Are Net FDI Flows and Reversals of Capital Flows a Result of Output Growth? *

Transcription

1 Are Ne FDI Flows and Reversals of Capial Flows a Resul of Oupu Growh? * Gilad David Aharonoviz a,b Washingon Sae Universiy, School of Economic Sciences and James Miller b Washingon Sae Universiy, School of Economic Sciences This version: Ocober 2009 Absrac Lieraure noes many facors as affecing capial flows, bu he effecs of hese flows over he recipien economies and he overall effec over growh are highly debaable. This sudy claims ha alhough capial flows may be required for he increase in oupu, oher forces are causing his growh and creaing he demand for capial. We consruc a model in which growh requires boh decisions of firms regarding raining of managers in order o expand, represening he absorpion capaciy of he firms, and capial for he firms expansion. The model shows ha in early sages of developmen, when oupu is low, capial inflows are increasing wih an increase in he oupu, bu are no he cause for he oupu growh. However, when oupu is higher, an increase in he oupu is associaed wih financial ouflows, since he local savings are increasing by more han he local demand for capial. We use cross-counry regressions o show ha capial flows are inline wih he predicions of he model, and manage o explain half of he variaions in ne FDI flows per capia using he sage of developmen. Many policies regarding capial flows may be, herefore, irrelevan. KEYWORDS: capial flows, reversals, FDI, economic developmen, firms, growh, on-he-job raining JEL CLASSIFICATION: F21, O41, O16, F43, J24 * We are graeful o paricipans a he Midwes Inernaional Economics meeing (Fall 2008) and Washingon Sae Universiy seminar for heir commens and suggesions. a Corresponding auhor, aharonoviz@wsu.edu, phone , fax b Washingon Sae Universiy, School of Economic Sciences, PO Box , Pullman, WA , USA.

2 1 Inroducion Research regarding capial flows menions several facors which are affecing hese flows: labor coss, marke size, labor availabiliy, preferenial policies, inellecual propery righs and oher facors (see, for example, Helpman, 1984, Markuzen, 1984, Lee and Mansfield, 1996, Gasange e al., 1998, Billingon, 1999, Cheng and Kwan, 2000, Bevan and Esrin, 2004, Javorcik, 2004, and Verdier, 2008). The effecs of he capial flows over he recipien economies are debaable, wih opposing argumens and findings in he lieraure. Effecs include, among ohers, possible echnological spillovers and produciviy, bu some claim ha here are no spillovers (Aiken and Harrison, 1999, Javorcik, 2002, Saggi, 2002, Javorcik and Spaareanu, 2008, and Bonfiglioli, 2008); posiive and negaive effecs over local invesmen (Henry, 2000, Agosin and Mahcado, 2005, Barrios e al., 2005, and Bonfiglioli, 2008); changes o he local wages (Lipsey, 2002, and Henry and Sasson, 2008); and financial developmen, alongside possible financial insabiliy (Beck e al., 2000, Sigliz, 2000, and Klein and Olivei, 2008). The overall effec of FDI and capial flows on he economic growh of he recipien economy is quesionable as well. Some claim ha here is a posiive shor-run effec over growh, which leads o a long run level-effec over he oupu (Quinn, 1997, Sigliz, 2000, Levine, 2001, Bekaer e al., 2005, and Hur e al., 2006). Ohers find mixed or negaive effecs, or claim ha he causal relaionship beween capial flows and oupu growh has no been proven (Rodrik, 1998, Lipsey, 2002, Prasad e al., 2003, and Eichengreen and Leblang, 2003). This sudy claims ha alhough capial flows may be required for he increase in oupu, hey are no he cause for his increase. Oher forces are enabling he growh and creaing he demand for capial, and capial flows are merely meeing his demand when capial is no locally available. Thus, hey enable he increase in oupu, bu do no cause i. We consruced a model in which growh requires boh decisions regarding raining of managers by firms in order o expand, represening he poenial of he firms o absorb capial and uilize he producion echnology, and addiional capial for he expansion of hese firms. While he laer is he neoclassical force behind he posiive effec of FDI over growh, he former is much less common. The lieraure dealing wih firms has long acknowledged he imporance of managers and he firms need o rain managers in order o expand (Coase, 1937, Penrose, 1959, Slaer, 1980, and Dias and McDermo, 2006), and he imporance of managers o economic performance has been acknowledged as well (Lucas, 1978, and Bursein and Monge-Naranjo, 2009), bu raining of managers by firms has rarely been aggregaed ino a 2

3 macroeconomic model. 1 An excepion is Aharonoviz (2009), which did no include capial or capial flows ha are included here, allowing for heoreical and empirical analysis of capial flows. In he model, firms are raining managers in order o expand and open addiional unis, an expansion ha also requires capial. Local supply of capial is creaed by local savings, and he difference beween local savings and local demand for capial is bridged by capial flows. The model shows ha in early sages of developmen, when oupu is low, capial inflows are increasing wih an increase in he oupu. These inflows are required o enable he increase in oupu, bu are no he cause or he consrain on oupu growh. In laer sages of developmen, when oupu is higher, an increase in he oupu is associaed wih financial ouflows, since he local savings are larger han he local demand for capial. On a broader perspecive, reaing he inernaional supply of capial as a echnology capial, capial ha embodies new echnology in i, or as inernaional availabiliy of capial and echnology, and he local supply of managers as he poenial of he economy o absorb he capial and he echnology, he model emphasizes he imporance of he local poenial compared o he exernal resources. The model predics an invered U-shaped relaionship beween capial flows used for invesmen and he sage of developmen of he economy. We characerized his relaionship in a large sample of counries, using ne FDI per capia as he measure for capial flow and GDP per capia of an earlier year as a measuremen for he sage of developmen. We found ha FDI per capia is increasing wih oupu in earlier sages of developmen bu decreasing and becoming negaive (capial ouflows) in laer sages, in line wih he predicion of he model. We managed o explain 50% of he variaions in FDI using he sage of developmen and wo conrol variables. These findings conribue o wo srands of he lieraure. Regarding he facors affecing financial flows and he conribuion of hese flows o economic growh menioned above, he model shows ha alhough he capial is imporan in enabling he increase in he amoun of producion unis, which is leading o he increase in oupu, i may no be he facor causing he increase. The cause may be he inernal developmen, leading o demand for exernal capial. The empirical evidence, showing ha he size of he flow is affeced by he sage of developmen of a given counry, suppors his view. This is in line wih Manova (2008), showing ha secors in need of credi are affeced more han oher secors by financial liberalizaion. Our claim is similar, bu deals wih he need of he 1 According o he 2005 World Bank Enerprise Survey for Easern Europe and Cenral Asia (Easern Europe and Cenral Asia Business Environmen and Enerprise Produciviy Survey (BEEPS), available a % of he firms in hese counries provide raining o heir workers, and providing raining and expansion of sales are posiively correlaed. However, he daase is no rich enough o analyze causaliy. 3

