Capital mobility in a second best world - moral hazard with costly financial intermediation*

Transcription

1 Forthcoing, REVIEW OF INTERNATIONAL ECONOMCS Capital obility in a second best world - oral hazard with costly financial interediation* Joshua Aizenan RRH: CAPITAL MOBILITY IN A SECOND BEST WORLD LRH: Joshua Aizenan Abstract This paper studies financial integration in the presence of oral hazard, where banks ay itigate excessive risk by costly onitoring. We show that a drop in banks' cost of funds, less efficient interediation technology, higher acroeconoic volatility, and a ore generous deposit insurance raise the riskiness of projects in a copetitive equilibriu. Overborrowing would arise even in the absence of deposit insurance in circustances where the cost of risk onitoring is high, the banks' cost of funds is relatively low, and acroeconoic volatility is high. Reforing an inefficient banking syste and iproving its operation is a precondition for successful financial integration. *Joshua Aizenan, Econoics Departent, Social Sciences I, Univ. of California, Santa Cruz, 95064, Tel: (831) , Fax: (831) , E-ail: jaizen@cats.ucsc.edu. Useful coents by an anonyous referee, Phil Brock, Rachel McCulloch, Assaf Razin, Yora Weiss, the participants in seinars at Brandeis University, Koc University, Tel-Aviv University, and University of Washington are gratefully acknowledged. Any errors are ine. JEL classification: F15, F2, F34 Abbreviations: Nuber of Figures: 5 Nuber of Tables: 0 Date: May 21, 2001

2 Introduction and suary The recent financial crises in eerging arket econoies have focused attention on the role of financial interediation in explaining the costs and benefits of capital arket integration. 1 Recent contributions stressed the tendency for overborrowing due to oral hazard considerations -- a phenoena coined "the Overborrowing Syndroe" [see McKinnon and Pill (1996), Dooley (2000), Krugan (1998); and Brock (1992) for an earlier analysis of loan guarantees and overborrowing]. 2 These studies identified the deposit insurance syste as the key echanis leading to overborrowing. Yet, several observers questioned the iportance of deposit insurance in explaining the crisis in the Far East. For exaple, Radelet and Sachs (1998, page 24) argue that "It is hard to ake the case, however, that foreign investors felt theselves in general way to be indenified against risk through the prospect of generous bailouts...thus, it is probably fairer to say that foreign investors thought too little about risk because they expected rapid growth and high profitability to continue, not because they expected a bailout." Our analysis will show that the Overborrowing hypothesis ay be relevant even if the above stateent is accurate. The iplicit insurance arguent ay have weak explanatory power for overborrowing due to other considerations. Ex-ante; it is hard to easure the extent of iplicit insurance. Ex-post, alost any realized path of borrowing can be 'explained' by the backward construction of expectations regarding iplicit insurance. Thus, it reains a challenge to explain the dependence of overborrowing on the underlying econoic structure -- why does the Overborrowing Syndroe see to atter for soe countries, whereas other countries anaged their borrowing ore prudently. Tracing the causes of overborrowing is pertinent to providing policy guidance. If the Overborrowing Syndroe is driven by the provision of iplicit insurance, alleviating it requires dealing with the tie inconsistency probles associated with public bailouts. One approach ay be to design better precoitent echaniss, in order to convince the private sector that the bailing out would not occur. This 'get tough' approach ay include budgetary steps that will ake bailing out prohibitive. If these policies are effective, one ay argue that there is no further need to restrict financial interediation. If the Overborrowing Syndroe were traced to inefficient financial interediation, overborrowing ay occur even in the absence of an iplicit deposit insurance. In these circustances one should design policies that iprove risk onitoring, and would shift the financial interediation fro the less efficient towards the ore efficient banks. As we will show in the paper, the root causes of overborrowing go beyond iplicit insurance, as overborrowing ay occur even in the absence of

3 - 3 - iplicit deposit insurance. In these circustances, prudent policies call for strengthening the efficiency of financial interediation -- penalizing borrowing if risk supervision is deficient, and encouraging financial interediation by ore efficient banks. 3 In order to address these issues, one should derive the Overborrowing Syndroe endogenously, in a ore fully specified econoic odel. The purpose of this paper is to construct such a odel, and to argue that the welfare effects of financial integration are ore involved than the ones suggested by the previous contributors. We show that the association between the depth of financial integration and welfare ay be non-onotonic. We point out that overborrowing would arise even in the absence of deposit insurance in circustances where the cost of risk onitoring is relatively high, the banks' cost of funds is relatively low, and acroeconoic volatility is high. Specifically, we propose a odel where the riskiness of investent supported by banks is endogenously deterined. Entrepreneurs rely on banks to finance investent, facing a trade off between risk and return. The liited liability associated with bank financing induces entrepreneurs to undertake excessive risk. We assue that banks ay control this risk by costly onitoring, where greater risk reduction requires ore resources devoted to project supervision. We characterize the copetitive equilibriu, where banks' rents are dissipated, and the arginal project earns a zero rent. We show that a drop in banks' cost of funds increases the risk tolerated by banks. Siilarly, a less efficient interediation technology (i.e., a ore costly risk onitoring), higher acroeconoic volatility, and a ore generous deposit insurance, all raise the risk in a copetitive equilibriu. Such an equilibriu tends to be inefficient -- a cobination of a low banks' cost of funds and a high enough cost of risk onitoring would iply a large distortion due to excessive risk taking. We construct the social welfare function, being the su of the expected surplus of all doestic agents. We use this welfare function to evaluate the consequences of financial integration for an econoy characterized by a relative scarcity of savings. For a large enough cost of risk onitoring, the dependence of welfare on banks' cost of funds has an inverted U shape. A drop in the banks' cost of funds due to financial liberalization would have two effects - the direct saving in financing costs of a given investent is welfare iproving, whereas the increase in the "excessive risk" distortion is welfare reducing. The "optial depth" of financial liberalization is reached when these two effects balance at the argin. Any further welfare gain fro financial liberalization would require iproveent in the efficiency of financial interediation. If the autarky banks' cost of funds is relatively large, it will curb the excessive risk distortion in autarky, iplying that partial financial liberalization would increase

