Costly Information, Entry, and Credit Access

Transcription

1 Universiy of Pennsylvania ScholarlyCommons inance Papers Wharon aculy Research Cosly Informaion, Enry, and Credi Access Todd A. Gormley Universiy of Pennsylvania ollow his and addiional works a: hp://reposiory.upenn.edu/fnce_papers Par of he Economic Theory Commons, and he inance and inancial Managemen Commons Recommended Ciaion Gormley, T. A. (2014). Cosly Informaion, Enry, and Credi Access. Journal of Economic Theory, hp://dx.doi.org/ /j.je This paper is posed a ScholarlyCommons. hp://reposiory.upenn.edu/fnce_papers/390 or more informaion, please conac reposiory@pobox.upenn.edu.

2 Cosly Informaion, Enry, and Credi Access Absrac Using a heoreical model ha incorporaes asymmeric informaion and differing comparaive advanages among lenders, his paper analyzes he impac of lender enry on credi access and aggregae ne oupu. The model shows ha lender enry has he poenial o creae a segmened marke ha increases credi access for hose firms argeed by he new lenders bu poenially reduces credi access for all oher firms. The overall impac on ne oupu depends on he disribuion of firms, he relaive coss of lenders, and he cos of acquiring informaion. The model provides new insighs ino he evidence regarding foreign lenders' enry ino emerging markes. Disciplines Economic Theory inance and inancial Managemen This journal aricle is available a ScholarlyCommons: hp://reposiory.upenn.edu/fnce_papers/390

3 Cosly Informaion, Enry, and Credi Access Todd A. Gormley Universiy of Pennsylvania This Draf: May 7, 2014 Absrac Using a heoreical model ha incorporaes asymmeric informaion and differing comparaive advanages among lenders, his paper analyzes he impac of lender enry on credi access and aggregae ne oupu. The model shows ha lender enry has he poenial o creae a segmened marke ha increases credi access for hose firms argeed by he new lenders bu poenially reduces credi access for all oher firms. The overall impac on ne oupu depends on he disribuion of firms, he relaive coss of lenders, and he cos of acquiring informaion. The model provides new insighs ino he evidence regarding foreign lenders enry ino emerging markes. Keywords: Asymmeric Informaion, Compeiion, Credi, inancial Liberalizaion JEL Classificaion: D82, 3, G2, O16, O19. The Wharon School, Universiy of Pennsylvania, 3620 Locus Walk, SH-DH Suie 2400, Philadelphia, PA gormley@wharon.upenn.edu. Telephone: 1 (215)

4 1. Inroducion By allowing financial insiuions in developed counries o lend direcly o firms in less developed counries (LDCs), open capial markes are generally hough o alleviae domesic liquidiy consrains, o improve he allocaion of credi, and hence o increase aggregae ne oupu. As a resul of hese poenial benefis, many LDCs opened heir capial markes in he 1980s and 1990s. These openings fosered foreign lenders enry ino heir economies and changed he local compeiive srucure of heir financial secors. Bu, he assumpion ha opening capial markes is beneficial has recenly come under serious doub, as empirical sudies have repeaedly failed o find a consisen relaion beween foreign lenders enry, credi access, and ne oupu in LDCs. 1 This lack of empirical evidence leads o his paper s cenral quesion: Why migh he enry of new lenders, as experienced in many LDCs, no increase credi access and aggregae ne oupu? In his paper, I show ha informaion asymmeries and compeiive ineracions beween lenders wih differing comparaive advanages provide an answer. This paper presens a heoreical framework ha explains how lender enry ino an already compeiive credi marke can affec firms access o credi when he enering lenders enjoy a differen cos of capial and abiliy o acquire informaion abou firms han incumben lenders. Specifically, he model assumes enering lenders have a lower cos of capial bu incumben lenders deermine firms qualiy a a lower fixed screening cos per firm. 2 Wihin his framework, i is possible o derive a number of novel predicions. irs, new lender enry has he poenial o induce a segmened credi marke ha reduces credi access for many firms. The inuiion is sraighforward. When he cos of acquiring informaion is sufficienly high, a compeiive, closed-economy equilibrium may occur in which incumben lenders pool all firms ogeher wih a uniform financial conrac raher han inves in he 1 or example, Rodrik (1998) and Edison, Levine, Ricci, and Sløk [25] find no effec of open capial markes and financial inegraion. See Eichengreen [27] for a more deailed review of his lieraure. More recen research focusing on he specific impac of foreign paricipaion in domesic equiy markes and foreign bank enry also reaches differing conclusions. or example, Bekaer, Harvey, and Lundblad [10] and Henry [36] find posiive correlaions beween equiy marke liberalizaion and economic performance, whereas Deragiache, Tressel, and Gupa [24] and Gormley [32] find foreign bank enry o be negaively relaed o overall domesic credi. 2 The comparaive advanage of enering lenders a higher cos of screening bu lower marginal cos of funds finds subsanial suppor in boh he heoreical and empirical lieraures on foreign lender enry ino LDCs (e.g., see Mian [42,43], Micco, Panizza, and Yañez [44], Sein [52]). This evidence is discussed in Secion

5 cosly screening echnology. Relaive o he firs-bes allocaion wihou informaion asymmeries, a pooling equilibrium overfunds low-reurn firms and underfunds high-reurn firms. The enrance of new lenders may break his pooling equilibrium. Because of heir lower cos of funds and he fixed naure of screening coss, enering lenders may find i worhwhile o acquire informaion abou firms ypes so as o offer more compeiive conracs o high-reurn firms capable of profiably invesing large amouns of capial a pracice commonly called cream skimming. While some firms benefi from cream skimming, he resuling separaing equilibrium may reduce credi access for oher firms by changing he se of financial conracs available o hem. This poenial decline in credi access leads o he model s second implicaion: Addiional lenders enry has he poenial o eiher increase or reduce ne oupu. Cream skimming by enering lenders increases ne oupu by eliminaing he underfinancing of high-reurn firms capable of profiably invesing large amouns of capial. The ne oupu of all oher firms, however, may decline. Because cream skimming reduces he average qualiy of firms ha accep pooling conracs, hese conracs may become more expensive, reducing he ne oupu of firms ha accep hem. In some cases, he pooling conrac will become unprofiable for lenders o offer, and he remaining firms will go unfunded enirely, furher reducing ne oupu, if neiher he incumben nor enering lenders find i cos-effecive o acquire he informaion necessary o idenify he remaining high-reurn firms. The model hus provides a relaively simple explanaion as o why open capial markes may no necessarily increase overall oupu in LDCs. In LDCs wih significan informaion acquisiion coss, he iniial domesic allocaion of credi may fail o achieve he firs-bes allocaion because domesic lenders opimally choose o pool risks and cross-subsidize losses on low-reurn firms wih gains on high-reurn firms raher han inves in cosly screening echnologies. This ype of lending paern is a sandard problem in emerging economies (Banerjee, Cole, and Duflo [6]). Because he enering foreign lenders have a differen cos srucure, hey will ener via cream skimming, which can boh redirec credi oward he mos profiable firms and reduce he credi access of oher firms ha are less profiable bu sill have posiive ne presen value (NPV) projecs. 2

6 The underlying mechanism by which ne oupu can decline is quie general. The model is robus o allowing for he renegoiaion of conracs afer screening reveals firms ypes and o allowing lenders o offer a very general se of financial conracs, including conracs ha pay low reurn firms o no implemen heir projec. The model is also robus o various assumpions regarding he disribuion of firms and assumpions regarding he correlaion beween firms produciviy and he riskiness of heir projecs. Insead, he key assumpion of he model is ha he enering lenders have a differen comparaive advanage han incumben lenders; his, combined wih imperfec informaion abou borrowers, is wha drives he poenial change in financial conracs available in he compeiive equilibrium ha includes boh lenders. I can also be shown ha he model generaes a decline in ne oupu under wide range of parameer spaces. By demonsraing how he impac of lender enry will depend on he disribuion of firms, he relaive coss of lenders, and he cos of acquiring informaion, he model sheds ligh on why he impac of lender enry migh vary across counries and over ime. A decline in ne oupu is more likely o occur when he cos of producing informaion in he local economy is greaer. Thus, counry-level facors ha migh affec lenders cos of screening, including he qualiy of he counry s local insiuions (e.g., weak enforcemen of accouning sandards or a lack of credi raing agencies), will be imporan. The model can also be exended o demonsrae ha cream skimming and a decline in ne oupu is less likely o occur when he comparaive advanages of lenders can be combined hrough a merger or syndicaed loan. If allowed, enering lenders will ofen prefer o expand heir screening capaciy upon enry via acquiring local lenders, and i can be shown ha such acquisiions will reduce he likelihood of a decline in ne oupu. The model also sheds ligh on how enry can affec incumben lenders invesmens in screening capaciy and experise; he arrival of a lender wih a differen comparaive advanage can increase incumben lenders incenive o inves in screening capaciy so as o mainain marke share in he open economy. The model can explain a number of he exising empirical findings regarding foreign lenders enry, credi access, and aggregae oupu in LDCs. or example, he model can explain why foreign 3

7 lenders ofen only arge he leas informaionally opaque, larges, and mos profiable firms (Berger, Klapper, and Udell [11], Clarke, Cull, Peria, and Sánchez [19], Gormley [32], Mian [43]) and why his cream skimming can be associaed wih an exi by domesic lenders and overall decline in credi (Beck, Demirguc-Kun, and Peria [8], Beck and Peria [9], Deragiache, Gupa, and Tressel [24], Gormley [32]). The model also provides an explanaion as o why foreign lender enry is no always associaed wih an increase in subsequen economic growh or why financial liberalizaion migh only be posiively associaed wih growh in counries in which screening coss are likely less, such as highincome counries or counries wih sronger local insiuions (e.g., Area, Eichengreen, and Wyplosz [5], Edwards [26], Galindo, Micco, and Ordoñez [29], Quinn [48]). The model can also explain why acquiring a domesic lender is a popular mode of enry in many LDCs or why an increase in growh is more likely o be observed when such acquisiions are allowed (e.g., Bruno and Hauswald [14], Giannei and Ongena [30]). Overall, he analysis provides new insighs abou he poenial consequences of financial liberalizaion and is relaed o four disinc lieraures. irs, he heoreical predicion ha lenders more efficien a financing cerain ypes of firms may exi following enry by oher lenders is similar o he argumen ha compeiion does no always resul in survival of he fies (Bolon and Scharfsein [12], Zingales [55]). The model exends his idea by demonsraing ha he exi of he seemingly more efficien lender can occur even when he surviving lenders are no shielded from poenial new enrans or when he exiing lender does no face direc compeiion in he marke in which i enjoys an efficiency advanage. Insead, he exi is driven by addiional enry making i difficul o offer cross-subsidized producs in a marke wih informaional asymmeries. Second, his paper is relaed o he heoreical lieraure on he effecs of compeiion in he presence of imperfec informaion. The poenial for posiive NPV firms o go unfinanced in compeiive credi markes is similar o ha of Sigliz and Weiss [54], while he poenial nonexisence of equilibrium and use of cream-skimming sraegies is similar o ha which can occur in models of insurance markes (e.g., Lewis and Sappingon [40], Rohschild and Sigliz [50]). In 4

8 conras o hese models, where screening occurs hrough agens self-revealing choices from he menu of offered conracs, his paper analyzes he effecs of compeiive screening in a seing where lenders are able o acquire and use privae informaion abou agens ypes o limi heir choice from he menu of offered conracs. urhermore, raher han analyze he effecs of compeiion among agens ha share he same screening echnology, his paper analyzes how he enry of agens wih a differen comparaive advanage can affec he compeiive equilibrium. In his regard, he paper is similar o Marin [41], which analyzes how he inroducion of a non-exclusive alernaive source of funds can affec he compeiive equilibrium in a marke wih asymmeric informaion. Third, his paper is relaed o he growing body of lieraure concerning he effecs of compeiion on lending relaionships and credi access (e.g., Boo and Thakor [13], Peersen and Rajan [46]). Raher han look a in increase in compeiion, however, his paper analyzes how he inroducion of lenders wih a differen comparaive advanage ino an already compeiive economy affecs equilibrium conracs. This is similar o Dell Ariccia and Marquez [22] and Sengupa [51], which demonsrae ha his ype of enry can induce segmened credi markes. However, by assuming ha incumben lenders have perfec informaion abou borrower ypes, neiher of hese papers is able o shed ligh on why segmened markes migh induce declines in credi access. 3 While his possibiliy is explored in Deragiache, Gupa, and Tressel [24], his paper differs in ha i can capure boh cases where all firms benefi from enry and cases where some firms do no. Addiionally, his paper explores how enry will affec lenders incenive o expand heir screening capaciy and how he impac on ne oupu will depend on he cos of acquiring informaion. inally, his paper is relaed o he growing body of lieraure on he impac of open capial markes and capial inflows. Despie growing empirical and anecdoal evidence o sugges a poenial dark side o capial inflows, he argumen is ofen made ha lowering enry barriers will be 3 However, in Dell Ariccia and Marquez [22], lender enry does increase incumben lenders loan porfolio risk, which in a more complee model wih cosly capial could cause a reduced lending capaciy for incumben lenders. Alhough his has he poenial o generae adverse effecs on credi similar o his paper, heir paper does no explore his possibiliy. 5