4 enire economy for credi, i.e., he demand for capial, which is se by he inernal developmen of he economy. Anoher srand is dealing wih sudden sops and capial flow reversals (see, for example, Calvo, 1998, Kaminsky and Reinhar, 1999, Bordo e al., 2001, Huchison and Noy, 2006, and Glick e al., 2006), which are considered o cause a decrease in oupu. Alhough our model canno explain a sharp reversal, we show ha gradual reversals are a naural evoluion of an economy from having a ne exernal deb (ne borrower) o ne exernal asses (ne credior). The decrease in growh rae associaed wih he reversal may be a naural one as well, caused by he end of a rapid cach-up of he economy wih he res of he world, leading o a maure growh rae and a slowdown in he demand for capial. When accompanied by high savings raes, i may lead an economy o become a capial credior raher han a debor. This mechanism can also be a facor conribuing o sharper reversals of flows, bu no heir sole cause. The paper proceeds as follows: secion 2 presens he model. Secion 3 analyzes he evoluion of he economy and of he capial flows based on he model. Secion 4 provides empirical analysis of capial flows. Secion 5 concludes. 2 The Model Assume a small open economy wih a populaion of N infiniely living agens. Each agen is employed every period in he radiional secor, as a worker in he manufacuring-echnological secor, or as a manager in ha secor. Denoe wih, and he quaniies of workers in he radiional secor, workers in he echnological secor and managers, such ha: F L M N (1) Producion of a single good, ha can serve as consumpion or capial good, akes place in wo secors, radiional and echnological. The radiional secor uses labor only and exhibis a consan marginal produciviy (which is normalized o 1). The echnological secor is divided ino producion unis (e.g. plans, firms). Each one of hose producion unis requires a manager in order o exis and has a producion funcion in he form of: 1 y f ( l, k ) A( l k ) (2) 4

5 where y is he oupu of he producion uni, A represens he produciviy, l is he number of employees including he manager, k is he capial and 0,1. 2 represens he span of conrol problems of a single manager, such ha a single manager canno simply increase k and l and increase he oupu by he same proporion, bu raher more managers are required in order o mainain he same produciviy when increasing he size of he firm. Accordingly, alhough he producion funcion of each uni is no homogeneous of degree 1, he echnological secor as a whole does exhibi homogeneiy of degree 1, when he producion facors are managers, workers and capial. Increasing he quaniy of managers, workers and capial in he same proporion leads o an idenical increase in he quaniy of he producion unis, and hus increases he oupu by he same proporion. More specifically, having m managers (i.e., m producion unis), addiional l-m workers (remember ha l includes he managers) and k unis of capial allows a firm, which is no raining any addiional managers, o produce / /, and homogeneiy of degree 1 is esablished since Thus, managers represen a cerain absorpion capaciy for labor and capial. Noe also ha having more managers in he economy increases he oupu, since labor and capial, which exhibi diminishing marginal produciviy wihin each uni, are divided among more producion unis. Capial does no depreciae. For simpliciy, we assume ha A=1. Labor markes are compeiive. Capial is freely flowing across counries, such ha each firm in he economy can ren capial a a cos of r, and he economy as a whole can borrow (or save) capial a he same ineres rae. The saving behavior of he individuals is characerized below. Training managers is he hear of his model. An agen can become a manager only hrough raining in an exising producion uni (firm). Each period, each exising producion uni (firm) is headed by a pre-rained manager and can rain new managers, a a cos. The newly rained managers can immediaely funcion as managers (i.e., o enable he exising producion uni o open new producion unis under i), alhough he new managers and he new producion unis canno rain addiional managers a he same period. The new managers are obliged o work a heir producion unis in ha same period (see below and secion 3.1 regarding a longer period), and hus raining is profiable for he firm, bu in laer periods hey are free o leave and head new producion unis (firms), which can rain new managers and hus open addiional producion unis, or say and head a producion uni ha can rain new managers and open addiional producion unis. The cos c of raining m new managers in a single producion uni in period is denoed by: 2 An addiional consrain on and is ha l 1, see subsecion

