Financia (Des)Integration. Enisse Kharroubi June 2005 Abstract This paper addresses the macroeconomic impact of internationa nancia integration. I rst provide empirica evidence that foreign banking penetration can be associated with a contraction of banking credit, especiay in countries with poor credit markets. Second I present a mode in which the presence or the absence of foreign enders endogenousy modi es rms credit constraints and hence the voume of credit extended in the economy. Speci cay, I show that foreign enders consider oans from domestic enders as a rm coatera. This impies that their ending suppy is positivey associated with the voume of capita a rm is abe to borrow from domestic enders. Two di erent cases are then possibe. If foreign enders are abe, in spite of the coatera e ect, to extend a arge voume of oans even when domestic enders ending capacity has shrinked (due to increased competition on the capita market), then the economy bene ts at the steady state both from a arge capita suppy and a ow cost of capita. Integration then raises the economy s growth rate. On the contrary, if foreign enders are not abe, due to the coatera e ect, to extend a arge voume of oans when domestic enders ending capacity has shrinked, then competition reduces domestic enders ending capacity and the coatera e ect prompts foreign enders to reduce their capita suppy. Integration then depresses the economy s growth rate, rms cost of capita and the voume of credit extended in the economy. Banque de France. Address: 1, rue de a Vriière 75049 Paris cedex 01. e-mai: rst name.surname(at)banque-france.fr. The views expressed herein are those of the author and do not necessariy re ect those of Banque de France. I thank Joshua Aizenman, Phiippe Martin Jean-Oivier Hairaut and participants at Paris I - Paris Jourdan Macroeconomics Seminar (Feb. 2005), In nity conference on Rea and Financia Aspects of Financia Integration (June 2005) and Geneva doctora conference in Internationa Economics and Finance (Jan. 2006) for hepfu discussions. Usua discaimers appy. 1
1. Introduction. In the ast years, a arge empirica iterature has emerged on the impact of foreign banking penetration (Caessens, Demirgüç-Kunt and Huizingua [1998], Godberg, Crysta and Dages [2002] or Bayraktar and Wang [2004]). Whie a consensus exists in this iterature on the positive e ects of foreign banking penetration on the behavior of domestic banks as we as on the functioning of the capita market, amost a these papers have focused on e ciency spiovers whie much ess attention have been devoted to ending behavior e ects. In this paper we try to shed ight on this point. We aim at understanding whether the presence of foreign nancia institutions contributes to increasing the goba voume of credit extended in the economy. In other words do oans from foreign enders substitute or compement oans from domestic enders and if there is substituabiity, how does the goba voume of oans react to changes in the presence of foreign nancia institutions? More generay does nancia integration contribute to reax or tighthen credit constraints? To expore these questions, we focus on the reationship between credit and foreign banking penetration 1. 180% 160% 140% 120% 100% 80% 60% 40% 20% 0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Foreign Banking Assets in tota Banking Assets vs. Private Banking Credit to GDP. 1 The credit variabe we use represents the caims on the private sector by deposit money banks divided by GDP for a given year in a given country. It isoates credit issued to the private sector as opposed to credit issued to governments and pubic enterprises. The foreign banking assets variabes is computed among foreign banks which are de ned as banks for which at east 50% of the equity is owned by foreigners. 2
At rst gance, it seems that if any, this correation is signi cant and negative. To investigate more deepy this reation, we run a number of regressions to estimate the determinants of private credit to GDP 2. From these estimations, we derive two main resuts. First, the negative correation between foreign banking penetration and banks private credit to GDP is signi cant and robust. Second, this correation between foreign banking penetration and banks private credit to GDP seems to catch a causa e ect from foreign banking ownership on the voume of credit the banking system deivers 3. In this paper we try to provide a theory for why foreign banking penetration may induce a reduction in the goba amount of credit extended in the economy and thereby on investment and growth 4. 1.1. The mechanism of the mode. The mode is based on two buiding bocks. First, an economy which opens to internationa capita ows undergoes a positive change in competition on its capita markets. Financia opening is therefore ikey to ower domestic enders pro ts and, because pro ts determine to a arge extent the resources avaiabe for ending 5, to consequenty reduce the ending capacity of domestic nancia intermediaries. This can indeed be positive for the economy as a whoe if foreign capita ows constitute a good substitute for domestic nance 6. The second buiding bock the mode is based upon, aims precisey at determining the conditions under which oans provided by foreign enders may be good substitutes to oans provided by domestic enders. To do so, we make two core assumptions. First, we assume the existence of two types of imperfections on 2 c.f. appendix for more detais about econometric estimations. We especiay aim at understanding whether this correation coud be spurious, the foreign ownership variabe catching other e ects such as nancia under-deveopment. 3 There is a third resut we do no stress due to its econometric fragiity: the contemporanous correation between foreign banking penetration and banking credit to GDP depends upon the goba size of the banking sector: it is negative in economies where banks assets constitute a sma fraction of GDP whie it is positive in economies where banks assets constitute a arge fraction of GDP. 4 Whie not addressing directy this question, Rodrik [1998] shows very intuitivey that foreign savings brought by nancia iberaization-integration are unikey to have a rst order e ect on investment and growth: "In practice, there have been few cases of high-investment countries, perhaps none at a, where foreign saving has accounted for more than 20 percent of investment over ong stretches of time. In an economy investing, say, 30 percent of its GDP, reying on foreign saving beyond this imit woud impy running a persistent current account de cit in excess of 6 percent of GDP, which woud be courting with disaster." 5 In the mode, we assume that the ony source of capita for domestic enders consists in their pro ts of the previous period. 6 If this is the case, internationa capita ows are positive because the decrease in domestic enders pro ts simpy re ects that foreign enders intermediation technoogy is more e cient than that of domestic enders. This means that foreign enders are abe to provide any quantity of capita domestic enders woud provide at a ower interest rate. 3
