11.91%. This is in line wih micro esimaes for 19 Following Sockman and Tesar (1995), he raded goods secor includes agriculural, manufacuring, mining, reail, and ransporaion secors. 20 Daa Source: BEA. 21 This is a lower bound of he firs raded commodiy s expendiure share in raded goods consumpion, because I assume ha he U.S. does no produce impored goods. A sensiiviy analysis found ha he main resuls are robus o he value of parameer η. 22 Daa Source: World Developmen Indicaors. 23 If κ is equal o one, he comparaive advanage disappears a he seady sae ( H S = κh N ), leading o muliple soluions. The parameer κ deermines he Souh s human capial a he seady sae. If we reduce he value of κ, he Souh s seady sae human capial decreases and he growh rae of is human capial declines because of he reduced difference beween he maximum benefi o human capial he Souh can receive by exporing o he more producive Norh and is curren human capial. All he qualiaive resuls are sill valid. 24 Daa Source: World Developmen Indicaors and Baniser (2005). 25 Daa Source: BEA, World Developmen Indicaors, and Baniser (2005). 26 Daa Source: Penn World Table. 15

16 China. 27 I use he same specificaion for he "no-policy" and "no-wto resricions" cases. The parameer values are summarized in Table 1. 4 Resuls This secion explains he quaniaive resuls and is organized as follows: Subsecion 4.1 explains he resuls of he benchmark model and compares hem o he observed daa paerns; Subsecions 4.2 and 4.3 presen he resuls of he "no-policy" economy and he "no-wto resricions" case in comparison wih hose of he benchmark, respecively. 4.1 Benchmark The period 0 corresponds o he year When he Souh governmen canno use an expor subsidy ( τ EX = 0 ) due o WTO resricions, he opimal policy for he Souh counry is o ax he non-raded good s consumpion and subsidize savings as shown in Figure 3. This shifs labor ino he radable secor and suppresses he Souh s overall consumpion o increase is expors. Figure 4 shows ha he Souh governmen iniially shifs labor ino he radable secor by suppressing consumpion of he non-raded commodiy. As he Souh s human capial grows hrough learning-by-exporing, i gradually raises is labor allocaion o he non-raded commodiy secor and herefore is consumpion. Figure 5 shows ha he Souh governmen suppresses consumpion of he expor good (Traded Commodiy 1) while reducing ha of he impor good (Traded Commodiy 2) for he iniial periods. This raises he Souh s expors, leading is human capial and real GDP o grow rapidly hrough learning-by-exporing as shown in Figures 4 and 6. The ransiion o he seady sae akes 112 periods, during which he Norh produces boh raded goods and he Souh produces he firs raded commodiy For insance, Park, Yang, Shi, and Jiang (2010) show ha if a Chinese firm experiences an exogenous 10% increase in expors, is produciviy rises by 11% o 13% in China. 28 As I describe in Subsecions 4.2, 4.3 and 5.2, he Souh produces he second raded commodiy during he laer par of he ransiion in he "No Policy", "No-WTO Resricions", and "Coordinaed Policy" cases. This is because iniially he Souh runs a subsanial curren accoun defici. In order o repay he 16