6 c c(m) (3) where c ( 0) 0, c(m) is coninuous and wice differeniable, / 0,, / 0, and c / m 0. These condiions resemble sandard m0 invesmen-cos funcion. The funcion, B>0, will serve as an example. The producion unis canno charge managers for heir raining, i.e., hey are paid heir marginal produciviy as workers, like every oher worker (noe ha l includes he manager). A longer period in which a manager has o work for he firm, or charging managers for heir raining, affec he pace of developmen, bu no is general form; see subsecion 3.1 for elaboraion. A ypical saving behavior is having a low saving rae a low levels of income, and a higher saving rae a higher levels of income. We have simplified his behavior by assuming a saving rae of zero for low levels of income, similar o he assumpion in he lieraure dealing wih povery raps, and a posiive saving rae for higher levels of income. Thus, individuals ha earn up o he wage of he radiional secor do no save, bu raher consume heir enire income. Individuals ha earn a labor income ha exceeds he wage of he radiional secor save s ou of he excess wage. The saving rae ou of he reurns o capial (r) is, and we denoe he reurns which are saved wih, such ha. Consumers in he economy canno borrow for consumpion purposes. Therefore, impors (and expors) are of capial goods. In every period exising managers (i.e., all hose ha worked as managers in he previous period, wheher hey were rained in ha period or before) are running (and working in) he producion unis (firms). Each of hese unis decides how many managers o rain (and herefore how many new unis o open) and how much capial and workers o hire. The res of he populaion, if here is any, works in he radiional secor. Producion, consumpion and savings ake place, and we move o he nex period. 3 Developmen Pah Prior o he developmen process all he oupu is semming from he radiional secor, which yields an oupu of N per period. Assume ha a period 1 one manager has been exogenously inroduced o he economy. One may consider, for example, a mulinaional corporaion which opens a plan a ha economy. We analyze he developmen pah of he economy, he evolvemen of he capial sock and he capial flows. 6

7 3.1 Firs phase The firs phase refers o he periods in which here are sill workers in he radiional secor, i.e., 0. As long as here are workers in ha secor, he quaniy of workers in every producion uni is se by equaing he marginal produciviy of he workers in he echnological secor o 1, he marginal produciviy in he radiional secor, which equals he wage (w=1). Capial is deermined by equaing is marginal produciviy o is cos, r. The decreasing marginal produciviy of boh facors ogeher ( 1) ensures ha his is achieved for a finie quaniy of workers and capial, noed wih and. For an explici soluion, one should derive (2) and solve, / 1 and, /, leading o: 1 (1 ) ( 1) 1 (1 ) 1 (1 ) l and r 1 (1 ) (1 ) 1 1 k. 3 r Each period, each firm decides how many new managers o rain. The profi of a firm, which is raining m new managers, excluding he wage of he already exising manager, is: ( m) ( m 1)( f ( l, k) w( l 1) rk) wm c( m) (4) where w is he wage of a worker, and w=1 as long as he radiional secor exiss. Noe ha he profi is simply he oupu minus he wages of he workers, he cos of capial, he wage of he newly rained managers (which are paid like workers) and he cos of raining. 4 Since he labor marke for pre-rained managers is compeiive, he wage of he pre-rained manager equals he marginal produc, which is he mere exisence of he firm, and herefore equals he enire profi of he firm, including is new producion unis. Consequenly, one may regard ha rained manager as he owner-enrepreneur as well. 3 (1 ) ( 1) 1 (1 ) 1/1 Since l includes he manager,, and r mus saisfy ( (1 ) / r ) 1. 4 Throughou he firs phase, while he radiional secor exiss, he creaion of a second firm (i.e., managers ha were rained by he firs manager and laer leaving o open heir own firms) does no affec he firs one, and herefore, raining oher managers does no affec fuure profis in his phase. I does bring upon, evenually, he second phase's compeiion, bu any reasonable discoun facor makes i negligible for a decision aken early in he firs period. Once here are many firms, each firm is small enough such ha is decision does no affec fuure compeiion. 7

8 Proposiion 1: The echnological secor exhibis consan growh rae during phase 1. Proof: given he wage of he pre-rained manager, firms maximize profis (equaion (4)) over m, i.e., /, /0. Since, and w=1, m is se according o: c( m) m f ( l, k ) l rk (5) which is equaing he marginal cos of raining one more manager o he marginal benefi one more producion uni minus he wage of is workers and he cos of is capial. 5 Since / is an increasing funcion, (5) is achieved for a consan value of m, denoed wih, and denoe accordingly he profis of a firm before paying he pre-rained manager wih. Therefore, every period each firm rains new managers and opens new producion unis, such ha i has 1 producion unis. In he following period all hese 1 managers are prerained, and every one of hem heads a firm-uni which rains new managers and opens new producion unis, ec. The oupu per uni is no changing, and herefore is he growh rae of he echnological secor. QED Walras' Law of Markes allows for omiing he analysis of he equilibrium in he marke for he oupu. Noe, however, ha equilibrium is easy o show. Firms have zero profis, since all he profis are paid as wage o he pre-rained manager. Paymens o capial are parly saved as capial and parly consumed. Wages o workers and managers are consumed or saved as capial based on he decision rule menioned above, hus he enire oupu is allocaed. Any capial requiremen above he accumulaed local savings is capial inflow, while any accumulaed local savings above he capial requiremen urns ino capial ouflow, see subsecion 3.3. Since in fuure periods firms will have o pay he managers hey are raining in he curren period a compeiive wage, which equals heir marginal produc, i.e., he exisence of he firm, and since all he populaion is equally alened, people would agree o pay he firms in order o be rained. We have assumed no paymen and one period in which he manager is obligaed o work for he firm. However, he model s oucome is invarian o hese assumpions, as is explained below. Firms can charge he price of raining in wo differen forms: (1) a firm can pay he manager in raining less han he wage of a worker, hus 5 The opimal m may be a fracion, while he quaniy of managers mus be an ineger. The inerpreaion here is ha some firms rain he ineger above m and some he ineger below, such ha m is he average over all he firms. 8