4 - 4 - welfare. For such an econoy, full financial integration would be welfare reducing relative to partial financial liberalization, as it leads to excessive risk-taking. Siilar to the case of iiserizing growth, it is the interaction between the initial distortion (excessive risk) and globalization of financial arkets that leads to these second best results. 4 Even in these circustances, the econoy will benefit by financial integration that is accopanied by the proper iproveents in the functioning of doestic banks. Furtherore, our paper suggests that financial integration and reforing the banking sector are copleentary policies, as the gain of each refor is agnified by the second. Before turning to the odel, it is constructive to put the paper in the broader context of the recent literature. Moral hazard was advanced as one of the possible causes of the crisis in the Far East in the late nineties [see, for exaple, Corsetti, Pesenti and Roubini (1999) and the references there]. 5 These authors highlighted the iplications of deposit insurance and other bailout schees in inducing overborrowing. According to this interpretation, a crisis is triggered when a large enough overborrowing threatens the credibility of the bailout schee due to fiscal constraints. Dooley (2000) and Chinn and Kletzer (1999) pointed out that, in the presence of governent guarantees, financial liberalization puts in otion the process leading to overborrowing. These observations are consistent with recent epirical studies [see Deirguc-Kunt and Detragiache (1998), and the review by Williason and Mahar (1998)]. The policy iplications oral hazard induced by deposit insurance are studied in Hellann, Murdock and Stigliz (2000). In their paper, banks deterine endogenously their risk exposure, and the prudential regulator learns the banks' riskiness only ex-post. The regulator's policies are the iposition of capital requireents and ceilings on deposit rate, and the ex-post closure of failed banks. They find that the optial prudential regulations should include both capital requireents and ceilings on deposit rate. Our paper shares the concern of the above papers about overborrowing. It suppleents these studies by showing that overborrowing ay exist even in the absence of deposit insurance, due to the banks' liited liability in circustances where onitoring is costly. The iplication of our study is that the bias towards overborrowing is even stronger than the one considered by the above literature, as it ay happen even in the absence of bailouts. Hence, one should adopt a cautious attitude towards financial integration even in the absence of concerns about deposit insurance and bailouts.

5 The odel All agents are risk neutral. Banks are copetitive, and there is no reserve requireent. Each project costs H, and is characterized by a probability of failure, and by a productivity index x. The productivity index is unobserved by banks, but is known to entrepreneurs. Henceforth we refer to as easuring the project's riskiness, and we assue that the entrepreneur deterines it ex-ante. Projects are independent, and are ordered by declining productivity -- a higher x is associated with a lower productivity. Failure iplies zero incoe, whereas success iplies incoe e = e( ; x), where e = h( x)( ) q, q > 0, h' < 0 (1) Hence, riskier projects that turned out to be successful are associated with higher output. Entrepreneurs ust finance the investent H by bank credit, at a real interest cost of r l. The expected gross incoe fro project type x, denoted by p, is p( ; x) = ( 1 - ) h( x)( ) q. (2) The entrepreneur's net expected incoe, G, is G( ; x) = p(, x) -( 1 - ) ( 1+ r) H = ( 1 - )[ h( x)( ) q -( 1 + r) H], (3) l Proposition 1: The liited liability of entrepreneurs induces excessive risk-taking in coparison to a self financed project -- entrepreneurs choose so that [ p ( )] < 0. Proof - See Appendix A Figure 1 suarizes the entrepreneur's behavior, drawn for the case where H = 1. Curve pp corresponds to the expected gross incoe, p( ). Curve cc is the expected financing cost, ( 1 - )( 1 + r l ). Point B deterines the optial risk undertaking fro the entrepreneurs point of view, whereas point D is the optial risk undertaking if the project is self financed INSERT FIGURE Banks ay engage in costly risk onitoring. Specifically, spending z per project allows the bank to verify that the project's probability of failure is. We assue that the onitoring cost increases with risk reduction -- l

