FDI Spillovers, Financial Markets, and Growth*

Transcription

1 Spllovers, Fnancal Markes, and Growh Laura Alfaro Harvard Busness School Areendam Chanda Norh Carolna Sae Unversy Sebnem Kaleml-Ozcan Unversy of Houson Seln Sayek Inernaonal Moneary Fund Ths verson: May 2002 Absrac Ths paper examnes he varous lnks among foregn drec nvesmen, fnancal markes and growh. We model an economy wh a connuum of agens ndexed by her level of ably. Agens can eher work for he foregn company n he secor; or hey can underake enrepreneural acves, whch are subjec o a fxed cos. Beer fnancal markes allow agens o ake advanage of knowledge spllovers from. Emprcal analyss, usng cross-counry daa beween , shows ha alone plays an ambguous role n conrbung o economc growh. However, counres wh well-developed fnancal markes gan sgnfcanly from. Key words: Foregn drec nvesmen, fnancal markes, economc growh. JEL Classfcaon: F23, F36, F43 We hank Ross Levne, Eduardo Borenzsen and Edmundo Murrugarra for her help wh he daa and Tm Kehoe, Ben Sorensen, and Alex Hoffmaser for valuable commens and suggesons. We receved valuable commens from he parcpans of he macro semnar a Brandes Unversy, Unversy of Norh Carolna- Chapel Hll, Unversy of Houson and he SED conference n Sockholm.. The vews expressed n hs paper are hose of he auhors and do no necessarly represen hose of he I.M.F.

2 Inroducon The 990s wnessed wo sgnfcan shfs n he role fnancal markes play n foregn drec nvesmen (). The frs, a shf brough abou by an ncreased share of n oal capal flows, changed he composon of capal flow self. The second became a paradgm shf n emphass among polcy-makers n developng counres n her effors o arac more. These shfs, whch concded wh he pos-980s deb crses, were renforced followng he Eas Asan fnancal crses. Durng hs perod, sudes focused on he relave volaly of dfferen ypes of capal and on how ha volaly nfluenced economc declne. The sudes conrbued o srenghenng he mporance of more long-erm nvesmen, ncludng, whch was much less volale compared o commercal bank loans and foregn porfolo flows. The raonale for ncreased effors o arac more sems from he poenal posve effecs provdes. They nclude producvy gans, echnology ransfers, new processes, manageral sklls and know-how o he domesc marke, employee ranng, nernaonal producon neworks, and access o markes. 2 If foregn frms nroduce new producs or processes o he domesc marke, domesc frms may benef from he acceleraed dffuson of new echnology. 3 In some cases, domesc frms may benef solely by observng hese foregn frms. 4 In oher cases, echnology dffuson may occur from labor urnover as domesc employees move from foregn o domesc frms. These benefs, ogeher wh he drec capal fnancng provdes, sugges ha can play an mporan role n modernzng a naonal economy and promong economc developmen. However, boh plan- and aggregae-level emprcal evdence remans ambguous, whch suggess ha posve knowledge spllovers canno be presumed. For example, lookng a plan See Fernandez-Aras e al. (2000) and World Developmen Repor (2000), showng he lower volaly of durng he perod See Caves (996) for a dscusson on echnology ransfers. 3 For echnology dffuson models see Grossman and Helpman (99) and Barro and Sala--Marn (995, 997). 4 See Blomsrom and Kokko (997).

3 level daa n Venezuela, Aken and Harrson (999) fnd ha he ne effec of on producvy s que small, where rases producvy whn plans ha receve he nvesmen whle lowerng ha of domescally owned plans. Naonal sudes by Borenszen, De Gregoro, and Lee (998) and Carkovc and Levne (2000), usng cross-counry growh regressons, also provde lle suppor ha has an exogenous posve effec on economc growh. Ths evdence suggess ha whle may seem naural o argue ha can convey greaer knowledge spllovers, a counry's capacy o ake advanage of hese exernales mgh be lmed by local condons. These condons nclude bu are no lmed o he polcy envronmen of he local counry, producve asses avalable, nfrasrucure, and nsuons. In lne wh he recen emphass on he role of nsuons n he growh leraure, we examne he nermedary role played by local fnancal nsuons n channelng he conrbuons of o economc developmen. In parcular, we argue ha he lack of developmen of local fnancal markes can adversely lm he economy's ably o ake advanage of poenal spllovers. The mporance of well-funconng fnancal nsuons n augmenng echnologcal nnovaon, capal accumulaon, and economc developmen has been recognzed and exensvely dscussed n he leraure. Goldsmh (969), McKnnon (973), Shaw (973), followed by Boyd and Presco (986), Greenwood and Jovanovc (990), and Kng and Levne (993b), among ohers, have shown ha well-funconng fnancal markes, by lowerng he coss of conducng ransacons, ensure capal s allocaed o he projecs ha yeld he hghes reurns and herefore enhances growh raes. Furhermore, as McKnnon (973) saed, he developmen of capal markes s necessary and suffcen o foser he adopon of bes-pracce echnologes and learnng by dong. In oher words, lmed access o cred markes resrcs enrepreneural developmen. If enrepreneurshp allows greaer assmlaon and adopon of bes echnologcal pracces made avalable by, hen he absence of well-developed fnancal markes lms he poenal posve exernales. 2

4 Alhough he benefs of, as a form of capal marke negraon, dd no receve much aenon unl recenly, emprcal evdence on he heorecal framework of he neracon beween fnancal markes and economc growh s ample. Provdng evdence a he counry level, Kng and Levne (993a,b), Beck, Levne and Loayza (2000a,b) sugges ha fnancal sysems are mporan for boh producvy growh and developmen. Levne and Zervos (998) show ha, alhough developmen of all fnancal ypes of fnancal nsuon posvely predc growh, he ype of fnancal nsuons (ncludng sock markes and banks) have dfferen magnudes of mpacs due o he provson of dfferen servces. A he ndusry level, Rajan and Zngales (998) fnd ha enhanced fnancal developmen reduces he cos of exernal fnance o frms, hereby promong growh. Combnng ndusry and counry level research, Wurgler (2000) shows ha even f fnancal developmen does no lead o hgher levels of nvesmen, seems o allocae he exsng nvesmen beer and hence promoes economc growh. Nowhsandng he wdespread agreemen abou he role of fnancal markes n enhancng spllovers and economc growh, he leraure on seems o have gnored s mporance alogeher. As Caves (999) noes, he consrans faced by local frms n reapng such spllovers are no menoned a all n he four volumes of The Handbook of Developmen Economcs. Generally speakng, he role of no jus fnancal markes bu oher facors, ncludng poenal shorages of sklls, knowledge, and nfrasrucure n he recpen counres, have been negleced n he developmen leraure. Only recenly have hese ssues been addressed. For example, Borenszen e al. (998) and Xu (2000), usng daa on flows from ndusralzed counres o 69 developng counres and daa on U.S. mulnaonal enerprses (MNEs) respecvely, show ha allows for ransferrng echnology and for hgher growh only when he hos counry has a mnmum hreshold sock of human capal. Mos LDCs do no mee hs hreshold. 3