9 subracing he price of he raining from he wage. Obviously, his price is bounded by he wage iself, since one does no find workers paying in order o work. While voluneering and inernships a low wage are common, negaive wage is unreasonable and is illegal in many places (one may also consider liquidiy consrains). A bounded price is increasing he righ-hand-side of equaion (5), bu since i is bounded he equaliy sill holds for a finie m, and hus he naure of he developmen process described in proposiion 1 remains unchanged. One can prove ha he opimal price of raining does no increase over ime (managers rained in laer periods enjoy heir raining for fewer periods), such ha he opimal m remains consan or decreases over ime, bu is obviously above he calculaed in proposiion 1. (2) Firms can reques ha a manager would work for he firm a a lower wage for several periods pas he raining period. However, firms canno resric he employmen of he newly rained managers for an excessively long duraion, since his is generally considered illegal. Similarly o he previous case, as long as his is a finie duraion, he righ-hand-side of equaion (5) is increasing by a finie amoun, and equaliy is achieved for a finie m, wihou changing he naure of he developmen process. Since calculaing he opimal m becomes much more complicaed, while no affecing he naure of he process, we have assumed ha he manager is obligaed o say for only one period. 6 For he explici example one ges: 2Bm (1 ) (1 ) ( 1) 1 (1 ) (1 ) 1 (1 ) 1 (1 ) 1 r (1 ) (1 ) 1 r r r, and is calculaed accordingly. Noe ha is consan over ime. The developmen pace of he whole economy is differen from he growh rae of he echnological secor, since in he iniial periods a significan porion of he counry's oupu is semming from he radiional secor. However, as he echnological secor is expanding on he expense of he radiional one, is effec over aggregae economic growh is increasing, up o he poin where he radiional secor disappears. Throughou his process oupu is growing, since he average produciviy of labor in he producion unis is greaer han ha of a worker in he radiional secor. As long as here are workers in he radiional secor (i.e., ), he developmen process occurs as described above. However, from he period in which firms will have o hire less han. This phase of developmen, he second phase, is analyzed below. 6 See also Aharonoviz (2009) for a more formal analysis in an economy wihou capial. 9

10 3.2 Second Phase As he radiional secor disappears, firms are hiring fewer workers per producion uni, and herefore paying hem a higher wage. This reflecs on he profiabiliy of he firm and accordingly on he number of producion unis each firm is opening and managers i is raining. Proposiion 2: second phase growh rae of he quaniy of producion unis is decreasing over ime. Proof: Since labor markes are compeiive, each period, afer he decision regarding m is made and is se, each producion uni hires / workers, hus having / workers including he manager, and he wage of a worker is he marginal produciviy of labor. Capial is se by equaing is marginal produciviy o is cos (ime index is omied for clariy of presenaion): 1 f ( l, k) (1 ) N k w l M (6) and f ( l, k) N (1 ) 1 (1 ) k r k M (7) Noe he quaniy of workers including he manager in he second phase wih, and he wage and capial solving (6) and (7), given, wih and. I is easy o see ha and are decreasing in M, and by subsiuion of (7) ino (6) ha is increasing in M. 1 where is he expeced number of new unis per firm. Since each firm is small, i can neglec is own effec over and herefore over and he subsequen, and when deciding how many new unis o open. Therefore, he maximizaion problem faced by a firm, given he wage of he pre-rained manager, is: ~ ~ ~ ~ Max{( m 1)( f ( l, k ) w~ ( l 1) rk ) wm ~ c( m)} (8) m ~ ~ ~ Accordingly, ( m) / m f ( l, k ) wl ~~ rk c( m) / m 0 and m is se such ha: 10

11 c ( m) ~ ~ ~ f ( l, k ) wl ~~ rk m (9) Each firm equaes he marginal cos of raining an addiional manager o he marginal benefi. Noe he soluion of (9) wih. As developmen proceeds, marginal benefi is decreasing (since each producion uni hires fewer workers and less capial, and pays a higher wage o each worker). Since / is increasing in m, is decreasing over ime and developmen pace slows down. Noe also ha, as before, he economy is in general equilibrium every period. QED Noe ha he oupu is growing hroughou his process, since having more managers implies dividing labor and capial among more producion unis, whereas boh labor and capial exhibi diminishing marginal produciviy wihin each uni. For he explici example in which / 2, (9) ranslaes ino, /2, and since and are decreasing in M and is increasing in M, decreases over ime. Noe also ha similar o subsecion 3.1, allowing workers o pay a bounded paymen for heir raining or allowing firms o reques a longer period in which managers would work for hem afer raining affecs he righ-hand-side of (9), bu no he naure of he developmen process. Developmen in his model ceases when (since c / m 0, m0 equaion (9) ensures a posiive amoun of rainees as long as ). 7 We regard he cessaion of his ype of developmen as reaching a phase ha requires a new one researching new knowledge. The counry has creaed he full capaciy for absorbing capial and improving produciviy wih he curren producion funcion, and fuure growh will depend on having a beer echnology (i.e., an increase of A). 3.3 Capial flows The posiion of a counry as having a ne exernal deb (ne debor) or a ne lender (credior) of capial, alongside he capial flows in or ou of he economy, is a funcion of he producion unis' demand for capial, compared o he sock of capial owned by he local agens. We focus iniially on he firs phase, assuming he populaion is large enough (such ha he firs phase is a leas wo-periods long), and proceed wih he analysis of he second phase. 7 One may also consider an assumpion of moraliy of he populaion, workers and managers, afer few periods. Workers are hus born every period and firms need o rain managers o replace hose who died. In his case raining never ceases. M, however, reaches a cerain seady sae level. 11