nancia markets, an ex post and an ex ante mora hazard probem 7. Second we assume that domestic nancia intermediaries are ony confronted to the ex post mora hazard probem whie foreign nancia intermediaries are confronted to both ex ante and ex post mora hazard probems. This ast assumption impies that domestic enders have an informationa advantage on foreign enders 8. Based on these two assumptions we show that foreign and domestic capita suppy are negativey reated for arge eves of domestic debt but positivey reated for ow eves of domestic debt. This means that in the atter case -ow domestic capita suppy-, a reduction in the domestic capita suppy is foowed by a drop in the foreign capita suppy whie in the former case -arge domestic capita suppy-, a reduction in the domestic capita suppy is o set by an increase in the foreign capita suppy. The intuition for these correations is straightforward: when the voume of domestic oans is ow then the ex ante mora hazard probem for foreign enders is binding because the costs for rms to adopt the ine cient technoogy increase with the voume of domestic oans. An increase in the voume of domestic oans deivered therefore reduces the incentives for rms to take the ine cient project and foreign enders can then increase their suppy of capita without destroying rms incentives to adopt in the e cient project. On the contrary for arge eves of domestic debt, the cost for rms to adopt ex ante the ine cient projects are so arge that the binding constraint is the ex post mora hazard constraint. In this case an increase in domestic ending prompts foreign enders to reduce their capita suppy because the costs for rms to defaut ex post decrease with the tota voume of oans. Given these two buiding bocks, openness to internationa capita ows rst has a direct positive e ect on rms pro ts because rms bene t from a new source of capita to nance their investments. Consequenty rms modify their debt portfoios (in favor of foreign oans). This reduces the demand for domestic oans and hence the equiibrium interest rate on domestic oans and hence domestic enders pro ts. Secondy however, because rms pro ts grow more rapidy, the demand for domestic oans aso increases more rapidy whie the suppy for domestic oans grows ess rapidy due to the decrease in the equiibrium interest rate on domestic 7 The ex post mora hazard probem impies that nancia contracts are imperfecty enforceabe: enders need to rey on incentives to have borrowers repay their debts. The ex ante mora hazard probem impies that enders cannot observe how borrowers use the funds they have borrowed. 8 This assumption is con rmed by Kaufman, Mehrez and Schmuker [2005] which provide empirica evidence that resident rms have an informationa advantage about the countries where they work. Simiary, Mian [2006] shows that greater cutura and geographica distance between a foreign bank s head quarter and the oca branches, eads it to further avoid ending to informationay di cut yet fundamentay sound rms. 4
oans. This dynamic e ect wi impy an increase in the interest rate on domestic oans. Then depending upon which of the direct or the dynamic e ect prevais, two di erent cases are possibe at the steady state. If the decrease in the demand for domestic oans is compensated by an increase in the growth rate of rms pro ts, then the interest rate on domestic oans increases and the economy s growth rate is arger under nancia integration. On the contrary, if the decrease in the demand for domestic oans is not compensated by a su cient increase in the growth rate of rms pro ts, then the interest rate on domestic oans decreases and the economy s growth rate is ower under nancia integration. Financia integration can possiby be growth decreasing here because rms which cannot borrow capita from domestic enders do not bene t from a arge foreign capita suppy. Therefore if these rms have a very imited access to foreign capita markets, then it is ikey that the decrease in the demand for domestic oans foowing integration be not compensated by a su cient increase in rms pro ts. This is why nancia integration is then unambigiousy growth reducing. Simiary it is easy to understand why the presence of foreign nancia intermediaries can contribute to reduce the goba voume of credit extended in the economy: due to the direct competition e ect, the suppy for domestic oans decreases compared to the cosed economy case. Moreover since the suppy for foreign oans depends positivey on the voume of domestic credit, the reduction in the suppy for domestic oans wi aso impy a decrease in the suppy for foreign oans. Financia integration eventuay eads to a desintegration of the credit market. From a theoretica point of view, our mode impies that nancia integration can generate increasing returns to scae between the growth rate of the economy and the size of the domestic nancia sector. Whie intuitivey, a arger domestic nancia sector woud impy ess productive investments at the margin and hence a ower growth rate, nancia integration reverses this reation due to the coatera e ect of domestic oans. When rms can borrow both from domestic and foreign enders, rms reduce their demand for domestic capita. This reduces the interest rate on domestic oans and the size of the domestic nancia system. Then due to the coatera e ect (the foreign ending suppy is positivey reated to the voume of domestic oans) the reduction in the size of the domestic nancia sector is foowed by a fa in the agregate oan suppy. This prompts a decrease in investment and thereby a decrease in growth. 5