17 The iniial paern of expors and impors of he Souh causes he counry o run a curren accoun surplus, accompanied by rapid growh in is real GDP as shown in Figure This resul shows ha he gains from disoring savings o encourage expors are even larger han he incenive o smooh consumpion. During he ransiional phase, he Souh s erms of rade say consan. Noe ha when one counry produces boh raded goods, he erms of rade are deermined by is produciviy raio beween he wo raded goods producion. Since he Norh produces boh raded goods for all periods, he Souh s erms of rade are ( p equal o he Norh s produciviy raio beween wo raded goods (1) = (HN ) 1+φ p (2) ). This H N is equal o he U.S. relaive labor produciviy of exporers in 1992 (1.169). The consan Souh s erms of rade makes he Souh s real exchange rae appreciae as is human capial grows. This is because he Souh s real exchange rae is a funcion of is erms of rade and ( ) )} ) 1 ψ relaive produciviy in he raded commodiy s producion e S. {( p(1) p (2) ( A S (1) A N (2) Simulaneous growh and real exchange rae appreciaion is consisen wih he Balassa (1964)-Samuelson (1964) hypohesis 30. Table 2 summarizes he average values of key aggregaes of U.S. and China from 1991 o 2007 and heir counerpars in he benchmark model. The model is calibraed o mach he average growh rae of China s real GDP per capia relaive o he U.S. (7.52%). As shown in Table 2, he Souh s policies generae he simulaneous fas growh and curren accoun surpluses observed in he daa. 31 In addiion, he benchmark model replicaes boh China s ineres on is deb, he Souh shifs more workers from non-radable secor o boh radable secors so ha i runs a rade surplus for he res of he ransiion periods. 29 As can be seen in Figure 2, China s curren accoun surplus has increased over ime. However, in his model, he Souh s curren accoun surplus is decreasing over ime. The Chinese governmen could have gradually implemened policies o ake advanage of learning-by-exporing. However, since his is a perfec foresigh model, he Souh does no gradually implemen policies. In order o mach he rend, I should inroduce some fricions like adjusmen coss ino his model. 30 Balassa (1964) and Samuelson (1964) argue ha economic growh driven by produciviy gains in he raded goods secor should accompany a real exchange rae appreciaion. 31 This model oversaes China s curren accoun surplus and undersaes he U.S. curren accoun defici. However, wha is imporan is ha he model can qualiaively replicae he sign of curren accouns of boh counries. In realiy, boh counries have oher rading parners han each oher. The discrepancy beween curren accoun generaed by he model and is daa counerpar may come from each counry s rade wih 17

18 expor over GDP and he appreciaion of is real exchange rae as in he daa. 4.2 No Policy Counerfacual In he "no-policy" economy, boh firms and workers know ha he Souh s human capial will grow over ime bu no one recognizes is law of moion. Therefore, he Souh governmen has no incenive o raise he Souh s expors in order o ake advanage of learning-by-exporing. If no policies were implemened in boh he Norh and Souh counries, he ransiion o he seady sae akes 118 periods as shown in Figure 7. This is 6 periods longer han ha of he benchmark economy because no one implemens policies o accelerae he Souh s growh hrough learning-by-exporing in he "no-policy" economy. During he ransiion, he paerns of specializaion in producion undergo hree sages: (i) he Norh produces boh raded goods and he Souh produces he firs raded commodiy for he firs 96 periods; (ii) boh counries are compleely specialized in period 97; and (iii) he Souh sars o produce he second raded commodiy, in addiion o he firs raded commodiy, in period 98. Figure 7 shows ha more Souh workers produce in he non-raded commodiy secor for he iniial 85 periods relaive o he benchmark case, because he Souh s labor produciviy in he firs raded commodiy s producion is much less han ha in he non-raded commodiy s producion over he same periods. This iniially suppresses he Souh s firs raded commodiy s expors alhough is consumpion for he firs raded commodiy is less han in he benchmark economy as shown in Figure 8, delaying he ake-off of is human capial relaive o he benchmark case. Since he Souh workers know ha heir income will grow in he fuure, hey wan o raise curren consumpion. The Souh consumes more non-raded goods in he firs 85 periods han in he benchmark economy, leading o a larger aggregae consumpion. This makes he Souh s curren accoun defici 32 increase for he same periods. The Souh workers move from he non-radable secor o boh radable secors for he res of he ransiion periods he res of he world. 32 The size of he curren accoun deficis is implausible. This is caused by he full commimen and perfec foresigh assumpions. 18