12 In he firs phase, he capial requiremens of he economy (demand), denoed wih, are a funcion of he number of unis operaing. Thus, in he firs period, in which here is one pre-rained manager and managers in raining, here are 1 producion unis, each requiring unis of capial, and herefore 1. Similarly, since hroughou he firs phase and are consan: K ( 1 m k (10) D ) The local supply of capial, denoed wih, is semming from he savings of he managers, who are he only ones receiving a wage higher han 1, and he saved ineres over he savings (he savings ou of he ren of he capial). Since consumpion and savings occur a he end of each period, savings affec he capial sock of he nex period. In he firs period he economy has no locallyowned capial, 0. The capial of he second period is he savings of he prerained manager of he firs period, i.e., 1. The local capial sock of he hird period is he savings of he pre-rained managers of he second period, plus he savings of he managers from he firs period, including he saved porion ou of he paymens from firms for using his capial. Where, one ges Similarly: 1 S 1 (1 ) ( 1)(1 ~ ) 1 i i K m s r, 1 (11) i1 For every, local sock of capial was creaed in he previous periods. Thus, i sands for he period of creaion, 1 is he number of pre-rained managers saving a his period, and 1 is he saved ineres accumulaed over hese savings ill period. The ne posiion of he economy as a borrower (or a credior) of capial, i.e., he ne exernal deb, noed wih, is, herefore: K NP K K (12) D S Where 0 represens having ne deb. Thus, 1, and: 1 i1 NP i1 (1 ) (1 ) ( 1)(1 ~ ) 1i K m k m s r, 1 (13) 12

13 The expression measuring a counry s ne exernal deb during he second phase is similar o (13), bu wih, and insead of, and. Proposiion 3: The economy is a ne borrower of capial saring from period 1 for a finie amoun of periods. The economy is a ne credior of capial saring from a cerain period. Thus, he economy experience capial inflows ha laer reverse o capial ouflows. Proof: Period 1 obviously involves being a ne borrower and hus capial inflows. Simplifying (13), he ne exernal deb for he remaining periods of he firs phase yields: 1 ~ 1i NP 2 1 r K (1 m) k (1 m) s( 1). i1 1 m Noe ha while 1 affecs he magniude of he borrowing or lending, he oher par of he expression ses he posiion iself, based on he sign of he expression, as well as affecing he magniude. While he lef erm of he laer expression () is consan hroughou he firs phase, he righ side is increasing over ime. If (depending on he raining cos funcion and on ) he expression on he righ is bounded, and hus small enough s and can allow for: 1 1i 2 1 r k (1 m) s( 1), i1 1 m such ha he counry is a ne borrower of capial hroughou he enire firs phase. Obviously, if 1 /1 and s are high enough, for a cerain, he posiion may change ino ne capial credior. If and he firs phase is long enough (i.e., he populaion is large enough), 1 /1 is increasing and unbounded, hus, here exiss a (and a populaion size ha prolongs he firs phase o a leas periods) for which: 1 1i 2 1 r k (1 m) s( 1), i1 1 m and he counry becomes ne credior during he firs phase. Noe ha once he inequaliy holds for a cerain and he posiion reverses, i canno reverse again, since he lef hand side of his expression is consan, while he righ hand side is increasing. ~ ~ 13

14 If he ne capial posiion does no reverse during he firs phase, i reverses during he second phase. During he second phase capial requiremens ( ) grow a a slower pace (since and ) and evenually, once he enire populaion is rained, ceases o grow. coninues o increase unboundedly (due o savings and reurns o capial), and hus, according o equaion (12), ne exernal deb urns negaive, and he counry becomes a ne credior. Since a ne exernal deb (ne borrowing, 0) is creaed in he firs period, and laer he counry changes o a ne credior ( 0), flows mus reverse as well. Thus, once sars decreasing, iniial capial inflow urns o a capial ouflow. QED Inuiively, he growh of he economy in he firs periods, and he fac ha savings join he capial sock only in following periods, are boh leading o an increase in he demand for capial relaive o he local supply, and hus o capial borrowing. The fac ha opimal k of a firm (including is unis headed by managers in raining) may be higher han he savings of a single manager is enhancing he need for capial borrowing. However, he accumulaion of capial and he savings ou of he reurns o capial, which are compounded over ime, affec in he opposie direcion, and may decrease and reverse he posiion over ime. Even if i does no do so in he firs phase, once he economy eners he second phase, and growh slows and evenually ceases wihou any echnological change, he savings increase relaive o he demand, and evenually reverses he ne posiion. Thus, he economy evenually experiences capial ouflows. We inerpre he growh in he number of managers as he absorpion capaciy of he economy. This capaciy is increasing over ime, leading o a greaer demand for capial. However, evenually he economy absorbs all he capial required for efficien producion using he given echnology, and since savings coninue o increase, he capial flow mus reverse. We do no regard he compleion of he absorpion process as he end of growh, bu raher as a need o urn o an alernaive source for growh, which is a echnological change (i.e., an increase in A in equaion (2)), and perpeual echnological changes lead o perpeual growh. This ype of growh increases he required capial as well ( iself is increasing), bu since his ype of growh is expeced o be relaively slow, i does no reverse he ouflows, bu merely slows hem down. Noe also ha if one considers drasic echnological innovaions, here may be a need o rerain he enire populaion of managers, leading, again, o a slow increase of. Since mos of he discussion regarding capial flows involves developing counries, in which financial inflows and reversals of flows play an imporan role in economic developmen, and since in hese counries growh is semming from adoping exising echnologies, we chose o absain from including echnological change in he model. 14