5 The World Bank s 200 edon of global developmen fnance dscusses he mporance of absorpve capaces and he success of. 5 Absorpve capaces here nclude macroeconomc managemen (as capured by nflaon and rade openness), nfrasrucure (elephone lnes and paved roads), and human capal (share of labor force wh secondary educaon and percenage of populaon wh access o sanaon). Fnancal markes are no menoned. Indeed, fgure, whch shows daa on and fnancal developmen, provdes movaon for ncludng fnancal markes among he relevan absorpve capaces. As fgure depcs, daa for he perod sugges a posve relaonshp beween, measured as a share of GDP, and fnancal markes, measured by fnancal developmen ndcaors nroduced by Beck e al. (2000). However, gven her neracon wh one anoher, s also apparen a wde varaon n boh varables exss. Indeed, f fnancal developmen plays an mporan role n nfluencng he effecs of on oupu, one can expec counres wh he same levels of o have very dfferen oucomes n erms of ncome levels. In hs paper, we formalze hs mechansm hrough whch he rckle down effec of depends on he exen of he developmen of he fnancal secor. We model an economy populaed by agens who are dfferenaed by her ably level. Agens have wo choces. They can work for he foregn company n he secor and use her nhered wealh o earn a reurn. Or hey can se up her own frm, whch wll benef from a spllover from foregn drec nvesmen. However, sarng a frm requres a seup cos, whch mus be fnanced parly hrough borrowng from fnancal nsuons. Due o neffcences n he fnancal secor, he borrowng rae s assumed o be hgher han he lendng rae. Under hs scenaro, more developed fnancal nsuons make easer for enrepreneurs o se up a busness. Ths no only spurs enrepreneural acvy bu, more mporanly, enables enrepreneurs o reap he spllovers from foregn drec nvesmen. Ths mples ha wll effec he local economy beyond he drec 5 The dscussons demonsrae how some counres wh low absorpve capaces, such as Morocco, Uruguay and Venezuela (he las based on Aken and Harrson, bd.), faled o reap spllovers; whereas 4

6 ncrease n capal from abroad. We fnd ha for hese posve effecs o be realzed he developmen level of local fnancal markes s crucal. Usng growh regressons, we sudy he mpac of he neracon of a range of fnancal marke varables ha exs n he leraure wh on economc growh. The resuls sugges ha alhough alone plays an ambguous role n conrbung o economc growh, well-developed fnancal markes aler he resuls sgnfcanly. Counres wh well-developed fnancal markes seem o gan sgnfcanly more from. Ths s conssen wh he resuls of Carkovc and Levne (2000). The res of he paper s organzed as follows. The benchmark model developed n Secon 2 s used o movae he emprcal esng. Daa are defned n Secon 3. Emprcal resuls are dscussed n Secon 4. Secon 5 concludes. 2 A Concepual Framework We presen a smple model o llusrae how mprovemens n he fnancal markes nfluence he effecs of on domesc producon. The model provdes a benchmark for emprcal analyss. We assume a small open economy wh no adjusmen coss. The economy s populaed wh a connuum of agens of oal mass. Each agen lves for one perod. There are wo secors n he economy, Y and Y DOM, he foregn producon and domesc producon secor respecvely. 2. Producon Secors are dsngushed by ownershp, echnology, and npus used. The foregn producon secor, denoed Y, s wholly owned by foregn nvesors and uses foregn capal and domescally suppled labor. We assume, followng Razn e al. (999), ha home nvesors who Malaysa and Tawan fared well wh hgher absorpve capaces. See World Bank (200) page 62. 5

7 lack access o foregn capal markes canno challenge he foregn drec nvesors n hs secor. 6 We assume ha s araced o he counry because of cos of capal dfferences. 7 Producon s perfecly compeve and echnology s assumed o be a Cobb-Douglas consan reurn o scale, ( K ) Y = AL () where 0<<, L denoes he domesc labor, K s he sock of foregn capal, and A s a producvy parameer. Opmaly, condons n he foregn producon secor mply ha foregn capal s pad s margnal produc, whch s gven by he nernaonal rae of neres, r, ( ) ( ) Rearrangng, we ge an expresson for he sock of foregn capal, r = AL K (2) K ( )A = L (3) r The foregn frm hres workers up o he pon where he margnal producvy of an exra worker equals he wage, w, ( ) w = A (4) r The second secor, Y DOM, s composed of a number of frms, each of whch s owned by a local enrepreneur. Producon n each frm requres a fxed capal nvesmen. In addon o he fxed capal nvesmen, oupu s posvely affeced by he enrepreneur's ably and by he amoun of foregn capal n he economy. 8 6 Alernavely, foregn frms can be hough of as ownng he echnology o produce n hs secor and because of s specal characerscs (asse specfc) hey choose o drecly produce n he counry raher han o lcense he echnology. 7 Our objecve s o undersand he effec of foregn producon on local oupu and he role of fnancal markes and no he decson o nves abroad. 8 The assumpon of enrepreneural ably s mporan o rule ou corner soluons as wll be evden laer. 6

8 Local enrepreneurs benef from spllovers from he secor. Ths assumpon s cenral o he model. 9 Poenal enrepreneurs can ake advanage of beer manageral pracces, neworks, access o markes, and oher spllovers from he foregn frms locaed n he domesc counry. The foregn frm does no nernalze hese posve effecs. Oupu n hs secor s gven by: Y DOM = Y d (5) where Y s he amoun produced by enrepreneur : γ ( K ) S Y =, (6) where 0<γ <, 0<<, s assocaed wh an enrepreneur of ably level, S s he fxed capal nvesmen. We assume ha he fxed nvesmen S exceeds he resources owned by any sngle ndvdual a any pon n me. Local enrepreneurs can borrow he dfference beween her endowmen and S n he local marke. 0 In he fnancal marke, here s a wedge - - beween he lendng rae, r, and he borrowng rae. The dfference reflecs he neffcences n he fnancal secor. As Kng and Levne (99) menon, hs wedge could reflec axes, neres celngs, requred reserve polces, or n general hgh nermedaon coss due o labor regulaon, hgh admnsraon coss, low echnology, ec. 2.2 Households and Occupaonal Choce The economy s populaed by a connuum of agens of oal mass, who lve for one me perod. Alhough hey are all endowed wh one un of labor, hey dffer n her ably level. For example, upon brh, ndvdual a me s endowed wh ably level, where s..d., 9 Blomsrom and Kokko (997) menon hree dfferen means of such spllovers from he MNF o he local markes. The frs one s due o lnkages n he marke. Backward and forwards lnkages beween he foregn and domesc frms creae an envronmen where he foregn processes can be easly learned by he domesc frms. Tranng of he local workers n he ndusry also allows such spllovers. Fnally, hey defne demonsraon effecs, whch are possble due o he compeon beween he domesc frms and he MNF. Smlar effecs are menoned n Aken and Harrson (999). 0 Due o nformaon asymmeres plus soveregn rsk s cosly or mpossble for enrepreneurs o fnance abroad or drecly hrough he sock marke. Galor and Zera (993) adop a smlar sraegy o allow for capal marke mperfecons. 7

