WORKING PAPER SERIES ON REGIONAL ECONOMIC INTEGRATION NO. 20. Welfare Implications of International Financial Integration.

Transcription

1 WORKING PAPER SERIES ON REGIONAL ECONOMIC INTEGRATION NO. 20 Welfare Implicaions of Inernaional Financial Inegraion November 2008 Jong-Wha Lee and Kwanho Shin

2

3 Welfare Implicaions of Inernaional Financial Inegraion Jong-Wha Lee + and Kwanho Shin ++ November 2008 We hank Donghyun Park for helpful suggesions. The views expressed in he paper are he auhors and do no necessarily reflec he views and policies of he Asian Developmen Bank (ADB) or is Board of Direcors, or he governmens ha ADB s members represen. + Head, Office of Regional Economic Inegraion, Asian Developmen Bank, 6 ADB Avenue, Mandaluyong Ciy, 1550 Mero Manila, Philippines. Tel: , Fax: , jwlee@adb.org. ++ Deparmen of Economics, Korea Universiy. khshin@korea. ac.kr.

4 The ADB Working Paper Series on Regional Economic Inegraion focuses on opics relaing o regional cooperaion and inegraion in he areas of infrasrucure and sofware, rade and invesmen, money and finance, and regional public goods. The Series is a quick-disseminaing, informal publicaion ha seeks o provide informaion, generae discussion, and elici commens. Working papers published under his Series may subsequenly be published elsewhere. Disclaimer: The views expressed in his paper are hose of he auhor and do no necessarily reflec he views and policies of he Asian Developmen Bank or is Board of Governors or he governmens hey represen. The Asian Developmen Bank does no guaranee he accuracy of he daa included in his publicaion and acceps no responsibiliy for any consequence of heir use. Use of he erm counry does no imply any judgmen by he auhors or he Asian Developmen Bank as o he legal or oher saus of any erriorial eniy. Unless oherwise noed, $ refers o US dollars by Asian Developmen Bank November 2008 Publicaion Sock No.

5 Conens Absrac 1 I. Inroducion 2 II. FDI and Technology Diffusion 3 A. Lieraure Review 3 B. Evoluion of FDI Flows 4 III. A Model of Economic Growh wih Technology Diffusion 5 A. Growh of a Reference Economy 5 B. Growh of a Developing Counry: The Case of a Closed Economy 8 C. Opening Capial Marke: Capial Inflows and Technology Cach-up 10 IV. Empirical Tess of he Effec of FDI on Technology Cach-up 12 A. Empirical Specificaion 12 B. Esimaion Resuls 14 C. Insrumenal Variables Esimaion 14 D. Esimaion wih FDI Socks 15 E. Checks for Robusness and Addiional Exercises (Summary of Findings Tha Are No Repored in he Tables) 16 V. Esimaion of he Welfare Effecs of Financial Inegraion 16 VI. Conclusion 20 References 21 ADB Working Paper Series on Regional Economic Inegraion 33 Tables 1. Descripive Saisics FDI and Rae of Growh of TFP FDI Socks and TFP Growh Rae: Fixed Effecs OLS and Insrumenal Variables Esimaion Welfare Implicaions of Financial Inegraion 32 Figures 1. The FDI Inflows (% of GDP) The FDI Inflows (% of GDP) by Income Level FDI Inflows (% of GDP) in Eas Asia Technology Cach-up 27 A. FDI Consiues 40% of Capial Inflows 27 B. FDI Consiues 100% of Capial Inflows 28

6 Absrac Focusing on echnology spillover from foreign direc invesmen (FDI) inflows, his paper invesigaes he welfare implicaions of financial inegraion. Calibraions of a neoclassical growh model wih inernaional echnology diffusion show ha when echnology cach-up due o FDI inflows is considered, he welfare gains from financial inegraion subsanially increase, which conrass wih he small gains from addiional, capial-accumulaion effecs of financial inegraion. The esimaes sugges ha by furher enhancing financial inegraion, emerging Asian economies, such as he People s Republic of China (PRC) and he larges four Associaion of Souheas Asian Naions (ASEAN) counries, will experience subsanial welfare gains. Keywords: Foreign direc invesmen, financial inegraion, echnology diffusion. JEL Classificaions: F21, F36, O33.

7 I. Inroducion As emerging economies have deregulaed heir financial markes over recen decades, here has been a rapid increase in cross-border capial ransacions. This coninuous financial opening has lef emerging marke economies more financially inegraed wih global markes. Despie he rapid financial developmens in emerging markes, few researchers have invesigaed he welfare implicaions of financial inegraion in he regions. This sudy aemps o fill his gap by esimaing he welfare gains from financial inegraion in emerging Asian economies. In general, here have been wo approaches in he lieraure aemping o quanify he welfare implicaions of financial inegraion. The firs approach focuses on he manner by which financial inegraion enables a pooling of risks across counries. According o his approach, welfare gains are measured by he degree of consumpion risk shared hrough financial inegraion (Cole and Obsfeld, 1991; Backus e al., 1992). The second approach emphasizes he producion side and measures welfare gains by he degree of acceleraion in growh due o foreign capial (Gourinchas and Jeanne, 2006). This approach shows in paricular ha financial inegraion speeds up capial accumulaion hrough he inflow of foreign capial. The welfare gains quanified by boh approaches are generally small. The gains, repored as he permanen percenage increase in expeced consumpion equivalen o he increase in welfare gains hrough financial inegraion, vary from eiher less han 0.1% or up o 1%. Repored gains are small largely because he approaches focus mainly on he direc impac of financial inegraion, such as he sharing of consumpion risk and capial accumulaion. However, he welfare gains can be larger if we include oher, indirec ways hrough which financial inegraion conribues o economic growh. In paricular, foreign direc invesmen is considered an imporan channel for produciviy growh hrough inernaional echnology diffusion. Recen empirical sudies show ha dispariies in he level of echnology and no in human or physical capial explain he bulk of he difference in income or oupu per worker across economies (Hall and Jones, 1999; Easerly and Levine, 2001). Theoreical and empirical sudies also show ha echnology diffusion plays an imporan role for echnologically lagging economies o cach up wih he global echnology fronier. In paricular, i is believed ha inward foreign direc invesmen (FDI) is an imporan channel for he diffusion of echnology across counries. Based on a simple calibraion exercise, Hoxha and Kalemli-Ozcan (2007) recenly showed ha if echnological improvemen is also aken ino consideraion, he impac of capial inegraion on welfare is abou wice as large as oherwise. The main goals of our sudy are wofold. Firs, by using aggregae measures of produciviy, we empirically es if FDI flows narrow he echnology gap relaive o he echnologically advanced counry. Second, by using our empirical esimaes, we documen he exen o which welfare in Asia is enhanced by financial inegraion. By assuming ha financial inegraion leads o welfare improvemen no only hrough he direc effec of he rapid accumulaion of capial bu also hrough he indirec effec of echnology diffusion he calibraion exercise is deeply rooed in our empirical findings. While Hoxha and Kalemli-Ozcan s calibraion is based on he findings of microeconomic sudies on he impac of FDI on he produciviy of firms, our calibraion exercise is rooed in he empirical specificaions used in our paper, which direcly capure he exen of echnology cach-up of a developing counry wih he fronier counry. 2