9 unambiguously beneficial o he growh of LDCs. 4 One possible reason for his apparen disconnec is ha here is lile heoreical undersanding as o how capial inflows migh adversely affec he local economy beyond heir poenial o reduce financial sabiliy (Agénor [3], Dell Ariccia and Marquez [23], Eichengreen and Leblang [27], Kaminsky and Schmukler [38], Sigliz [53]). This paper formalizes a heory for why capial inflows may adversely affec he local economy, even in he absence of reduced financial sabiliy. The model demonsraes his channel o be quie robus o assumpions abou local compeiion, firms, and lenders, while also providing guidance on exacly when fosering enry ino financial markes will be beneficial. The resuling policy implicaions of his analysis are quie differen han hose ha focus on financial sabiliy. The remainder of his paper proceeds as follows. Secion 2 provides he basic seup and assumpions of he model. Secion 3 discusses he possible equilibria prior o he new lenders enry, and Secion 4 describes he possible equilibria following enry. Secion 5 hen analyzes he facors ha deermine he impac of lender enry on ne oupu. Secion 6 demonsraes he robusness of he models findings and discusses possible exensions. Secion 7 discusses empirical evidence regarding he model s key assumpions and esable predicions. inally, Secion 8 concludes. 2. The Basic Model 2.1 Agens and Technology There are wo ypes of agens: firms and lenders. All agens are risk neural, and because of limied liabiliy, no firm can end wih a negaive amoun of cash. The real secor consiss of hree ypes of firms, i { A, B, C}, and a coninuum, i, of each ype, where A B C is normalized o equal one. Each ype of firm has he abiliy o implemen one projec of size I {1, }, where 1. If successfully implemened, he projec yields a verifiable 4 or example, in a memo o he World Trade Organizaion on June 6, 2005, delegaions from Japan, he Unied Saes, and European Union argued ha Policies ha impede compeiion, such as enry resricions and resricions on foreign banks, have been shown o raise he cos of financial services and hur economic performance. WTO Documen #

10 reurn, RI r I, where r is an exogenous cos of capial. or simpliciy, all firms have zero wealh and mus borrow he enire amoun I from lenders o implemen he projec. Among he hree ypes of firms, here will be one ype ha lenders always wan o finance, C (he cream ), anoher ype hey never wan o finance, B (he bad ), and a hird ype ha hey only wan o finance for small projecs, A (he average ). This is formally esablished by having he hree ypes differ in heir abiliy o implemen projecs successfully. If financed, he cream firms always succeed wih probabiliy 1, regardless of projec size, whereas bad firms only succeed wih probabiliy p. Projecs ha only succeed wih probabiliy p have a negaive ne expeced reurn given he cos of funds, r, such ha pr r. Average firms, however, implemen he smaller projec of size 1 wih cerain success, whereas larger projecs only succeed wih probabiliy p. Given his seup, he firs-bes allocaion of credi is achieved and ne oupu is maximized when cream firms are financed for projecs of size, average firms for projecs of size 1, and bad firms are no financed. 5 The concep of cream firms should be inerpreed broadly. Their abiliy o successfully implemen he projec of size 1 serves o represen high-reurn firms capable of profiably invesing large amouns of capial. This includes firms able o inves larger amouns of capial oday or firms able o inves in more fuure projecs. Hence, cream firms are no necessarily larger in size oday or able o inves in larger projecs. The financial secor consiss of many perfecly compeiive lenders willing o exend capial in he amoun of I {1, }. Wihou cosly invesmens in he producion of informaion abou firms ypes, lenders are unable o idenify a firm s ype, hus providing he source of informaion asymmery in he model. Lenders, however, may inves in a screening echnology ha perfecly idenifies a firm s ype. The cos of his screening echnology will capure he severiy of he asymmeric informaion 5 This seup is a specific case of he more general framework, where bad firms succeed wih probabiliy p L, cream firms succeed wih probabiliy p H > p L, and p H R > r > p L R. Assuming ha p H = 1 and p L < 1 only helps simplify expressions, and all subsequen findings and inuiion hold in he more general case, which is provided in he online Appendix. Therefore, he implici negaive correlaion beween risk (variance of oupu) and produciviy in his specific case is no necessary for he model s findings. In he general case, he implici correlaion beween risk and produciviy can go eiher way. 7

11 problem. There will be wo ypes of lenders: domesic and foreign. Domesic lenders will be he incumben lenders, whereas foreign lenders will be he poenial new enrans ino he economy. oreign and domesic lenders will differ in wo key ways: Domesic lenders will find i less cosly o produce informaion abou firms ypes, whereas foreign lenders will enjoy a lower cos of funds. Specifically, domesic lenders can screen a cos 0 per firm, whereas foreign lenders mus pay. 6 The lower screening cos for domesic lenders will reflec heir prior experience wih lending o firms in he local economy. Regarding he cos of funds, foreign lenders have access o an unlimied supply of funds a cos, r 0, whereas domesic lenders have access o an unlimied funds a a cos, r, where r r. The lower cos of funds for foreign lenders will reflec some operaional and echnological advanage of he new enran over ha of he incumben lenders. The differences in coss provide each lender wih a poenial comparaive advanage. Domesic lenders have an informaion producion advanage per firm, whereas foreign lenders have a cos of capial advanage per dollar invesed. Thus, for firms wih large enough credi needs, a foreign lender will have a compeiive advanage regardless of wheher informaion producion is necessary. To formalize his comparaive advanage and he above resricions on parameer values, he following assumpions are made: r r, (A1) r r (A2) Assumpion (A1) formalizes ha domesic lenders have a higher cos of funds and a lower screening cos. Assumpion (A2) ensures ha projecs of size are sufficienly large o provide foreign lenders he compeiive advanage in financing hese projecs. The assumed comparaive advanages of domesic and foreign lenders appear o fi well in he conex of inernaional capial markes and cross-border lending. This evidence is discussed in Secion The assumpion of a uniform, per firm screening coss grealy reduces he analysis, bu is no essenial. All subsequen findings will hold in a more general seing in which screening coss are allowed o vary wih he scale of expeced lending o a firm so long as he screening cos does no increase 1-1 wih he amoun of expeced lending. 8

12 2.2 Timing of Evens There is no discouning beween periods, and he iming of evens is as follows: 0 : firms discover heir ype, i, 1: lenders choose heir menu of financial conracs, ; firms apply for financing, 2 : lenders screen applicans and provide capial, I, o successful applicans, 3 : projec oucomes are realized; financial conracs are seled. The basic idea of his ime line is he following: Lenders iniially choose which menu of financial conracs hey wish o offer firms. In doing his, hey will choose boh which ype of financial conracs o offer and o which firms hey will offer hese conracs. irms hen approach lenders and apply for heir preferred financial conrac from he menu of available conracs. If he conrac is designaed for firms of a specific ype, he lenders inves in informaion producion and screen applican firms o verify heir ype, and financing is provided o successful applicans. inally, projec oucomes are realized and all financial conracs are seled. 2.3 inancial Conracs and Sraegies Le j represen he menu of conracs offered by lender j, where denoes a financial conrac from lender j in amoun I designaed for firms of ype k {0, A, B, C}. When k 0, he conrac is unscreened and available o all firms, regardless of ype, bu for k 0, he lender acquires informaion abou firms ypes and he conrac is only available o firms for which screening by he lender reveals i k. 7 Each conrac is a mapping of he observable oupu from he projec ino a paymen for he firm. Specifically, :{0, RI}. Each ype of conrac maps ino a nonnegaive paymen because firms have no iniial wealh and canno receive a negaive paymen. Moreover, i is imporan o noe ha his mapping spans he universe of poenial conracs, and hence he concep of a lender used here is very general and encompasses deb, equiy, or any mixure hereof. Ik, j 7 The analysis and subsequen findings are qualiaively similar if lenders are allowed o offer nondeerminisic screening conracs in which screening only occurs wih some probabiliy. 9

13 A sraegy configuraion in his economy consiss of he se of conracs j for each lender j L, and he conrac choice, f () i, for each firm i E. A firm s choice is limied o he se of conracs offered by lenders,, or a choice of no conrac, f() i. The equilibrium concep used is subgame perfec, and a sraegy configuraion will be an equilibrium if each lender j and each firm i is maximizing is expeced profis given he sraegies of all oher agens in he economy. The expeced profi of a firm i wih financial conrac can be expressed as i pi IRI 1 p( i I) 0, where pi Iis he probabiliy of success for a firm of ype i wih a projec of size I, which is deermined by he amoun of financing associaed wih he finance conrac,. Likewise, he expeced profis of lender j lending o firm i wih conrac is j ( i ) p( i I) R r( j) I ( ) ( j) S, where r( j ) and ( j ) represen he cos of funds and screening for lender j ; I represens he amoun of financing associaed wih conrac ; and S 0 for unscreened conracs and equals one oherwise. inally, le (, ) be he se of firm ypes ha accep he conrac offer when he se of available financial conracs is. In oher words, i (, ) if and only if f () i. Given his, he economy s equilibrium is formally defined as Definiion of Equilibrium: A sraegy configuraion, f () i for each firm i E and implied by j for each lender j L, consiues equilibrium if and only if (1) given, each firm i E chooses f() i o maximize ( f ); (2) each lender j L chooses j o maximize given by condiion 1; and, j( i j) di, where i (, ) i j j is (3) because of free enry, each lender makes zero profis, i (, ) j j ( i ) di 0. j The inuiion for he equilibrium is as follows. The firs condiion saes ha, given he se of all available conracs offered by lenders, each firm in he economy is choosing he financial conrac 10

14 ha maximizes heir expeced profis. The second condiion saes ha, given each firm s opimal conrac choice from he available menu of conracs offered by all lenders, each lender is offering a menu of conracs ha maximizes heir own expeced profis. In oher words, no individual lender can improve heir own profiabiliy by deviaing and offering a differen se of conracs o firms. The hird condiion arises from free enry; all lenders make zero expeced profis in equilibrium. Before solving he equilibrium, i is firs worh noing he wo implici assumpions being made in he model. These assumpions simplify he iniial analysis bu are no crucial o resuls. irs, I am assuming ha all firms implemen he projec if hey receive financing from a lender. In he absence of his assumpion, lenders migh have an incenive o offer a conrac ha acually pays bad firms o no implemen he projec. Whereas a conrac ha pays bad firms o do nohing can never be an equilibrium conrac, because each individual lender could improve profiabiliy by dropping he conrac, his ype of conrac migh be a profiable deviaion for lenders in an equilibrium in which bad firms accep unscreened conracs and implemen projecs. Paying bad firms o do nohing may be less cosly han allowing hem o implemen projecs. In realiy, his deviaion is unlikely o be profiable because such paymens for doing nohing would induce all individuals wihou projecs o seek he same payoff. This can be easily capured by inroducing a fourh ype of firm ha has no projec. So long as he mass of hese firms is sufficienly large, an unscreened conrac ha pays a posiive amoun o borrowers o ake no acion will no be a profiable deviaion. A screened conrac ha pays bad firms o no implemen a projec will also no be a profiable deviaion so long as he cos of screening exceeds he expeced loss on bad firms. 8 Second, I am implicily assuming ha lenders can fully commi o heir financial conracs in wo ways. (1) Lenders will always acquire informaion abou firms ypes and screen financial conracs of ype k 0. This eliminaes lenders from deerring bad borrowers by declaring ha all conracs will be screened, bu no acually screening hem. (2) Lenders can fully commi o he iniial 8 Acemoglu [1] uses a similar mehod o eliminae hese unrealisic ypes of conracs, and an exension of he model ha relaxes his assumpion abou implemening projecs is available in he online Appendix. 11

15 erms of any conrac,, and heir iniial menu of conracs, j. In oher words, here is no possibiliy of renegoiaion beween lenders and firms afer screening reveals a firm s ype, and hence firms will have no incenive o misrepresen heir ype when applying for a screened conrac. Wih a few exensions of he model, i can be shown ha full commimen by lenders is an equilibrium sraegy in a repeaed game. In a repeaed game, full commimen can be accomplished by assuming ha firms observe wheher lenders have violaed full commimen in he pas and by assuming ha firms assign a nonzero probabiliy of such lenders doing so again in he fuure. Wih hese assumpions, deviaions from full commimen, which can yield immediae gains, will arac applicans in he fuure ha are ex ane unprofiable for he lender o do business wih. The fuure cos of screening hese unwaned applicans will exceed he immediae gains and preven lenders from deviaing. or example, consider an equilibrium in which bad firms are no financed and average and cream firms are offered screened conracs. If a lender deviaes from full commimen and does no acually screen is average and cream applicans, i gains immediaely by avoiding he screening coss. The cos, however, is ha all bad firms will apply for he lender s screened financial conrac in he fuure because hey assign a nonzero probabiliy of he lender shirking again and heir ouside opion is zero. Screening and urning away hese bad firms in he fuure is cosly, and he immediae gains from deviaing from full commimen will be offse by hese expeced losses Equilibrium prior o Enry In an economy ha consiss of only domesic lenders, he domesic cos of screening,, will deermine wheher a pooling or separaing equilibrium exiss. Domesic lenders can always offer cream firms a lucraive, screened conrac of size ha provides expeced profis of ( Rr) o he firm. The cos of screening,, reduces he firm s expeced profi since he lender will only offer a conrac ha allows i o recoup is coss and breakeven in expecaion. Alhough his conrac clearly dominaes any screened conrac of size 1 for a cream firm, i may no dominae an unscreened 9 This repeaed game exension is available in he online Appendix. 12