19 so ha he Souh runs a rade surplus in order o repay he ineres on is deb. 33 Since more workers produce in boh raded commodiy secors in he Souh counry relaive o he benchmark economy beginning in period 86, he level of Souh real GDP becomes greaer han ha in he benchmark case as Figure 9 presens. This is because labor produciviy in he raded goods producion is higher han ha in non-raded commodiy producion over he same period. Figure 9 shows ha he Souh erms of rade sar o deeriorae in period 97 when boh counries are compleely specialized. 34 Beginning in period 98 when he Souh produces boh raded goods, is erms of rade, which are equal o he Souh ( ) (H S produciviy raio ) 1+φ, improve as is human capial grows. For he same period, H S he real exchange rae appreciaes following he Souh s human capial growh. 4.3 No WTO Resricions If here were no WTO resricions ( τ EX 0 ), he Souh counry can direcly subsidize expors. Figure 10 shows ha he ransiion o he seady sae akes 71 periods bu mos of he cach-up akes place in he firs 50 periods. During ransiion, he Norh produces boh raded goods and he Souh produces he firs raded commodiy for he firs 26 periods. As he Souh s human capial grows, i gradually expands he world marke share of he firs raded commodiy. In period 27, he Souh compleely akes over he marke for he firs raded commodiy, which leads o complee specializaion of boh counries unil period 35. The Souh sars o produce he second raded commodiy in period 36. As shown in he lef panel of Figure 11, he Souh governmen suppresses consumpion of he expor good (Traded Commodiy 1) o a larger degree during he ransiion relaive o benchmark oucomes. This raises he Souh s expors, making is human capial grow a an even faser rae hrough learning-by-exporing relaive o he benchmark economy. Figure 10 shows ha, compared o he benchmark, he Souh governmen shifs more workers from he non-raded commodiy secor o boh raded goods secors. Since labor produciviy in boh 33 The righ panel of Figure 8 shows ha he Souh expors even he second raded commodiy from period 100 despie a comparaive disadvanage. 34 The Souh erms of rade in he benchmark model is normalized o 1 in Figures 9, 12, and

20 raded goods producion is higher han ha in he non-raded commodiy s producion, he level of Souh real GDP in "no-wto resricions" case is greaer han ha in he benchmark economy beginning in period 11. The righ panel of Figure 11 shows ha he Souh subsiues he alernaive impors by raising consumpion of he impor good (Traded Commodiy 2) subsanially for he iniial periods. As he Souh accumulaes human capial hrough learning-by-exporing, is impors decline o a level which is even below ha in he benchmark afer period 36 when he Souh sars o produce he second raded commodiy. The Souh s increased impors iniially cause he counry o run a large curren accoun defici. As he Souh s impors decline, he curren accoun defici also goes down and ulimaely becomes balanced in he seady sae. The Souh governmen s expor subsidy ) generaes a deerioraion in is erms of rade beginning in period 27 when boh ( p(1) p (2) counries sar o be compleely specialized, as shown in he lower panel of Figure 12. When he Norh produces boh raded goods for he iniial 26 periods, he Souh s erms of rade ( (HN ) 1+φ are equal o he Norh s produciviy raio beween wo raded goods ), which H N is ime-invarian. When he Souh produces boh raded goods beginning in period 36, is ( ) (H S erms of rade are equivalen o is produciviy raio ) 1+φ, hus boh rise wih an increase in he Souh s human capial. Figure 13 shows ha if here were no WTO resricions, he Souh counry will use an expor subsidy for he iniial 35 periods insead of a non-raded commodiy consumpion ax, and i no longer relies heavily on he savings subsidy o promoe expors. The swich of sign in he curren accoun from he benchmark o he "no-wto resricions" case implies ha, ideally, he Souh would like o manipulae is erms of rade raher han is curren accoun. However, if he abiliy o explicily subsidize expors is absen due o WTO resricions, i mus "over" disor boh he ineremporal margin and he non-raded margin. H S 20