8 The remainder of he paper is organized as follows. In secion 2, we briefly review he lieraure on he relaion beween FDI and echnology diffusion and illusrae he main feaures of he evoluion of FDI flows in he daa. Secion 3 inroduces a model of echnology diffusion and secion 4 empirically ess he effec of FDI on echnology cach-up in Asia. Based on he empirical findings in secion 4, we calibrae a model in secion 5 o invesigae he welfare implicaions of financial inegraion in Asia. Concluding remarks are in secion 6. II. FDI and Technology Diffusion A. Lieraure Review An economy s echnological advancemen depends on innovaions of is own echnology and he adopion of echnologies produced elsewhere. For counries low on he echnology ladder, he imiaion and adopion of new echnologies provide opporuniy for caching up wih echnologically more advanced counries. This advanage of backwardness (Gerschenkron, 1962) implies ha he larger he gap in he echnology level relaive o he advanced counry, he faser he lagging economy can cach up wih he global echnology fronier. Recen growh models, such as hose of Grossman and Helpman (1991, Chapers 11 and 12), and Barro and Sala-i-Marin (2004, chaper 8), incorporae he role of echnology diffusion in economic growh. While echnology can diffuse hrough a variey of channels, FDI by mulinaional corporaions (MNCs) is ofen regarded as one of he mos imporan channels for inernaional echnology diffusion. MNCs hemselves accoun for a subsanial par of he world s research and developmen (R&D) invesmen, and can play a major role in he diffusion of new echnology. Recen work on economic growh highlighs he role of FDI in he echnology diffusion process. Findlay (1978) posulaes ha FDI increases he rae of echnical progress in he hos counry hrough a conagion effec from he more advanced echnology, managemen pracices, and oher hings ec., used by he foreign firms. Wang (1990) incorporaes his idea ino a neoclassical growh framework by modeling he increase in knowledge as a funcion of FDI. Borenszein, De Gregorio, and Lee (1998) provide a simple, endogenous growh model in which he more advanced knowledge embodied in FDI allows hos counries o inroduce new varieies of capial goods a lower cos, hereby leading o a higher rae of echnology progress. The adopion of advanced echnologies by less-advanced counries is no free and requires effor and capabiliy. Abramoviz (1986) uses he phrase absorpive capaciy o denoe domesic capabiliy for absorbing foreign echnology spillover. In paricular, he lack of human capaciy for adoping new echnologies is considered a crucial facor ha limis he absorpive capabiliy of a naion. Nelson and Phelps (1966) consruc a model in which he faciliaion of new knowledge is only possible when a sufficien level of human capial is presen in a developing counry. Many recen models also highligh he complemenary effecs beween human capial and echnology, as boh human capial and echnology invesmen are endogenous choices of sociey. Redding (1996), for insance, assumes ha boh forms of invesmens in human capial and echnology (R&D) exhibi pecuniary exernaliies and are sraegic complemens. In his model, he incenives o inves in each are inerdependen and, hus, muliple equilibria exis: an economy can eiher fall ino a low-educaion, low-echnology posiion or achieve a higheducaion, high-echnology posiion. Acemoglu and Ziliboi (2001) poin ou ha echnologies invened in advanced counries are more skill-complemenary. Hence, he mismach beween skills and echnology leads o differences in produciviy even when all counries have equal 3