16 conrac. Unscreened conracs avoid he cos of screening,, bu ineviably finance some negaive NPV projecs. When he cos of screening,, is sufficienly high, cream firms will prefer unscreened conracs being offered by domesic lenders, resuling in a pooling equilibrium in which all firms accep he same unscreened conrac. And, when is sufficienly low, cream firms prefer screened conracs, resuling in a separaing equilibrium. To simplify he equilibrium, I will assume here is relaively small number of cream firms, such ha lenders can never profiably pool jus cream and bad firms ogeher on an unscreened conrac. This reduces he number of possible pooling equilibriums bu does no qualiaively affec he subsequen resuls. 10 This is accomplished wih he following assumpion: B C ( R r) (A3) r pr This assumpion ensures ha for any unscreened conrac, he ne loss per uni of invesmen for bad firms, ( r pr ), exceeds he ne gain per uni of invesmen for cream firms, ( R r ). This will B hold whenever here is a significanly large raio of bad o cream firms. Wih he above assumpion, he only possible pooling conrac will be one ha pools all firms ono he smaller projec. The highes expeced profis ha such a conrac can provide o cream C firms is Rr/[1 (1 p ) ]. Thus, when, where is defined by equaion (1), he economy B can exhibi a pooling equilibrium in which all firms prefer o accep a small unscreened conrac of size 1. And, when, he larger screened conrac, which provides a payou, ( Rr ), is preferred by cream firms, resuling in a separaing equilibrium in which cream firms prefer o ake screened conracs for he larger invesmen Absen assumpion (A3) and when λ is sufficienly large, here will exis an unscreened conrac ha pools firms ono he larger projec. Bu, similar o he smaller pooling conrac, his larger pooling conrac only exiss in equilibrium if he cos of screening is sufficienly high ha lenders canno profiably offer a large, screened conrac ha is preferred by cream firms. or his reason, subsequen findings are similar when he closed economy sars from such a pooling equilibrium; foreign enry sill increases he likelihood of a separaing equilibrium where credi access and ne oupu boh decline. However, one difference wih saring from his pooling equilibrium is ha foreign enry has less poenial o increase he ne oupu of cream firms. 11 Because hey can successfully inves larger amouns of capial, cream firms, raher han average firms, are always he firms ha lenders can mos easily enice o ake a screened conrac, hus iniiaing a separaing equilibrium. 13

17 r ( R r) R (1) 1 (1 p) B The range of screening coss when a separaing equilibrium occurs,, will be higher when he amoun of capial,, and reurn, R, of a cream firm s invesmen is larger. This will increase he araciveness of a screened conrac o cream firms. An increase in he number of bad firms, B, or a reducion in heir probabiliy of success, p, will increase he cos of he pooling conrac, also increasing he chance of a separaing equilibrium. The oucome for average and bad firms in a separaing equilibrium will depend on wheher an unscreened conrac ha pools jus bad and average firms is feasible or wheher i is feasible for lenders o screen average firms. The more inriguing equilibrium is he poenial pooling equilibrium for. The pooling equilibrium always fails o achieve he firs-bes allocaion because cream firms fail o ake on larger projecs, and bad firms are financed for negaive NPV invesmens. unds divered away from bad firms oward larger projecs for cream firms would increase ne oupu, and here is a poenial for he enry of new lenders, wih a differen comparaive advanage, o increase overall oupu. The pooling equilibrium will exis if domesic lenders can profiably pool all borrowers, which is rue when r /[1 (1 p) ] R, and here does no exis any oher conrac capable of enicing cream firms B away from he unscreened conrac (i.e., ). This equilibrium is described in Proposiion 1. Proposiion 1. In an economy wih only domesic lenders, where and r/1 (1 p) B R, here is an unique equilibrium in which all firms accep an unscreened financial conrac of size I=1 wih payoffs (Y ) R r /1 (1 p) B if Y R. 0 oherwise The equilibrium conrac can be inerpreed as a deb conrac. irms receive nohing in failure bu receive a posiive payoff in success, wih an implici equilibrium ineres rae of r r / 1 (1 p) B r. This ineres rae is jus enough o offse lenders expeced losses on he fracion (1 p) B of projecs ha will be aken by bad firms and subsequenly fail. A proof of Proposiion 1 is found in he Appendix. 14

18 In he conex of opening capial markes, he pooling equilibrium appears o capure economic characerisics ofen used o moivae financial liberalizaion in LDCs. There is an overfinancing of bad firms and underfinancing of good firms. Moreover, he pooling equilibrium occurs when he cos of acquiring informaion is high, which is a common characerisic of emerging economies (Aleem [4]). Empirical evidence also suggess a lack of informaion producion done by domesic lenders in many emerging markes. 12 Given his, I will now analyze he impac of allowing foreign lenders o ener an economy ha exhibis a pooling equilibrium. 4. Equilibrium afer Enry The equilibrium wih boh foreign and domesic lenders also depends on he cos of screening borrowers, bu i now depends on boh he foreign and domesic cos of screening. oreign enry has no effec on he pooling equilibrium allocaion of credi described in Proposiion 1 if foreign lenders cos of producing informaion is prohibiively expensive, such ha, where r ( R r ) R. (2) 1 (1 p) B This hreshold is similar o ha of he economy wih only domesic lenders, bu now he hreshold is deermined by foreign lenders cos of funds, r, and screening,, because, by assumpion (A2), hey enjoy a comparaive advanage in financing cream firms. When foreign lenders cos of screening is sufficienly low, such ha, foreign lenders induce cream firms in a domesic pooling equilibrium o underake larger projecs by offering hem more compeiive conracs for hose projecs. They can accomplish his despie heir higher cos of screening because of heir lower marginal cos of funds and he fixed naure of screening coss. This resul is saed formally in Proposiion 2, and he proof is provided in he Appendix. Proposiion 2. In an economy wih boh foreign and domesic lenders, where and, foreign 12 or an example involving banks in India, see Banerjee, Cole, and Duflo [6]. Gormley, Johnson, and Rhee [33] also provide suggesive evidence ha Korean bond holders did no screen heir invesmens in

19 enry causes a swich from a pooling equilibrium o a separaing equilibrium in which all cream firms accep large screened conracs of size offered by foreign lenders. If, only a pooling equilibrium exiss. While he enry of foreign lenders and a swich from a pooling equilibrium o a separaing equilibrium benefis cream firms wih more lucraive conracs and increases heir ne oupu, average firms may be worse off. If average firms are financed in a separaing equilibrium, hen his will eiher occur hrough a domesic or foreign screened conrac or a foreign unscreened conrac ha pools average and bad firms. Average firms expeced profis and ne oupu under eiher conrac, however, may be lower. or example, if average and bad firms coninue o choose a pooling conrac, he equilibrium payoff of his conrac, R r ( )/( p ), may be lower han ha of he A B A B closed economy payoff, R r/(1 (1 p)θ B ), since he average qualiy of firms being pooled declines and lenders mus charge a higher ineres rae o coninue breaking even. Moreover, i is possible ha neiher a screening nor pooling conrac will be feasible. If 0 max{ R r, R r }, hen neiher ype of lender can profiably screen average firms, and if ( r pr) ( R r ), foreign lenders canno profiably offer an unscreened conrac. The expeced profis from average firms, B A ( ), would no be enough o offse he expeced losses on bad firms, A R r B( r pr ). If boh hese condiions hold, only cream firms will be financed in he separaing equilibrium. The overall impac of foreign enry on ne oupu will depend on he relaive gains and losses of cream, average, and bad firms. or example, in a separaing equilibrium in which neiher average nor bad firms are financed, he enry of foreign lenders will enice cream firms o ake on larger projecs. This increases heir ne oupu by R r R r ( ) C, where r is he equilibrium ineres rae in he closed economy ha exhibis a pooling equilibrium. Bu, he inabiliy of average and bad firms o obain financing causes a loss in ne oupu of R r p ( ) A B. This suggess a 16

20 possible decline in oupu, which is described in Proposiion 3 and proven in he Appendix. 13 Proposiion 3. In an economy ha swiches from he pooling equilibrium wih domesic lenders o he separaing equilibrium wih foreign lenders and no financing of average and bad firms, ne oupu will decline when ( R r ) p ( R r ) R r, where r r p A B C / 1 (1 ) B. The poenial drop in oupu can be considerable, as illusraed in he following numerical example: Suppose successful projecs yield a 15% reurn (R = 1.15) and ha cream firms are able o implemen projecs four imes as large ( 4 ). Cream firms represen one-fifh of he firms ( C 1/5), whereas he oher firms are spli equally beween average and bad ( 2/5) A B. Projecs of bad firms only succeed wih 75% probabiliy (p = 0.75). Domesic lenders cos of funds is 3% (r = 1.03), whereas foreign lenders cos of funds is only 2% (r = 1.02). Under his seup, jus a small difference in screening coss for he wo ypes of lenders will generae differing comparaive advanages and a drop in ne oupu. or example, if = 0.48 and = 0.50 (which would imply breakeven lending raes on large, screened loans from domesic and foreign lenders of 15% and 14.5%, respecively), foreign enry will cause a shif from a pooling equilibrium o a separaing equilibrium, and ne oupu will decline by 20% Comparaive Analysis and Implicaions The model provides a relaively simple explanaion as o why enry by addiional lenders may no necessarily increase overall oupu. In markes wih significan coss o producing informaion abou firms ypes, lenders may choose o pool risks and cross-subsidize losses on low-reurn firms 13 When a decline in ne oupu occurs in he open economy, he separaing equilibrium allocaion is always consrained inefficien, and he pooling equilibrium is always consrained efficien. Because of his, i is possible o use a mechanism design approach o analyze poenial welfare-maximizing policies for economies ha experience a decline in ne oupu afer addiional lender enry. In paricular, i is possible o show ha here exiss a revenueneural policy ha will improve ne oupu by subsidizing he cos of capial len for projecs in he economy and axing he reurns on hese projecs. Taxing projec reurns can be used o implemen he consrained efficien pooling equilibrium by providing a relaive disincenive for cream firms o accep a larger screened conrac. 14 The implied equilibrium lending raes of his numerical example are on par wih ineres raes observed in many credi markes, especially emerging markes. or example, Banerjee and Duflo [7] find ha he average ineres rae of an India bank in heir sample was 16%, Mian (2006) finds ha he average ineres rae of domesic and foreign banks in Pakisan was 12.75% and 10.75%, respecively, and Giannei and Ongena [30] esimae ha he average ineres rae of firms in heir Easern European sample was 23%. 17

21 wih gains on high-reurn firms raher han inves in cosly screening echnologies. Whereas new lenders may be even less effecive a producing informaion, a comparaive advanage in funding coss may allow hem o offer a more compeiive conrac o firms capable of invesing large amouns of capial. Therefore, heir enry can increase ne oupu by inducing hese firms o ake on larger projecs, bu a he same ime, ne oupu may be declining for oher firms as he financial conracs available o hese firms will differ in a separaing equilibrium. This poenial for a decline in credi access and oupu is no found in exising models of compeiion beween lenders wih differen comparaive advanages in screening; hese models find ha enry will improve credi access for all firms (Dell Ariccia and Marquez [22], Sengupa [51]). 15 A he same ime, he model suggess ha he inconclusive evidence peraining o financial liberalizaion may also be he consequence of differences in he underlying fundamenals. or example, ne oupu can decline if foreign enry resuls in a swich from a pooling equilibrium o a separaing equilibrium where boh average and bad firms go unfinanced. This can occur if he cos of obaining informaion is sufficienly high for boh domesic and foreign lenders, bu foreign lenders cos of funds advanage is sufficien enough o faciliae cream skimming. The parameer space where his occurs is described in Proposiion 4 and proven in he Appendix. Proposiion 4. Under he following condiions, foreign enry causes a swich from a pooling equilibrium ha finances all firms o a separaing equilibrium where only cream firms are financed and ne oupu falls: (a) r > r (b) 0 A( R r ) B( r pr), (c) max, R r (d), and R 1 (1 p) B r max R r, ( R r ), 1 A B where and are defined by equaions (1) and (2), respecively. 15 The key reason for his differen oucome is heir assumpion ha incumben lenders have perfec informaion abou borrower ypes, whereas enering lenders have no informaion. The imporance of hese assumpions can be illusraed using a modified version of he model, where, similar o Dell Ariccia and Marquez [22], screening is always ineffecual for a fracion α of he bad and average firms. Versions of Proposiions 1 3 will sill hold in his modified model, bu as goes o zero and goes o infiniy, he impac of foreign enry insead resembles ha of Dell Ariccia and Marquez [22] in ha no firm is worse off because of segmenaion. 18

22 Condiion (a) of Proposiion 4 jus resaes he assumpion ha foreign lenders enjoy a cos of funds advanage, which is necessary o make cream skimming possible, while condiion (b) ensures i is unprofiable o pool average and bad firms in a separaing equilibrium. Condiion (c) ensures he domesic cos of screening is high enough ha he closed economy exhibis a pooling equilibrium ( ) and ha domesic lenders find i unprofiable o screen average firms in a separaing equilibrium ( R r ). inally, condiion (d) ensures ha he foreign cos of screening is low enough o break he pooling equilibrium ( ) bu high enough ha foreign lenders canno profiably screen average firms in a separaing equilibrium ( R r ) and ne oupu declines ( ( R r ) ( R 1 (1 p) r)/(1 )). To characerize he se of parameers for B A B which enry breaks he pooling equilibrium, average firms are shu ou of he financial marke, bu ne oupu does no decline, one simply drops he second lower bound in condiion (d). 16 The condiions necessary for a decline in ne oupu highligh he imporance of informaion acquisiion coss in he local economy. If domesic lenders cos of producing informaion,, is no oo high, hen average firms will be screened and financed in a separaing equilibrium, hus reducing he likelihood of enry reducing aggregae ne oupu. This suggess ha indusries in which i is easier for lenders o assess a borrower s poenial (i.e., low ), are more likely o experience an increase in ne oupu afer addiional lender enry. This migh include maure indusries, indusries ha rely less heavily on inangible asses, and indusries wih less uncerain growh prospecs. Low screening coss migh also be driven by counry-level facors. In counries wih ransparen accouning rules or srong audiing enforcemen sandards, he cos of screening is likely less because lenders can rely more on he informaion conained in firm s financial saemens and avoid engaging in he cosly collecion of addiional informaion hrough oher channels. The imporance of informaion acquisiion coss for foreign lenders is more nuanced. Again, a sufficienly high screening cos,, is necessary o ensure i is no feasible for average firms o be 16 Ineresingly, he second par of Assumpion (A1), ha foreign lenders have a higher cos of screening, ĸ > ĸ, is no necessary for a decline in ne oupu. There exis parameer spaces for ĸ < ĸ where foreign enry and marke segmenaion cause a decline in ne oupu; examples of his are discussed laer. 19