1.2. Reated iterature: Internationa nancia integration styized facts. This paper is reated to three di erent strands of iterature. It rst reates to the iterature deaing with the e ect of nancia openness and capita ows on domestic savings and investment. In their semina paper, Fedstein and Horioka [1980] show that among OECD countries, the correation between investment and domestic savings is very arge and hence di cut to reconcie with a view of capita being higy mobie. Rodrik [1998] shows that foreign savings cannot account for a arge share of investment even in countries with a arge nancia openness degree. More generay a number of papers have tried to determine the e ect of nancia integration on domestic savings and investment (Obstfed [1998], Bosworth and Coins [1999] or Razin Sadka and Yuen [1999]). Simiary, Cabaerro and Krishnamurthy [2001] focuses on the e ects of exogenousy given domestic and internationa borrowing constraints on rea and nancia variabes. In this paper, we try to go one step further to show how openness to internationa capita ows can endogenousy modify rms credit constraints based on the interaction between competition and coatera e ects. We aso show that nancia integration is ikey to increase the dependence of investment and aggregate credit and domestic savings due to the coatera e ect. Secondy this paper reates to the iterature on the cost of capita e ects of nancia iberaization. Bekaert, Harvey and Lundbad [2001], Bekaert, Harvey and Lumsdaine [2002] or Bair Henry [2003] a show that nancia iberaization-integration reduces signi canty the cost of capita for rms, this being a rst order channe to account for the increase in investment and growth that can foow nancia integration. Kose, Prasad and Terrones [2003] show for instance that nancia integration has positive growth e ects in deveoped countries. Bekaert, Harvey and Lumbad [2004] show that equity market iberaization is foowed by a decrease in growth voatiity. Here the contribution of the paper consists in showing that the decrease in the cost of capita is not necessariy an indicator of nancia integration success since it can happen with a decrease in the goba voume of credit and investment. Finay this paper reates to the empirica iterature on foreign banking penetration. Caessens, Demirgüç- Kunt, and Huizinga [2001] show that it can prompt a reduction in pro tabiity and margins. Beck [2000] asserts that foreign banking penetration increases the voatiity of capita ows. Finay banking competition 6
is shown to be positive for growth ony in deveoped nancia system (Caessens and Laeven [2003]). 1.3. Road map of the paper. The paper is organized as foows. The foowing section describes how the credit market operates when rms can borrow from both domestic and foreign enders. Section 3 ays down rms nancia optima choices. In section 4 we set up the macroeconomic mode main assumptions. Section 5 describes the cosed economy and section 6 the open economy. Finay concusions are drawn in section 7. 2. The capita market. The capita market is characterized by two types of imperfections. First nancia contracts are not perfecty enforceabe. Borrowers can defaut strategicay on their iabiities. We mode this possibiity of strategic defaut as an ex post mora hazard probem. Second borrowers face an ex ante mora hazard probem, they can choose to invest in two di erent technoogies, one of the two producing ony private bene ts and aowing borrowers to defaut on their iabiities at no cost. Entrepreneur make financia and investment choices. Entrepreneurs reap the benefit of their investments and pay back their oans. Figure 1: Timing of the mode. To determine incentive compatibe contracts, et us consider an entrepreneur with one unit of own capita (equity) who borrows units of capita from domestic enders at a gross interest rate r, f units of capita from foreign enders at a gross interest rate r f and invests in the production technoogy with a margina return noted R. Then the entrepreneur s end-of-period pro t is equa to = 1 + + f R r r f f (2.1) 7
On the contrary if the entrepreneur decides to defaut 9 on his iabiities then his end-of-period pro t then writes as 1 = 1 + + f (R r1 ) p r p f r f f (2.2) where r 1 represents the margina cost to defaut and p (resp. p f ) is the proportion of oans domestic (resp. foreign) enders manage to be reimbursed upon when the defauting entrepreneur has invested in the production technoogy. The entrepreneur can aso decide to invest ex ante in the private bene t technoogy whose margina return is equa to R r 2. With this technoogy he can defaut on his iabiities at no cost. In this case his end-of-period pro t writes as 2 = 1 + + f (R r2 ) q r q f r f f (2.3) where q (resp. q f ) is the proportion of oans domestic (resp. foreign) enders manage to be reimbursed upon when the defauting entrepreneur has invested in the private bene t technoogy. Then we make three assumptions: 1. There is ex ante mora hazard: R r 1 < R r 2 < R. If an entrepreneur does not (resp. does) defaut, then the production technoogy is more (resp. ess) e cient than the private bene t technoogy. 2. Domestic (resp. foreign) enders technoogy to recover defauted caims on entrepreneurs who have invested in the production technoogy is such that recovering a proportion p of caims of size L costs c n (1 p ) L (resp. c f n (1 p f ) L) with r 1 < min fc ; c f g. When an entrepreneur invests in the production technoogy and defauts, then the margina cost to recover defauted caims for domestic and foreign enders is aways arger than the margina cost to defaut for the entrepreneur 10. 3. Domestic (resp. foreign) enders technoogy to recover defauted caims on entrepreneurs who have 9 In this case it can be shown that defauting on both types of iabiities (domestic and foreign) is equivaent to defauting on ony one type of iabiities as regards incentive compatibiity constraints. In other words, there is no oss of generaity in considering that borrowers defaut on both types of iabiities. 10 There is no particuar consequence to the assumption that enders technoogies to recover caims write as c i n (1 p i ) apart from the fact it yieds borrowing constraints which do not depend upon the interest rate. This heps simpifying the anaysis. A di erent assumption (if for instance interest rates made borrowing constraints stronger) woud certainy reinforce the mechanism of the mode. 8