9 access o new echnologies. A subsanial number of recen papers empirically invesigae echnology spillover from he presence of foreign firms a he firm, indusry or economy levels (see he survey of Lim, 2001). The evidence based on firm- or indusry-level daa is no ye fully conclusive. Blomsrom (1986) and Kokko (1994) find a posiive effec of FDI on produciviy a he secoral level in Mexico. In conras, based on firm-level daa from Venezuela, Aiken and Harrison (1999) find no evidence of a posiive spillover from foreign o domesic firms. Similarly, Djankov and Hoekman (2000) find no significan echnology spillover beween foreign and domesic firms in he Czech Republic. While hese sudies focus on he effec of FDI on horizonal spillover, ha is, he effec mulinaional firms have on domesic firms in he same secor, recen sudies examine verical spillover from he presence of mulinaionals o heir local suppliers or cusomers. For Indonesian and Lihuanian firms, Blalock and Gerler (2004) and Javorcik (2004) find evidence of posiive FDI spillover hrough backward links, ha is, spillover from he presence of mulinaionals in upsream secors o domesic suppliers of inermediae inpus. A he economy-wide level, Blomsrom, Lipsey, and Zejan (1994) find FDI has a significan posiive effec on growh, bu he effec is confined o high-income developing counries. Bu Blonigen and Wang (2004) provide conrasing evidence ha FDI has a significanly posiive effec only in less-developed counries. Some sudies find evidence ha o have posiive echnology spillover hrough FDI he hos counry mus have an absorpive capaciy for new echnologies ha manifes in FDI. Using a panel daase over 69 developing counries beween 1970 and 1989, Borenszein e al. (1998) show ha FDI conribues o produciviy growh when he hos economy has a minimum hreshold sock of educaed workers as an absorpive capaciy for FDI echnology spillover. Likewise, Xu (2000), using daa on he affiliaes of. mulinaionals in he Unied Saes (US), finds posiive spillover from US mulinaionals on produciviy growh in hos counries ha have a minimum sock of human capial. Subsequen empirical sudies also esablish he imporance of various means of domesic absorpive capaciies, oher han human capial, for echnology spillover from FDI. Durham (2002) and Alfaro, Chandra, Kalemli-Ozcan, and Sayek (2004) find ha for a broad cross-secion of counries, financial or insiuional developmen in hos economies of FDI plays an imporan role in absorpive capaciy for echnology spillover hrough FDI and economic growh. B. Evoluion of FDI Flows Figure 1 illusraes ne FDI inflows as percenage of GDP for he world, Asia, and Lain America. The FDI daa are colleced from he World Developmen Indicaor (WDI). The WDI repors FDI daa for en Eas Asian economies: China; Hong Kong; Indonesia; Japan; Korea; Malaysia; Philippines;Singapore;Taipei,China;and Thailand. The WDI s FDI daa for Lain America also includes 31 counries. We use simple averages o represen each region. Unil 1985, he world average of ne FDI inflows did no vary much, being around 1 2% of GDP. Afer 1985, i sared o increase, reaching abou 4% in 1992 and, following he European Exchange Rae Mechanism (ERM) crisis, plummeed o below 2% in FDI inflows sared o increase again and exploded in he mid-1990s. The world average reached over 7% in 2002 and remained over 6% in

10 The regional averages of Eas Asia and Lain America also rack he world average quie closely, excep during regional crises. When a crisis hi eiher of he regions, FDI inflows slowed, bu evenually bounced back. Figure 2 illusraes ne FDI inflows as a percenage of GDP for subgroups classified by level of income. The WDI classified counries ino four groups based on he per-capia level of income: low (50 counries); low-middle (50 counries); upper-middle (35 counries); and high (34 counries). I is ineresing o noe ha while FDI inflows generally expanded around he mid-1990s, he increase for high-income counries was markedly greaer han for oher counries. Figure 3 shows FDI inflows as a percenage of GDP for each Eas Asian economy. The hree bars represen average values for he 1980s, 1990s, and 2000s, respecively. FDI inflows o Hong Kong, China and Singapore, which ranged beween 10 and 16%, by far surpass hose of oher counries. The PRC, Malaysia, and Thailand showed on average, FDI inflows of 2 4 %. In conras, FDI inflows o Indonesia, Japan, Korea, and he Philippines were much lower and less han 2% hroughou he whole period. III. A Model of Economic Growh wih Technology Diffusion We inroduce a neoclassical growh model wih inernaional echnology diffusion. We measure he dispariies in he level of echnology across economies and see how hey evolve over ime when FDI inflows from an advanced economy have an impac on he echnology cach-up of a developing economy. A. Growh of a Reference Economy We use a growh-accouning framework o decompose he level of oupu per worker ino he levels of inpus and labor-augmened produciviy. In his framework, he labor augmened produciviy measures he level of echnology. We assume a Cobb-Douglas producion funcion, ha is, Y = K 1 α ( AhL) α (1) In Eq. (1), K denoes he sock of physical capial, h is he amoun of human capial per labor uni, L is he number of workers, and A denoes a measure of labor-augmened produciviy. We assume ha each worker inelasically supplies one uni of labor. The producion funcion is rewrien in erms of he oupu per worker, y=y/l, as: y = k 1 α ( Ah) α (2) In Eq. (2), k (=K/L) is he physical capial per worker. Le us assume ha he labor share, α, is consan across counries. Now, we disinguish beween he reference counry, which is denoed by an aserisk, and a represenaive developing counry by a subscrip i. Then, he raio of he oupu per worker beween counry i and he reference counry is wrien as: 5

11 y / y = ( k i i / k) 1 α ( h i / h) α ( A / A) i α (3) Eq. (3) allows us o decompose he differences in he GDP per worker beween he wo counries ino differences in: he capial-labor raio; human capial, and labor-augmened produciviy. By aking logarihms of boh sides of Eq. (3), he gap in oupu per worker beween counry i and counry is wrien as an addiive sum of hree componens, as shown in Eq. (4). ln( y i / y ) = (1 α )ln( k / k ) + α ln( h / h ) + α ln( A / A ) ] (4) i Human capial per worker is assumed o be relaed o years of schooling as follows. 1 h i φ (E) = e (5) In Eq. (5), φ(e) measures he efficiency of a uni of labor wih E years of educaion relaive o one wih no schooling. The derivaive, φ (E), is he marginal reurn from an addiional year of schooling. To invesigae he dynamics of he reference economy, we explicily inroduce he ime dimension in he subscrip as follows. Y 1 α α = K ( A L h ) (6) i Now, we assume ha labor grows a an annual rae n and he annual labor-augmened produciviy growh is x. In oher words, L L e n = 0 (7) and x A = A e (8) 0 The producion funcion can be rewrien as: Y (1 α ) ( n+ x ) α α ( A0 L0e ) h = K (9) To work wih variables ha are consan in he long run, we define: ~ Z z = (10) ( n+ x ) e In Eq. (10), he symbol ~ denoes ha he variable is being measured in erms of efficiency 1 See Klenow and Rodriguez-Clare (1997), and Hall and Jones (1999). 6