23 screened by foreign lenders in a separaing equilibrium and for ne oupu o decline. However, as shown in condiion (d) of Proposiion 4, i mus also be he case ha foreign lenders cos of screening is no so high ha i offses heir cos of funds advanage and prohibis heir abiliy o break he pooling equilibrium by offering more compeiive conracs o cream firms. If closed economy exhibis a pooling equilibrium, hen foreign enry will no break he pooling equilibrium and ne oupu increases because of foreign lenders lower cos of funds. 20 and he or a given se of parameers A,, rr B,,, p, and R where condiions (a) and (b) of Proposiion 4 hold, one can use he condiions (c) and (d) of Proposiion 4 o map ou he possible equilibria for every possible combinaion of screening coss, ĸ and ĸ. An example of his is provided in igure 1, where possible equilibria are described for ĸ and ĸ such ha foreign lenders are disadvanaged in screening (i.e., ĸ > ĸ), bu can offer more compeiive conracs o cream firms. 17 As shown in igure 1, when, he equilibrium will be a separaing equilibrium in he closed economy, bu for, he closed economy exhibis a pooling equilibrium. Wheher foreign enry causes a swich o a separaing equilibrium and a decline in ne oupu for depends on ĸ. or, where ( R r ) ( R(1 (1 p) ) r)/(1 ), enry causes a swich from B A B he pooling equilibrium o a separaing equilibrium where only cream firms obain financing and ne oupu declines. 18 When R r, cream firms remain he only firms financed in he separaing equilibrium, bu ne oupu increases. or financed in he separaing equilibrium and ne oupu rises. R r, boh average and cream firms are igure 2 provides a numerical example of igure 1 under he following assumpions: successful projecs yield a 15% reurn (R = 1.15); bad firms only succeed wih 80% probabiliy (p = 0.8) and represen one-fifh of he firms wih he remaining firms being spli evenly beween average and cream (θ B = 1/5, θ A = 2/5); cream firms are able o implemen projecs wo and a half imes as 17 or illusraive purposes, igure 1 assumes ha R r and R r are no he binding lower bounds in condiions (c) and (d) of Proposiion 4. I can be shown, however, ha hese will never be he binding consrains when domesic lenders have a compeiive advanage in screening he smaller projec, such ha ĸ + r > ĸ + r. 18 While igure 1 only describes regions where foreign lenders have a higher cos of funds, ĸ > ĸ, ne oupu also declines for even when ĸ < ĸ.

24 large (λ = 2.5); and domesic lenders cos of funds is 9% (r = 1.09), whereas foreign lenders cos of funds is 7.5% (r = 1.075). 19 Under hese assumpions, he possible equilibria resemble ha of igure 1, and igure 2 graphs he possible equilibria in he range of screening coss ĸ and ĸ ha capure he swich in equilibrium. As shown in igure 2, a pooling equilibrium will occur in he closed economy when ĸ is greaer han abou or example, ĸ = would yield a pooling equilibrium in he closed economy wih an equilibrium unscreened ineres rae of 13.5% and breakeven lending raes on he wo (non-equilibrium) screened loans of size 1 and 2.5 equal o 23% and 14.5% respecively. A foreign lender wih a higher cos of screening, bu one ha is below abou 0.16, however, could break his equilibrium. or example, a foreign lender wih a cos of screening equal o would ener he economy and offered screened conracs of size 2.5 o cream firms wih an ineres rae of abou 13.75%. This enry would break he pooling equilibrium, causing average firms o be shu ou from financing, and leading o a reducion in ne oupu of abou 7%. Toal ne oupu can also decline even when average firms are no shu ou of he credi marke following foreign enry. or example, if he pooling conrac remains feasible in he separaing equilibrium, such ha ( R r ) ( r pr) 0, boh average and bad firms may A B coninue o be financed, bu on worse erms han in he closed economy. The equilibrium lending rae on he conrac ha pools bad and average firms will be r r (1 )/(1 (1 p) ), C C B whereas he equilibrium lending rae in he closed economy was r r /(1 (1 p) B ). The equilibrium lending rae for he pooling conrac can rise in he open economy because cream skimming by oher lenders increases he raio of firms acceping he pooling conrac ha will fail. This increase in borrowing coss can reduce he ne oupu of average and bad firms, and his decline in heir ne oupu can exceed he increase in ne oupu from cream firms aking on larger projecs. The parameer space where his occurs is described in Proposiion 5 and proven in he Appendix. 19 I choose a differen se of parameers here han in Secion 4 o demonsrae robusness of he model o choosing a higher cos of funds and making screening coss a lower share of lenders oal cos. 21

25 Proposiion 5. Under he following condiions, foreign enry causes a swich from a pooling equilibrium ha finances all firms o a separaing equilibrium where cream firms are screened and financed for large projecs, average and bad firms coninue o accep he pooling conrac of size 1, and ne oupu falls: (a) r > r (b), A( R r ) B( r pr) 0 (c) max, r ( A B) / ( A pb) r, and (d) max r (1 p) B / ( A pb),( 1)( R r ) ( r r ) / (1 A B), where and are defined by equaions (1) and (2), respecively. Condiion (a) is he same as in Proposiion 4, while condiion (b) is reversed o ensure i is feasible o pool average and bad firms. Condiion (c) ensures he domesic cos of screening is high enough ha he closed economy exhibis a pooling equilibrium ( ) and average firms find he pooling conrac in he open economy more aracive han a domesic screened conrac ( r ( )/( p ) r ). inally, condiion (d) ensures ha he foreign screening cos is low A B A B enough o break he pooling equilibrium ( ) bu high enough ha average firms prefer he open economy pooling conrac ( r (1 p) /( p )) over ha of a screened conrac and ne B A B oupu declines ( ( 1)( R r ) ( r r )/(1 )). While ne oupu can decline even when average firms are financed in he separaing equilibrium, he range of A B where his occurs is smaller han in he scenario where boh average and bad firms are shu ou enirely. The hreshold level of above which ne oupu declines in Proposiion 5 is greaer han he hreshold when average firms are shu ou enirely in Proposiion 4. The inuiion is sraighforward; he ne oupu of cream firms in he open economy, C [ ( R r ) ], is decreasing in he screening cos,, and remains he same in boh scenarios, bu he decline in ne oupu for bad and average firms is greaer when hey lose financing enirely The likelihood of a decline in ne oupu in his seing is even greaer when foreign lenders supply of capial is limied such hey are only able screen and finance a fracion α of he cream firms. In his seing, i can be shown ha here exiss an open economy equilibrium where αθ C cream firms are screened and financed by he foreign lenders, all oher firms coninue o be pooled by domesic lenders, and ne oupu declines when ĸ > (λ 1)(R r ) + (r r ). The range of ĸ where ne oupu declines is greaer in his consrained seing since average and bad firms no longer benefi from foreign lenders lower cos of funds and fewer cream firms increase heir ne oupu. 22

26 Similar o before, i is easy o see ha he parameer space described in Proposiion 5 is nonempy. or example, foreign enry will cause a shif from he pooling equilibrium o a separaing equilibrium where average and bad firms coninue o be pooled bu ne oupu declines by abou 4 percen under he following se of assumpions: successful projecs yield a 16% reurn (R = 1.16); bad firms only succeed wih 80% probabiliy (p = 0.8) and represen one-fifh of he firms wih he remaining firms being spli evenly beween average and cream (θ B = 1/5, θ A = 2/5); cream firms are able o implemen projecs hree imes as large (λ = 3); and domesic lenders cos of funds is 9% (r = 1.09), whereas foreign lenders cos of funds is 7.5% (r = 1.075) Robusness and Exensions This secion discusses he robusness of he model s main implicaions. irs, I exend he model by allowing lenders o inves in greaer screening experise, and second, I exend he model by allowing foreign lenders o improve heir screening capaciy via acquisiions. While providing furher insighs on how enry affecs lenders incenive o inves in informaion producion, neiher exension affecs he main findings. Third, I will show ha he findings are robus o more general assumpions regarding he disribuion of firms, projec sizes, and expeced reurns. 6.1 Invesmens in Screening Experise To analyze how foreign enry migh affec lenders incenives o inves in screening capaciy, I now exend he model o allow for such invesmens. Specifically, I assume ha a domesic lender s screening cos is given by e 0, where is he baseline screening cos, and e is he lender s experise in screening. Likewise a foreign lender s cos of screening is given by e. While baseline screening coss, and, are sill fixed, lenders can lower heir per-firm screening cos by invesing in experise, e. A lender s cos of obaining experise e is increasing and convex in e and given by 21 I is also possible o show ha here exiss a non-empy parameer space where foreign enry causes a shif o a separaing equilibrium where boh average and cream firms accep screened conracs bu ne oupu declines. The inuiion is similar; ne oupu of average firms can decline when he small, screened conrac is more expensive han he closed economy conrac ha pooled all firms on he small projec. 23

27 c(e), where c(0)=c (0)=0, c (e)>0, c (e)>0 for e > 0, and equals he number of firms he lender expecs o screen (i.e., a bank ha plans o screen and acquire informaion abou more firms will need o rain more loan officers, ec.). Wih his change, a sraegy configuraion in his economy now consiss of he se of conracs j and choice of e j and j for each lender j profis of lender j lending o firm i wih conrac is now given by j ( i ) p( i I) R r( j) I ( ) ( j) e( j) S L, and he expeced where r( j ) and ( j ) represen he cos of funds and baseline screening cos for lender j ; I represens he amoun of financing associaed wih conrac ; and S 0 for unscreened conracs and equals one oherwise. Lender j s oal expeced profis is given by i ( j, ) ( i ) di c( e ). j j j j Using he same noaion as in Secion 2, he economy s equilibrium is formally defined as Definiion of Equilibrium: A sraegy configuraion, f () i for each firm i E, e j, j, and implied by j for each lender j L, consiues equilibrium if and only if (1) given, each firm i E chooses f() i o maximize ( f ); (2) each lender j L chooses e j, j,and o maximize where i ( j, ) is given by condiion 1; and j i j, j( i j ) di jc() e, (3) because of free enry, each lender makes zero profis, i ( j, ) ( i ) di c( e ) 0. j j j j To preserve he comparaive advanage of domesic lenders in screening and o simplify he analysis, I will coninue o assume ha foreign lenders baseline cos of screening,, is higher han ha of domesic lenders bu ha boh lenders have access o he same cos srucure for invesmens in screening experise. All oher assumpions of he model remain he same. In his exended model, each lender now makes wo choices. irs, a lender mus decide wheher hey will inves in screening some firms by offering screened conracs. Second, if hey offer screened conracs, hen hey mus decide wha level of screening experise o inves in. 24

28 In equilibrium, i will urn ou ha if lenders find i worhwhile o inves in he screening firms, hen compeiion among lenders will induce hem o inves in experise, e, up o he poin, e, which is he level of experise where he reducion in he screening cos from an addiional invesmen in experise is equal o he marginal cos of ha invesmen. A lender ha offers screened conracs bu only invess in experise e < e can always be undercu by a lender ha invess e. While providing insighs on how foreign enry migh affec lenders incenive o inves in screening capaciy, his exension does no qualiaively change he possible equilibria. Similar o before, a pooling equilibrium will occur in he closed economy if domesic lenders baseline cos,, is sufficienly high, and foreign enry can cause a separaing equilibrium ha reduces ne oupu when falls wihin a cerain range. or example, we have he following proposiion, which is proven in he Appendix: Proposiion 6. Wih foreign enry and he abiliy o inves in screening experise e a cos c(e), a swich from a pooling equilibrium ha finances all firms o a separaing equilibrium where only cream firms are financed and ne oupu falls will occur when: (a) r > r (b) 0 A( R r ) B( r pr), (c) max e c( e ),R r e c( e ), and (d) R 1 (1 p) B r ( e c( e )) max R r, ( R r ). 1 A B where e is given by c '( e ) 1 and and are defined in equaions (1) and (2). Comparing Proposiions (4) and (6), we see ha he abiliy o inves in screening experise increases he necessary o cause a pooling equilibrium in he closed economy and shifs upward he range of necessary o cause a separaing equilibrium where ne oupu falls. or e c( e ), a pooling equilibrium in he closed economy will no longer be susainable since lenders can inves in screening experise and improve he araciveness of heir screened conracs, bu for e c( e ), he pooling equilibrium persiss. Similar dynamics play ou when foreign 25

29 enry occurs, leading o an upward shif in he range of ha cause a shif o he separaing equilibrium and decline in ne oupu. While he poenial decline in ne oupu remains, he exended model highlighs how foreign enry can induce greaer invesmen in screening capaciy by domesic lenders. In he separaing equilibrium, domesic lenders may inves in screening experise so as o reain a compeiive advanage among average firms, and mainaining his compeiive advanage will be possible when foreign lenders disadvanage in screening is sufficienly large, such ha r r. 22 If he baseline cos of screening,, is less han R r e c( e ), i will be feasible for domesic lenders o inves in screening capaciy e and offer screened conracs o average firms following foreign enry. oreign lenders lower cos of funds, however, migh help hem offse domesic lenders invesmens in screening experise. or example, if invesing in screening capaciy requires lenders o borrow funds up fron, hen foreign lenders lower cos of funds can provide hem an advanage in he race o expand screening experise. Specifically, if he cos of invesmen e in screening experise is insead given by r c() e for domesic lenders and r c() e for foreign lenders, hen foreign lenders will choose o inves more in screening experise since r < r. Specifically, hey will inves e where e is given by c'( e ) 1/ r, while domesic lenders, if hey inves a all, will inves e e where e is given by c'( e ) 1/ r. If foreign lenders cos of funds advanage is sufficienly large, heir invesmens in experise may be enough o offse he screening advanage of domesic lenders. 6.2 Invesmen in Screening Capaciy via Mergers Anoher roue by which foreign lenders migh aemp o improve heir screening capaciy is by acquiring a domesic lender. If domesic lenders lower screening cos is driven by heir loan 22 In his regard, one can hink of hese invesmens in screening capaciy as an arms race similar o ha of Glode, Green, and Lowery [31], hough he arms race here is quie differen in ha one lender s invesmen in informaion is no made superfluous by oher lenders invesmens since each firm is screened by a mos one lender. Wheher hese invesmens reflec an overinvesmen in screening capaciy depends on he parameer space. In he parameer space described in Proposiion 6, one could view foreign lenders equilibrium invesmen in screening as an overinvesmen since i leads o a consrained inefficien equilibrium, whereas he pooling equilibrium (wihou screening) is consrained efficien. Bu if eiher condiion (c) or (d) of Proposiion 6 were violaed, hen he equilibrium invesmens in screening capaciy would increase ne oupu. 26