invested in the private bene t technoogy is such that recovering a proportion p of caims of size L costs b n (1 q ) L (resp. b f n (1 q f ) L) with b < r 2 < b f. When an entrepreneur invests in the private bene t technoogy and defauts, then the margina cost to recover defauted caims for foreign (resp. domestic) enders is aways arger (resp. ower) than the margina cost to defaut for entrepreneurs 11. Proposition 1. Noting the domestic debt equity ratio and f the foreign debt equity ratio for a given entrepreneur, then domestic and foreign enders suppy capita to this entrepreneur in a way consistent with the conditions ( f 1) f + ( 1) 1 f 1 f (1 ) 1 (2.4) where = c =r 1, f = c f =r 1, = b =r 2 and f = b f =r 2. Proof. In the case of the production technoogy, domestic and foreign enders wi respectivey choose p and p f such that (1 p ) r = c and (1 p f ) r f = c f whie in the case of the private bene t technoogy, enders wi choose q and q f such that (1 q ) r = b and (1 q f ) r f = b f. Pugging these equaities in expressions (2.2) and (2.3), and soving the incentive constraints 1 and 2 we end up with (2.4). 11 Here this cost corresponds to r 2 since the cost to defaut on the private bene t technoogy is simpy zero. 9
µ f µ Figure 2: Entrepreneurs borrowing constraints. There are two remarks about conditions (2.4). First domestic and foreign oans are negativey correated at the frontier of the constraint which soves the ex post mora hazard probem. This is because the soution to the ex post mora hazard probem consists in a imit on the overa amount of capita entrepreneurs can borrow. If entrepreneurs can borrow arge amounts of capita, they wi defaut, irrespective of the identity of the ender. On the contrary in the second constraint, domestic and foreign oans are positivey correated at the frontier. Since domestic enders are reativey e cient in recovering their caims from entrepreneurs who choose the private bene t technoogy, entrepreneurs who borrow arge amounts of capita from domestic enders wi incur arge osses if they invest in the private bene t technoogy and choose to defaut. This impies that an entrepreneur who borrows arge amounts of capita from domestic enders is more ikey to choose the production technoogy and ess ikey to defaut on his debts. Therefore foreign enders can suppy arger amounts capita without destroying entrepreneurs incentives to repay their oans. This expains the positive correation which can be interpreted as evidence that foreign enders consider domestic oans as some form of informationa coatera 12. 12 Within this framework, one can note that the voume of capita foreign enders accept to extend to rms is a quasi-concave function of the voume of capita domestic rms borrow from domestic enders. This characteristic pays a key roe in the stabiity of the open economy. Removing it or aowing for a oca quasi-convexity opens the door to the examination of credit cyces and capita ows reversas. 10
3. Agents decisions. 3.1. Entrepreneur s optima borrowing choices. An entrepreneur s choices consist in choosing a debt portfoio ; f which maximizes the pro t function (2.1) given the constraints (2.4) on nancia capita suppy. Entrepreneurs program therefore writes as s.t. max (R r ) + (R r f ) f ; 8 f >< ( f 1) f + ( 1) 1 >: f 1 f (1 ) 1 (3.1) where r (resp. r f ) is the expected gross interest rate on domestic (resp. foreign) debt a rm faces. Let us then note = [ 1] 1, f = [ f 1] 1 and (resp. f ) the optima amount of capita borrowed from domestic (resp. foreign) enders. We then have the foowing resut. Proposition 1. Assuming that r f < R and f < f, there exists a threshod vaue q (R) = R f (R r f ) such that: Entrepreneurs borrow excusivey from domestic enders: = and f = 0 if and ony if r q (R). Entrepreneurs borrow from foreign and domestic enders: f +(1 ) f [ +(1 ) f ] if and ony if q (R) < r R = f f +(1 ) f [ +(1 ) f ] and f = Entrepreneurs borrow excusivey from foreign enders: = 0 and f = f f 1 1 if and ony if r > R. Proof. Evident. These resuts are competey standard. Since the program of the rm and the constraints the rm faces are a inear, we simpy need to compute a quantity-pro tabiity trade-o : rms borrow from the most e cient source of capita. Moreover since borrowing excusivey from foreign enders may not be optima, there are situations (q (R) < r R) where entrepreneurs prefer to borrow from both types of enders. 11
4. The mode. 4.1. Main assumptions. We consider a singe good competitive economy with non overapping generations. In each generation, there are two types of agents. One haf of the popuation are entrepreneurs and the other haf are (domestic) nancia intermediaries. A agents ive for one period 13. At the end on their ives, agents make a bequest to their o -spring and take a consumption decision. Preferences of agents born at time t write as u t = b t+1 c1 where b t+1 is the bequest made at time t + 1 and c t+1 is the consumption at time t + 1. Entrepreneur i can invest his capita in a (production) technoogy with a margina return R i or in a private bene t technoogy with a margina return B i. We assume that R i is uniformy distributed among entrepreneurs on R; R. We note m = R+R 2, = R R 2 and F the cumuative function of R i. Entrepreneur i can defaut on his iabiities: the cost to defaut on the production technoogy is equa to r 1 and the cost to defaut on the t+1 private bene t technoogy is zero. Moreover we assume that both the cost to adopt the private bene t technoogy r 2 R i B i and the cost to defaut on the production technoogy r 1 to be constant among entrepreneurs 14. Entrepreneurs can borrow capita from domestic nancia intermediaries at a domestic gross interest rate r. They can aso borrow capita from foreign nancia intermediaries at a domestic gross interest rate r f. To simpify the exposition of the mode, we assume that r f R. This impies that a rms have access to internationa capita markets 15. The capita market is exacty simiar to that of section 2. 