9 unformly dsrbued over (0, ). s realzed a he begnnng of he perod. Agens are all endowed wh some nal wealh b -, on whch hey can earn he nernaonal reurn r a he end of he perod should hey choose o nves n capal markes. Inally, we assume ha b -=b - for all. All ndvduals consume and leave bequess b -, a he end of he perod. A he begnnng of perod, each agen faces wo choces. They can work for he foregn frm n he secor or become an enrepreneur and work n he domesc-producon secor. If agens choose o work for he foregn company, hey nelascally supply her labor endowmen and earn a wage w. Wage w s ndependen of he agen's ably level. Agens who choose o work for he foregn company earn an ncome sream equal o he wage plus he reurn on her level of asses, w ( r )b. Those who choose o produce mus pay her loans a he end of he perod and herefore earn a ne ncome of Y ( )( S b ). An ndvdual chooses o work for he foregn company, f he ncome he or she earns s hgher han ha n he enrepreneural secor: w ( r )b > Y ( )( S b ) (7) alernavely, he or she chooses o sar a frm f w ( r )b < Y ( )( S b ) (8) An agen s ndfferen beween workng for he foregn frm and sarng s own frm f w ( r )b = Y ( )( S b ) (9) The above equaon characerzes he break-even level of ably and herefore also he measure of people ha work for he foregn secor. Subsung for oupu, Y, usng equaon (3), we oban: fd γ ( K ) S ( )( S b ) w ( r )b =, B (0) Wh a unform dsrbuon for he ably level, we oban he followng expresson for,, ( )( S b ) w ( r )b = () B γ ( K ) S 8

10 Le denoe he value of ably ha sasfes he above condon. Ths means ha he oal amoun of labor employed n he secor wll be, = L d = 0, (2) From equaons (2) and (2), we can rewre he amoun of foregn capal as K ( )A = (3) r Subsung hs expresson no equaon () and rearrangng provdes an explc form for he hreshold level of enrepreneural ably, ( ( )( S b ) A ( r) b r = A( ) γ B S r (4) 2.3 Comparave Sacs The above model allows us o undersand how wll mpac oupu, and how hs effec depends on he local fnancal marke condons. More generaes hgher oupu va wo channels: ncreased producon n he secor and ncreased producon n he domesc secor. Noe ha he oal oupu n he economy s: Y = Y Y d (5) whch mples, γ ( K ) S Y = Y ( ) B (6) The oal effec of on oupu s herefore he sum of he prvae margnal produc of n s own secor plus he dfference beween he socal and he prvae margnal produc: Y K Y = K γ [( ) B( K ) S ] K (7) 9

11 whch s equvalen o, ( ) ( ) 0 S K B r K Y > = γ (8) Fnancal nermedaon n hs model affecs he socal margnal produc of. The oal effec herefore s, ( ) ( ) ( )( ) γ γ = 2 T 2 K S K B S K B K Y (9) equvalenly, ( ) ( )( ) 0 f 0 S K B K Y T 2 > < = γ (20) In order o fully analyze he above equaon, we need o sudy how fnancal marke neffcences affec he decson of becomng an enrepreneur. Observe ha hgher he value of he parameer, whch denoes hgher neffcences n he fnancal secor, he less aracve enrepreneural acvy becomes. Usng he defnon of, = r, we can rewre equaon (8) as, γ = S r ) A( B )b r ( r ( A ) b S )( r ( (2) From hs expresson, follows ha 0 >. As expeced, hgher fnancal coss reduce he number of enrepreneurs. Conversely, an mprovemen n he effcency of he fnancal secor ends o reduce he hreshold level of enrepreneurshp, hereby leadng o an ncrease n he number of enrepreneurs n socey. Ths mples ha an mprovemen n he effcency of he fnancal secor ncreases he socal margnal produc of foregn capal. As s evden n equaon (9), essenally wo effecs are workng here. Frs, an mprovemen n he fnancal secor ncreases he number of enrepreneurs n socey. Ths ncrease rases he socal margnal produc 0

12 of snce he wo are complemenary. Second, here s a drec effec of he number of enrepreneurs on he amoun of foregn capal sock. As he number of enrepreneurs rse, he number of laborers fall. The declne n he number of laborers mples ha he sock of foregn capal wll declne. Overall, hgher effcency of he local markes rases he socal margnal produc of foregn capal n he domesc secor. 2 The model above shows, n a very smplfed form, how beer fnancal markes can lead o greaer effecs of foregn drec nvesmen on oupu. In pracce, however, fnancal markes affec no only he fnancng of nvesmen bu also he day-o-day conduc of busness. Ths mporan channel s mporan bu ofen negleced n he leraure. In he appendx, we work ou a more complee verson of he model, where such effecs also are ncorporaed and where he qualave resuls of he model do no change. The appendx also ncludes a numercal analyss of he complee verson of he model. I shows ha for reasonable parameer value, a beer-developed fnancal marke encourages ncreased enrepreneural acvy, domesc oupu, and aracs more ;- alogeher showng ha beer fnancal markes magnfy he effecs of on oupu. The overall resuls sugges ha hgher levels of generae more oupu n he economy, and he magnude of hs posve effec s nfluenced by he fnancal marke neffcences. In he followng secons, an emprcal analyss based on hese fndngs s carred ou n order o undersand he neracon beween, local capal markes, and growh. 3 Daa 2 Improvemens n he fnancal markes have a posve effec on he margnal produc of capal for - > 0. The assumpon of < clearly helps remove any ambguy n hs condon. However, hs resrcon s necessary only for = 0. Furher, emprcally, a value of > seems hghly mplausble. For example, esmaes of R&D spllovers by Coe, Helpman and Hoffmaser (995) show ha a $00 ncrease n he U.S. or Japanese R&D domesc capal sock ncreases real GDP of developng counres by almos 25 dollars.