30 officers greaer knowledge abou local firms or he abiliy of hese loan officers o beer parse local firms financial saemens, hen he acquisiion of an incumben lender may allow he enering lender o lower is screening cos and reduce is informaional disadvanage. To formalize his possibiliy and analyze he poenial role of acquisiions in how lenders enhance heir screening capaciy, I now exend he base model o provide foreign lenders he choice on wheher o acquire a domesic lender. Specifically, I assume ha if a foreign lender acquires a domesic lender, i is able o lower is cos of screening o, bu his comes a he cos of aking on some of he inheren organizaional or echnological disadvanages of he incumben, such ha he foreign lenders cos of funds rises o r for some Under his seup, i can be shown ha he merged lender will be able o offer he mos compeiive screened conrac of size 1 when min{, r r } since ensures ha he merged lender can offer a more compeiive screened conrac han foreign lenders and r r ensures he merged lender can offer a more compeiive screened conrac han he domesic lender. urhermore, if ( )/, he merged lender will also be able o offer he mos compeiive screened conrac of size. The merged lender will never offer a pooling conrac since he foreign lender can always offer he mos compeiive pooling conrac because of is lower cos of funds. Wih his exension, i is easy o see ha he abiliy o merge can reduce he parameer space under which a decline in ne oupu can occur. or example, average firms will only go unfinanced in a separaing equilibrium when he pooling conrac is infeasible and 0 max R r, R r, R ( r ), such ha no lender domesic, foreign, or merged finds i feasible o screen average firms. When min{, r r }, he laer condiion peraining o he merged lender will be he mos binding consrain, which reduces he range of where average firms go unfinanced. The use of acquisiions o expand screening capaciy, however, does no eliminae he 23 In pracice, such a cos srucure migh be also accomplished hrough syndicaed lending or if incumben lenders gain access o he lower cos of capial. This migh occur if foreign enry coincides wih oher reforms ha foser domesic lenders access o inernaional capial markes or if enry resuls in a ransfer of echnology o incumben lenders, as suggesed by Levine [39]. 27

31 possibiliy of a decline in oupu. To see his, consider he exreme example where 0. In his seing, acquisiion of a domesic lender allows he foreign lender o compleely capialize on is lower cos of funds and he incumben lender s lower cos of screening. Because a merged lender can always offer he mos compeiive conracs in his special case, foreign lenders will always acquire domesic lenders upon enry, and only merged lenders will offer conracs in equilibrium. Wheher ne oupu declines will now only depend on. In paricular, we have he following proposiion, which is proven in he Appendix: Proposiion 7. Wih foreign enry ha allows for acquisiions of domesic lenders and 0, a swich from a pooling equilibrium ha finances all firms o a separaing equilibrium where only cream firms are financed and ne oupu falls will occur when: (a) r > r (b) R r r pr, 0 A( ) B( ) R 1 (1 p) B r (c) max, R r, ( R r ) 1 A B where and are defined in equaions (1) and (2). Comparing Proposiions 4 and 7, we see ha he range of where ne oupu declines is reduced. I can also be shown ha he range of implied by condiion (c) of Proposiion 7 is nonempy. or example, in he numerical example from Secion 4, here would sill be a drop in ne oupu for (0.495, ). Given he parameers from he firs numerical example in Secion 5, however, here is no longer a screening cos,, ha can saisfy condiion (c) of Proposiion The poenial merger of lender aribues also yields dynamic implicaions. While he above model is saic (such ha lenders merge a ime of enry), one could easily imagine ha such a process migh only occur slowly over ime. or example, he enering lenders cos of screening,, migh 24 Ineresingly, if lenders screening coss, ĸ and ĸ, occur on a per firm basis raher han a per projec basis, hen firms will also have an incenive o merge. So long as lenders screening coss do no scale up one-o-one wih he amoun of capial, I, ha can be successfully invesed, firms will have an incenive o merge so as o obain more lucraive screened financial conracs. To preven such mergers from occurring unil only one large firm remains, one would need o exend o model o include an organizaional cos of running such conglomeraes. In such an exension, firms would only merge up o he poin where he marginal benefi of furher mergers, via lower borrowing coss, equals he marginal cos, via greaer organizaional coss. By changing he se of equilibrium conracs, foreign enry may increase he marginal benefis of such mergers and hus increase he creaion of conglomeraes afer enry. 28

32 sar high bu decline wih ime (eiher hrough mergers or hrough he accumulaion of local knowledge). In such a case, one migh sar off in parameer space where foreign lenders are limied o cream skimming and ne oupu falls (similar o Proposiion 4) and evenually end up in a parameer space where foreign lenders arge a larger subse of firms and ne oupu increases. 6.3 Disribuion of irms and Projec Sizes The basic mechanisms of he model are also robus o allowing for a richer disribuion of firms wih varying projec sizes,, and reurns, R. In such a model, he screening cos hresholds, and, would simply become firm specific. or insance, a cream firm, i, wih a projec of size () i and reurn Ri (), such ha () i, would be screened and financed fully in he economy wihou foreign lenders. And, all cream firms wih smaller projecs or reurns, such ha () i, will be pooled wih average and bad firms. Again, foreign enry has he poenial o unravel he pooling equilibrium as foreign lenders lower cos of funds migh allow hem o arge a larger se of cream firms and reduce he number of firms pooled by domesic lenders. Allowing for differen reurns across projec sizes also does no affec which firms foreign lenders will arge in he open economy. oreign lenders lower cos of funds provides hem a compeiive advanage per dollar invesed, and his compeiive advanage does no depend on expeced reurns. Specifically, he highes payoff a foreign lender can offer on screened conracs of size I is (R r )I, while he highes payoff a domesic lender can offer is (R r )I. Therefore, foreign lenders will have a compeiive advanage in offering screened conracs o any firms wih a projec of size I r r ( )/( ) Empirical Evidence This secion discusses he empirical evidence underlying he model s key assumpions abou he comparaive advanages of foreign and domesic lenders. The secion also discusses he model s 25 Mainaining he higher average reurn of cream firms on he larger projec, bu allowing for he possibiliy ha average firms produce an even higher reurn wih some probabiliy also does no affec he equilibrium. So long as he expeced reurn of average firms on he large projec is lower han he cos of funds, no lender will offer large, screened conracs o average firms. 29

33 esable predicions and how hey relae o he exising empirical evidence regarding foreign lender enry ino LDCs and subsequen changes in local credi and ne oupu. 7.1 Empirical Suppor for Assumpions The key assumpion of he model is ha foreign lender enry coincides wih he arrival of a new lending cos srucure ha has he poenial o break a domesic pooling equilibrium. In he model, his is accomplished hrough he enry of a foreign lender ha is assumed o have a higher cos of screening bu a lower marginal cos of funds han he incumben lenders. The assumpion ha screening coss are higher for foreign lenders is widely suppored by exising empirical evidence. A greaer disance beween lender and borrower where disance is broadly defined o include hierarchical, geographical, and culural disance can increase a lenders cos of acquiring informaion and is a key feaure of foreign lending (Berger, Klapper, and Udell [11]). or example, Sein [52] demonsraes ha he greaer hierarchical srucure of foreign lenders can make i more cosly for hem o use he sof informaion necessary o screen firms, and Peersen and Rajan [47] noe ha he cos of acquiring informaion abou borrowers likely increases wih he geographical disance beween he lender and borrower. Consisen wih informaional coss associaed wih disance being paricularly salien for foreign lenders, Mian [43] finds evidence ha disance barriers for foreign banks operaing in Pakisan are sufficienly large o exclude hem from cerain secors of he economy enirely, and Buch [15] finds a negaive correlaion beween disance and he inernaional banking aciviies of banks locaed in rance, Germany, Ialy, Unied Kingdom, and he Unied Saes. Recen work on lending relaionships and loan prices in Belgium, Ialy, and he Unied Saes also sugges ha greaer lending disances are associaed wih increased ransporaion and informaional coss for lenders (Agarwal and Hauswald [2], Degryse and Ongena[21], Misrulli and Casolaro [45]). The second assumpion ha foreign lenders enjoy a lower cos of funds is also suppored by exising evidence. Wihin-counry comparisons sugges ha foreign banks have, on average, lower ineres expenses, overhead coss, and oal employmen per uni of asses relaive o heir domesic 30

34 counerpars, paricularly in LDCs (Mian [42], Micco, Panizza, and Yañez [44]). There are a variey of reasons why foreign banks may enjoy a cos of funds advanage. oreign lenders are ofen less beholden o local laws and labor unions han domesic lenders, making i less cosly for hem o expand operaions and raise addiional funds in he domesic economy. 26 oreign banks migh also enjoy a comparaive advanage in raising capial. oreign banks may be able o raise capial locally a a lower cos because invesors in LDCs perceive foreign banks as safer because hey are backed by a large, foreign affiliae (Mian [42]) and less likely o make poliical loans (Micco, Panizza, and Yañez [44]). Well-developed securiies markes and beer insiuions in he home counries of foreign lenders may also provide hem access o cheaper sources of capial. 7.2 Tesable Implicaions and Evidence The model generaes a number of esable predicions regarding he impac of foreign enry. (P1) oreign enry can induce a segmened credi marke where foreign lenders only arge he larges, mos profiable, and leas-informaionally opaque firms (see Proposiion 4). (P2) oreign enry will coincide wih a decline in credi from domesic lenders and a poenial decline in overall credi and ne oupu (see Proposiions 4 and 5). (P3) A decline in ne oupu is less likely o occur when screening coss in he closed economy are low (see Proposiions 4 and 5). In pracice, many facors, like srong counry-level insiuions (e.g., counries wih ransparen accouning rules and srong audiing enforcemen sandards), migh reduce he screening cos of local lenders. (P4) Condiional on a swich from a pooling o separaing equilibrium where foreign lenders cream skim, a broad curailmen of lending by domesic lenders (and a decline in ne oupu) is less likely o occur when any of he following condiions hold: a. he domesic cos of screening is lower (see Proposiion 4c); for example, in counries wih srong insiuions or in indusries wih easier o value asses, 26 or example, by sidesepping local unions in India, foreign banks are able o hire fewer workers and pay a lower average wage bill per deposi colleced relaive o domesic banks (Hanson [35]). This provides hem a compeiive advanage in esablishing addiional branches from which hey can raise new deposis. 31

35 b. raio of bad o average firms is lower (see Proposiion 4b); for example, in older, more esablished indusries in which he number of likely failures is lower, and (P5) To improve heir abiliy o produce informaion and obain a compeiive advanage, foreign lenders may prefer o merge wih domesic lenders, and a decline in ne oupu is less likely o occur when such mergers occur (see Secion 6.2). Many of hese predicions map closely o he broad, exising empirical evidence regarding he effecs of financial liberalizaion on growh and oupu, whereas oher predicions have ye o be formally sudied. The remainder of his paper discusses his evidence and proposes areas ha are promising direcions for fuure empirical research on financial liberalizaion Evidence on Segmenaion and Acquisiions There is broad empirical suppor for he model s predicion ha foreign lenders cream skim he leas informaionally opaque, larges, and mos profiable firms (Predicion P1). Mian [43] finds ha foreign banks in Pakisan end o avoid loans ha are ypically associaed wih acquiring sof informaion, such as loans o small firms and firs-ime borrowers, whereas Gormley [32] finds ha foreign banks in India only len o a small subse of he larges, mos profiable firms. In paricular, only he op 10% of firms, in erms of profiabiliy, appear o experience an increase in bank loans following foreign lender enry in India. Oher papers find ha foreign banks are less likely o lend o small, informaionally opaque firms in Lain America (Berger, Klapper and Udell [11], Clarke, Cull, Peria, and Sánchez [19]), ha small firms in Easern Europe appear o benefi less from foreign enry (Giannei and Ongena [30]), and ha foreign lenders shy away from lower-qualiy firms wih pas delinquencies (Berger, Klapper, and Udell [11]). There is also evidence o suppor he predicion ha, by lowering foreign lenders cos of screening, acquisiions of domesic lenders will be a preferred mode of enry and allow foreign banks o arge a larger share of he lending marke (Predicion P5). Anecdoally, counries ha allow foreign banks o acquire domesic banks end o experience a subsequen wave of acquisiions. or example, Mexico firs allowed foreign banks o purchase conrolling sakes in is larges banks in 32