6 - 6 - z = z( ' " ), z < 0, z > 0; z( 1) = 0. (4) The bank's expected surplus with onitoring, per project, is B( ) = ( 1 - )( 1 + rl ) H -z( ) -( 1 + rc ) H (5) where r c is the bank's cost of funds. Henceforth, we assue a non prohibitive cost of onitoring -- i.e., banks are better off onitoring. 6, 7 Each bank controls a large nuber of independent projects, diversifying away the idiosyncratic risk. Copetition aong banks induces rent dissipation. Proposition 2: The interest rate and the projects' risk in a copetitive equilibriu are characterized by ( 1 + rl ) H = -z'( ) (6a) z( )[ t - 1] = ( 1 + rc ) H (6b) where t = dlog z/ dlog( 1 - ) is the elasticity of supervision cost z with respect to the project's probability of success, 1 -. Proof See Appendix A The resultant equilibriu is characterized by Figure 2, drawn for the case where H = 1. The bold curve is the bank's cost of a 1$ loan [i.e., z( ) + ( 1 + r c ) 1]. The downward sloping line, cc, is the expected repayent per unit loan. Free entry and optial onitoring iplies an equilibriu at the tangency of the expected repayent line and the bank's unit cost line, deterining INSERT FIGURE The productivity of the arginal project, denoted by x, is deterined by the rent dissipation condition, Applying (3) and (A2) to (7) we infer that G( x, ) = 0 (7) x -1 Ïz( ) + ( 1 + r H = h c) Ì Ó ( 1 - ) q (8)

7 - 7 - In a copetitive equilibriu entrepreneurs will finance all the projects characterized by x x (recall that a higher x is associated with lower productivity). Equations (6a), (6b) and (7) for a syste of 3 equations, the solution of which deterines ( x; ; r l ) as a function of the bank's cost of funds, r c. Proposition 3: Less efficient financial interediation or lower banks' cost of funds increase the projects' risk in a copetitive equilibriu. Proof See Appendix A Exaple: Consider the case where onitoring technology and productivity are characterized by constant elasticities, z( ) = k( 1 - ) t ; t > 1, hx ( ) = h0 ( x) ; 1 > f > 0, (9) -f where h 0, f, t and k are constants. In these circustances (6) and (7) yield 1/ t ÈH( 1 + rc ) = 1 - Í Î ( t -1) k q 1/ È h 0( ) x = Í t -1 Ît( 1 - ) k f (10a) (10b) Consuers All agents are risk neutral, and their utility is C C (11) 1 + r Soe agents have access to an outside incoe in period 1, denoted by Y. These agents supply their saving, S, deanding real interest rate r c = r for S Y. 3. Welfare and financial integration We consider now the iplications of financial integration. We start the discussion with the characterization of the social welfare function, being the su of the expected surplus of all doestic agents -- producers, banks and savers. The welfare contribution of project x is obtained by suing (3) and (5), resulting in

8 - 8 - Wx = ( 1 - ) { h( x)( ) q -( 1 + rl) H} + {( 1 - )( 1 + rl) H -z( ) -( 1 + rc ) H} = ( 1 - ) hx ( )( ) q -z( ) -( 1 + rc ) H (12) We assue an internal solution in autarky, where the deand for investent is satisfied by the supply of saving at r c = r, and therefor savers' surplus is zero. We consider a continuous version of the odel, where the 'nuber' of projects of productivity x is easured by f( x) [i.e., the ass of projects the productivity index of which is between x and x + e is f( x)e ]. The social welfare function is the expected surplus aggregated across all the realized projects -- x SW = Ú [ p (, x) -z( ) -( 1 + rc ) H] f ( x) dx 0 (13) The social welfare for the constant elasticities exaple considered earlier, assuing a unifor distribution of x, f( x) = 1, is SW - ( - ) q 1 f 1 h x = 0( ) -k( - t 1 ) x -( 1 + rc ) Hx. (14) 1 - f Financial integration allows doestic banks access to the global pool of savings, offering funds at a cost of r o. We assue that the autarky banks' cost of funds exceeds the global risk free interest rate [ r > r o ]. Hence, financial integration is viewed as a process that reduces the banks' cost of funds to the global level. In these circustances the patterns of risk undertaking and investent are suarized by (10), where the banks' cost of funds, r c, drops fro r to r o. The welfare contribution of doestic investent with financial integration is obtained by (13), evaluated for the investent ( x ) and risk ( ) that correspond to rc = ro. To siplify discussion, we assue that doestic consuers can not borrow against future incoe, hence their saving is zero when rc = ro. With these assuptions the savers' surplus is zero, and (13) is the exact welfare function. We can apply (13) to identify the socially optial level of risk. This would correspond to an equilibriu where banks break even, investors finance all projects offering non-negative expected rents, and the riskiness is deterined by a policy aker who axiizes (13). Proposition 4: The socially optial risk ( S ) and investent ( x S ) for the constant elasticities exaple, with f( x) = 1, is characterized by

9 - 9 - È Ï t S = 1- ÍH( 1 + rc)/ Ì( -1) k Î Ó W x = W - h ( ) q t / t( 1 - ) 1 k 0 S S S 1/ t [ { }] 1/ f (15a) q( S) where W = S. (15b) 1 - f Proof See Appendix A Note that the copetitive equilibriu conditions (10) are a special case of (15), for W =1. Hence, the copetitive equilibriu is associated with excessive risk if W>1. This is likely to hold the higher the supervision cost k is relative to the start up cost H( 1 + r c ). 8 The welfare effects of financial integration are found by evaluating the iplications of a drop in banks' cost of funds. Applying (13) it follows that dsw drc d x p(, x) x = { -kz f x dx - Hf x dx dr Ú [ '( )] ( ) } Ú ( ) c 0 0 (16) In general, the effect of a drop in banks' cost of funds is abiguous, as the sign of d x p(, x) [ kz'( )] f ( x) dx dr Ú - c 0 { } ay be positive. This ter easures the net welfare effect of the increase in project's risk induced by a lower bank's cost of funds (recall that d dr c < 0). It equals d dr c ties the su of the arginal effect of risk on project's expected incoe, inus the arginal ipact of higher risk on the onitoring cost, p (, x) - kz'( ). This su is zero when the risk allocation is socially optial, and is negative if excessive risk is undertaken, and positive if too little risk is x p(, x) undertaken. Hence, -{ Ú [ - kz'( )] f ( x) dx} easures the distortion associated with excessive 0 risk-taking. There are 2 useful benchark cases where the "excessive risk" distortion is absent, and thus (16) is unabiguously negative. First, if all projects are self financed, p (, x ) = 0 and no resources are spent on onitoring, thus - kz'( ) = 0. In these circustances, (16) is negative, and a lower interest rate r c unabiguously raises welfare. Second, in the absence of "deep pockets," if policies are used to induce