21 5 Coordinaed Policy Problem In his secion, I consider a "coordinaed policy" problem in which he Norh and Souh could coordinae policies in order o obain he firs-bes oucome for he world. I assume ha here is a ficiious world planner who maximizes he weighed average of boh he Norh and Souh uiliies by aking advanage of he Souh counry s learning-by-exporing. The world planner solves max β { µu(c S ) + (1 µ) u(c N ) }, =0 subjec o H S +1 = H S + ( κh N H S ) { ( 1 exp EXS α c i = n i ; N i = n i + n i (1) + n i (2), i {N, S} ; )} ; c S (z) + c N (z) = A S (z) n S (z) + A N (z) n N (z); n S (z) 0; n N (z) 0, z {1, 2}, where µ [0, 1] is he Souh counry s Pareo weigh. If he Souh counry expors he raded commodiy z {1, 2}, he world uiliy maximizing behavior of he planner implies he following firs-order condiions: µu c S (z) λ ( ) κh N H S ( ) exp EXS = (1 µ) u α α c N (z); (2) u c S (z) A S (z) = u c S, (3) where λ is a muliplier for he law of moion for Souh human capial, u c i (z) is he counry i s marginal uiliy of consumpion for he raded commodiy z, and u c i is he counry i s marginal uiliy of consumpion for is non-raded commodiy. Noe ha he second ( ) erm in he lef side of he condiion (2), λ(κhn H S ) exp EXS, which appears due o α α learning-by-exporing, is posiive. This implies ha when he Souh counry expors he raded commodiy z, he world planner reduces he Souh counry s consumpion for he exporing good c S (z) in order o ake advanage of learning-by-exporing. Condiion (3) 21

22 shows ha here is no disorion beween he Souh s consumpion for he expor good and ha of non-raded good. If a worker shifs from he non-raded commodiy secor o he raded commodiy z secor in he Souh counry, his reduces one uni of he non-raded commodiy. Thus, he welfare loss is he marginal uiliy of consumpion for he non-raded commodiy. However, he worker produces A S (z) unis of he raded commodiy z. By consuming hese unis, he worker can enjoy heir marginal uiliy of consumpion for he commodiy imes heir marginal produc ( A S (z) ). Since he Souh counry does no expor he raded commodiy z bu consumes i, here is no addiional welfare gain from learning-byexporing. 35 The condiions (2) and (3) imply ha he planner decreases no only he Souh counry s consumpion for he expor good bu also ha for non-raded good. This means ha he planner raises he Souh s expors by reducing is consumpion of domesically produced goods and increasing consumpion of he impor good. 5.1 Decenralizaion In his subsecion, I explain he way I find prices and wedges, which imply he firs-bes allocaion for he world. The Norh counry s non-raded commodiy price is normalized o one. Since he Norh s relaive consumpion across goods is undisored, I use heir marginal rae of subsiuion beween non-raded and each respecive raded good as world prices. Tha is, he world price p (z) of raded commodiy z is defined by p (z) (z), z {1, 2}, where (z) is he Norh counry s marginal uiliy of consumpion for he raded commodiy z and is he Norh s marginal uiliy of consumpion for is non-raded commodiy. The 35 Noe ha a non-raded goods consumpion ax is needed in he presence of WTO resricions. If a worker shifs from he non-raded commodiy secor o he expor commodiy secor in he Souh counry, his reduces one uni of he non-raded commodiy. Thus, he welfare loss is marginal uiliy of consumpion for he non-raded commodiy. However, he worker produces he marginal produc of he expor commodiy, and consumes par of ha while exporing he remaining par o increase heir consumpion of he impor commodiy. This way, he erms of rade are kep consan. Since he Souh counry sill expors some of he addiional expor commodiy, here is addiional welfare gain from learning-by-exporing. 22