4.2. Inter-tempora decisions and dynamics of the economy. Given the assumption about agents preferences, agents spend a proportion 1 of their end-of-ife pro ts in consumption and a proportion in bequests to the next generation. Therefore if e;t represents entrepreneurs 13 The resuts of the mode are not dependent on the assumptions that there is an equa number of entrepreneurs and nancia intermediaries in each period, nor that agents ive for one period. The one period ife assumption coud for instance be repaced by an in nite horizon assumption where agents hod preferences of the form X s n (c s) without any quaitative change in the s resuts of the mode. 14 With this assumption, we restrict rms heteregeneity to dimensions where enders inabiity to observe individua productivities R i does not produce any market faiure. 15 We take this view to show that even in the impausibe case of a "universa" access to internationa capita markets, capita market, integration may have negative e ects. As a consequence of this assumption, capita suppied by foreign enders is aways cheaper than capita suppied by domestic enders. 12
pro ts at date t and fi;t domestic nancia intermediaries pro ts at date t then we have fi;t+1 = r fi;t and e;t+1 e;t Z = 1 + (R) + f (R) R r f (R) r f f (R) df (R) where (R) (resp. f (R)) is the optima domestic (resp. foreign) debt equity ratio for an entrepreneur of productivity R. Finay noting y t = fi;t 1 = e;t 1 the dynamics of the economy writes as y t+1 = r (y t ) y t R h 1 + (R) + f (R) R r (y t ) f (R) r f f (R) i df (R) (4.1) 5. The cosed economy. 5.1. Equiibrium of the capita market. Let us consider that the economy is cosed to foreign capita ows (enders). Then, the demand for domestic oans at any date t in the economy L d t writes as L d t = [1 F (r )] e;t 1 This demand for capita is competey standard: it is a decreasing function of the domestic gross interest rate r. The suppy for domestic oans 16 at any date t in the economy L s t writes as L s t = fi;t 1 Then the equiibrium of the capita market de nes a gross interest rate r. This yieds 8 >< m + r (y t ) = >: m 2 yt if y t if y t > 16 The domestic capita suppy is non eastic to the domestic interest rate r. Whie not crucia, this is an important assumption. The crucia assumption is that the foreign capita suppy be more eastic than the domestic one. 13
Entrepreneurs whose productivity R i is arger (resp. ower) than m + 2 yt have a debt equity ratio equa to (resp. 0). 5.2. Dynamics of the cosed economy. Using the expression (4.1) indicating how nancia intermediaries pro ts evove reativey to those of entrepreneurs, the dynamics of the cosed economy foows the aw of motion r (y t ) y t y t+1 = R R<r (y t) (R) dp (R) + R R>r (y t) (R) dp (R) with (R) = R and (R) = (1 + ) R r. Simpifying the ast expression yieds the foowing dynamics 8 >< y t+1 = >: [(m+) 2y t]y t m +yt 2 (m ) m+ if y t otherwise The cosed economy has two steady states. In the rst one, y = 0, a the capita stock beongs to entrepreneurs. This is a steady state because domestic enders cannot generate pro ts if they have no own capita (equity). However, any positive change in the amount of enders own capita woud get the economy out of the steady state y = 0, because the margina productivity of rms woud then be equa to m whie the margina productivity of domestic enders r woud be equa to m +. The steady state y = 0 is therefore unstabe since by de nition > 0. 14
y t+1 y t Figure 3: Dynamics of the cosed economy. A second steady state exists for yc = h1 + p i 1 1 + This steady state is stabe and therefore represents the ony ong run situation of the cosed economy. The steady state interest rate r is r = m + 2y c = The interest rate r is the steady state growth rate of entrepreneurs and domestic enders pro ts and the steady state growth rate of the economy whie y c = is the proportion of rms which are abe to borrow from domestic capita markets. As is cear domestic capita market deveopment understood as an increase in has two e ects. First rms which are abe to borrow from domestic capita markets make arger investments because their borrowing capacity is arger. This has a positive e ect on growth. Second an increase in raises the domestic interest rate r because it raises the demand for capita and therefore reduces the proportion of rms which are e ectivey abe to borrow 17. This has a negative e ect on growth. However the rst positive e ect aways dominates and domestic nancia deveopment is aways associated with higher growth when the economy is cosed to internationa capita ows athough the margina return to domestic 17 The share of rms whose productivity is arger than the steady state interest rate is equa to y c= and is a decreasing function of. 15
nancia deveopment on growth decreases with the eve of domestic nancia deveopment. 6. The open economy. 6.1. Equiibrium of the capita market. Let us now consider the case of an economy opened to internationa capita ows. To simpify the exposition of this section et us suppose that < f. This ast assumption and the assumption that r f m impy that foreign enders are more e cient than domestic ones on both quantities and prices. Then in based on section 3.1, the demand for domestic capita L d t is such that L d t = e;t 1 [1 F [r ]] The suppy for domestic capita at any date t L s t writes as L s t = fi;t 1 Let us as previousy note y t = fi;t 1 = e;t 1, the equiibrium of the capita market then de nes a gross interest rate r such that y t = [1 F [r ]] The interest rate r on domestic oans is then equa to 8 >< m + 2 yt if y t r (y t ) = (6.1) >: m if y t > Entrepreneurs whose productivity R i is arger than m + 2 yt have a debt equity ratio equa to + f whie entrepreneurs whose productivity R i is ower than m + 2 yt have a debt equity ratio equa to f. The dynamics of the economy then writes as r (y t ) y t y t+1 = R R<r (y t) f (R) dp (R) + R R>r (y t) f (R) dp (R) 16