13 In Appendx 2, daa used n he emprcal analyss are descrbed n deal. In hs secon, however, daa for he wo mos sgnfcan varables are dscussed: he measures of foregn drec nvesmen and fnancal marke developmen. An mporan source for he daa s he IMF publcaon Inernaonal Fnancal Sascs (IFS), whch repors he Balance of Paymens sascs on. The ne nflows, repored n IFS, measure he ne nflows of nvesmen o acqure a lasng managemen neres (0 percen or more of vong sock) n an enerprse operang n an economy oher han ha of he nvesor. I s he sum of equy capal, renvesmen of earnngs, oher long-erm capal, and shor-erm capal as shown n he balance of paymens. The gross fgures reflec he sum of he absolue values of nflows and ouflows accouned n he balance of paymens fnancal accouns. Our model focuses on he nflows o he economy; herefore, we prefer usng he ne nflow measure. 3 Alernave daa sources nclude UNCTAD and OECD publcaons; however, he IMF daa allow a more comprehensve analyss by avalably of daa for a larger se of counres. Followng Kng and Levne (993), Levne and Zervos (998), and Levne e al. (2000a), we consruc several fnancal marke seres, rangng from he sock marke o he rae of monezaon of he economy. I s very dffcul o consruc accurae and comparable measures of fnancal servces daa for a broad cross-secon of counres over several decades. Followng Levne and Zervos (998), sock marke lqudy s measured as he value of sock radng relave o he sze of he economy, labeled as value raded (henceforh SVALT). In order o capure he relave sze of he sock marke, we use he average value of lsed domesc shares on domesc exchanges n a year as a share of he sze of he economy (he GDP), n smlar ven o Levne and Zervos (998). Ths seres s labeled as capalzaon (henceforh SCAPT). 3 Ths s he daa used by Soo (2000) as well, whle Borenzsen e al (998) use nflows only from OECD counres. 2

14 We furher draw on varables nroduced by Levne e al. (2000), whch n urn buld on Kng and Levne (993a). The daa assocaed wh he former are avalable from he World Bank Fnancal Srucure Daabase. 4 Four varables are ncluded n our work. Frs, Lqud Lables of he Fnancal Sysem (henceforh, LLY): equals currency plus demand and neres-bearng lables of banks and nonfnancal nermedares dvded by GDP. I s he broades measure of fnancal nermedaon and ncludes hree ypes of fnancal nsuons: he cenral bank, depos money banks, and oher fnancal nsuons. Hence, LLY provdes a measure for he overall sze of he fnancal secor whou dsngushng beween dfferen fnancal secors. Second, Commercal-Cenral Bank Asses (henceforh, BTOT): equals he rao of commercal bank asses dvded by commercal bank plus cenral bank asses. BTOT measures he degree o whch commercal banks versus he cenral bank allocae socey s savngs. Kng and Levne (993a) and Levne e al. (2000), as well as ohers, have used hs measure, whch provdes a relave sze ndcaor,.e., he mporance of he dfferen fnancal nsuons and secors relave o each oher. Thrd, Prvae Cred (henceforh, PRIVCR): equals he value of creds by fnancal nermedares o he prvae secor dvded by GDP. The wo prevous measures do no dfferenae beween he end users of he clams of fnancal nermedares,.e., wheher he clams are n he publc or he prvae secor. Ths and he followng measure focus solely on he clams on he prvae secor. Fourh, Bank Cred (henceforh, BANKCR): equals he creds by depos money banks o he prvae secor as a share of GDP ( does no nclude nonbank creds o he prvae secor and herefore may be less comprehensve han PRIVCR for some counres). The number of counres for whch we have hese fnancal marke varables and shares s 7. 5 These varables are classfed no wo broad caegores: hose relang o he bankng secor (or loosely, cred markes) and hose relang o he sock marke (or equy markes). The frs 4 The URL for he daabase s hp:// daabase.hm. 5 In keepng wh he leraure, we use he logarhm of he fnancal secor varables. 3

15 daa se, relang o he cred marke ndcaors, ncludes 20 OECD counres and 5 non- OECD counres. The second daa se, concenrang on equy marke ndcaors consss of 20 OECD counres and 29 non-oecd counres. 6 4 Emprcal Analyss Table 2 presens descrpve sascs for nvesmen, growh, and fnancal developmen daa. The varaon n he share of n GDP across counres ranges from -0.5% n Serra Leone ( ) o 0% n Sngapore ( ). GDP growh also shows varaon, rangng from - 4% for Guyana o 7% for Korea (boh for ). Fnancal developmen varables also range exensvely: capalzaon of he sock marke ranges from % for Uruguay o 26% for Souh Afrca and value raded ranges from close o 0% for Uruguay o 30% for Swzerland. Fnally, he lqudy measure (M2/GDP) ranges from 6% for Argenna o 6% for Japan. The prvae cred varable ranges from 3% for Ghana o 64% for Swzerland. Ghana and Swzerland also form he wo ends of he specrum for he bank cred varable. Ghana also has he lowes value for he share of Commercal-Cenral Bank Asses; Ausra records he hghes. 4. Growh and : Fnancal Markes as a Channel As shown n secon 2, he purpose of our emprcal analyss s o examne he fnancal markes channel hrough whch may be benefcal for growh. The heorecal model shows ha mprovemens n fnancal markes ncrease oupu by ncreasng he margnal produc of. Ths resul suggess ha one should observe ransonal growh effecs. In an nfluenal paper, Mankw, Romer and Wel (992) (MRW) derved an emprcal specfcaon based on he assumpon ha counres were unlkely o be a her seady saes and herefore ransonal dynamcs should be more mporan. We employ a specfcaon smlar o hers. We look a he drec effec of on economc growh and, based on MRW, esmae he followng equaon: 6 Here OECD counres refer o hose ha were early members and herefore exclude newer members, such as Mexco and Korea among ohers. For Value Traded, we also have Ireland n he sample, akng he number of OECD counres o 2. 4

16 GROWTH = INITIAL GDP CONTROLS ν (22) Table 3 shows resuls for a selecon of conrol varables ha nclude nal ncome, human capal, populaon growh, governmen consumpon, and a sub-saharan Afrca dummy varable. For he sample of 7 counres, s clear ha s no sgnfcan a all, whereas n he smaller sample clearly s. The resuls could be drven by he composon of he wo samples; approxmaely 28 % of he frs sample (column ()) and approxmaely 4 % of he second sample (column (3)) conss of developed counres. In columns (2) and (4), we have an expanded se of conrol varables ha nclude he black marke premum, nsuonal qualy (capured by he ICRG measure called rsk of expropraon ), rae of nflaon, and rade volume. Now he share no longer s sgnfcan n eher of he samples. Ths ncely summarzes he problem ha exss n he leraure: whereas on heorecal grounds here s a srong bass for expecng o have a posve role n growh, he emprcal evdence s fragle, o say he leas. 7 Ths ambguous effec of s wha forms par of he movaon for hs research. The regressons n Table 4 examne he role of on growh hrough fnancal markes. Followng he model developed n secon 2, we nerac wh fnancal markes and use hs as a regressor. To ensure ha he neracon erm does no proxy for or he level of developmen of fnancal markes, boh of he laer varables also were ndependenly ncluded n he regresson. Thus, we run he followng regresson: GROWTH ' 0 ' ' 2 ' ' ( FINANCE ) 3FINANCE 4CONTROLS ν = (23) As shown n Table 4, he neracon erm urns ou o be posve and sgnfcan n all columns. Each regresson uses a dfferen ndcaor for fnancal marke developmen and, hence, samples may dffer from one regresson o anoher. Column () uses BTOT, column (2) uses BANKCR, column (3) uses LLY, column (4) uses PRIVCR, column (5) uses SCAPT and column 5