12 unis. We normalize he iniial labor and level of echnology o one, ha is, A 1 and L 1. For simpliciy, for he momen, we assume ha human capial per worker, h, is fixed over ime and normalized o one, ha is, h = 1. Then, he producion funcion Eq. (1) becomes: 0 = 0 = y = (11) ~ ~ 1 α k In his economy, he sock of capial in erms of efficiency unis evolves by new invesmen less he porions ha are due o: populaion growh (a a rae n); he increasing efficiency of labor (a a rae x); and physical depreciaion (a a raeδ ). In oher words, ( n + x ) k ~ & k + ~ ~ = i δ (12) We assume ha his economy is self-financed (auarky) and already has reached he seady sae. Then, an equilibrium in his economy is characerized by a balanced-growh pah in which oupu, consumpion, and per-capia capial grow a a rae x. Assuming a CRRA uiliy funcion wih he relaive risk-aversion coefficien, γ, we can show ha, ~ from an opimaliy condiion for consumers, he consumpion per efficiency-uni, c, saisfies he following Euler equaion: c~ ( (1 )) 1/ γ (1 ) ~ = + r + x c + 1 β (13) In Eq. (13), β is he discoun facor and r is he real ineres rae a ime. Hence, a seady sae, he real ineres is derived as follows. r = (1 + x ) γ / β (14) The profi-maximizaion condiion of he firm implies ha he real ineres also should be equal o he marginal produciviy of capial less he depreciaion rae. ( α α ) r = 1 ) k δ (15) ( In seady sae, since he capial per efficiency-uni is consan, we ascerain he seady-sae capial per efficiency-uni as follows. ~ α 1 1 / α k = ( ) (16) r + δ We can also derive he seady-sae oupu per efficiency-uni as follows. ~ 1 α (1 α )/ α ( ) y = (17) r + δ 7

13 From Eq. (12), he seady-sae invesmen per efficiency-uni is given by: ~ ~ 1 α 1/ α i = ( n + δ + x ) k = ( n + δ + x )( ) (18) r + δ Finally, he seady-sae consumpion per efficiency-uni is: c~ = ~ y ~ i 1 = ( r + δ 1 α 1 α ( n + δ + x )( ) α (1 α )/ α ) / r + δ (19) In erms of he efficiency-unis, his economy does no grow in he long run. Using lower-case leers for per-capia variables and uppercase leers for aggregae variables, we can easily verify ha in he long run, g = g = g = x (20) k y c and g = g = g = n + x (21) K Y C Since we assume ha he reference economy has already reached he seady sae, he growh raes per worker and he aggregae growh raes are deermined by Eqs. (20) and (21), respecively. In oher words, he reference economy follows a balanced-growh pah, ha is, a siuaion where each variable grows a a consan rae. B. Growh of a Developing Counry: The Case of a Closed Economy Now, we urn o he growh of a developing counry. Since we consider a represenaive developing counry, we suppress he counry-index i in he subscrip and insead explicily inroduce he ime-dimension,. We assume ha he counry s rae of growh of laboraugmened echnology asympoically approaches he growh-rae of he reference counry, ha is, Lim x = x (22) This assumpion is commonly made in he empirical lieraure on growh. Since he advancemen of knowledge is no counry-specific in he long run, i is naural o assume ha, in he limi, he echnology in every counry improves a he same rae as ha in he reference counry. If no, he enire global oupu will evenually be aken up by a paricular counry whose produciviy grows permanenly faser han oher counries, which is no likely o happen. I is imporan o noe ha while he limiing growh-rae of echnology is he same, he level of echnology can differ across counries. Differences in he level of echnology in he long run reflec oher differences, which are no explici in he model, such as hose of insiuions, resource endowmen, climae, and so on. 2 However, we will laer assume ha his long-run level of echnology can change as echnology diffuses across counries, especially from he 2 See, among ohers, Hall and Jones (1999) for reasons ha lead o differences in he level of echnology across counries in he long run. 8

14 fronier counry o a developing counry. While echnology diffusion causes echnology o improve faser in he shor run, he more imporan impac occurs in he long run. We also assume ha he seady-sae values of all variables in per-efficiency unis are he same across counries. This assumpion implies ha, in addiion o he same rae of echnological growh, he preference, he rae of populaion growh, and he rae of depreciaion are he same across counries. If a developing counry is closed, he only difference in growh performance beween he reference counry and he developing counry is ha he developing counry grows faser as long as i approaches he seady sae. In he ransiion, he capial sock in erms of efficiencyunis evolves according o he following equaion. 3 ~ & k ~ ~ = ( n + x + δ ) α( k k ) (23) Hence, in he viciniy of he balanced-growh pah, k ~ ~ approaches k a a rae ha is ~ proporional o is disance from k. Using he previous expressions for per-capia growh, during he ransiion period, we have: ~ ~ k k = ( δ ) α (24) g k x n + x + ~ k By considering ha y=ak 1-α h α, we can easily show ha k ~ ~ moves oward k. Tha is, y~ moves oward ~ y a he same rae as ~ ( ~ y & = ( n + x + δ ) α y ~ y ) (25) In a similar way, we can show ha during he ransiion period, he rae of growh of per-capia oupu is: g y = x ( n + x + ) α ~ y ~ y ~ y δ (26) Thus, Eqs. (25) and (26) imply condiional convergence: provided ha economies converge o he same seady sae, less-developed counries should grow faser. The greaer he disance of an economy from he long-run capial per effecive labor, he faser he economy grows. Noe ha he model allows he level of echnology o differ across counries even a he seady sae. In oher words, while he developing counry caches up wih he reference counry by accumulaing capial, he model does no consider anoher imporan avenue for caching-up hrough echnology spillover from advanced counries. 3 This is derived by a firs-order Taylor series approximaion of (2004) and Gourinchas and Jeanne (2006). k &~ around he seady sae, ~ k. See Barro and Xala-i-Marin 9