where f (R) = 1 + f R r f f, f (R) = 1 + + f R r r f f. Simpifying the ast expression, and noting f = f f we obtain the foowing aw of motion: 8 >< y t+1 = >: (m+ 2y t= )y t m+ f (m r f )+(m+ r f ) f (y t= )+[ f ](y t= ) 2 if y t (6.2) (m ) m+ + f (m r f ) if y t > 6.2. E cient nancia integration. Proposition 1. The economy has a unique steady state if and ony if f > m r f. Moreover if f > m then nancia openness increases wefare and growth and decreases the rm cost of capita. r f Proof. A xed point to the aw of motion of the economy (6.2) veri es y = 0 or i 2 h f y= + 2 + (m + rf ) f y= + f (m r f ) = 0 (6.3) i As is cear this equation has no soution for y 2 h0; if and ony if f (m r f ) > 0. Moreover y = 0 is a stabe xed point if and ony if f (m r f ) > 0. Consequenty the economy has a unique steady state for y = 0 if and ony if f > m r f. At the steady state, the growth rate of the economy is given by g o = 1 + f m r f f The cost of domestic capita is equa to m + and the rm cost of capita is equa to r f. Comparing this case to the cosed economy where growth and the rm cost of capita are both equa to g c = m + [ p 1 + + 1] 2 it is straightforward to note that nancia openness reduces rms cost of capita since by de nition r f m and m < m + [ p 1 + + 1] 2. Moreover if f > m r f then g o > g f. Openness therefore raises growth and therefore wefare. 17
When the economy opens to internationa capita ows, there is rst a signi cant decrease in the cost of domestic capita due to a sharp reduction in the demand for domestic oans: The most productive rms switch party to foreign oans and the east productive rms turn to foreign enders excusivey. The decrease in the cost of domestic capita prompts a decrease in domestic enders pro ts and an increase in domestic rms pro ts. Consequenty at the next period the ending capacity of domestic enders is reduced reativey to rms demand for domestic oans. This has two di erent e ects: rst the cost of domestic capita (the interest rate on domestic oans) increases. Second entrepreneurs which are excuded from the domestic capita market face a reduction in foreign enders capita suppy. However because f (entrepreneurs owest borrowing capacity w.r.t. foreign enders) is su cienty arge, entrepreneurs pro ts sti increase on average at a faster pace than those of domestic enders. In other words the negative externaity produced by the increase in the domestic interest rate on entrepreneurs access to internationa capita market is su cienty sma. As a resut, during the transition to the steady state, entrepreneurs pro ts grow duraby faster than those of domestic enders. At the steady state, the interest rate on domestic oans r is equa to m +, rms borrow excusivey from foreign enders and the growth rate of rms pro ts (which is aso the steady state growth rate of the economy) is equa to m + f (m r f ). y t+1 y t Figure 4: Dynamics of the open economy with a unique steady state. 18
This case rests on the assumption that f is su cienty arge. This impies that, every thing ese equa, the di erence f f is sma which means that entrepreneurs access to domestic capita markets does not have a arge impact on entrepreneurs borrowing capacity vis-à-vis foreign enders. In other words, domestic and foreign nance are reativey good substitutes. In this case, given that foreign enders have been assumed to be more e cient than domestic enders, it is straightforward that the economy is better-o under nancia integration. Openness basicay provides a cheaper source of capita whie negative externaity e ects are reativey sma. Foreign enders e ciency advantage compensates for their informationa disadvantage. However it is not cear that a arge degree of substituabiity between domestic and foreign capita is the most accurate description of what happens in a number of emerging market economies. We therefore examine the ow substituabiity case in more detai in the next paragraph. 6.3. The case for ine cient nancia integration. Proposition 2. If f < m r f then the open economy has a steady state for y = 0 and a steady state y de ned by y o = f (m r f ) + m+ r f 2 f 2 v 6 4 1 + u t1 + h 4 f (m r f ) (m + rf ) f + 2 2 3 i f 7 5 1 Proof. Soving for the xed points to equation (6.2) yieds proposition 2. When y = 0, i.e. the reative suppy for domestic capita is zero, then the interest rate on domestic oans r is equa to m+. In this case a entrepreneurs borrow from foreign enders f per unit of own capita. The growth rate of entrepreneurs pro ts is equa to m + f (m r f ) whie the growth rate of domestic enders pro ts is equa to m +. As is cear f < m r f is equivaent to m + f (m r f ) < m +. This impies that when the reative domestic capita suppy y is zero, it stricty increases with time since the suppy for domestic capita increases at a faster pace than the demand for domestic capita. The steady state y = 0 is therefore unstabe and we can disregard this case. At the non degenerate steady state y, the interest rate on domestic oans is equa to r = m + 2 y o 19
It aso represents the steady state growth rate of domestic enders and domestic entrepreneurs pro ts. We then have the foowing proposition. Proposition 3. In the open economy, growth and wefare are arger at the steady state compared to the cosed economy if and ony if is su cienty sma and/or f is su cienty arge and/or is su cienty arge. Proof. capita. Financia integration is wefare improving if and ony if it raises the steady state cost of domestic At the steady state, the cost of domestic capita is arger in the open economy if and ony if y o = < y c = which simpi es as 1 + p 1 + < + m+ r f 2 f f (m r f ) 2 v 6 4 1 + u t1 + h 4 f (m r f ) (m + rf ) f + 2 2 3 i f 7 5 The LHS of this inequaity is a increasing function of whie the RHS is an increasing function of, f and f. This condition is therefore more ikey to be satis ed when is ow, is arge and/or f is arge. At the steady state, nancia integration has two opposite e ects on the interest rate on domestic oans r. It rst has a negative e ect because entrepreneurs reduce their demand for domestic oans (from to ). Secondy it has a positive e ect because it raises entrepreneurs pro ts by o ering them new nancing opportunities and therefore increasing the growth rate of the demand for domestic oans. In the case where is arge, the drop in the demand for domestic oans foowing nancia integration is arge. Then to be Pareto improving, there shoud be a arge increase in entrepreneurs pro ts foowing nancia integration. Likewise the growth rate of rms demand for domestic oans woud be arge and thereby raise the domestic interest rate. However since f is assumed to be su cienty sma, ( f < m r f ) a entrepreneurs whose productivity is not su cienty high to borrow from domestic capita markets have a very imited access to internationa capita markets. Consequenty whie entrepreneurs pro ts do increase foowing nancia integration, the increase remains modest and may be insu cient to compensate for the decrease in the demand for domestic 20