17 (6) uses SVALT. 8 The man resul s ha he neracon erm s sgnfcan a he 0% level for he enre range of fnancal secor varables used. Moreover, he neracons wh LLY, PRIVCR, and BANKCR are sgnfcan a he % level. On he oher hand, fnancal marke ndcaors by hemselves are nsgnfcan and even negave for he non sock marke varables. 9 Ths may n par be due o he neracon erm capurng an mporan allocaon funcon ha he fnancal secor performs havng a well-developed fnancal secor s a means o an end and no an end n self. Ineresngly, he coeffcen of dsplays consderable varaon even whn he same sample of counres as he fnancal secor varable changes clearly makng he case for lookng a he range of fnancal secor varables raher han a few. Table 4 also repors a) he jon sgnfcance es of fnancal markes wh he neracon erm and b) he jon sgnfcance es of foregn drec nvesmen wh he neracon erm. For mos fnancal marke varables, he ess confrm he mporance of boh fnancal markes and. The hypohess ha he coeffcens of boh and he neracon beween and fnancal markes are zero canno be rejeced ourgh a he 0% level only n he case of BTOT and SVALT. No surprsngly he coeffcens of he neracon erms n hese wo regressons also repor he lowes -sascs compared o he counerpars n he oher columns. The hypohess ha he coeffcens of boh fnancal markes and he neracon beween and fnancal markes are zero s rejeced n all regressons. To ge an esmae of how mporan he fnancal secor has been n enhancng he growh effecs of, one can ask he hypohecal queson of how much a one sandard devaon ncrease n he fnancal developmen varable would enhance he growh rae of a counry 7 We repeaed hese regressons by addng he fnancal marke varables as well. Alhough hese varables were sgnfcan and posve, hey dd no aler he nsgnfcance of. 8 See he daa secon for dealed defnons. 9 The leraure ha ess he effecs of fnancal developmen on growh has no consdered and s neracon erm wh fnancal markes, hus lmng comparsons. 6

18 recevng he mean level of n he sample. 20 If we use he PRIVCR varable (.e., column (4)), urns ou ha havng beer fnancal markes would have allowed counres o experence an annual growh rae ncrease of 0.60 percenage pons durng he 20-year perod Concluson The wdespread belef ha posve exernales and spllovers due o new producs and processes made possble by and dssemnaed by learnng-by-dong or observng, neworkng and ranng of he labor force, accompaned by he search for he more sable form of capal flows followng he urmol n he emergng markes snce he lae 990s, has nduced a change n he audes of many counres owards. Several governmens n boh developng and developed counres have nroduced boh fscal and fnancal ncenves o acvely pursue foregn nvesmens. Addonally, hey have esablshed nvesmen agences o specfcally arge mulnaonal frms and have seeked o mprove he local regulaory envronmen and he cos of dong busness (See UNCTAD 999). Such polces however, lmed o aracng, do no guaranee realzaon of he poenal benefs of ha go beyond he capal ransfers o he hos counry and he wages generaes. Local condons n he recpen counry can pose bndng consrans on such spllovers. In hs paper, we sudy he consrans ha could be posed by local fnancal markes, one among many condons, n allowng posve spllovers from foregn drec nvesmen o economc developmen. In parcular, we model how poorly developed fnancal markes can adversely lm he economy s ably o ake advanage of such poenal benefs. Cross- 20 The mean value for s.003% n he 7-counry sample. Noe ha he fnancal developmen varable here s he log of he fnancal marke ndcaor. 2 The ne effec beng measured here s ( mean 2 σ ) log( prvcr σ ) 3 log( prvcr. Here mean s ).003% as menoned n he earler foonoe. σ s he sandard devaon of log PRIVCR and s equal o log( prvcr) 7

19 counry analyss provdes supporng emprcal evdence ha he level of developmen of fnancal markes s crucal for he posve effecs of on oupu o be realzed. Two relevan polcy resuls emerge from he analyss. Alhough capal flows are less volale han shor-erm capal flows, her full benefs may no accrue o he recpen counry n he absence of well-funconng fnancal markes. Hence, as much as bad fnancal markes render a counry weak o cope wh unregulaed shor-erm capal flows, hey also lm he benefs ha could accrue from long-erm sable flows. Second, he ne benef of polces dreced a aracng mgh be sgnfcanly lowered due he level of developmen of local condons. Therefore, counres should wegh he coss of nvesmen ncenves argeed a aracng MNF versus he cos of mprovng he local condons. A. Techncal Appendx In hs secon, we expand he model worked ou n he ex o ncorporae he effecs of fnancal secor effcency no producon. The foregn secor's producon funcon now can be rewren as, fd ( ) L ( K ) Y = A (A.) Followng, Roubn and Sala--Marn (992), A() s fnancal secor effcency parameer, where reflecs he neffcences n he fnancal secor. We can hnk ha he foregn frm pays wages, local npus and oher local coss as well as receves paymens hrough he local bank sysem. Thus, he more auomac eller machnes, he beer and faser he servce a he branches; and, n general he more fnancally developed he economy, he hgher he producvy of he secor. Consequenly, we assume ha A () < 0, A () > 0. Usng L = d = 0 (A.2) equaon (A.) can be rewren as: 8

20 fd ( )( ) ( K ) The amoun of foregn capal n he counry s gven by: Y = A (A.3) K fd ( ) A( )( ) = r (A.4) The amoun of foregn labor used n he secor s gven by: ( ) ( ) fd The producon funcon n he enrepreneural secor s gven by: A = k (A.5) w Y fd γ ( )( K ) S = B (A.6) B (), as n he secor, s a fnancal secor effcency parameer. We assume B ( )<0, B ()>0. A. Sac Equlbrum An agen s ndfferen beween workng for he foregn frm or sarng s own frm f Subsung he oupu, ( r) b = Y ( )( S ) w (A.7) 0 b0 Y equaon (A.6), and he wage, equaon (A.5), we oban: fd fd γ ( )( ) ( K ) ( r) b0 = B( )( K ) S ( )( S b0 A ) (A.8) Wh unform dsrbuon, we oban he followng expresson for = fd ( )( S b0 ) A( )( ) ( K ) ( r) fd γ B( )( K ) S b 0 (A.9) For he ndvdual who s ndfferen beween secors, = = c (A.0) Subsung he equaon for wages and capal: ( r,,b,,,,s ) = f 0 γ (A.) 9