23 world ineres rae r +1 is defined by r +1 β Since he Souh counry has a comparaive advanage in he firs raded commodiy s producion, i produces he firs raded commodiy and he Norh produces he second raded commodiy. Therefore, he Souh s domesic wage is defined by w S ( ) p (1) p (2) ( ) A S (1) = p S A N. (2) A wedge τ r +1 in he Souh counry s domesic ineres rae, a wedge τ EX domesic relaive price of he expor good o impor good, and a wedge τ NT domesic relaive price of he expor good o non-raded good are defined by τ r +1 w S +1 u c S r +1 w S β u c S +1 1 r +1 1 τ EX p (2) u c S (1) p (1) u c S (2) τ NT 5.2 Resuls u c S A S (1) u c S (1) 1 u c S (1) u c S (2) = 1 u c S (1) u c S = in he Souh s in he Souh s ws +1 u cs 1 = ( ) 1 + τ r w S +1 r+1 ; β u c S ( ) τ EX p (1) ; p (2) p (1) (1 + τ NT ) p S = 1 (1 + τ NT ) A S (1). I use he same parameers for he "coordinaed policy" case as in he benchmark economy, excep for he Souh Pareo weigh, µ = , which is chosen so ha he model maches he balanced seady sae curren accoun. 36 When boh he Norh and Souh coordinae policies o achieve he world bes allocaion, he ransiion o he seady sae akes 64 periods as Figure 14 presens. This is 7 periods less han ha of "no-wto resricions" economy because he world planner faciliaes growh hrough erms of rade disorion even more. During he ransiion, he paerns of specializaion in producion undergo hree sages, as before: (i) he Norh produces boh raded goods and he Souh produces he firs raded commodiy for he firs 25 periods; (ii) 36 If µ is greaer han , he Souh runs a curren accoun defici a he seady sae. If µ is less han , i runs a curren accoun surplus a he seady sae. 23

24 boh counries are compleely specialized from period 26 o 35; and (iii) he Souh sars o produce he second raded commodiy in period 36. The firs sage ges shorer relaive o he "no-wto resricions" case. The lef panel of Figure 15 shows ha he world planner reduces he Souh s consumpion of he expor good (Traded Commodiy 1) even more from period 16 han in he "no-wto resricions" economy. This increases he Souh s expors, making is human capial grow more rapidly hrough learning-by-exporing han in he "no-wto resricions" case. As can be seen in Figure 14, he world planner moves more workers from he non-raded commodiy secor o boh raded commodiy secors in he Souh counry relaive o he "no-wto resricions" economy beginning in period 16. This makes he level of Souh real GDP greaer han ha in he "no-wto resricions" case over he same period because labor produciviy in he raded goods producion is higher han ha in he non-raded commodiy s producion. The righ panel of Figure 15 shows ha, for iniial periods, he world planner raises he Souh s impors by 42% of is GDP relaive o he "no-wto resricions" world by increasing is consumpion of he impor good (Traded Commodiy 2). This difference in he Souh s impors beween "coordinaed policy" and "no-wto resricions" economies iniially leads o a larger curren accoun defici (42% of he Souh GDP) relaive o he "no-wto resricions" case. Figure 16 shows ha beginning in period 26, when boh counries sar o be compleely specialized, he increase of he Souh s expors deeriorae is erms of rade much more han in he "no-wto ) resricions" ( p(1) p (2) economy. This large deerioraion of he Souh s erms of rade causes is real exchange rae o be undervalued from period 26 o 35, as shown in Figure 16. Over he same period, he Souh counry s real GDP grows rapidly due o he fas growh of is human capial, which implies a rapid produciviy improvemen in raded goods secor. Growh accompanied wih real exchange rae undervaluaion conrass wih he predicion of he Balassa (1964)-Samuelson (1964) hypohesis. When he Souh produces boh raded goods beginning in period 36, is erms of rade (1) = (HS ) 1+φ ( ) p p (2) improve due o he growh of Souh H S human capial. For he same period, he real exchange rae appreciaes following he real 24