oans. The steady state interest rate on domestic oans r then ends up being ower when the economy is opened to internationa capita ows. In this case integration reduces the growth rate of the economy and is wefare decreasing. For the same reasons it is easy to understand why nancia integration raises the interest rate on domestic oans r and thereby increases growth and wefare in the case where and/or f is arge. In the rst case, a arger vaue for reduces the decrease in the demand for domestic oans foowing integration and it raises entrepreneurs pro ts when entrepreneurs are abe to borrow from the domestic capita market. For these two reasons, it is ikey that integration is Pareto improving when is arge. In the second case, a arger vaue for f increases the borrowing capacity of rms which can borrow from both domestic and foreign enders. Hence, a arger f increases rms pro ts and the interest rate on domestic oans is more ikey to increase foowing nancia integration due a arger di erence in pro t growth between rms and domestic enders. y t+1 y t Figure 5: Dynamics of the open economy. In this mode, nancia integration modi es both the reative size of the domestic nancia sector y and the growth rate of the economy r. Intuitivey, nancia integration, because it raises competition on nancia markets, shoud bene t the non nancia sector, reduce the size of the domestic nancia sector and thereby increase the growth rate of the economy. This happens when a rms have an identica access to internationa capita markets. More precisey in the absence of the coatera e ect, and under the assumptions made up to 21
now (r f m and f ) the steady state growth rate of the open economy is equa to 1 + f m rf f whie the reative size of the domestic nancia sector is equa to 0. Then integration raises both the growth rate of the economy and reduces the reative size of the domestic nancia sector. One can then observe that there exists a negative reationship between the growth rate of the economy and the reative size of its domestic nancia sector y when comparing the cosed to the open economy. However in the presence of the coatera e ect of domestic oans on the foreign capita suppy, there may be on the contrary a positive reationship, i.e. nancia integration can reduce both the size of the domestic nancia system and the growth rate of the economy. 7. Back to the empirica puzze. In the introduction we motivated this paper through the existence of a negative correation between the voume of credit extended in the economy and the share of foreign assets in the goba banking assets of the economy. In this section our aim is to show that this negative correation can be accounted for within the framework we have considered. At any time t, the goba voume of credit enders accept to extend to rms is given by t = + f yt + f 1 y t! where y t = is the measure of rms which are abe to borrow from both domestic and foreign enders at time t. As is cear from the ast expression, t is an increasing function of y t because an increase in the reative suppy in domestic credit wi reduce the interest rate on domestic oans, thereby aowing some new rms to enter on the market for domestic oans which wi eventuay raise the borrowing capacity of these rms vis-à-vis foreign enders. At time t the share of foreign oans in the tota voume of oans t is given by t = f y t + f 1 + f yt + f 1 y t y t As is cear from this expression, t is a decreasing function of the reative domestic capita suppy y t. Firms debt portfoio incorporates proportionay more foreign oans when the size of the goba debt portfoio is 22
smaer. For instance when rms are unabe to borrow from domestic enders, then their debt portfoio is made excusivey of foreign debt. On the contrary when a rms are abe to borrow from domestic enders, then the share of foreign debt in their debt portfoio is equa to f + f. Therefore as the reative domestic capita suppy y t increases the share of domestic rms which can borrow from foreign enders increases and the share of foreign oans in the goba voume of credit decreases. We therefore end up with a negative reation between the goba voume of credit and the share of foreign oans in the goba voume of credit extended. This means that countries which rey more on foreign enders to nance domestic rms do not bene t from a arger voume of credit. On the contrary, the arger the share of foreign enders in the capita market, the ower the voume of credit extended. Thus the coatera e ect is a good candidate to account for the empirica reationship raised in the introduction. Simiary one can study the e ect of a change in the reative domestic capita suppy on growth. Noting g t the growth rate of the economy (i.e. the growth rate of aggregate weath) at time t writes as m + f (m g t = r f ) + yt h (m + ) + (m + r f ) f i 1 + y t yt 2 h + f i Then an increase in the reative domestic capita suppy y t raises the growth rate of the economy if and ony if s y t < 1 + + f m r f 1 f + 1 + m r f f 1 This can be easiy expained: for ow vaues of y an increase in the reative domestic capita suppy is reduces domestic enders productivity since the interest rate on domestic oans decreases. However the increase in the reative domestic capita suppy has a arge positive impact on rms pro ts since it raises the measure of rms which can borrow from domestic capita markets and thereby raises the agregate borrowing capacity of the economy w.r.t. foreign enders. We can therefore concude that there is a positive correation between the extent to which investment is domesticay nanced and the growth rate of the economy when domestic nance represents a sma share of GDP. However when y is arge then the negative e ect on domestic enders productivity becomes arger than the positive e ect on rms pro ts since domestic enders are much arger 23