21 = ( )( S b ) A( ) B 0 ( ) ( ) r ( ) A( ) r S γ ( r) b 0 (A.2) A.2 Comparave Sacs The model allows us o sudy several aspecs of an recpen economy. The ulmae objecve s o show how wll mpac economc oupu and how he magnude of hs mpac depends on local fnancal marke condons. To fnd hese resuls, we frs need o solve for how he effcency of he fnancal markes affecs he crcal level of enrepreneurshp and he level of foregn capal (). Boh channels conrbue o he full effec on oupu. Under ceran parameer condons, beer fnancal markes allow ncreased oupu n he enrepreneur secor as well as more, and more mporanly, hgher margnal producvy of capal and hence more producon n he foregn-producon secor. The frs lnk we are neresed n s he effec of changes n he effcency of he fnancal secor on he allocaon of ndvduals across secors. In oher words, we are neresed n he sgn of. As menoned above, hs wll allow us o fnd he full effecs of local fnancal markes on oupu n he economy. Rewrng (A.), we defne F as ( ) A( ) ( r,,b,,, γ,s ) = f ( r,,s,b,,, γ ) 0 F 0 0 = (A.3) γ F = ( ) B( ) S A( ) ( r )( S b0 ) ( r) b0 (A.4) r r We can use he mplc funcon o fnd he effec of changes n he effcency of he fnancal secor n he allocaon of ndvduals across secors : ( r,,b,,, γ,s ) ( r,,b,,, γ,s ) F 0 0 = F( r,,b0,,, γ,s ) (A.5) 20

22 ( ) ( )( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )( ) ( ) ( ) ( ) S r A B r A A A A A B B S r A B b S 0 γ γ = (A.6) In he numeraor, he erm (S-b 0 ) s greaer han zero by assumpon, and capures he margnal ncrease n he cos of borrowng due o hgher nermedaon coss. As hs cos ncreases, fewer agens become enrepreneurs and more people work for he foregn frm. The erm ( )( ) ) B( ) B ( S r ) )A( ( B γ reflecs how hgher fnancal coss affec he enrepreneural secor producon funcon. I s negave by assumpon, ( ) < 0. If hgher nermedary coss affec he enrepreneural secor, fewer people become enrepreneurs and more agens go o he secor. The erm ( )( ) ) A( ) A ( S r ) )A( ( B γ s also negave snce A () < 0. Hgher fnancal coss negavely affec he secor, hus lowerng he level of foregn capal and hus he exernales n favor of he enrepreneural secor. Gven hs effec, fewer agens become enrepreneurs. The las erm ( ) ( ) ( ) ( ) ( ) r A A A reflecs how hgher fnancal coss affec he secor, and s negave by assumpon. If becomes more cosly o produce n he secor, and wages fall and, herefore, he enrepreneural secor becomes more aracve. The denomnaor n equaon (A.6) reflecs he ne effec of changes n n he enrepreneural secor and s posve. The followng expresson summarzes he effecs: 2

23 : cos o sar a frm : workers o secor = : enrep. secor less producve : workers o frm enrep secor : less eff., less exern. : workers o secor : less effcen : workers o frm The effec of hgher nermedary coss on, fd K s: K fd ( r,,b γ ) ( ) ( ) 0,,,,S A A ( ) = r A( ) (A.7) The frs erm s negave, snce we assumed A () < 0. The value of he second erm depends on. Thus, he value of K fd depends on he sum of a negave effec due o hgher nermedary coss plus a posve effec due o a hgher number of workers. Ths resul depends, of course, on he assumpon ha s araced by cos of capal consderaons. We are also neresed n how changes n he nermedary coss affec he fnal oupu. Toal oupu s he sum of he oupu n each secor: Y = Y Y d (A.8) Subsung (A.) and (A.6) no (A.8) and usng (A.2): Y fd fd γ ( r,,b,,, γ,s ) = A( )( ) ( K ) ( ) B( )( K ) S 0 (A.9) The effec of hgher nermedary coss on he oupu level, ( r,,b,,, γ, S) Y 0 s gven by Y ( r,,b,,, γ,s) = g( r,,b,,,, S) (A.20) 0 γ 0 Y = A γ ( )( ) ( K ) ( ) ( )( ) fd fd B K S ( ) ( )( ) ( ) ( ) ( )( ) fd fd A K B K γ S fd K fd fd γ [ A( )( ) ( K ) B( )( K ) S ] (A.2) The frs squared bracke erm s he loss n producon due o he hgher neffcences n he fnancal secor, whch has a negave sgn. The second squared bracke erm s posve and 22

24 K mulples fd. The value of he hrd erm depends on he effec of hgher nermedaon cos on he allocaon of workers, and he erm on he hrd bracke. A fd fd γ ( )( ) ( K ) B( )( K ) S (A.22) fd γ Usng (A.), (A.22) s negave f Y B( )( K ) S <. Thus, f he producon n he ndfferen agen s hgher han n he enrepreneural secor, hen he value of (A.22) s negave. Our man objecve s o undersand how he developmen of fnancal markes affecs he role wll have on growh. Techncally hs means ha we need neresed on how fnancal marke neffcency affecs he margnal produc of foregn capal (). The margnal produc of capal s gven by: = MPK K Y = fd fd γ ( ) A( )( ) ( K ) ( ) B( )( K ) S (A.23) Ths equaon shows boh effecs of foregn capal on oupu. The frs effec s he drec effec on producon, whch s he frs erm. The second effec, whch s capured by he second erm, s he spllover effec of foregn capal on he enrepreneural secor. Boh erms are posve, whch says has a posve overall effec on oupu. For our man objecve as menoned above, we need o show he effec of fnancal developmen on margnal producvy of. The effec of hgher nermedaon coss on he margnal produc of capal s gven by: MPK = h( r,,b,,, γ, S ) (A.24) 0 γ [ S ] ( ) fd fd = ( ) A ( )( ) ( K ) ( ) B ( )( K ) MPK 23