36 1997, and he foreign ownership of banking asses quickly increased from 16% in 1997 o 82% in 2004 (Haber and Musacchio [34]). A more limied ype of enry ends o occur when counries prohibi such acquisiions. or example, when India allowed he enry of new foreign banks in 1994, enry was largely limied o green field invesmens, and as of 2009, foreign banks only owned abou 5% of he banking asses in India (Gormley [32]). 27 And consisen wih acquisiions lowering he cos of screening and allowing foreign lenders o arge more firms, Degryse, Havrylchyk, Jurzyk, and Kozak [20] find ha foreign banks ha ener via acquisiion finance more informaionally-opaque firms relaive o foreign banks ha ener via greenfield invesmens Evidence on Heerogeneiy of oreign Enry s Impac The model s predicion regarding he poenial negaive impac of foreign enry on overall oupu and growh (Predicion P2) fis well in he conex of he exising empirical lieraure. To dae, many sudies find no clear impac of foreign lender enry on overall credi, oupu, and growh (e.g., Area, Eichengreen, and Wyplosz [5], Edison, Levine, Ricci, and Sløk [25], Rodrik [49]) while ohers find he impac is negaive (e.g., Beck and Peria [9], Deragiache, Gupa, and Tressel [24], Gormley [32]). Moreover, in counry-specific sudies in which foreign enry was found o reduce overall credi, he evidence seems o confirm he model s predicion ha his will ofen coincide wih cream skimming by foreign lenders and a decline in lending by domesic banks. Analyzing foreign lender enry ino India following is liberalizaion in 1994, Gormley [32] found ha he new foreign banks only len o a small subse of he mos profiable firms and ha domesic lenders responded o heir enry by sharply curailing heir lending o all firms, no jus he mos profiable firms argeed by he new foreign lenders. On ne, Gormley found here was a decline in overall credi in Indian disrics in which foreign bank enry occurred relaive o disrics in which no enry occurred. Counry-specific sudies also provide suggesive suppor o he model s predicion ha mergers beween foreign and domesic lenders will increase he probabiliy of posiive impac on oupu and credi (Predicion P5). In a sudy of foreign lenders enry in Easern Europe, where 27 These ypes of resricions are ofen pu in place because domesic poliicians worry abou preserving financial sabiliy and abou allowing a majoriy of he counry s banking asses o be suddenly acquired by foreigners. 33

37 foreign acquisiions were allowed and widespread, Giannei and Ongena [30] found a posiive effec on growh, whereas in a sudy of foreign lender enry in India, where such acquisiions were prohibied, Gormley [32] found evidence of a decline in credi access for many firms. Using variaion across boh indusries and counries, Bruno and Hauswald [14] provide even more evidence regarding he poenial imporance of acquisiions. Comparing oucomes across indusries ha are more- and less-dependen on exernal financing, hey find ha increases in foreign bank ownership driven by acquisiions are posiively relaed o a relaively larger increase in economic growh in indusries more dependen on exernal financing; increases in ownership driven by greenfield invesmens are no associaed wih changes in growh Qualiy of Local Insiuions and Impac of Liberalizaion The model s predicion regarding he imporance of local insiuions (Predicions P3 and P4a) also has he poenial o explain a number of exising empirical paerns regarding he impac of financial liberalizaion. Capial accoun liberalizaion ends o be posiively associaed wih subsequen economic growh in high-income counries, where qualiy of accouning and audiing sandards is likely greaer, while negaively relaed o growh in low-income counries, where accouning sandards are likely weaker (e.g., see Edwards [26], Quinn [48]). Likewise, foreign bank enry in low-income counries is ofen associaed wih a decline in privae credi and lending by domesic insiuions (Deragiache, Tressel, and Gupa [24], Gormley [32]). A number of sudies also find ha opening capial markes and domesic financial liberalizaion is only associaed wih a posiive impac on growh in counries wih greaer law and order radiions or beer legal proecions for crediors, boh of which are likely posiively correlaed wih sronger enforcemen of accouning sandards (e.g., see Area, Eichengreen, and Wyplosz [5], Galindo, Micco, and Ordoñez [29]). There is also direc evidence regarding he poenial imporance of accouning sandards. In an analysis of equiy marke liberalizaions, Bekaer, Harvey, and Lundblad [10] find a large increase in economic growh for counries wih an above-average accouning qualiy, bu no increase in growh for counries wih below average accouning sandards. 34

38 7.2.4 Scope for uure Empirical Research Whereas a number of he model s predicions are consisen wih he exising empirical lieraure, here are oher predicions ha have ye o be exensively sudied. or example, if foreign lenders cos of obaining informaion declines wih ime, hen he model suggess some dynamic implicaions where foreign lenders arge more informaionally-opaque firms over ime, which can cause furher changes in ne oupu. There is evidence ha foreign bank profiabiliy is higher he longer he bank has operaed in a counry (Claessens and van Horen [17]), ha cream-skimming-ype behavior is less likely o occur when a foreign lender expands is exising operaions wihin a counry (Gormley [32]), and ha foreign banks lend o more informaionallyopaque borrowers as ime passes (Degryse, Havrylchyk, Jurzyk, and Kozak [20]), bu, o he auhor s knowledge, here is no direc analysis of wheher he impac of foreign lender enry on overall credi, oupu, and growh changes wih ime afer iniial enry occurs. The model also provides a number of predicions regarding how he effec of liberalizaion may vary across indusries (Predicions P4a and P4b). or example, indusries in which i is more difficul o screen he qualiy of projecs or in which he likelihood of failure is greaer, such as newer indusries or indusries wih fewer angible asses, migh be more likely o be adversely affeced by foreign lender enry. Wih he excepion of Gormley [32], who found ha foreign enry in India was more likely o be associaed wih a subsequen decline in sales growh for indusries wih fewer angible asses, here is very lile exising evidence on wheher financial liberalizaion has heerogeneous effecs across indusries. 8. Concluding Remarks Emerging economies are ofen criicized for having financial secors ha seem o over finance low-reurn projecs and under finance high-reurn projecs. or his reason, and many ohers, i is ypically argued ha opening capial markes would improve credi access and overall oupu in hese economies. However, he heory developed in his paper suggess ha his ype of domesic credi allocaion may occur when informaion asymmeries are large and domesic lenders choose o pool risks raher han inves in cosly screening echnologies. 35

39 If rue, foreign enry may ake he form of cream skimming and adversely affec overall credi access. oreign lenders may use heir lower cos of funds o offer more compeiive financial conracs o firms capable of profiably invesing large amouns of capial. This ype of enry and he resuling separaing equilibrium may boh redirec credi oward he larges, mos profiable firms in he economy and reduce he credi access of informaionally opaque firms by changing he se of conracs available o hem. As a resul, he overall ne oupu may decrease afer foreign enry when informaion asymmeries are sufficienly cosly o overcome. The poenial decline in oupu provides new insighs o he inconclusive relaion beween foreign lender enry and aggregae oupu. More generally, he model illusraes a possible dark side o liberalizaion ha has been suggesed by empirical evidence bu is no well undersood heoreically. The model is also able o generae predicions of when a new lender s enry will adversely affec credi access and ne oupu. The impac of he lender s enry will depend on he disribuion of firms, he comparaive advanages of compeing lenders, he severiy of informaion asymmeries, wheher lenders are allowed o merge, and he qualiy of local insiuions. This yields a number of esable hypoheses on how he impac of lender enry may vary by indusry and counry. Many of hese predicions find subsanial suppor in he daa, whereas ohers provide ineresing avenues for fuure empirical research. The model also provides an explanaion for why exising empirical sudies on he opening of capial markes, which assume a uniform impac across counries and indusries, fail o find consisen evidence. The model also exends our undersanding of how compeiion in markes wih asymmeric informaion can lead o unfavorable oucomes for many agens. In conras many exising models, where screening occurs hrough agens self-revealing choices from he menu of offered conracs, his paper analyzes he effecs of compeiion in a seing where lenders are able o acquire and use privae informaion abou agens ypes o limi heir choice from he menu of offered conracs. The model also broadens he se of scenarios in which increased compeiion can faciliae he exi of seemingly more efficien lenders. The model suggess his exi can occur even in already compeiive markes or even when an incumben lender does no face direc compeiion for borrowers for which i enjoys a compeiive advanage in financing. 36

40 Appendix A. Proof abou shape of equilibrium conracs or all financial conracs where projecs are implemened, i is sufficien o consider only conracs wih RI ( ) 0 and (0) 0 as long as here are many lenders offering idenical conracs in equilibrium. This is proven in Lemma 1. Lemma 1: or all financial conracs of size I {1, } and ype k{0, A, B, C } i is sufficien o consider only equilibrium conracs wih offering conracs in equilibrium. Ik, ( RI) 0 and Ik, (0) 0 when here are n 2 lenders or each financial conrac, lenders mus provide a non-negaive paymen in each sae of he world when projecs are implemened. This implies some paymen RI ( ) 0 for successful projecs and (0) 0 for failures. or financial conracs where k 0, his yields an expeced profi of Ik, Ik, Ik, ( i k) p( k I) ( RI) [1 p( k I)] (0) for he firm and an expeced Ik profi of, Ik ( k ) [ p( k I) R r( j)] I (, k) ( j ) for he lender. Since all j j firms acceping his conrac will be of ype k, he expeced profis can always be replicaed for each agen involved by using a conrac where Ik, (0) 0and Ik, RI p k I ( ) / ( ). or financial conracs where k 0 and all borrowers acceping i in equilibrium have he same probabiliy of success, pi ( ) I, a similar reasoning holds. A paymen of RI ( ) / pi ( I) in success and zero oherwise can always replicae he expeced paymen of conracs ha pay a non-zero amoun in failure. or financial conracs where k 0 and all borrowers acceping he conrac in equilibrium do no have he same probabiliy of success, p ( ) i I, he expeced paymen for all agens canno be replicaed using a conrac wih Ik, (0) 0. However, Ik, i can be shown ha a conrac wih (0) 0 canno exis in equilibrium when 37

41 k 0 and no all borrowers acceping he conrac have he same probabiliy of success. Consider he case where a lender offers a conrac wih I,0 ( RI) G0and I,0 (0) H 0. If a coninuum 1 of enrepreneurs accep he conrac where a fracion only succeed wih probabiliy p, he expeced reurn for his lender is given by [1 (1 p)]( RI G) (1 p) H riand his mus equal zero in equilibrium. If anoher lender offered a conrac where I,0 ( RI) G and I,0 (0) 0 for some 1 p/ ph,0, however, i would make profis of (1 )( RI G ri ) because only firms wih probabiliy of success 1 will ake his new conrac. And, for (1 )( RI G ri ) 0 his conrac will be profiable. Bu, since [1 (1 p)]( RI G) (1 p) H ri 0 in any equilibrium, i mus be rue ha RI G ri when H 0. Therefore, here exiss some sufficienly small such ha (1 )( RI G ri ) 0. Therefore, conracs wih k 0 and H 0 can never be an equilibrium conrac. QED Appendix B. Proof of Proposiion 1 Given he seup, here are eigh differen ypes of financial conracs ha domesic lenders could offer: 1, k, k, k{0, A, B, C }. The proof ha he equilibrium of Proposiion 1 exiss and is he unique allocaion will be done in five pars. In pars 1-3, I will show ha 5 of he 8 financial conracs canno be equilibrium conracs. In par 4, I will derive he condiions under which he hree remaining financial conracs can co-exis in equilibrium. This will be sufficien o prove he allocaion of Proposiion 1 exiss and is unique when. inally, in par 5, I will prove ha none of he non-equilibrium conracs can be used o break he equilibrium in Proposiion 1. Par 1 When here are n 2 lenders offering he same conracs in equilibrium, any financial conrac Ik, yielding negaive expeced profis for he lender a 1 canno be an equilibrium conrac as any individual lender could increase profis by dropping he conrac. This 38

42 allows me o exclude financial conracs ha are ex-ane unprofiable for he lender if any firm were o accep he conrac. Those conracs are:, A 1, B, B,, and. Because conracs ake he form of RI ( ) 0 and (0) 0, as shown in Lemma 1, and pr r, he, and, A 1, B, B conracs always yield a negaive reurn for he lender and canno be equilibrium conracs. Par 2 Suppose ha,0 was an equilibrium conrac. By assumpion (A3) and pr r, his conrac can only be profiable if cream firms accep i, and will never be profiable if boh cream and bad firms accep i. If also choose,0 since Par 1 proves ha is no also an equilibrium conrac, however, hen all bad firms will 1, B, B and canno be equilibrium conracs. Therefore,,0 can only exis in equilibrium if also exiss and bad firms choose i. Bu if cream firms accep,0, hen i mus be ha,0 ( R) ( R ), which implies ha bad firms mus also prefer his conrac since (0) 0. Therefore,,0 can never be an equilibrium conrac. Par 3 In order for he 1,C conrac o be an equilibrium conrac, i mus be ha lenders receive non-negaive profis from offering i, such ha 1, C ( R) Rr, and ha cream firms do no prefer any oher conrac. Bu if his conrac is feasible, hen anoher lender could always feasibly offer he conrac, C ( RI) ( Rr ), and cream firms would prefer he his larger conrac since is payou exceeds he maximum possible payou of screened conrac for he smaller projec, 1,C. Therefore, 1,C canno be an equilibrium conrac. Par 4 rom Pars 1-3, we know here are only hree possible ypes of equilibrium conracs: 1,,, A C and. Therefore, lenders eiher offer an unscreened conrac for small projecs, a screened conrac for average firms, or a large screened conrac for cream borrowers. Moreover, by Lemma 1, i is sufficien o consider only conracs wih RI ( ) 0and (0) 0. In order for he 1,A conrac o be an equilibrium conrac, such ha lenders have nonnegaive profis from offering i, such ha 1, A ( R) Rr. Likewise, i mus be ha, C ( R) ( Rr ). Therefore, hese are he maximum expeced profis ha hese conracs 39

43 can provide o average and cream firms respecively. Average or cream firms will prefer he pooling conrac,, if is payou, cream prefer he pooling conrac, If ( R ), exceeds he maximum payou of 1, A and,c. Moreover, if, hen average firms mus also prefer he pooling conrac. is an equilibrium conrac, hen i mus be he case ha bad borrowers choose i since here is no oher conrac available o bad firms. In order for he conrac o be feasible for lenders when all firms selec i, i mus be ha R R r p. When possible payoff,c equilibrium conrac. Likewise, ( ) / 1 (1 ) B does no exceed he maximum possible payoff of, and,c 1, A is no an equilibrium conrac. This means ha, he maximum will no be an is he unique possible equilibrium conrac when. This conrac, however, is only feasible when r/1 (1 p) B R. Oherwise, lenders can never offer a non-negaive payoff o firms, ( R ) 0, and also make non-negaive profis. And, compeiion and lenders zero profi condiion ensures ha ( R ) R r / 1 (1 p ). B Par 5 To prove his is in fac an equilibrium financial conrac, i mus now be shown ha none of he oher non-equilibrium conracs can offer a poenial profiable deviaion for agens. Consider he case where, and all firms are pooled on he small projec. I can never be a profiable deviaion for lenders o offer IB, conracs since bad firms would sill implemen heir projec a a loss and he lender would now ake a larger loss because i screens he bad firms. Similarly, i is never profiable o offer,a since he conrac will always lose money. And, 1,C canno be profiable deviaions since a lender since ensures ha neiher can be greaer han 1,A or 1,A or 1,C (i.e. be preferred by average or cream firms) and be a profiable conrac for he lender. The,0 conrac will also by unprofiable by assumpion (A3) and he fac ha bad will,c always prefer he conrac if cream borrowers do. This leaves only. However, implies ha lenders can never profiably induce cream firms o ake a larger conrac wih screening. Therefore, is an equilibrium conrac for and r /1 (1 p) B R. QED 40