10 x the first best risk allocation, 0 = Ú - 0 [ p(, x) kz'( )] f ( x) dx, and again (16) is negative. The excessive risk distortion is eliinated in these benchark cases, either due to the "full liability" associated with self financing, or due to the optial design of policies. The dependence of welfare on the banks' cost of funds ay be non-onotonic, characterized by an inverted U shape curve. This follows fro Proposition 3 -- recall that a drop in the banks' cost of funds and a less efficient interediation technology increase the risk tolerated by banks. Hence, a cobination of low banks' cost of funds and a high enough cost of risk onitoring would increase the distortion associated with excessive risk taking, and ay induce dsw > 0. In these econoies, a drop in dr c the bank's cost of funds would lead to a 'perverse' outcoe, reducing welfare. The reverse applies if the banks' cost of funds increases. For a high enough banks' cost of funds, the excessive risk distortion would be sall enough so that the sign of (16) is reversed - further increase in the banks' cost of funds reduces welfare [hence, (16) will becoe negative]. We confir this intuition with the help of a siulation. Figure 3 reports the dependency of welfare on the banks' cost of funds. The four curves are obtained by increasing sequentially the cost of risk onitoring by increents of 20%, and their relative position corresponds inversely to the cost of risk onitoring. The siulation confirs the presence of an inverted U shape, and reveals that a higher cost of risk onitoring shifts the curves downwards and to the right, increasing thereby the 'welfare axiizing' interest rate. Hence, for an efficient enough risk onitoring technology, financial integration is unabiguously welfare enhancing, whereas for highly inefficient risk onitoring technology, financial integration is welfare reducing [as will be the case if the autarky banks' cost of funds is below the welfare axiizing' interest rate]. For interediate cases, the effect of financial liberalization is abiguous. If the autarky banks' cost of funds is high, the first stages of financial liberalization are beneficial, but the latter stages ay be welfare reducing INSERT FIGURE Further insight is obtained by Figure 4 tracing the excessive risk distortion, easured by the gap between the socially optial and the copetitive risk levels [i.e., the gap between (15a) and (10a)]. Panel I of Figure 4 corresponds to relatively inefficient interediation (using the paraeter values associated with the bold curve in Figure 3). Panel II of Figure 4 corresponds to the case of relatively efficient interediation (using the paraeter values associated with the solid, top curve in Figure 3).

11 Curve CC traces the projects' risk in the copetitive equilibriu, and curve OO corresponds to the projects' risk in the optial allocation. As our previous discussion suggested, a cobination of low banks' cost of funds and a high risk onitoring cost would lead to a large excessive risk. This situation is depicted by Figure 4, Panel I, where the excessive risk is about 10% for low interest rates. In these circustances higher banks' cost of funds increases welfare, as is depicted by the bold curve in Figure 3. A by product of the higher banks' cost of funds is that the "excessive risk" distortion shrinks gradually, iplying that for a high enough banks' cost of funds the welfare effects of further increase in the banks' cost of funds are reversed. Conversely, in a relatively efficient syste the gap between the optial and the actual riskiness is relatively sall, as is depicted by panel II in Figure 4. In these circustances financial integration is welfare enhancing, as is indicated by the top curve in Figure INSERT FIGURE Our analysis points out that an eerging arket characterized by financial autarky and inefficient financial interediation would benefit fro the gradual relaxation of restrictions on capital obility. An ideal transition should be characterized by sequential reoval of capital controls, as part of a coprehensive progra designed to iprove the efficiency of financial interediation. 10 In ters of Figure 3, suppose that the eerging arket is characterized by financial autarky, where the doestic real interest rate is higher than the global one, and the process of financial interediation is relatively inefficient, so that the cost is k = 1.4k 0. The optial sequencing should target the reoval of capital controls so that the doestic interest rate ay follow overtie a path along line AB. This would correspond to an interest rate that, for the given efficiency of financial interediation, axiizes the expected welfare. Political econoy considerations ay iply that existing banks would have the incentive to backtrack the process, in order to protect their quasi-rents. This suggests that it would be preferable to announce a pre-set path of reoving capital control as part of a liberalization process that would lead to iproveent in the efficiency of banks. This process would allow overtie access of doestic banks to foreign funds, as well as the operation of foreign banks in the doestic econoy Extensions - Macro shocks, deposit insurance The odel outlined above should be viewed as a benchark fraework, and one can extend it to reflect other concerns. In this section we review the ipact of acro shocks and deposit insurance. We show that ore generous deposit insurance and higher acroeconoic volatility agnify the "excessive risk" distortion. The presence of acro shocks and deposit insurance does not change the socially