15 C. Opening Capial Marke: Capial Inflows and Technology Cach-up In he previous case, where he economy s capial marke is closed, he developing counry grows faser han he reference counry as i accumulaes capial faser over he period in which i approaches he seady sae. Since he economy s capial marke is closed, i should approach he seady sae by accumulaing capial hrough is own savings. Once i reaches he seady sae, however, i grows a he same rae as he rae of labor-augmened echnological progress, which is common across counries. Also, noe ha we allowed he level of echnology o differ across counries even a he seady sae. Now we urn o he effecs of inernaional financial inegraion on he economy s growh and welfare. We assume ha here are no impedimens o financial flows across counries under financial inegraion. When he developing counry opens is capial marke o foreign counries, he evoluion of he economy experiences wo imporan changes. Firs, if financial inegraion is in effec, since he reurn o capial is greaer when he economy is away from he seady sae, foreign capial coninuously flows in unil he economy moves o is seady sae. We assume ha he capial movemens are fully and immediaely made. To he exen ha his ransiion period is shorened, he economy grows faser and is welfare improves. 4 Following Hoxa and Kalemli-Ozcan (2007), we call his firs channel of financial inegraion he capial-accumulaion effec. Noe ha while he level of echnology differs, since capial in efficiency-unis is he same across counries a he seady sae, no furher movemens of capial will occur once he economy reaches he seady sae. Second, he developing counry grows faser as i narrows he echnology gap beween iself and he reference counry, which is he echnologically mos enhanced counry. Here, we focus on he role of foreign direc invesmen (FDI), and assume ha i is he main channel hrough which echnology diffuses across counries. As capial flows in from he reference counry o he developing counry, especially in he form of FDI, workers in he developing counry have he opporuniy o learn he advanced echnology. When he labor-augmened echnology improves, he growh rae of he economy rises due o no only faser produciviy growh bu also addiional capial inflows, as he level of echnology permanenly increases. We will call his second channel of financial inegraion he echnology cach-up effec. Recenly, Hoxha and Kalemli-Ozcan (2007) emphasized his second channel while esimaing he welfare effec of financial inegraion. How large is he echnology cach-up effec? We believe ha he answer is enirely an empirical maer. For empirically invesigaing he exen of he echnology cach-up effec, while Hoxha and Kalemli-Ozcan (2007) adoped he value ha was esimaed from firm-level micro-daa, we model he dynamics of he aggregae echnology cach-up in he following form. A& FDI x = + x G ln( A / A ),, h A K (27) In Eq. (27), G is assumed o approach 0 as (or Lim x = x ). Since learning akes ime, we assume ha he growh rae of labor-augmened echnology is greaer han x in he shor run and gradually approaches x unil echnology cach-up is compleed. 4 This is exacly he approach followed by Gourinchas and Jeanne (2006). However, hey did no incorporae oher benefis of financial inegraion such as echnological diffusion; his lowered heir esimae of welfare improvemen. 10

16 According o Eq. (27), he exen o which he growh-rae of he labor-augmened echnology of he developing counry exceeds he seady-sae growh rae of he labor-augmened echnology of he reference counry depends upon he echnology gap, ln( A / A ), he proporion of he FDI aggregae capial ha he FDI capial accouns for,, and he per-capia human capial of K he developing counry, h. In general, we expec ha he following inequaliies hold. and G (ln( A/ A )) < G > 0 FDI ( ) K 0 (28) (29) G h > 0 (30) Eq. (28) implies ha, ceeris paribus, he wider he gap, he faser he echnology gap narrows. Noe ha our specificaion allows he echnology gap o narrow even wihou FDI inflows. This is clear from Eqs. (27) and (28) ha allow he rae of echnology improvemen o depend upon he echnology gap even when FDI=0. We believe ha i is more realisic o assume ha even a closed economy caches up wih he counry ha is a he fronier of echnology. For example, if he level of echnology is low, i is relaively easier for he closed economy o creae innovaions by iself wihou he ransmission of echnology from oher foreign counries. Hence, i is likely ha echnology improves faser when he level of echnology is lower. In oher words, he echnology growh rae is sill negaively associaed wih he echnology gap. However, as will be shown below, he exen o which he gap narrows in he seady sae depends only upon he proporion of FDI capial and he level of human capial. Eq. (29) implies ha he rae a which he echnology gap closes also depends on he proporion of FDI capial. Finally, Eq. (30) suggess ha as he level of human capial of he developing counry rises, he counry learns faser and, herefore, he echnology gap narrows faser. We will furher assume ha he following inequaliy holds. 2 G > 0 FDI ( ) h K (31) This inequaliy suggess ha he same increase in FDI capial has a larger impac on he echnology cach-up if he level of human capial in he developing counry is greaer. Noe ha, in general, he echnology cach-up may no be complee. According o Eq. (27), echnology cach-up ceases as G approaches 0 in he limi. If he echnology cach-up ceases, 11

17 he echnology of he developing counry advances a he rae x, which is he rae a which he reference counry s echnology improves. Given he amoun of FDI inflows needed in he seady sae and he level of human capial, we can derive he echnology gap in he seady sae by solving for G=0. FDI FDI G ln( A / A ),, h = 0 ln( A / A ) = g ( ), h K K (32) In Eq. (32), FDI ( K ) is he proporion of FDI capial in he seady sae and g is he explici soluion for ln( A / A ), as a funcion of FDI and h. Eq. (32) clearly shows ha he exen of K echnology cach-up in he long run depends upon he level of human capial as well as capial inflows. Unlike he capial-accumulaion effec, since he echnology cach-up effec requires some ime for compleely blowing up, he growh rae of oupu does no immediaely reach he seady sae and he shor run dynamics of echnology are characerized as follows. FDI x = x + G(ln( A / A ),( ), h ) (33) K IV. Empirical Tess of he Effec of FDI on Technology Cach-up A. Empirical Specificaion For he empirical invesigaion, we se up he following basic specificaion. 5 d logtfp = α + α1 Gap( TFP ) + α 2 H + α3fdi + α 4 H FDI + η + ε i o i i i i i i i (34) In Eq. (34), d log TFPi is he (average annual) growh rae of he oal facor produciviy for counry i a ime, Gap( TFPi ) is he iniial gap of he level of echnology of counry i relaive o he US, H is he iniial sock of human capial, which is measured by he iniial-year level of he average number of years of secondary and eriary educaion for he populaion ha is aged 15 and above, FDI is a measure of FDI inflows, η is an unobserved counry-specific effec, ε is he error erm, and he subscrips i and represen counries and ime-periods. We also include period dummies o conrol for unobserved period-specific fixed effecs. Since we have assumed he Cobb-Douglas producion funcion, he labor-augmened echnology, A, is relaed o he oal facor produciviy, TFP, in he following way. i i 5 This specificaion closely follows Borenszein e al. (1998), who es a complemenary effec beween FDI and human capial in echnological diffusion and economic growh. However, while hey used GDP growh as an indicaor for echnological progress, we use direc measures of he echnological gap and echnological growh by measuring he oal facor produciviy. 12