15 Figure 1 demonsraes he process. Ne exernal deb is posiive (capial borrowing) and increasing in he beginning, bu laer sars decreasing and evenually urns negaive (i.e., o ne credior). Once he exernal deb sars decreasing, capial flow changes from posiive (inflows) o negaive (ouflows). The parameers of Figure 1 allow all hese changes o occur during he firs phase. Figure 2 presens he capial flows as a funcion of he oupu, showing he increase of he inflows in firs sages of oupu growh, he reversal and he decrease in inflows, which urns negaive, in laer sages of oupu growh. Figure 1: Ne Exernal Deb and Ne Capial Flows over Time Ploed for f ( l, k) ( l k ), c( m) 2m, r=0.1, s=0.2, ~ s 1 and N>62,261 Ne Exernal Deb ime Ne Exernal Deb (lef scale) Ne Capial Inflows (righ scale) Ne Capial Inflows Capial, herefore, is flowing ou of he developed (high oupu per capia) counries o he oher counries. However, he major recipiens of capial are no he counries wih he lowes levels of oupu (since hese counries have only few producion unis, and herefore few newly rained managers and few new producion unis), bu raher he counries wih he inermediae level of oupu (per capia), in which many new managers are rained and many new producion unis are esablished. Noe ha he oupu growh rae is no sufficien o esablish he direcion of he capial flows, since he mos developed counries, which have compleed he raining process, have low growh rae and significan capial ouflows, while he leas developed counries, in which mos of he oupu is sill semming from he radiional secor, have low growh raes as well bu posiive 15

16 (ye low) capial inflows. However, in he model, he oupu (per capia) iself provides sufficien informaion, as demonsraed by figures 1 and 2. Ploed for Figure 2: Ne Capial Flows and Oupu f ( l, k) ( l k ), c( m) 2m, r=0.1, s=0.2, ~ s 1 and N>62, Ne Capial Inflows firms' oupu Ne Capial Inflows 4 Empirical Analysis Empirical analysis validaing he model explicily requires daa regarding raining in a large sample of counries, which is unavailable. However, we are able o show ha he predicions regarding capial flow, i.e., ha capial flows in during early sages of developmen, when oupu per capia is low, bu flows ou in laer sages, when oupu per capia is higher, are consisen wih he daa regarding capial flows. 4.1 The esimaed equaions We use he lagged oupu, Y, as a measuremen for he sage of developmen, wriing he predicion of he model as, where he lef hand side is he ne capial flow ino a given counry and he righ hand side includes he oupu, oupu squared and conrol variables (X), including educaion, qualiy of governance and poliical sabiliy, religion, ec., of ha counry. In order for he predicion of he model o be consisen wih he daa, we need o show ha is posiive and is negaive. 16

17 Since our model refers o capial flows for producion purposes, as opposed o consumpion purposes, and due o daa availabiliy, we have used ne FDI flows as he capial flows variable. In order o overcome he size effec, i.e., he problem ha a larger counry migh arac more FDI, as if i is comprised of several small counries, alhough educaion, policy and oher variables may be he same, we examined per capia variables. In order o preven problems of causaliy, we have used he oupu and conrols of years prior o ha of he capial flow, such ha he flow canno be causal o oupu. Since annual daa of capial flow is exremely volaile, we have used averages over several years. Accordingly, he firs esimaed equaion is of he form: FDIpc1990_ 2003i a1 GDPpc1989 i a2gdppc1989sqri a3 X i i (14) We esimae he average FDI per capia in every counry i (over he years ) as a funcion of he GDP per capia a year before, and he squared GDP per capia a year before. We regressed many conrol variables (see below), and repor he regressions wih significan conrols (FDI per capia for served as a conrol as well; mos of he conrols common in he lieraure were found o be insignifican once conrolling for size, using per capia variables, and for he sage of developmen, i.e., he oupu). A second se of regressions includes regressions of he same srucure, bu for differen years (FDI per capia for , as a funcion of he 1979 GDP per capia). 4.2 The daa Our poin of deparure is Sala-i-Marin (1997) daase, including daa of oupu, educaion, governance, ec. for 134 counries. We used he WDI (2006) for he FDI daa, o updae some of Sala-i-Marin (1997) variables and for addiional conrols. 8 Some of he educaion variables were aken from he Barro and Lee (1994) daase. We were forced o omi Wes Germany, Taiwan and Yugoslavia due o geopoliical changes ha caused hese counries no o appear on he WDI. The main variables we use are GDP per capia for he years 1979 and 1989, and FDI per capia (average for he periods and ). Conrol variables included FDI of prior years, capial formaion a 1980, laiude squared, various educaion variables, various oupu composiion variables, indicaors for corrupion, regime, religious, ec. Table 1 provides a shor descripion of he variables and a complee lis of he conrol variables. 8 World Bank, World Developmen Indicaors 2006 (CD Rom). World Bank, Washingon DC. 17

18 Table 1 - Variables Variable Descripion Source Mean Sd. Dev. FDIpc1980_1989 FDIpc1990_2003 Average of Foreign Direc Invesmen, ne (consan 2000 US $) divided by oal populaion for Average of Foreign Direc Invesmen, ne (consan 2000 US $) divided by oal populaion for WDI 2006 WDI 2006 GDPpc1979 GDP per capia (housands of consan 2000 US $), WDI 2006 GDPpc1989 GDP per capia (housands of consan 2000 WDI US $), GDPpc1979sqr GDPpc1979 squared. WDI 2006 GDPpc1989sqr GDPpc1989 squared. WDI 2006 CapialForm1980 Gross fixed capial formaion (gross domesic WDI fixed invesmen) of 1980, % of GDP CapialForm1980 sqr CapialForm1980 squared. WDI 2006 Laiudesqr Absolue laiude, squared. Sala-i- Marin (1997) Addiional Domesic savings rae (1980), Domesic WDI conrols savings rae (1990), Life expecancy (1980) 2006 Addiional conrols Addiional conrols Absolue laiude, Lain America dummy, Sub- Saharan Africa dummy, Fracion Proesans, Fracion Caholics, Fracion Confucians, Fracion Muslims, Fracion Buddhiss, Former Spanish colony, Rule of law, Poliical righs, Civil libery, Degree of capialism, Revoluions and coups, War dummy, Fracion GDP from Mining, Fracion of primary producs in oal expors, Black Marke Premium (sandard deviaion), Exchange rae disorion, Years open (a measure of openness for ), Life expecancy (1960), Primary schooling (1960). Average schooling years (aduls, 1975), Average schooling years (aduls, 1980), Male primary school compleed (1975), Male primary school compleed (1980), Male secondary school aained (1975), Male secondary school aained (1980). Sala-i- Marin (1997) Barro Lee (1994)