12 optial risk, yet both would encourage banks to tolerate greater risk, increasing thereby the range where financial liberalization is welfare reducing. 4.1 Deposit insurance We illustrate now the ipact of changing the "generosity" of a deposit insurance schee. For siplicity of exposition, we follow the assuptions of Section 2, where only idiosyncratic risk is present. Suppose that banks anticipate a partial bailout. Specifically, suppose that banks expect that, if their net incoe is negative, a fraction Y of the non-perforing loans will be repaid by the public sector. In these circustances banks' expected profits (5) are odified to (17) Ï[ Y + ( 1 - )]( 1 + rl ) H -z( ) -( 1 + rc ) H if ( 1 - )( 1 + rl ) H -z( ) -( 1 + rc ) H < 0 Rx (, Ô ) = Ì Ô Ó( 1 - )( 1 + rl ) H -z( ) -( 1 + rc ) H if ( 1 - )( 1 + rl ) H -z( ) -( 1 + rc ) H 0 If deposit insurance is expected to be used, the first order condition deterining a project's risk is [ 1 - Y]( 1 + rl ) H = -z'( ) (18a) y 1 + z( 1 - )[ t - 1] = ( 1 + rc ) H 1 - Y (18b) Deposit insurance reduces the arginal cost of risk fro the bank's point of view [the LHS of (18a)], encouraging thereby risk undertaking. While deposit insurance does not ipact the expected output of a given project, it "socializes" part of the risk. Hence, banks would increase the project's risk tolerated. 12 The net effect is increasing the excessive risk distortion. This point is exeplified in Figure 5, panel II, tracing the excessive risk as a function of the bank's cost of funds. The curve is drawn for the paraeter values used in Figure 4, panel I. The efficient risk is traced by curve OO. The top curve corresponds to a copetitive equilibriu, where Y =005., and the iddle curve corresponds to the copetitive equilibriu in the absence of deposit insurance ( Y =0). As suggested by (18a), deposit insurance increases the risk tolerated by banks, agnifying the excessive risk distortion. Consequently,

13 as in to the previous discussion, deposit insurance increases the range where financial liberalization would be welfare reducing INSERT FIGURE Macro shocks We preserve all the previous assuptions about the idiosyncratic characteristic, x, assuing a odified production function, where projects are also subject to acro shocks e = ( 1 + d) h( x)( ) q, q 0, h' < 0 (19) and d is a acro shock, following a binoial distribution Ïe d = Ì Ó -e (20) each state ay occur with probability Let us denote by R(, x) the repayent on project type x, the risk of which is. We assue that the realized output is public inforation. If the contractual repayent exceeds the realized output, the bank gets all output. 14 Hence, Ï0 probability Ô R(, x) = ÌMin[( 1 - e) h( x)( ) q ;( 1 + rl ) H] probability ( - ) Ô ÓÔ Min[( 1 + e) h( x)( ) q ; ( 1 + rl ) H] probability ( - ) (21) Unlike our pervious analysis, with acro shocks soe producers would default partially, iplying that the realized bank repayent will differ across producers. In the appendix we show that the equilibriu is characterized by x Ú { ER [ (, x)] - ( 1 + rc ) H - z} f( xdx ) = 0 0 x { } (22a) dú E[ R(, x)] -( 1 + rc ) H -z f( x) dx 0 = 0 (22b) d -f q h0( x) ( 1- ) - E[ R(, x)] = 0 (22c)

14 where E is the expectation operator. Condition (22a) states that banks break even ex-ante. Recall that the projects' productivity index is unobservable ex-ante, hence the banks' expected revenue is obtained by averaging it across all projects. Condition (22b) is the optial risk onitoring, and condition (22c) is the brake-even condition for the arginal entrepreneur. The presence of the acro shock does not odify the expected output, and the social welfare function continues to be (13). Consequently, result (16) regarding the abiguous welfare effects of a drop in the banks' cost of funds continues to hold. In Appendix A we show that, for constant elasticity, the unifor distribution exaple considered before, an extended version of Proposition 3 holds -- Proposition 3': Higher volatility of the acro shock, less efficient risk onitoring or lower banks' cost of funds increase the projects' risk in a copetitive equilibriu. Proof -- see Appendix A. Macroeconoic volatility increases the distortion associated with excessive risk. Higher acroeconoic volatility induces ore frequent partial defaults. This in turn leads banks to increase both the lending interest rate and the project's risk tolerated. The net effect is a rise in "excessive risk". The econoic rationale is that the repayent in bad states of nature is capped by partial default. Hence, banks will benefit by increasing the risk tolerated and the lending interest rate in the presence of a ore volatile acro shock. The greater risk will increase the realized output in good states of nature, whereas the higher lending interest rate charged by banks will shift the repayent towards the good states of nature. This point is exeplified in Figure 5, panel I, tracing the risk as a function of the bank's cost of funds. The curve is drawn for the paraeter values used in Figure 4, panel I. The efficient risk is traced by curve OO. The top 2 curves plot the copetitive equilibriu for varying acroeconoic volatility. The top curve corresponds to e = 025., and the iddle curve corresponds to e = 0 (the absence of acroeconoic volatility). Applying Proposition 3' to result (16) we conclude that, as in the previous discussion, for a high enough cost of risk onitoring, a low banks cost of funds, and significant enough acro volatility, a drop in the bank's cost of funds is welfare reducing. Macroeconoic volatility agnifies the excessive risk distortion, increasing thereby the range where financial liberalization would be welfare reducing. 5. Discussion and concluding rearks Our paper considered the case where oral hazard can be controlled by costly risk onitoring undertaken by banks. While the details of the equilibriu in the credit arket are odel specific, the