reative to the rm sector. This impies that for arge vaues of y, an increase in the reative domestic capita suppy reduces the growth rate of the economy. g µ y Figure 6: Domestic capita suppy and Growth. 8. Concusion. We have buit a mode of nancia integration where the main idea consists in noting that under imperfect capita markets, oans from domestic enders can act as coatera to foreign enders when rms try to borrow from internationa capita markets. We have shown that this property can be derived in a simpe capita market mode where domestic enders have an informationa advantage over foreign enders. Then based on this mechanism we have shown that nancia integration creates opposite forces: one the one hand, it reduces the cost of domestic capita because nancia integration brings new nancing sources to rms. On the other hand it deteriorates domestic enders pro ts which in a dynamic framework reduces domestic enders capita suppy and through the coatera e ect aso reduces foreign enders capita suppy. We have shown that this framework can account for the fact that nancia integration may enhance or reduce growth depending upon the intensity of credit constraints. In particuar, there are cases where nancia integration reduces both the suppy of credit from domestic enders and the interest rate on domestic oans which depresses economic growth. With this mode, we have aso shown that we can account for the negative empirica reationship 24
between the voume of credit nancia intermediaries accept to extend and the share of oans coming from foreign enders. This mode therefore iustrates two ideas: rst the decrease in the cost of capita foowing nancia integration (iberaization) may not come for free as it reduces domestic enders capita suppy which creates a negative informationa externaity on foreign enders and can thereby transate into a reduction in the goba voume of capita enders accept to end. Second, because rms with di erent productivities do not have the same borrowing capacity w.r.t. foreign enders, nancia integration has a quaitative e ect on the reation between the growth rate of the economy and the size of its nancia sector. We have shown that economies may end up with smaer domestic nancia systems as a resut of increased competition and nancia integration whie not reaping the bene ts in terms of an enhanced growth performance. 9. Appendix. To investigate the reationship between the voume of credit (private credit by banks to GDP) and foreign banking penetration (the share of foreign bank assets in tota banking sector assets). we run a number of regressions to estimate the determinants of private credit to GDP. More precsiey we estimate the foowing equation pc i;t = i + t + x i;t + fb i;t + fb i;t x i;t + " i;t Banking private credit to GDP is pc, x is a vector of contro variabes, fb is an indicator of foreign banking penetration and and are respectivey time and xed e ects. As contro variabes, we incude measures of nancia markets size and macroeconomic indicators. Finay, we add the share of foreign banking assets in tota banking assets and an interaction term between an indicator of the banking sector size and the share of foreign assets in tota banking assets. The next tabes summarize the resuts we obtain. 25
Tabe 2a. Dependent variabe: Private Credit by banks to GDP bagdp 0.91 a 0.89 a 0.63 a 0.60 a 0.63 a cbagdp -0.00 o agdp 0.03 b 0.04 a gdp 0.50 a -0.04-0.06 fba -0.12 a -0.11 b -0.14 b -0.06 b -0.14 a -0.15 a fbabagdp 0.12 b 0.14 b contros yes yes yes yes yes yes xed E ects yes yes yes yes yes yes time E ects yes yes yes yes yes yes N obs. 168 175 293 293 293 293 Note: bagdp stands for banking assets to GDP, cbagdp stands for centra bank assets to GDP, o agdp stands for other nancia intermediaries assets to GDP, gdp stands for iquid iabiities to GDP. The share of foreign banking assets in tota banking assets is fba and fbabagdp is an interaction term between fba and bagdp. Contros incude the og of GDP per worker, the GDP growth rate, the og of popuation, the og of CPI in ation and the openness to trade ratio measured as the sum of imports and exports to GDP. A the estimations contain xed and time e ects and have been carried out under the assumption of heteroscedactic residuas. A data are drawn from Beck, Demirgüç- Kunt, and Levine [1999] "A New Database on Financia Deveopment and Structure" for nancia variabes and from Word Bank Word Deveopment Indicators for macroeconomic variabes. The time period of the sampe is 1990-1997. Countries in the sampe: Argentina, Austraia, Austria, Begium, Brasi, Canada, Switzerand, Chii, China, Costa Rica, Germany, Ecuador, Egypt, Spain, France, United Kingdom, Greece, Honk-Kong, Indonesia, India, Itay, Japan, Netherands, Norway, Pakistan, Peru, Phiippines, Poand, Portuga, Singapore, Sweden, Thaiand, Tunisia, Turkey, Uruguay, USA, Venezuea, South Africa. 26
Tabe 2b. Dependent variabe: Private Credit by banks to GDP pcogdp 0.05 0.06 0.13 a 0.07 a 0.07 b 0.07 a bagdp 0.90 a 0.88 a 0.62 a 0.60 a 0.62 a cbagdp 0.06 o agdp -0.01-0.05 gdp 0.49 a -0.01-0.06 fba -0.11 a -0.11 a -0.14 a -0.06 b -0.14 a -0.14 a fbabagdp 0.11 b 0.14 b contros yes yes yes yes yes yes Fixed E ects yes yes yes yes yes yes Time E ects yes yes yes yes yes yes N.obs. 168 175 293 293 293 293 Note: pcogdp stands for private credit to GDP by non-bank nancia intermediaries, bagdp stands for banking assets to GDP, cbagdp stands for centra bank assets to GDP, o agdp stands for other nancia intermediaries assets to GDP, gdp stands for iquid iabiities to GDP. The share of foreign banking assets in tota banking assets is fba and fbabagdp is an interaction term between fba and bagdp. Contros incude the og of GDP per worker, the GDP growth rate, the og of popuation, the og of CPI in ation and the openness to trade ratio measured as the sum of imports and exports to GDP. A the estimations contain xed and time e ects and have been carried out under the assumption of heteroscedactic residuas. A data are drawn from Beck, Demirgüç-Kunt, and Levine [1999] "A New Database on Financia Deveopment and Structure" for nancia variabes and from Word Bank Word Deveopment Indicators for macroeconomic variabes. The time period of the sampe is 1990-1997. Countries in the sampe: Argentina, Austraia, Austria, Begium, Brasi, Canada, Switzerand, Chii, China, Costa Rica, Germany, Ecuador, Egypt, Spain, France, United Kingdom, Greece, Honk-Kong, Indonesia, India, Itay, Japan, Netherands, Norway, Pakistan, Peru, Phiippines, Poand, Portuga, Singapore, Sweden, Thaiand, Tunisia, Turkey, Uruguay, USA, Venezuea, South Africa. 27