18 α A i = TFP i α d log A = d logtfp (35) i i ( Ai ) Gap( TFPi α Gap = ) The measure of echnological growh is he average annual rae of oal facor produciviy. To conduc he decomposiion of he oupu, we measure he oupu per worker by he level of GDP per working-age populaion aged beween 15 and 64. GDP daa are aken from he Penn- World Tables (PWT) 6.1 of Heson e al. (2002). We have used he working-age populaion as a measure of he number of workers, considering ha he available cross-counry sources of labor forces or employees are less reliable han hose of working-age populaions. We use he daa on physical capial sock, consruced by Lee (2005), by using he perpeual invenory mehod ha is based on aggregae invesmen daa in he Penn-World Tables (PWT) 6.1. For he laborshare parameer, α, we assume a value of To consruc a measure of he sock of human capial, we assume ha in Eq. (5), φ(e) is linear and he average marginal reurn from an addiional year of schooling is 7%, as in Bernanke and Gurkaynak (2001) and Bosworh and Collins (2003). 6 The years of schooling are measured by he average years of schooling for he populaion aged 15 and over, as consruced by Barro and Lee (2001). The gap in echnology is measured by he log difference in he level of oal facor produciviy beween each counry and he US, as consruced by Lee (2005). For our preferred approach of represening FDI flows as a proporion of he aggregae capial, we would need o conver one of hese values ino he unis of he oher. Insead, in he empirical seing, we characerize FDI flows as a proporion of GDP, averaged over a five-year period. Furher, noe ha he heory presened in he previous secion implies ha he exen of echnology spillover is more associaed wih he raio of FDI capial sock, no FDI flows, o oal capial sock. Hence, we need o swich FDI flows o FDI sock as well. The raio of FDI flows o GDP can be convered o FDI sock as a share of he oal capial sock in he following way. FDI k = FDI y F FDI FDI F y k FDI FDI = ( ) ( F y FDI r + δ )( ) 1 α (36) In Eq. (36), FDI and equaliy holds a seady sae. F FDI are FDI sock and FDI flows a ime, respecively. The second FDI is he raio of FDI sock o FDI flows a seady sae. F FDI Our regression of specificaion (34) applies o a panel se of cross-counry daa over six fiveyear periods from 1970 o 2000, which correspond o he inervals, , , , , , and Some sudies have used annual observaions bu he highfrequency variaions of TFP (oal facor produciviy) are more likely o resul from business flucuaions han from echnological change. Ohers have used cross-secional regressions, using period-average daa, bu hey are more likely subjec o bias from omied counry-specific variables. Our panel consiss of 491 observaions ranging from a minimum of 68 counries in he period o a maximum of 88 counries in and Table 1 summarizes descripive saisics of he daa. 6 The resuls are no sensiive o alernaive assumpions. 13

19 The empirical echnique we use here is he fixed-effecs esimaion o conrol for unobservable counry fixed-effecs. The esimaion also uses insrumenal-variable (IV) regressions o conrol for he poenial endogeneiy bias of FDI. B. Esimaion Resuls Columns (2.1)~(2.3) of Table 2 summarize he regression resuls from he fixed-effecs OLS panel esimaion. Regression 2.1 shows ha he coefficien of he echnology gap is saisically significan, wih he expeced negaive sign, indicaing ha more echnologically backward counries experienced faser echnological progress over ime. The average years of schooling a secondary and eriary levels has a posiive, hough no srong, effec on TFP growh. FDI also has a posiive effec on echnology growh, bu he coefficien is only marginally significan a he 10% level. If viewed causally, he esimaed coefficien, (a a sandard error of 0.055), implies ha an increase of per year in he raio of FDI o GDP (abou one sandard deviaion) leads o an increase in he growh rae of TFP of abou 0.2 percenage poins per year. The heory oulined above suggess ha he effec of FDI on echnological growh ineracs wih he level of human capial sock. The specificaion in regression 2.2 replaces he FDI variable by he ineracive erm beween FDI and human capial. The ineracive erm is posiive and saisically significan. Regression 2.3 exends his specificaion by adding he FDI variable. In his way, we es if FDI affecs echnological growh by iself or hrough is ineracion erm. The resul shows ha he ineracion erm is posiive and saisically significan, while he coefficien of FDI is negaive and insignifican. The regression coefficiens indicae ha all counries wih pos-primary school aainmen above 0.32 years will benefi from FDI. In our sample, 362 ou of he oal 491 (counry-year) observaions saisfy his hreshold. The esimaes indicae ha, for insance, in an economy wih a pos-secondary aainmen of 2.2 years (which is he average value for he sample counries in 2000), an increase of per year in he raio of FDI o GDP (equivalen o one sandard deviaion) raises he growh rae of TFP of he hos economy by abou 0.5 percenage poins per year. C. Insrumenal Variables Esimaion While he esimaion resuls in secion B are indicaive, hey do no necessarily mean ha more FDI raises he growh rae of TFP; i is possible ha he causaliy runs in he opposie direcion, ha is, FDI flows ino hose counries ha exhibi fas echnological improvemen. To conrol for his endogeneiy problem, we use insrumenal variables (IV) panel esimaion by using as insrumens for FDI he lagged value of FDI, he log value of GDP, and a measure of nominal exchange-rae volailiy. 7 The resuls of he IV esimaion are repored in Columns (2.4)~(2.6) of Table 2. In regressions 2.4 and 2.5, he FDI variable and he produc of he FDI and human-capial erms individually ener posiively and are saisically significan. The coefficiens are an order of magniude larger in IV panel esimaion han in he fixed-effecs panel. The specificaion of regression 2.6 includes wo variables ogeher. The resul shows ha he ineracion erm is posiive, hough marginally significan, while he coefficien of FDI is insignifican. The FDI variable and is ineracion erm wih human capial are joinly significan (p-value = 0.004). Hence, FDI has a 7 Mos previous sudies use lagged values of FDI flows as insrumens. Wei (2001) finds ha a measure of nominal exchangerae volailiy has a significan explanaory power on he variaion of FDI inflows. The nominal exchange-rae volailiy is measured by he coefficien of variaion of he nominal exchange-rae over he ime-inerval. 14