15 logic of the second best described in our paper should apply to other odels as well - if the equilibriu is characterized by excessive risk, financial integration ay agnify this distortion. This paper provides another exaple of iiserizing growth, this tie due to "excessive risk" induced by the cobination of low banks' cost of funds and costly financial integration. In these circustances liited liability induces a distortion, leading frequently to overborrowing. In autarky, the daaging effect of the distortion is confined by the liited availability of doestic savings, which act both to restrict investent and to reduce the size of the distortion. Financial integration would agnify the cost of the "excessive risk" distortion, both by increasing the distortion and by increasing the volue of investent. The essage of the odel is that sequencing atters -- efficient doestic banking is a pre condition for successful financial integration. 15 Our paper suggests that financial integration and reforing the banking sector are copleentary policies, as the gain of each refor is agnified by the second. If one starts with a highly inefficient banking syste, reforing it and iproving its operation is a precondition for successful financial integration.

16 Appendix A Appendix A suarizes the proofs of all the propositions in the paper. Proposition 1 Proposition 1 follows iediately fro (3) -- the entrepreneur's optial risk is deterined by [ p( )] = -( 1 + rl ) H. (A1) Had the project been self financed, the entrepreneur's optial risk would be deterined by [ p ( )] = 0. 1 ( 1 + rl ) H For exaple, if q = 1, entrepreneurs choose = + ; denoting the entrepreneur's optial 2 2h risk. In these circustances, a risk neutral entrepreneur who self finances the project would choose ' = 1 2. Proposition 2 Equation (6a) follows directly fro axiizing the bank's expected incoe, (5), with respect to the project's risk,. It equates the arginal benefit of risk reduction (the LHS of (6a)) with its arginal cost (the RHS of (6b)). Copetition aong banks induce rent dissipation. Hence, the borrowing interest rate is deterined by Equation (6b) is inferred by applying (6a) and (A2). Proposition 3 0 = ( 1 - )( 1 + rl ) H -z( ) -( 1 + rc ) H. (A2) Let the cost of banks' onitoring be kz( ), k being a shift paraeter easuring the efficiency of financial interediation (i.e., the cost of risk onitoring). Equations (6a) and (A2) iply that 0 = ( 1 - ) kz'( ) + kz( ) + ( 1 + rc ) H, (A3) fro which it follows that d ( 1 )'( z ) z( ) ( rc ) H/ k = = dk ( 1 - ) kz"( ) ( 1 - ) kz"( ) d H > 0; = - < 0. (A4) drc ( 1 - ) kz"( )

17 In ters of Figure 2, a less efficient onitoring technology rotates the bold curve clockwise around A, to AB, leading to higher risk in the copetitive equilibriu. Siilarly, a lower banks' cost of fund shifts the bold curve upward, by the extra cost, increasing thereby the risk undertaken in a copetitive equilibriu. 16 Note that (A3) also iplies that d > 0. Hence, we conclude that a higher dk [ /{( 1 + rc ) H}] supervision/start up costs ratio, k/{( 1 + rc ) H}, increases the projects' risk in a copetitive equilibriu. Proposition 4 The socially optial risk undertaking is deterined by SW = 0, iplying that ( ) [ ] ( ) q -f 1 - q 1 - h ( ) 1-0 x = ( 1 - t k t ). (A5) 1 - f equation (15b) is obtained by solving (A5) for x. Recall that x is deterined by the brake-even condition in the copetitive allocation -- all the projects offering non negative expected rents are financed. In these circustances (A2) and (7) continue to hold, and consequently ( ) q -f 1 ( ) ( 1 t - h0 x = k - ) + ( 1 + rc ) H. Equation (15a) is obtained by applying this condition to (A5), solving for. Proposition 3' We consider the siplest exaple, of two states of nature. Suppose that no producer would default in the good state of nature, as would be the case if ( 1 + e) hx ( )( ) q ( 1 + rl ) H(our discussion can be extended to cover the case where the weakest producers would default in all state of nature, without odifying the key results). Let us denote by x * the productivity index inducing arginal default in the bad state, defined by - ( 1 - e) h0 ( x*) f ( ) q = ( 1 + rl ) H. (A6) The scale of investent x is deterined by the rent dissipation of the arginal producer, hence

18 ( 1 - )[ ( - e) h0( x) f ( ) q ( + r) H] = ( 1 - ) h0( x) f ( ) q l. (A7) This equation corresponds to (22c). It is equivalent to - ( 1 + e) h0 ( x) f ( ) q = ( 1 + r l ) H. (A8) Cobining (A6) and (A8) we get The bank's break even condition iplies x * x 1/ - = È1 e ÎÍ 1 + e f. (A9) Èx* x - Í Ú ( 1 + r) ( ) + ( 1 - e) Ú 0( ) ( ) q l Hf x dx h x f f( x) dx x Í ( 1 - ) * x Í x - Ú {( 1 + rc ) H + z} f( x) dx = 0. (A10) Í + ( 1 + ) ( ) 0 Í Ú rl Hf x dx Î 0 This is the exact for of (22a). For the case of our constant elasticity exaple, we get È * - ( ) - ( *) Í( 1 + e) 0( ) ( ) q + ( 1 - e) 0( ) ( ) q f f f x f x x h x h x ( 1 - ) 0. 5Í x x( 1 - f) = ( 1 + rc ) H + z Í Í Î + ( 1 + r) H l. (A10a) Applying (A6) and (A7) to (A10a), collecting ters, we infer that È Ê - ˆ e q / f f e e.( ) h ( x) ( ) Í f + - f Ë1 + f ÎÍ e e = ( + rc ) H + z. (A11) The bank deterines the project's risk by axiizing (A11), leading to 17 ( t -1) z -( 1 + rc ) H 1 - ( )/ q ( ). ( e) ( ) ( ) q f f f e È Ê ˆ h0 x Í 1 - f Ë + e 1 ÎÍ 1 = 0. (A12) Applying (A12) to (A11), collecting ters, we infer that