44 Appendix C. Proof of Proposiion 2 To differeniae conracs offered by foreign lenders, I will express heir conracs as. Ik, Using he same logic as in pars 1-3 of he proof of Proposiion 1, here are only hree poenial foreign lender conracs ha can be equilibrium conracs, 1, A, and, C, and i is sufficien o consider conracs of he form ( RI) 0and (0) 0. In an economy wih boh domesic and foreign lenders, he domesic lender conrac,,c domesic lenders because of assumpion (A2), and, can no longer be an equilibrium conrac for canno be an equilibrium conrac since r r. Therefore, here are only four possible equilibrium conracs:, 1, A, 1, A, and., C Similar o pars 4-5 of Proposiion 1, i can be shown ha, C only exiss and is preferred by cream firms over he pooling conrac heir zero profi condiion will ensure ha for. Compeiion among foreign lenders and, C R Rr, which exceeds he maximum ( ) ( ) possible payoff o cream firms wih he pooling conrac,, when. QED Appendix D. Proof of Proposiion 3 In he pooling equilibrium wih domesic lenders, ne oupu is( p )( R r ), where A B C r r /(1 (1 p) B ), while in he separaing equilibrium where only cream firms accep projecs from foreign lenders, he ne oupu is C [ ( R r ) ]. Thus, a decrease in ne oupu will occur when ( R r) p ( R r ) R r A B is rue. QED C Appendix E. Proof of Proposiion 4 As shown in Proposiion 1, he closed economy exhibis a pooling equilibrium when, and as shown in Proposiion 2, he open economy exhibis a separaing equilibrium when. Moreover, as shown in he proof of Proposiion 2, here are only hree possible equilibrium conracs available o average firms in he open economy:, 1, A, and 1, A. If exiss, i mus be 41

45 aken by bad firms, since i is he only conrac available o hem, and i is never feasible if average firms don also choose his conrac in equilibrium. Given his, only exiss if boh bad and average firms ake he conrac, and he maximum payou ha lenders can offer wih such a conrac is R r A B A pb [( )/( )]. The maximum payou ha domesic lenders can offer for he screened conrac is 1, A Rr, and he maximum payou he foreign lenders can offer for he small, screened conrac is R r. If boh 0 max R r, R r 1, A and Rr [( )/( p ) 0, hen none of hese oher conracs provide a posiive payoff o A B A B firms, and, C is he only equilibrium conrac. Below is he lis of assumpions and condiions given by Proposiions 1-3 ha ensure a swich from a pooling equilibrium in he closed economy o a separaing equilibrium in he open economy where average and bad firms are no financed and ne oupu declines:, r r ; Assumpion (A1) (E.1) r r ; Assumpion (A2) (E.2) r ( R r) R ; pooling equilibrium in closed economy 1 (1 p) B (E.3) r R ; pooling equilibrium in closed economy is feasible 1 (1 p) B (E.4) r ( R r ) R ; foreign enry causes separaing equilibrium 1 (1 p) B (E.5) 0 ( R r ) ( r pr); infeasible o pool average and bad firms (E.6) A B 0 max R r, R r ; infeasible o screen average firms (E.7) (E.8) ( R r) A p B ( R r ) R r C; decline in ne oupu when swich The above condiions, however, can be grealy simplified. or example, equaion (E.2) can be eliminaed since i mus always holds when 1 and equaions (E.1), (E.3), (E.5), and (E.7) all hold. To see his, noice ha equaions (E.3) and (E.7) boh place lower bounds on ĸ, such ha max, R r, while equaion (E.5) places an upper bound on ĸ. One can quickly show ha for λ > 1, r > r, hese bounds ensure ha equaion (E.2) holds. The firs half of equaion (E.1) can also 42

46 be eliminaed since i is no acually necessary for a decline in ne oupu; a decline in ne oupu can occur even when so long as equaions (E.3)-(E.8) all hold. Equaion (E.4) can also be eliminaed since i always hold if equaions (E.5) and (E.8) boh hold. And, plugging in for r r /(1 (1 p) B ), equaions(e.5), (E.7), and(e.8), can be combined o creae bounds on ĸ, such ha R r R r R p B r A B max, ( ) ( (1 (1 ) ) ) / (1 ). We are hen lef wih he four condiions lised in Proposiion 4. QED Appendix. Proof of Proposiion 5 As shown in Proposiion 1, he closed economy exhibis a pooling equilibrium when, and as shown in Proposiion 2, he open economy exhibis a separaing equilibrium when. Moreover, as shown in he proof of Proposiion 2, here are only hree possible equilibrium conracs available o average firms in he open economy:, 1, A, and 1, A. If exiss, i mus be aken by bad firms, since i is he only conrac available o hem, and i is never feasible if average firms don also choose his conrac in equilibrium. Given his, only exiss if boh bad and average firms ake he conrac, and he maximum payou ha lenders can offer wih such a conrac is R r A B A pb [( )/( )]. The maximum payou ha domesic lenders can offer for he screened conrac is 1, A Rr, and he maximum payou he foreign lenders can offer for he small, screened conrac is R r. If R r [( ) / ( p ) 0 and 1, A A B A B A B A B R r [( ) / ( p )] max R r, R r hen he pooling conrac is boh feasible and preferred by average firms over a screened conrac. In order for ne oupu o decline in his scenario, i mus be ha ne oupu in he closed economy, ( p )( R r /(1 (1 p) ), exceeds ne oupu in he open economy, A B C B p R r R r. This will be rue when ( A B)[ ( A B)/( A B)] C[ ( ) ] R r r r A B ( 1)( ) ( )/(1 ). 43

47 Below is he lis of assumpions and condiions ha ensure a swich from a pooling equilibrium in he closed economy o a separaing equilibrium in he open economy where average and bad firms are pooled and ne oupu declines:, r r ; Assumpion (A1) (.1) r r ; Assumpion (A2) (.2) r ( R r) R ; pooling equilibrium in closed economy 1 (1 p) B (.3) r R ; pooling equilibrium in closed economy is feasible 1 (1 p) B (.4) r ( R r ) R ; foreign enry causes separaing equilibrium 1 (1 p) B (.5) ( R r ) ( r pr) 0; feasible o pool average and bad firms (.6) A B A B A B R r [( ) / ( p )] max R r, R r ; average firms prefer pooling (.7) R r r r A B ( 1)( ) ( ) / (1 ); decline in ne oupu when swich (.8) The above condiions, however, can be grealy simplified. or example, equaion (.2) can be eliminaed since i mus always holds when 1 and equaions (.1), (.3), (.5), and (.7) all hold. To see his, noice ha equaions (.3) and boh place lower bounds on ĸ, such ha max, ( A B) / ( A B) r p r, while equaion (.5) places an upper bound on ĸ. One can quickly show ha for λ > 1, r > r, hese bounds ensure ha equaion (.2) holds. The firs half of equaion (.1) can also be eliminaed since i is no acually necessary for a decline in ne oupu; a decline in ne oupu can occur even when so long as equaions (.3)-(.8) all hold. Equaion (.4) can also be eliminaed since i always hold if equaions (.5) and (.8) boh hold. And equaions (.5), (.7), and (.8), can be combined o creae bounds on ĸ, such ha max r (1 p) / ( p ),( 1)( R r ) ( r r ) / (1 ). We are hen lef wih B A B A B he four condiions lised in Proposiion 5. QED Appendix G. Proof of Proposiion 6 This proof will proceed in four pars. irs, I will prove ha a lender ha offers a screened conrac will always inves an amoun ce ( ) in screening experise, where is he number of firms 44

48 screened by he lender in equilibrium and e is given by c'( e ) 1. Second, I will prove ha he condiions in Proposiion 6 ensure he closed economy exhibis a pooling equilibrium wih conrac. This proof will parallel ha of Proposiion 1. Third, I will prove ha ha hese condiions also ensure ha foreign enry resuls in a separaing equilibrium. This proof will parallel ha of Proposiion 2. inally, I will prove ha only cream firms are financed and ne oupu declines in he separaing equilibrium ha occurs afer foreign enry. This proof will parallel ha of Proposiion 3. Par 1 Using he same logic as in Proposiion 1, i can be shown ha here are only hree possible ypes of equilibrium conracs: 1,,, A C and. Therefore, if lenders offer a screened conrac, i will only do so o a firm ha succeeds wih probabiliy 1. Moreover, by Lemma 1, i is sufficien o consider only conracs wih RI ( ) 0and (0) 0. experise e I can hen be shown ha a screened conrac from a lender ha invess in screening e, where e is given by c'( e ) 1, canno be an equilibrium conrac. or example, consider a lender of ype j ha offers screened conracs bu makes zero invesmen in screening experise. The bes expeced payoff, (RI), ha such a lender can offer wih on a screened conrac for firms of ype i, loan size I and sill breakeven is [ R r( j)] I ( j). However, a lender of he same ype j ha invess ce () in order o obain screening experise e (0, e ) for screened loans will always be offer a more compeiive screened conrac. The oal profis of such a lender ha screens and finances firms of ype i is given by [ R r( j)] I [ ( j) e] ( RI) c( e), where RI ( ) is he paymen provided o firms ha accep his conrac, and hence, he bes expeced payoff such a lender can offer firms is RI ( ) [ Rr( j)] I[ ( j) e] ce ( ). I can be shown ha RI ( ) RI ( ) ece ( ) 0since c(0) c'(0) 0, c'() e 0, c''() e 0for e > 0, and c'( e) 1 for e e. Therefore, here will exis an 0 such ha a lender of ype j could always iniiae a profiable deviaion by invesing in experise e(0, e ) and offering screened conracs ha provide a payoff RI ( ) o firms of ype i. Using a similar logic, i can hen be shown ha a screened conrac from a lender ha invess e (0, e ) canno be an equilibrium conrac since here will exis a profiable deviaion for a lender ha invess in screening experise e. 45

49 inally, i can be shown ha invesing ce ( ) o obain he screening experise e necessary o fund firms, where is greaer han he acual number of firms screened and financed by he lender in equilibrium, canno be an equilibrium oucome. A lender could always make a profiable deviaion by invesing a smaller amoun ce ( ) where exacly maches he number of firms screened and financed by ha lender. Par 2 Par (c) of Proposiion 6 ensures ha he pooling conrac ( R ) R r / 1 (1 p ) is he equilibrium conrac in he closed economy. Given lenders B ha offer screened conracs mus inves in experise e, he maximum expeced payoffs ha domesic screened conracs 1,A and,c can provide o firms in he closed economy are R r ( e ) c( e ) and ( R r) ( e ) c( e ), respecively. Cream firms will prefer he pooling conrac,, if is payou, ( R ), exceeds he maximum payou of,c, and if cream prefer he pooling conrac, hen average firms mus also prefer i. ollowing he same logic as in Pars 4-5 of Proposiion 1, i can hen be shown ha when e c( e ), he maximum possible payoff,c does no exceed he maximum possible payoff of, and hence,,c and 1, A will no be equilibrium conracs. This conrac is feasible when r/1 (1 p) B R, which similar o he proof of Proposiion 4, mus hold if condiion (d) of Proposiion 6 holds. ollowing he same logic as in Pars 4-5 of Proposiion 1, i can hen be shown ha is he only equilibrium conrac for e c( e ) and since par (c) of Proposiion 6 ensures ha e c( e ), i mus be ha R R r p ( ) / 1 (1 ) B is he only equilibrium conrac he closed economy. Par 3 Using he same logic as in he proof of Proposiion 2, i can be shown ha condiion (d) of Proposiion 6 ensures ha he open economy exhibis a separaing equilibrium. To differeniae conracs offered by foreign lenders, I will express heir conracs as Ik,. Using he same logic as in pars 1-3 of he proof of Proposiion 1, here are only hree poenial foreign lender 46

50 conracs ha can be equilibrium conracs, 1, A, and, C, and i is sufficien o consider conracs of he form ( RI) 0and (0) 0. In an economy wih boh domesic and foreign lenders, he domesic lender conrac,,c lenders because of assumpion (A2), and, can no longer be an equilibrium conrac for domesic canno be an equilibrium conrac since r r. Therefore, here are only four possible equilibrium conracs:, 1, A, 1, A, and., C Similar o pars 4-5 of Proposiion 1, i can be shown ha, C only exiss and is preferred by cream firms over he pooling conrac for lenders and heir zero profi condiion will ensure ha e c( e ). Compeiion among foreign R R r e c e,, C ( ) ( ) ( ) ( ) which exceeds he maximum possible payoff o cream firms wih he pooling conrac,, when e c( e ). Par (d) of Proposiion 6, however, ensures ha e c( e ), such ha he open economy mus exhibi a separaing equilibrium where cream firms ake, C conracs. Par 4 The only conrac offered in a separaing equilibrium wih domesic and foreign lenders will be, C conracs. By condiions (c) and (d) of Proposiion 6, i mus be ha R r e c( e ) and lenders can feasibly offer screened conracs R r e c( e ), which ensures ha neiher domesic or foreign 1, A and o average firms in a separaing 1, A equilibrium, and Condiion (b) ensures ha foreign lender canno profiably offer. Using he same logic as in Proposiion 3, one can hen show ha condiion (d) in Proposiion 6 ensures ha he separaing equilibrium in he open economy resuls in a decline in ne oupu. In he pooling equilibrium wih domesic lenders, ne oupu is( p )( R r ), where A B C r r /(1 (1 p) B ), while in he separaing equilibrium where only cream firms accep projecs from foreign lenders, he ne oupu is C [ ( R r ) ( e ) c( e )]. Thus, a decrease in ne oupu will occur when ( p )( R r ) [ ( R r ) ( e ) c( e )] is rue. Bu, condiion (d) of A B C C Proposiion 6 ensures ha his mus hold. QED 47