20 posiive effec on TFP growh and he magniude of he effec depends on he sock of human capial ha is available in he hos economy. The poin esimae of he ineracive erm indicaes ha in a counry wih an average sock of human capial of 2.2 years, an increase of per year in he raio of FDI o GDP raises he growh rae of TFP by abou 1.1 percenage poins per annum year, which is abou wice as large as in he OLS case. D. Esimaion wih FDI Socks The model oulined in Secion 3 suggess ha he effec of FDI on produciviy growh is capured by FDI socks raher han FDI flows. In fac, i is no expeced ha echnology spillover is posiively relaed o he curren FDI flows alone. Given ha i akes ime o learn new echnology, he exising FDI sock is also expeced o coninuously ransmi new echnology. Our empirical specificaions so far are inconsisen wih he model in his sense because we have adoped FDI flows as an explanaory variable. In his secion, we remedy his problem by replacing he FDI-flow variable wih he FDI-sock variable in he empirical analyses. In Table 3, we repor he resuls. Regression resuls are based on OLS esimaion and are based on IV esimaion. In regression 3.1, which shows he basic esimaion resuls, he echnology gap eners saisically significanly, wih he expeced sign. The coefficien of schooling is marginally significan a he 10% level, bu he coefficien of he FDI-sock variable is no saisically significan. The specificaion in regression 3.2 replaces he FDI variable by he ineracive erm beween FDI and human capial. The ineracive erm is posiive and saisically significan a he 5% level. Regression 3.3 exends his specificaion by adding he FDI-sock variable. In his way, we es if FDI affecs produciviy growh by iself or hrough is ineracion erm. The resul shows ha he ineracion erm is posiive and saisically significan a he 10% level, while he coefficien of FDI is negaive and insignifican. Though he evidence is somewha weaker, we reconfirm he hypohesis ha FDI conribues o produciviy growh by ineracing wih he sock of human capial. The esimaed coefficiens imply ha in a counry wih an average sock of human capial of 2.3 years, an increase of 0.15 per annum in he raio of FDI-sock o GDP raises he growh rae of TFP by abou 0.4 percenage poins per annum in he period. In regressions , we repor he esimaion resuls wih he same specificaion as , bu wih he` insrumenal-variables echnique. The insrumenal variables are consruced in he same way as in he IV esimaion wih he FDI-flow variable, repored in Table 3. The resuls are generally consisen wih hose of he OLS esimaion. When eiher he FDI sock or he ineracive erm beween FDI sock and human capial is included, i is significan a he 5% level. However, when boh are included in regression 3.6, he associaed coefficiens are boh posiive bu insignificanly differen from zero. Overall, we conclude ha our resuls do no change even if we use he FDI-sock variable raher han he FDI-flow variable. Ineresingly, he esimaed coefficiens of oher variables change lile, wih only he coefficien of he FDI-sock variable scaled down by a facor of approximaely 10. This reflecs he fac ha he FDI-flow variable is approximaely one enh of he FDI-sock variable in he daa. 15

21 E. Checks for Robusness and Addiional Exercises (Summary of Findings Tha Are No Repored in he Tables) We have carried ou a broad se of empirical exercises ha invesigae he robusness of our resuls. We briefly summarize he resuls. 8 (i) Consideraion of only he years The resuls are sronger. The ineracive erm becomes significan a a 1% level in IV esimaion. If we use a more balanced daase, for example., a sample of counries ha has complee TFP daa, he resuls are he same. (ii) Resricion o Organisaion for Economic Co-operaion and Developmen (OECD) counries wihou high incomes. The resuls are qualiaively he same, alhough hey become weaker. (iii) Effecs in Eas Asia. We have examined wheher here were differences for Eas Asia in he effecs of FDI on echnology progress by including muliplicaive erms of FDI variables wih Eas Asia region-dummies. There is no evidence ha he effecs are differen for Eas Asia. (iv) Addiion of oher missing variables. The possible deerminans of TFP include governmen consumpion, rade openness, insiuions, and financial deph. When we add all hese variables o he regressions, he main resuls are qualiaively he same. However, he sample size shrinks. (v) Consideraion of differing absorpive capaciies. We ry oher measures such as insiuional qualiy (a measure of propery righs or he invesmen climae ), openness, and financial deph. There is evidence ha insiuional qualiyfdi has some significance. However, here are some problems; in paricular, he daa are available only from V. Esimaion of he Welfare Effecs of Financial Inegraion Based on he growh model and esimaion resuls of he previous secions, we can underake a calibraion exercise o assess he welfare effecs of financial inegraion. As explained in secion 3, if a counry opens up he capial marke, he welfare of he counry increases hrough wo channels: he capial-accumulaion effec and he echnology cach-up effec. In his secion, we apply our mehodology o Eas Asian counries and invesigae how he welfare level improves as financial markes are compleely inegraed wih he financial marke of an advanced economy. Benchmark case In he benchmark case, we assume ha financial markes are compleely closed. Since he economy relies only on domesic savings, i akes ime for he economy o reach he seadysae. We can hink of wo versions of he benchmark case. In he firs version, here is no echnology cach-up, besides he closed naure of financial markes. This version of he 8 The resuls are available upon reques. 16