19 The sample includes 131 counries. Lack of daa for some counries for some of he variables resriced us o 69 o 98 observaions (counries) for he various regressions. Two ouliers, for which he observed FDI per capia was exremely high, were omied from he regressions (Singapore for and Ireland for ). I should be noed ha he regression s predicion for hese wo observaions is posiive and relaively large (capial inflow), bu falls far below he acual value. 4.3 Resuls Table 2 describes regressions based on equaion (14). Column (1) refers o FDI per capia for , as a funcion of GDP per capia for 1989 and GDP per capia squared. The coefficien for he firs variable is posiive (27.1, i.e., every $1,000 increase in GDP per capia of he baseline year increases ne FDI flow by $27.1) and for he second is negaive, and boh are significan a 1%. The firs coefficien is posiive and he second is negaive, yielding an invered U shaped relaionship beween oupu and FDI flows, in line wih he model s predicion. Column (2) adds conrol variables o column (1). The r-squared is 0.49, i.e., we are able o explain half of he variaions in capial flows using hese variables. We have esed many oher conrol variables, including a measure of openness, rule of law, poliical righs, civil liberies, revoluions and coups, a war dummy variable, exchange rae disorion, fracion of GDP from mining, fracions of various religions, Spanish colony, life expecancy and several educaion variables, bu hey were all found o be insignifican. 9 The only addiional significan variable we found is FDI per capia of he previous decade (column (3)). Including his variable in he regressions of column (1) or (2) makes all he oher variables insignifican, probably since many of he variables affecing FDI are changing very slowly, and hus pas FDI may include informaion on he sage of developmen alongside oher variables. 10 In order o furher analyze his issue we have divided he sample o wo groups of idenical size based on heir 1989 oupu per capia (he less developed half and he more developed half), and repeaed regression (2) for each group separaely (columns (4) and (5)). We have found ha for he less developed half of he sample (column (4)) 1989 GDP per 9 Noe ha he educaion variables are no relaed o he raining mechanism presened in he model, since educaion is a decision of he individual agen, while raining is a decision of he firms, and depends on he availabiliy of firms. Noe also ha educaion variables are ofen insignifican or exhibi a negaive sign in cross counry growh regressions (see, for example, Bils and Klenow, 2000, Fernandez e al., 2001, and Eicher e al., 2007), hus moivaing our decision o use raining raher han educaion in he model. 10 According o he model s predicion, in all he periods besides hose of reversal from having a ne exernal deb o being a ne credior, inflows are increasing wih GDP and ouflows are increasing wih GDP, hus explaining a coefficien of slighly more han 1. 19

20 capia is posiive and significan, while for he more developed half (column (5)) i is negaive, alhough insignifican. 11 Therefore, for he less developed counries, greaer oupu a he baseline year, i.e., being more developed, is associaed wih greaer capial inflows, while for he more developed counries i may be associaed wih ouflows, conrolling for pas FDI. Table 2 FDI Regression Analysis Dependen Variable: FDIpc1990_2003 (1) (2) (3) (4) a (5) b GDPpc (3.13)** (4.00)** (5.29)** (1.09) GDPpc1989sqr (4.43)** (5.03)** CapialForm (2.04)* (0.68) (1.33) CapialForm1980sqr (2.65)** (0.80) (1.63) Laiudesqr (2.49)* (2.46)* (0.60) FDIpc1980_ (9.18)** (0.64) (3.13)** Consan (0.45) (1.56) (1.17) (0.57) (1.32) R-square Observaions The absolue values of he -saisics appear in parenheses. a Sample: counries wih 1989 GDP per capia below $3,000 (he lower half of he b complee sample). Sample: counries wih 1989 GDP per capia above $3,000 (he upper half of he complee sample). * Significan a 5%. ** Significan a 1%. The regressions are in line wih he predicion of he model. Capial inflows are posiive and increasing wih GDP a low levels of developmen, bu once GDP is higher, capial ouflows are occurring. Thus, he changes in oupu are generaing he capial flows. The negaive sign of capial formaion (column 11 As menioned is subsecion 3.3, he model does no explicily include echnological changes, and herefore predics he flows o he less developed counries beer han he flows o he developed counries. Including slow echnological changes, however, would slow he ouflows from developed counries, bu would no reverse he process. 20

21 (2)) implies ha when capial is already available (given he oupu), he demand for addiional capial decreases. Noice ha we do no ge his resul for he less developed counries in he sample (column (4)), supporing his inerpreaion. These facors explain abou half of he variaions in ne FDI per capia flows (column (2)), and herefore diminish he imporance of policy, various FDI aracing agencies, oher local characerisics, ec. 12 Figure 3 presens he acual values of FDI per capia and he fied values based on he regression column (1) of able 2. Noice ha he fied values demonsrae he same paern of reversal ha is prediced by he model. Noice also he similariy beween his figure o figure 2, which is presening he model s predicion. Figure 3: FDI per Capia and Oupu per Capia FDIpc1990_2003/Fied values GDPpc_1989 FDIpc1990_2003 Fied values As a robusness analysis we have esimaed similar regressions o hose of column (1) and (2) of able 2, for he years Table 3 repors he resuls. All he coefficiens are of he same sign as in able 2, and he coefficiens of he 12 We did no use pas growh, since he model links he capial flow o he sage of developmen, which is a funcion of oupu per capia. The effec of curren growh over he supply of capial varies by he level of oupu, and herefore he model does no link growh o ne capial flows. Noe also ha in he model growh is associaed wih oupu such ha counries wih inermediae levels of oupu per capia experience fas growh, while counries wih eiher very high oupu or very low oupu experience slow growh. 21