25 fd fd γ ( ( ) A( )( ) ( K ) B( )( K ) S ) fd fd 2 γ ( ( ) ( )( ) ( ) ( )( ) ( )( ) ) K A K B K S (A.25) The frs-squared erm s negave snce we assumed ha A () < 0, B ( ) < 0. The sgn of he second erm depends on parameer values and he value of. Fnally, he hrd erm nsde he brackes s negave. A.3 Numercal Analyss Snce equaons K fd,, MPK Y, ( ) do no have smple close form soluons, we ploed equaons (A.6), (A.7), (A.2) and (A.25) agans dfferen parameer values. The objecve of hs secon s no o compleely characerze he doman of hese equaons, bu o show ha for reasonable values > 0, K fd < 0, MPK Y < 0, and ( ) < 0. Fgure 2, shows smulaons for he followng benchmark case:. For he fnancal secor effcency funcons, A ( ), B( ) S = ; = 0.3; = 0.6; r = 0.0; γ = a b assumed A ( ) =, a = ; B( ) =, b =, we. Fgure 2.a plos he change n he allocaon of workers n he foregn secor due o changes n he nermedaon cos (labeled de/dd n he graph) agans dfferen values of he exernaly parameer. For he benchmark parameer values, s always posve. Fgure 2.a plos as well he change n he margnal cos of foregn capal due o changes n he nermedaon cos (labeled dmpk/dd n he graph) agans. For he benchmark parameer values, s always negave. Fgure 2.b plos he change n oal oupu due o changes n he nermedaon cos (labeled dy/dd n he graph), whch s always negave for dfferen values of. Fnally, he change n he level of capal due o changes 24

26 n he nermedaon cos (labeled dk/dd n he graph) s always negave for dfferen values of, as seen n Fgure 2.b Fgure 2.2a plos he change n he allocaon of workers n he foregn secor due o changes n he nermedaon cos agans dfferen values of b 0, he nal asse level. For he benchmark parameer values, s always posve. The same fgure shows ha K fd s always Y MPK negave. Fgure 2.2b shows ha boh and vary nversely wh changes n he nal asse level. Fgures 2.3a and 2.3b plo (A.6), (A.7), (A.2) and (A.25) agans dfferen values of b, whch s he coeffcen n he fnancal effcency parameer n he foregn producon funcon. For he benchmark parameer values, s always posve and K fd, Y, MPK s always negave. Fnally, Fgures 2.4a and 2.4b plo he dfferen equaons agans dfferen values of a, whch s he coeffcen n he fnancal effcency parameer n he enrepreneural secor. For he benchmark parameer values, s always posve and K fd, Y, ( ) MPK s agan always negave. Fgure 3, shows a dfferen smulaon for he followng benchmark case: S = ; = 0.; = 0.67; a = ; b = 3; r = 0.; γ =. We ploed each equaon agans, b 0, a and b. K We obaned he desred sgns for all equaons. In general, he value of fd,, Y, and ( MPK ) depend on he parameer values chosen. Therefore, we feel ha an economerc analyss s essenal o undersand he neracon beween, local capal markes, and growh. 25

27 A2. Daa Appendx A2. Counres n he Samples. Sample of 7 counres for whch daa on cred markes are avalable (BANKCR, BTOT, PRIVCR, LLY). 2. Sample of 49 counres for whch daa on SCAPT and SVALT are avalable. 3. Sample of 53 counres for whch SVALT was avalable bu SCAPT was no: Sample of 49 plus Cosa Rca, Honduras, Ireland, and Panama. Ls: Algera (), Argenna (,2), Ausrala (,2), Ausra (,2), Bangladesh (2), Belgum (,2), Bolva (), Brazl (,2), Cameroon (), Canada (,2), Chle (,2), Colomba (,2), Congo (), Cosa Rca (), Cyprus (,2), Denmark (,2), Domncan Republc (), Ecuador (), Egyp (,2), El Salvador (), Fnland (,2), France (,2), Gamba (), Germany (,2), Ghana (,2), Greece (,2), Guaemala (), Guyana (), Ha (), Honduras (), Inda (,2), Indonesa (,2), Iran (), Ireland (), Israel (,2), Ialy (,2), Jamaca (,2), Japan (,2), Jordan (2), Kenya (,2), Korea (,2), Mala (), Malaw(), Malaysa (,2), Mexco (,2), Neherlands (,2), New Zealand (,2), Ncaragua (), Nger (), Norway (,2), Paksan (,2), Panama (), Papua New Gunea (), Paraguay (), Peru (,2), Phlppnes (,2), Porugal (,2), Senegal (), Serra Leone (), Sngapore (2), Souh Afrca (,2), Span (,2), Sr Lanka (,2), Sudan (), Sweden (,2), Swzerland (,2), Syra (), Thaland (,2), Togo (), Trndad Tobago (,2), Turkey (2), Uned Kngdom (,2), Uned Saes (,2), Uruguay (,2), Venezuela (,2), Zmbabwe (,2). A2.2 Daa Sources and Descrpons Foregn Drec Invesmen: The ne nflows measure he ne nflows of nvesmen o acqure a lasng managemen neres (0 percen or more of vong sock) n an enerprse operang n an economy oher han ha of he nvesor. I s he sum of equy capal, 26

28 renvesmen of earnngs, oher long-erm capal, and shor-erm capal as shown n he balance of paymens. Source: IMF Inernaonal Fnancal Sascs. Oupu levels and growh: Oupu level and growh daa are he growh of real per capa GDP, consan dollars. Source: World Developmen Indcaors (WDI), World Bank (2000). Value raded: Value of sock radng relave o he sze of he economy. Source: World Bank Fnancal Srucure Daabase. (hp:// /daabase.hm) Capalzaon: Capures he sze of he sock marke and measures he average value of lsed domesc shares on domesc exchanges n a year as a share of he sze of he economy (he GDP). Source: World Bank Fnancal Srucure Daabase. Lqudy (LLY): Lqud Lables of he fnancal sysem (currency plus demand and neres bearng lables of he fnancal nermedares and nonblank fnancal nermedares) dvded by GDP. Source: World Bank Fnancal Srucure Daabase. Prvae cred (PRIVCR): The value of creds by fnancal nermedares o he prvae secor dvded by GDP. I excludes creds ssued by cenral and developmen banks. Furhermore, excludes cred o he publc secor and cross clams of one group of nermedares on anoher. Source: World Bank Fnancal Srucure Daabase. Bank Cred (BANKCR): Cred by depos money banks o he prvae secor as a share of GDP. Source: World Bank Fnancal Srucure Daabase. Commercal-Cenral Bank (BTOT): Rao of commercal bank domesc asses dvded by cenral bank plus commercal bank domesc asses. Source: World Bank Fnancal Srucure Daabase. Domesc Invesmen: Gross domesc nvesmen s a measure of he oulays on addons o he fxed asses of he economy plus ne changes n he level of nvenores. Source: WDI (2000). Inflaon: Percenage changes n he GDP deflaor. Source: WDI (2000). 27