22 economy was described in deail in secion 3.B. However, as explained earlier, i is more realisic o assume ha even a closed economy caches up wih he counry a he echnology fronier. In he second version of he benchmark case, he echnology is assumed o evolve following he same dynamics ha are esimaed by Eq. (34). where he level of FDI is se o zero. As explained in secion 3, if a counry opens is capial marke, he welfare level of he counry increases hrough wo channels: he capial-accumulaion effec and he echnology cach-up effec. We consider hree cases of full financial inegraion. Case 1 (he capial-accumulaion effec) We assume ha capial movemens from foreign counries are immediaely made. However, here is no addiional echnology cach-up due o he inflow of foreign capial. Case 2 (he echnology cach-up effec wih 40% FDI inflows) We allow addiional echnology cach-up hrough FDI inflows. However, he FDI sock is assumed o consiue only 40% of he oal foreign capial. 9 Hence, he exen of echnology cach-up is low. Case 3 (he echnology cach-up effec wih 100% FDI flows) As in Case 2, we allow addiional echnology cach-up hrough FDI inflows. The FDI sock is assumed o consiue 100% of he oal foreign capial. Hence, he exen of echnology cachup is high. In Table 4, we repor he calibraion resuls for eigh Asian counries. In calibraing he model, he esimaed coefficiens repored in regression specificaion 2.3 in Table 2 are used o se he parameer values, excep ha he coefficien of FDI flows is scaled down by a facor of 10. This modificaion reflecs he fac ha while he empirical specificaion in Table 2 is based on FDI flows, he model specificaion is based on he FDI-sock variable raher han he FDI-flow variable (see Eq. (36)). 10 While we devised wo versions of he benchmark case, he welfare implicaion of addiional echnology improvemen due o financial inegraion is he same, irrespecive of which version we use as a benchmark. However, for evaluaing Case 1, i is easier o use he firs version of he benchmark. In oher words, we focus on he capial-accumulaion effec wihou worrying abou echnology changes by assuming he absence of echnology cach-up boh before and afer financial inegraion. In conras, i is easier o use he second version of he benchmark case while evaluaing Cases 2 and 3. In his insance, he addiional echnology cach-up due o financial inegraion is defined by comparing he echnology cach-up under complee financial inegraion o he echnology cach-up ha prevails in he absence of foreign capial inflow. The exen of echnology cach-up when capial inegraion is complee can be esimaed from Eq. (34) by subsiuing he appropriae value of FDI ha is supposed o flow in. On he oher hand, he exen of echnology cach-up when here is no financial inegraion is obained by seing FDI o zero. According o Eq. (34), a low-echnology counry caches up wih he counry a he echnology fronier as long as he echnology gap is less han 1. Therefore, when we evaluae Cases 2 and 3, he addiional welfare improvemen due o financial 9 10 As noed in Hoxha and Kalemli-Ozcan (2007), FDI flows consiue abou 40% of oal capial flows. The reason why we repor he calibraion resuls on he basis of esimaes of he flow specificaion is ha he FDI-sock values are much more subjec o measuremen errors. Our resuls do no significanly change if we use he esimaed coefficiens of he regression specificaion 4.3 where he FDI-sock variable is used, in which case he coefficiens do no need o be scaled down. 17

23 inegraion is capured by using he second version of he benchmark case. I is imporan o noe ha as long as echnology improves, addiional FDI coninuously flows in. For example, in Korea, he iniial capial per efficiency-uni is 93% of is seady-sae value. This is because, while he per-capia iniial capial is low, since he level of echnology is also low, he iniial capial per efficiency-uni becomes higher a he level close o is seady-sae value. Hence, he iniial inflow of FDI is small. However, as echnology subsequenly improves, he capial per efficiency-uni is lowered. In oher words, he capial per efficiency-uni becomes lower han is seady-sae value. In his case, we assume ha foreign capial immediaely flows in again so ha he capial per efficiency-uni is equalized o is seady-sae value. In oher words, we assume ha foreign capial flows in no only a he beginning of capial liberalizaion, bu also coninuously flows in as long as echnology subsequenly improves. This addiional FDI furher improves echnology and he improved echnology induces furher FDI inflows and so on. The calibraion resuls are repored in Table 4. For he calibraion of he model, we assume ha any counry s seady-sae values of k ~ and ~ y are equivalen o hose of he US. The levels of per-capia oupu and capial for each counry in 2000, as a raio of he counry s level o ha of he US, are repored in rows 1 and 2. In erms of he oupu per-capia, Hong Kong, China (0.73) and Singapore (0.72) are he mos advanced counries, followed by Korea (0.47) and Malaysia (0.32). Ineresingly, however, he level of per-capia capial is much higher in Singapore (1.10) han in Hong Kong, China (0.80). In fac, i is higher in Singapore han even he US We conver he raio of per-capia capial ino he raio of capial per efficiency-uni in he following way. ~ k ~ k k / A k A k TFP = (37) i i i i i 1/α = = ( ) k / A k Ai k TFPi In row 3 of Table 4, we repor he raio of he capial per efficiency-uni for each counry. If we ake ino consideraion he produciviy level (0.88), he level of capial per efficiency-uni in Singapore becomes much higher han ha in he US. In Korea, while he per-capia capial (0.53) is much lower, since he produciviy level is also low (0.57, as repored in row 5), is capial per efficiency-uni is higher han ha in he US We inerpre his as evidence ha capial markes in hese wo counries are compleely open o global markes. Since no addiional foreign capial is supposed o flow in hese wo counries, he benefi of financial inegraion is nil. In he oher counries, he level of capial per efficiency-uni is lower han ha in he US If hese counries furher open heir capial markes, we will assume ha he raio becomes one, which is he source of welfare improvemens ha arise from financial inegraion. The level of human capial in 2000 is repored in row 4. I is highes in Korea, followed by Hong Kong, China and he Philippines. For he calibraion of he model, we assume ha he level of human capial is fixed hroughou a he level of The iniial level of TFP ( Iniial gap ), which is consruced as he raio of each counry s TFP level o ha of he US, is repored in row 5. The iniial TFP level is also he highes in Hong Kong, China and Singapore, followed by Malaysia and Korea. As explained earlier, in he second version of he benchmark case, he developing counry can cach up wih he reference counry in erms of echnology, wihou FDI flows. In his case, he 18