19 - 19-1/ f 1/ f q t e e e e f f ( ) [ z ÈÊ1 - ˆ ] = Á - - È Í / ( 1 + ) Ë1 + e Ê - ˆ Í - Á Î 1 + e Î 1 + e Ë1 + e z rc H. Note that equation (6b) corresponds to (A13) for the case of zero acroeconoic volatility. Proposition 3' follows fro (A13). (A13) Appendix B The purpose of Appendix B is to characterize an equilibriu where banks have the incentive to onitor. This will be the case, if in a copetitive equilibriu with no supervision, a bank that supervises would increase its expected profits. Note that, with free entry, a copetitive equilibriu with no onitoring is characterized by [ p( )] = -( 1 + rl ) H (B1) ( 1 - )( 1 + rl ) = 1 + rc. (B2) Equation (B1) describes the entrepreneurs optial risk undertaking (equation (4)). Equation (B2) is the rent dissipation condition. Banks would refrain fro onitoring if, at that equilibriu, onitoring is reducing expected profits. This will be the case if Applying (6), B( ) > 0. (B3) = B( ) = = -( 1 + r) H -z'( ). (B4) l Applying (A2) to (A4), we infer that B( ) 1 rc = H - z '( ) = [ t z -( 1 + r c) H ] (B5) = where t = dlog z/ dlog( 1 - ). Hence, if t z > ( 1 + rc ) H, banks will refrain fro onitoring, as their costs are too high to justify the benefit associated with risk reduction. Otherwise, banks have the incentive to onitor, and the corner solution with zero onitoring is unattainable. Henceforth we assue that the onitoring cost is not prohibitively high, so that t z < ( 1 + rc ) H. In these circustances, the equilibriu onitoring is characterized by Proposition 2.

20 References Bernanke Ben. S. and Mark Gertler, "Agency costs, net worth and business fluctuations," Aerican Econoic Review, 79, (1989): Bhagwati, Jagdish, "Iiserizing growth: a geoetrical note," Review of Econoic Studies, 25 (3), June, (1958): Brock, Philip, "External shocks and financial collapse: foreign loan guarantees and interteporal substitution of investent in Texas and Chile, " Aerican Econoic Review 82, May, (1992): Chinn, Menzie D. and Kenneth M. Kletzer, "International Capital Flows, Doestic Financial Interediation, and Financial Crises Under Iperfect Inforation," Glick, R. R. Moreno and M. Spiegel, eds. Financial Crises in Eerging Markets, Cabridge Press, (2001): Corsetti, Giancarlo; Paolo Pesenti and Nouriel Roubini, "Paper Tigers? A Model of the Asian Crisis " European Econoic Review; 43(7), June, (1999): Deirguc-Kunt, Asli and Enrica Detragiache, "Financial liberalization and Financial Fragility, " in Pleskovic B. and J. E. Stiglitz, eds, Annual World Bank Conference on Developing Countries, World Bank, (1998): Dooley, P. Michael, "A odel of crises in eerging arkets," Econoic Journal 110, (2000): Edwards Sebastian and Carlos A. Vegh, "Banks and Macroeconoics Disturbances under Predeterined Exchange Rates," NBER Working Paper No. 5977, Edwards Sebastian, and Sweder Van Wijnbergen, "The Welfare Effects of Trade and Capital Market Liberalization," International Econoic Review; 27(1), February (1986): Hellann, Thoas; Kevin Murdock, and Joseph Stiglitz, "Liberalization, Moral Hazard in Banking, and Prudential Regulation: are Capital Requireent Enough?" Aerican Econoic Review, March, 90 (2000): Jaffee, Dwight and Joseph E. Stiglitz, "Credit Rationing," in Handbook of onetary econoics. Volue 2, Friedan B. and F. Hahn, eds., North Holland, (1990): Krugan, Paul, "Fire-sale FDI," NBER Conference on Capital Flows to Eerging Markets, Feb , McKinnon, Ronald I. and Hugh Pill, "The Overborrowing Syndroe," Chapter 1 in Financial deregulation and integration in East Asia, Takatoshi Ito and Anne O. Krueger (eds.), University of Chicago Press, 1996.