51 Appendix H. Proof of Proposiion 7 This proof will proceed in hree pars. irs, I prove ha he condiions in Proposiion 7 ensure he closed economy exhibis a pooling equilibrium wih conrac. This proof parallels ha of Proposiion 1. Second, I prove ha ha hese condiions also ensure ha enry of a merged lender resuls in a separaing equilibrium. This proof parallels ha of Proposiion 2. inally, I prove ha only cream firms are financed and ne oupu declines in he separaing equilibrium ha occurs afer enry of merged lenders. This proof parallels ha of Proposiion 3. Par 1 Par (c) of Proposiion 7 ensures ha, which following he same logic as Proposiion 1 implies ha he pooling conrac R R r p ( ) / 1 (1 ) B is he only possible equilibrium conrac in he closed economy, and his will be an equilibrium conrac so long as / 1 (1 ) 0. And, ( R r ) R1 (1 p) r/ 1 R r p B by condiion (c) of Proposiion 7, ensures ha B A B, as given R r / 1 (1 p) B 0mus be rue. Par 2 Since merged lenders can always offer he mos compeiive financial conracs (because hey have boh he lowes cos of screening and lowes cos of funds), i is only necessary o consider equilibria where merged lenders offer financial conracs. Using he same logic as in he proof of Proposiion 2, i can be hen shown ha, as implied by par(c) of Proposiion 7, ensures ha he open economy exhibis a separaing equilibrium. Par 3 Condiion (b) and (c) of Proposiion 7 ensure ha neiher a pooling conrac wih only average and bad firms or a screened conrac for average firms is feasible. Therefore, only cream firms will be financed in he separaing equilibrium. Using he same logic as in Proposiion 3, one can hen show ha ( R r ) R1 (1 p) r/ 1 B A B, as implied by par (c) of Proposiion 7, ensures ha he separaing equilibrium in he open economy resuls in a decline in ne oupu. QED 48

52 Acknowledgmens or heir many useful commens, I am graeful o Daron Acemoglu, Abhiji Banerjee, Marin Brown, Iay Goldsein, Radha Gopalan, Simon Johnson, David Masa, Guy Michaels, Todd Milbourn, Anjan Thakor, and Andrey Ukhov, as well as seminar paricipans a he CEPR-EBC-UA Conference on Compeiion in Banking (Anwerp, Belgium), he Simon School of Business (Universiy of Rocheser), Olin Business School (Washingon Universiy in S. Louis), New York ederal Reserve Bank, ederal Reserve Board of Governors, Kelley School of Business (Indiana Universiy) and uqua School of Business (Duke Universiy). All errors and omissions are my own. 49

53 References [1] Acemoglu, Daron (1998) Credi Marke Imperfecions and he Separaion of Ownership from Conrol, Journal of Economic Theory, 78, [2] Agarwal, Sumi and Rober Hauswald (2010) Disance and Privae Informaion in Lending, Review of inancial Sudies, 23(7), [3] Agénor, Pierre-Richard (2003) Benefis and Coss of Inernaional inancial Inegraion: Theory and acs, World Economy, 26(8), [4] Aleem, Irfan (1990) Imperfec Informaion, Screening, and he Coss of Informal Lending: A Sudy of a Rural Credi Marke in Pakisan, World Bank Economic Review, 4(3), [5] Area, Carlos, Barry Eichengreen and Charles Wyplosz (2001) When Does Capial Accoun Liberalizaion Help More han I Hurs? NBER working paper, [6] Banerjee, Abhiji V., Shawn Cole and Esher Duflo (2005) Bank inancing in India, in India s and China s Recen Experience wih Reform and Growh, eds. Wanda Tseng and David Cowen, Hampshire, UK and New York: IM and Palgrave-Macmillan, [7] Banerjee, Abhiji V., and Esher Duflo (in press) Do irms Wan o Borrow More? Tesing Credi Consrains Using a Direced Lending Program, Review of Economic Sudies. [8] Beck, Thorsen, Asli Demirguc-Kun and Maria Soledad Marinez Peria (2007) Reaching Ou: Access o and Use of Banking Services across Counries, Journal of inancial Economics 85, [9] Beck, Thorsen and Maria Soledad Marinez Peria (2010) oreign Bank Paricipaion and Oureach: Evidence from Mexico, Journal of inancial Inermediaion 19, [10] Bekaer, Geer, Campbell R. Harvey and Chrisian Lundblad (2005) Does inancial Liberalizaion Spur Growh? Journal of inancial Economics, 77(1), [11] Berger, Allen N., Leora. Klapper and Gregory. Udell (2001) The Abiliy of Banks o Lend o Informaionally Opaque Small Businesses, Journal of Banking & inance, 25, [12] Bolon, Parick and David S. Scharfsein (1990) A Theory of Predaion Based on Agency Problems in inance Conracing, American Economic Review, 80(1), [13] Boo, Arnoud W.A. and Anjan V. Thakor (2000), Can Relaionship Banking Survive Compeiion? Journal of inance, 55(2), [14] Bruno, Valenina and Rober Hauswald (2011), The Real Effec of oreign Banks, working paper, American Universiy. [15] Buch, Claudia M. (2005) Disance and Inernaional Banking, Review of Inernaional Economics, 13(4),

54 [16] Claessens, Sijn and Neelje van Horen (2012a) oreign Banks: Trends, Impac and inancial Sabiliy, IM working paper WP/12/10. [17] Claessens, Sijn and Neelje van Horen (2012b) Being a oreigner Among Domesic Banks: Asse or Liabiliy? Journal of Banking & inance, 36, [18] Clarke, George R. G., Rober Cull and Maria Soledad Marinez Peria (2001) Does foreign bank peneraion reduce access o credi in developing counries? Evidence from asking borrowers, World Bank, working paper. [19] Clarke, George R. G., Rober Cull, Maria Soledad Marinez Peria, and Susana M. Sánchez (2005) Bank Lending o Small Businesses in Lain America: Does Bank Origin Maer?, Journal of Money, Credi and Banking, 37(1), [20] Degryse, Hans, Olena Havrylchyk, Emilia Jurzyk, and Sylweser Kozak (2012) oreign Bank Enry, Credi Allocaion and Lending Raes in Emerging Markes: Empirical Evidence from Poland, Journal of Banking & inance, 36, [21] Degryse, Hans and Seven Ongena (2005) Disance, Lending Relaionships, and Compeiion, Journal of inance, 60(1), [22] Dell Ariccia, Giovanni and Rober Marquez (2004) Informaion and bank credi allocaion, Journal of inancial Economics, 72, [23] Dell Ariccia, Giovanni and Rober Marquez (2006) Lending Booms and Lending Sandards, Journal of inance, 61(5), [24] Deragiache, Enrica, Poonam Gupa and Thierry Tressel (2008) oreign Banks in Poor Counries: Theory and Evidence, Journal of inance, 63(5). [25] Edison, Hali J., Ross Levine, Luca Ricci, and Torsen Sløk (2002) Inernaional inancial Inegraion and Economic Growh, Journal of Inernaional Money and inance, 21, [26] Edwards, Sebasian (2001) Capial Mobiliy and Economic Performance: Are Emerging Economies Differen? NBER working paper, [27] Eichengreen, Barry (2001) Capial Accoun Liberalizaion: Wha Do Cross-Counry Sudies Tell Us? The World Bank Economic Review, 15(3), [28] Eichengreen, Barry and David LeBalang (2003) Capial Accoun Liberalizaion and Growh: Was Mr. Mahahir Righ? Inernaional Journal of inance and Economics, 8, [29] Galindo, Aruro, Alejandro Micco, and Guillermor Ordoñez (2002) inancial Liberalizaion: Does i Pay o Join he Pary? Economía, 3(1), [30] Giannei, Mariassuna and Seven Ongena (2009) inancial Inegraion and irm Performance: Evidence from oreign Bank Enry in Emerging Markes, Review of inance, 13(2), [31] Glode, Vincen, Richard C. Green, and Richard Lowery. (2012) inancial Experise as an Arms Race, Journal of inance, 67(5), pp

55 [32] Gormley, Todd A. (2010) The Impac of oreign Bank Enry in Emerging Markes: Evidence from India, Journal of inancial Inermediaion, 19(1), pp [33] Gormley, Todd A., Simon Johnson and Changyong Rhee (2011) Ending Too Big o ail : Governmen Policy vs. Invesor Percepions, NBER working paper [34] Haber, Sephen and Aldo Musacchio (2011) oreign Banks and he Mexican Economy, , working paper, Sanford Universiy and Harvard Business School. [35] Hanson, James A. (2003) Indian Banking: Marke Liberalizaion and he Pressures for Insiuional and Marke ramework Reform in Anne O. Krueger and Sajjid Z. Chinoy eds., Reforming India s Exernal, inancial, and iscal Policies, Sanford Universiy Press, California, [36] Henry, Peer Blair (2000) Do Sock Marke Liberalizaions Cause Invesmen Booms? Journal of inancial Economics, 58, [37] Holmsrom, Beng and Jean Tirole (1997) inancial Inermediaion, Loanable unds, and he Real Secor, Quarerly Journal of Economics, 112(3), [38] Kaminsky, Graciela and Sergio Schmukler (2008) Shor-Run Pain, Long-Run Gain: The Effecs of inancial Liberalizaion, Review of inance, 12(2), [39] Levine, Ross (1996) oreign Banks, inancial Developmen, and Economic Growh, in Claude E. Barfield ed., Inernaional inancial Markes: Harmonizaion versus Compeiion, AEI Press, Washingon, DC, [40] Lewis, Tracy R. and David E. M. Sappingon (1995) Insurance, Adverse Selecion, and Cream- Skimming, Journal of Economic Theory, 65, [41] Marin, Albero (2010) Adverse Selecion, Credi, and Efficiency: he Case of he Missing Marke working paper, Universia Pompeu abra. [42] Mian, Aif (2003) oreign, Privae Domesic, and Governmen Banks: New Evidence from Emerging Markes, working paper, Universiy of Chicago [43] Mian, Aif (2006) Disance Consrains: The Limis of oreign Lending in Poor Economies, Journal of inance, 61(3), [44] Micco, Alejandro, Ugo Panizza, and Monica Yañez (2007) Bank ownership and performance. Does poliics maer? Journal of Banking & inance, 31(1), [45] Misrulli, Paolo Emilio and Luca Casolaro (2008) Disance, Lending Technologies and Ineres Raes working paper, Bank of Ialy. [46] Peersen, Michell A. and Raghuram G. Rajan (1995) The Effec of Credi Marke Compeiion on Lending Relaionships, Quarerly Journal of Economics, 110(2), [47] Peersen, Michell A. and Raghuram G. Rajan (2002) Does Disance Sill Maer? The 52

56 Informaion Revoluion in Small Business Lending, Journal of inance, 57, [48] Quinn, Dennis P. (2000) Democracy and Inernaional inancial Liberalizaion, working paper, Georgeown Universiy. [49] Rodrik, Dani (1998) Who Needs Capial-Accoun Converibliy? Princeon Essays in Inernaional inance, 207, [50] Rohschild, Michael and Joseph Sigliz (1976) Equilibrium in Compeiive Insurance Markes: An Essay on he Economics of Imperfec Informaion, Quarerly Journal of Economics, 90(4), [51] Sengupa, Rajdeep (2007) oreign Enry and Bank Compeiion, Journal of inancial Economics, 84(2), [52] Sein, Jeremy C. (2002) Informaion Producion and Capial Allocaion: Decenralized versus Hierarchical irms, Journal of inance, 57(5), [53] Sigliz, Joseph E. (2000) Capial Marke Liberalizaion, Economic Growh and Insabiliy, World Developmen, 28(6), [54] Sigliz, Joseph E. and Andrew Weiss (1981) Credi Raioning in Markes wih Imperfec Informaion, American Economic Review, 71(3), [55] Zingales, Luigi (1998) Survival of he ies of he aes? Exi and inancing in he Trucking Indusry, Journal of inance, 53(3),

57 or κ below his line, enran bank has comparaive advanage for projecs of size λ Pooling equilibrium in boh closed and open economies λ(r-r ) R-r Separaing equilibrium in closed economy Pooling o separaing equilibrium swich; only cream financed; ne oupu falls Pooling o separaing equilibrium swich; only cream financed, bu ne oupu increases κ > κ above his line R-r irs bes allocaion; boh average and cream firms are screened and financed and ne oupu increases igure 1. Example of possible equilibria for given screening coss, and. This figure provides an illusraive example of possible equilibria in boh he open and closed economy for a given,, rr A B,,, p, and R, where 0 A( R r ) B( r pr), r > r, ( R r) R r /(1 (1 p) B, ( R r ) R r (1 (1 p) B, and ( R r ) ( R(1 (1 p) ) r)/(1 ). B A B 54

58 Separaing equilibrium in boh closed and open economy Pooling equilibrium in boh closed and open economies Pooling o separaing equilibrium swich; only cream financed, ne oupu falls 0.15 Pooling o separaing equilibrium swich; only cream financed, bu ne oupu increases igure 2. Numerical example of possible equilibria for given screening coss, and. This figure maps he possible equilibria in boh he open and closed economy for a range of κ and κ whena 0.4, B 0.2, r 1.09, r 1.075, 2.5, p 0.8, and R=