25 GDP growh. This implies ha he world planner pospones he Balassa (1964)-Samuelson (1964) effec by deerioraing he Souh erms of rade. Figure 17 shows ha he world planner uses a bigger expor subsidy and a less saving subsidy relaive o he "no-wto resricions" case. This implies ha he world planner calls for more erms of rade manipulaion han suggesed by No-WTO policies. 6 Welfare Analysis This paper explores opimal policies in he presence of learning-by-exporing in various environmens: wih WTO resricions, wih no resricions, and under policy coordinaion. An ineresing quesion o ask is wha implicaions hese policies have for developing and developed counries welfare. In order o answer his quesion, I measure he welfare changes resuling from moving from he benchmark economy wih WTO resricions o an alernaive case. This is accomplished by finding he percenage change in per-period consumpion ha I should give o a worker in each counry in he benchmark such ha he worker is indifferen beween he wo environmens. In he "no-policy" economy, he Souh counry is no aware of he posiive exernaliy from learning-by-exporing, hus no policy is implemened o faciliae expor-led growh. On he oher hand, in he benchmark model, he Souh governmen recognizes learning-byexporing and has an incenive o implemen policies o increase expors. Since i canno direcly subsidize expors due o WTO resricions, i alernaively axes he non-raded good s consumpion and subsidizes savings. This policy enables he Souh counry o grow faser, benefiing is workers, han in he "no-policy" world. Conversely, moving from he "no-policy" world o he benchmark economy slighly decreases he welfare of he Norh counry. As can be seen in Table 3, moving from he benchmark o a "no-policy" economy resuls in welfare changes equivalen o a 19.12% decline and 0.06% increase in per-period consumpion of he Souh and he Norh, respecively. If he Souh counry is allowed o manipulae is erms of rade ("no-wto resricions" 25

26 case), boh he Norh and Souh benefi from welfare improvemen relaive o he benchmark economy. Wihou resricions on policies, he Souh counry subsidizes expors o boh reduce is consumpion of he expor good and increase consumpion of he impor good. This policy generaes a large deerioraion in he Souh s erms of rade and a curren accoun defici. This policy makes he Souh grow faser wihou raising savings heavily o promoe expors, which improves he welfare of he Souh relaive o he benchmark economy. In he "no-wto resricions" world, he Norh s welfare also increases due o improved erms of rade compared wih he benchmark economy. This "win-win" oucome hrough erms of rade manipulaion is refleced in he posiive welfare gains of 14.05% and 0.45% in perperiod consumpion of he Souh and Norh, respecively. 37 This is conrary o Bagwell and Saiger (1999) s view ha he WTO improves world welfare by prevening zero-sum erms of rade manipulaion. 38 As shown in Table 3, if boh counries coordinae policies o achieve he firs-bes allocaion, boh counries welfare rises significanly compared wih he benchmark economy. The ficiious world planner manipulaes he erms of rade of boh counries so ha he Souh grows faser hrough learning-by-exporing. This leads o increased welfare gains equivalen o 13.16% and 1.22% in per-period consumpion of he Souh and Norh, respecively. However, moving from he "no-wto resricions" economy o he "coordinaed policy" world, he Norh is beer off whereas he Souh is worse off. The world planner pulls down he relaive price of he Souh s expor good o is impor good even furher han he "no-wto resricions" economy. Even hough his makes he Souh counry grow faser, he larger deerioraion of he Souh s erms of rade hurs is welfare. On he oher hand, he Norh benefis from is improved erms of rade. However, he welfare changes of boh counries 37 Alhough he Souh counry is a small open economy ha canno affec world prices, he Souh governmen can manipulae is inernal erms of rade (he relaive price of is exporing goods o imporing goods in he Souh domesic markes) and grow rapidly hrough learning-by-exporing. However, he Norh s welfare does no increase because is erms of rade are no changed by he Souh s economic policies. 38 There may be oher reasons ha WTO resricions improve he welfare of developing counries. For insance, Bajona and Chu (2010) claim ha WTO resricions increase China s welfare by reducing subsidies o he ineffi cien sae-owned secor. 26