29 Governmen Consumpon: Toal expendure of he cenral governmen as a share of GDP. I ncludes boh curren and capal (developmen) expendures and excludes lendng mnus repaymens. Sources: WDI (2000). Trade Volume: Expors plus mpors as a share of GDP. Source: WDI (2000). Schoolng: Human capal measured as he average years of secondary schoolng n oal populaon. Source: Barro and Lee (994). Updaed verson downloadable from: hp:// Bureaucrac qualy: The nsuonal srengh of he economy. Hgh levels of qualy mply ha he bureaucracy has he srengh and experse o govern whou drasc changes n polcy or nerrupon o publc servces. Source: Inernaonal Counry Rsk Gude (ICRG). Rsk of expropraon: The probably ha he governmen may exproprae prvae propery. Source: ICRG. Black marke premum: I s calculaed as he premum n he parallel exchange marke relave o he offcal marke (.e., he formula s (parallel exchange rae/offcal exchange rae- )00). The values for ndusral counres are added as zero. Source: World Bank. (hp:// 28

30 References Aken, B.J. and A. Harrson (999), Do Domesc Frms Benef from Drec Foregn Invesmen? Evdence from Venezuela, Amercan Economc Revew, 89:3, Barro, R. and J-W Lee (996), Inernaonal Measures of Schoolng Years and Schoolng Qualy, Amercan Economc Revew, 86(2), Barro, R. and Sala--Marn (995), Economc Growh, McGraw-Hll Inc. Barro, R. and Sala--Marn, X. (997), Technology dffuson, convergence and growh, Journal of Economc Growh, 2, -26. Beck, Thorsen, R. Levne and N. Loayza (2000b), Fnance and he Sources of Growh, Journal of Fnancal Economcs, 58: Beck, T., Levne, R., Loayza, N. (2000), Fnance and he Sources of Growh. Journal of Fnancal Economcs 58, Blomsrom, M. and A. Kokko (997), The Impac of Foregn Invesmen on Hos Counres: a Revew of he Evdence, World Bank Research Paper, 745. Borenszen, E. J. De Gregoro and J-W. Lee (998), How Does Foregn Drec Invesmen Affec Economc Growh?, Journal of Inernaonal Economcs, 45, Boyd, John H. and E. Presco (986), Fnancal Inermedary-Coalons, Journal of Economc Theory, vol. 38, no.2. Carkovc, M., Levne, R. (2000), Does Foregn Drec Invesmen Accelerae Economc Growh? Unversy of Mnnesoa, Workng Paper. Caves, R. (996), Mulnaonal Enerprse and Economc Analyss, Cambrdge Unversy Press. Caves, R. (999), Spllovers from Mulnaonals n Developng Counres: Some Mechansms a Work. Manuscrp prepared for he Wllam Davdson Conference on The Impac of Foregn Invesmen on Emergng Markes, Unversy of Mchgan, Ann Arbor, July

31 Coe, Davd T., Helpman, E. and A. Hoffmaser (995), Norh-Souh R&D Spllovers, NBER Workng Paper, Fernandez-Aras, E. and R. Hausman (2000), Foregn Drec Invesmen: Good Choleserol?, IADB Workng Paper, 47. Galor, O. and J. Zera (993), Income Dsrbuon and Macroeconomcs, Revew of Economc Sudes, 60, Goldsmh, R. W. (969), Fnancal Srucure and Developmen, New Haven, Conneccu: Yale Unversy Press. Greenwood, J. and B. Jovanovc (990), Fnancal Developmen, Growh and Dsrbuon of Income, Journal of Polcal Economy, vol 98, no.5. Grossman, G., Helpman, E. (99), Innovaon and Growh n he Global Economy, Cambrdge: MIT Press. Inernaonal Moneary Fund (2000), Inernaonal Fnancal Sascs on CD-ROM. Washngon, DC. Kng, R. and R. Levne (993a), Fnance and Growh: Schumpeer Mgh be Rgh, Quarerly Journal of Economcs, 08, Kng, R. and R. Levne (993b), Fnance, Enrepreneurshp and Growh: Theory and Evdence, Journal of Moneary Economcs, 32, Levne, R., N. Loayza and T. Beck (2000a), Fnancal Inermedaon and Growh: Causaly and Causes, Journal of Moneary Economcs, 46:, Levne, R. and S. Zervos (998), Sock Markes, Banks and Economc Growh, Amercan Economc Revew, 88, Mankw, Gregory N., Davd Romer and Davd N. Wel (992), A Conrbuon o he Emprcs of Economc Growh, Quarerly Journal of Economcs, May: McKnnon, R.I. (973), Money and Capal n Economc Developmen, Washngon: Brookngs Insue. 30

32 Rajan, R. and L. Zngales (998), Fnancal Dependence and Growh, Amercan Economc Revew, 88, Razn, A., E. Sadka and C. Yuen (999), An Informaon-Based model of : he Gans From Trade Revsed, NBER Workng Paper 6884,. Roubn, N. and Sala--Marn, X. (992), Fnancal Represson and Economc Growh, Journal of Developmen Economcs, 39(), Shaw, E. S. (973), Fnancal Deepenng and Economc Developmen. New York: Oxford Unversy Press. Soo, Marcelo (2000), "Capal Flows and Growh n Developng Counres: Recen Emprcal Evdence," OECD Developmen Cenre, Techncal Papers No. 60. UNCTAD (999), Foregn Drec Invesmen and Developmen. UNCTAD (2000), World Invesmen Repor. Uned Naons, New York. World Bank (2000), World Bank Developmen Repor. The World Bank, Washngon, DC. World Bank (2000), World Developmen Indcaors on CD-ROM. The World Bank, Washngon, DC. World Bank (200), Global Developmen Fnance Repor. The World Bank, Washngon, DC. Wurgler, J. (2000), Fnancal Markes and he Allocaon of Capal, Journal of Fnancal Economcs, 58, Xu, Bn (2000), Mulnaonal Enerprses, Technology Dffuson, and Hos Counry Producvy Growh, Journal of Developmen Economcs,

33 Fgure Counres n hs plo are all he 7 counres for whch all accompanyng daa are avalable and form he frs sample n Table 2. 32

34 Fgure hea de/dd dk/dd dy/dd e de/dd dk/dd dy/dd Fgure 2a Fgure 2b de/dk dk/dd de/dd dk/dd dy/dd dy/dd b a Fgure 2c Fgure 2d Benchmark Parameer values: S-b 0 =; =0.; =0.67; r=, b=3; S γ =; =0.5 A()=a/; a=3, =; B()=b/;. 33

35 Fgure de/dd dk/dd dy/dd de/dd dk/dd dy/dd Thea e Fgure 3a Fgure 3b de/dk dk/dd dy/dd de/dd dk/dd dy/dd b a Fgure 3c Fgure 3d Benchmark Parameer values: S-b 0 =; =0.; =0.67; r=0.; S γ =; =0.5. A()=a/; a=, =; B()=b/; b=. 34