22 GDP and GDP squared are significan a 1%, supporing he findings of able 2. Therefore, he evidence sugges ha capial flows exhibi a paern similar o he one prediced by he model. 5 Discussion Table 3 FDI Regression Analysis Dependen Variable: FDIpc1980_1989 (1) (2) GDPpc (2.91)** (4.28)** GDPpc1979sqr (4.15)** (4.90)** CapialForm (0.57) CapialForm1980sqr (0.99) Laiudesqr (3.53)** Consan (0.49) (0.35) R-square Observaions The absolue values of he -saisics appear in parenheses. * Significan a 5%. ** Significan a 1%. We have presened a model in which growh required increasing he absorpion capaciy of he firms and capial. In he model, exernal capial canno creae growh if here is no inernal capaciy. The model yielded an invered U shaped relaionship beween he sage of developmen and ne capial inflows, due o changes in he gap beween demanded capial and local savings. Alhough we lack he daa required o es he model iself, we found ha he empirical relaionship beween ne FDI flows and he sage of developmen is in-line wih he predicion of he model. Accordingly, his sudy emphasizes he imporance of he inernal growh forces wihin he economy, over exernal availabiliy of capial. While financial liberalizaion may be required o make he exernal capial available, he demand for capial is no creaed if here are no inernal forces leading o growh. The policy conclusion is sraighforward. Capial is required for growh, bu i is flowing a some sages in which an economy is growing, no causing i o 22

23 grow. Therefore, policies and agencies aimed a direcly increasing he flow would have lile effec on he growh. Insead, counries should focus on policies aimed a creaing inernal growh, and he capial flows would follow. Such policies may include educaion and raining, beer legal environmen for operaion and expansion of firms, ax incenives for new enerprises, proper infrasrucure, ec. The posiive exernaliies of raining, i.e., he fac ha firms consider only he period (or periods) in which a manager works for he firm, while he economy is enjoying he manager s abiliy for he remaining periods, is enhancing he gains from such policies even more. While his sudy is no rying o idenify policies ha foser growh, i does sugges ha such policies do no include direcly argeing FDI, alhough a successful policy would lead a lowincome counry o experience FDI inflow. Obviously, we do no claim ha curren accoun liberalizaion is no necessary. I is required, in order o allow for he occurrence of hese capial inflows, once local demand is creaed. A similar conclusion is relevan regarding capial flow reversals. A gradual reversal may simply be a signal of mauriy of an economy, leading o a slowdown in demand relaive o he supply of capial, and in such a case, here is no need for any couner-policy direcly argeing he ouflows, bu raher for a policy fosering oher growh facors. As menioned before, we absained from adding echnological changes ino he model and used raining as he growh engine, making he model more relevan for less developed counries, which have o adop exising echnologies, and in which capial flows play an imporan role. Technological progress in developed counries would increase he local demand for capial, bu no enough o generae capial inflows, and herefore would no affec he policy conclusions menioned above. Including echnological changes, alongside oher exensions, such as a research secor and moraliy of he agens, may prove useful in improving he model s predicion regarding he long-run capial flows of he developed counries. References Agosin, M. R., and R. Machado, Foreign Invesmen in Developing Counries: Does i Crowd in Domesic Invesmen?. Oxford Developmen Sudies 33(2), Aharonoviz, G. D., Why Canno Poor Counries Uilize Exising Knowledge? Expansion of Firms and Human Capial Accumulaion by Training. Forhcoming, Economic Inquiry. 23

24 Aiken, B. J., and A. E. Harrison, Do Domesic Firms Benefi from Direc Foreign Invesmen? Evidence from Venezuela. American Economic Review 89(3), Barrios, S., H. Gorg, and E. Srobl, Foreign Direc Invesmen, Compeiion and Indusrial Developmen in he Hos Counry. European Economic Review 49(7), Barro, R. J., and J. W. Lee, Daa Se for a Panel of 138 Counries. Harvard Universiy. Available a hp:// Beck, T., R. Levine, and N. Loayza, Finance and he Sources of Growh. Journal of Financial Economics 58(1 2), Bekaer, G., C. R. Harvey, and C. Lundblad, Does Financial Liberalizaion Spur Growh? Journal of Financial Economics 77(1), Bevan, A. A., and S. Esrin, The Deerminans of Foreign Direc Invesmen ino European Transiion Economies. Journal of Comparaive Economics 32(4), Billingon, N., The Locaion of Foreign Direc Invesmen: an Empirical Analysis. Applied Economics 31(1), Bils, M., and P. J. Klenow, Does Schooling Cause Growh? American Economic Review 90(5), Bonfiglioli, A., Financial Inegraion, Produciviy and Capial Accumulaion. Journal of Inernaional Economics 76(2), Bordo, M., B. Eichengreen, D. Klingebiel, and M. S. Marinez-Peria, Is he Crisis Problem Growing More Severe? Economic Policy 16(32), Bursein, A., and A. Monge-Naranjo, Foreign Know-How, Firm Conrol, and he Income of Developing Counries. Quarerly Journal of Economics 124(1), Calvo, G., Capial Flows and Capial-Marke Crises: he Simple Economics of Sudden Sops. Journal of Applied Economics 1(1), Cheng, L. K., and Y. K. Kwan, Wha are he Deerminans of he Locaion of Foreign Direc Invesmen? The Chinese Experience. Journal of Inernaional Economics 51(2), Coase, R. H., The Naure of he Firm. Economica 4(16),