21 Radelet, Steven C. and Jeffrey D. Sachs, "The East Asian Financial Crisis: Diagnosis, Reedies, Prospects," anuscript, HIID, Razin, Assaf; Efrai Sadka and Chi-Wa Yuen, "An inforation-based odel of FDI: the gains fro trade revisited, anuscript," Tel-Aviv University, Rubini, Nouriel, Asia Hoe Page, Sachs Jeffrey D.; Aaron Tornell and Andres Velasco, 1996, "Financial crises in eerging arkets: The Lessons fro 1995: NBER Working Paper No Stiglitz, Joseph E. and Andrew Weiss, "Credit rationing in arkets with iperfect inforation," Aerican Econoic Review; June, (1981): Townsend, Robert M.,"Optial contracts and copetitive arkets with costly state verification," Journal of Econoic Theory, 21 (1979): Williason, John and Molly Mahar, "A survey of financial liberalization," Essays in International Finance, no Princeton: Princeton University, 1998.

22 c D B p 1+r c l p 1 Figure 1: Risk undertaking in the absence of onitoring

23 z+(1+r c ) B c 1+r l 1+r c c A ~ 1 1+r l Figure 2: Copetitive equilibriu and the interest rate

24 SW B 2.4 k=k 0 k=1.2k A k=1.4k k=1.6k r c Figure 3: Financial integration and welfare The siulation assues t = 25.; H = 006. ; q = 1; f = 09.; f() x = 1 The solid curve corresponds to k0= 0.34, the dotted curve to k = 1.2k0, the broken curve to k = 1.4k0, the bold curve to k = 1.6k0

25 C O C O C O O C r c I II k = k = r r r c Figure 4: Risk undertaking and the banks' cost of funds The siulation assues t = 25.; H = ; q = 1; f = 0.; 9 f() x = 1; h = 1 0 Curve CC corresponds to the copetitive equilibriu, curve OO to the optial riskiness of the arginal project in the efficient allocation

26 e = 0 e = Y = 0 Y = O O 0.5 O O r c r I Macroeconoic volatility r II Deposit insurance r c Figure 5: Risk and banks' cost of funds -- the effects of acroeconoic volatility and deposit insurance The siulation assues t = 25.; H = ; q = 1; f = 09.; f() x = 1; h = 1; k = Curve e = 0 corresponds to the copetitive equilibriu, in the absence of acroeconoic volatility. Curve e = 025. corresponds to the copetitive equilibriu, for the case where e = Curve OO traces the optial riskiness of the arginal project in the efficient allocation.

27 Following the Tequila period, its after-effects in Latin Aerica and the ore recent events in East Asia, the effect of volatility on eerging arket econoies has becoe an iportant topic of research. In any of these papers, the doestic financial interediation process is advanced as one of the ost iportant transission echaniss for volatility effects (see Sachs et. al (1996), Edwards and Vegh (1997), and the references in Rubini (1998)). At the sae tie there has been continued interest in issues related to iperfect inforation and rationing in credit arkets (see the seinal article by Stiglitz and Weiss (1981), the review by Jaffee and Stiglitz (1990) and the any references therein). The thees in this literature include also agency costs and costly state verification (see Townsend (1979) and Bernanke and Gertler (1989)). 2 McKinnon and Pill (1996, page 27) suarized this syndroe stating "...overborrowing arises when doestic residents becoe excessively optiistic about the econoy's prospects following the ipleentation of refor. These optiistic expectations are generated by arket failure of soe for - in our case, that induced in the banking syste by deposit insurance- but the existence of such failures is obscured until it is too late for their effect to be accounted for." 3 The U.S. Saving and Loan debacle is a painful exaple of the iportance of adequate risk supervision, and the costs associated with the lack of it. 4 On iiserizing growth due to trade distortions see Bhagwati (1958). See Razin, Sadka and Yuen (1998) for asyetric inforation odel, where the developent level of the doestic capital arket deterines the welfare effects of FDI. 5 The recent literature is too voluinous to be reviewed coprehensively in this paper. A coprehensive collection of related papers can be accessed via Roubini's Web Page, the Asian Crisis. 6 In Appendix B we show that this condition is satisfied if t z( ) < ( 1 + rc ) H, where t = dlog z/ dlog( 1 - ), and denoting the entrepreneur's optial risk. 7 We assue an internal solution to the bank's optial onitoring proble at the optial, the producer's surplus in the good state of nature suffices to cover the contractual debt. It can be shown that

28 this will be the case if h( 0)( ) q >- z'( ). A sufficient condition assuring it is a high enough productivity of the best project [i.e., a large enough h( 0 )]. 8 This follows fro Proposition 3, and fro the observation that 1 > f > 0 and that W>1 for close enough to 1. 9 k 10 An exaple of this possibility is depicted by the dotted curve in Figure 3 (corresponding to = 12. k 0 ), for r 0 = 0, r = While we do not odel in the paper the process leading to iproveent in financial interediation, recent banking developents in Argentina, the Czech Republic and other countries illustrate the role of the policy aker in this process. Financial interediation would be enhanced by iproving the collection and the sharing of credit records and credit history of potential borrowers. These iproveents ay involve investent in inforation infrastructure by the central bank, to be financed partially by the private banks, which ay tap this inforation in the future. Strealining bankruptcy procedures, and strengthening law and order would further strengthen financial interediation. 11 This procedure ay be preferable to a refor that projects the future reovals of capital controls as a process contingent on the efficiency of banks. Such an alternative refor would generate perverse incentives -- existing banks would have the incentive to delay efficiency iproveents, in order to delay the tiing of the reoval of capital control, thus prolonging the protection of their quasi-rents. Our fraework is not rich enough to account for the political econoy considerations that are involved in such a process, and these issues are left for future research. 12 Note that (18b) iplies that d dy > Hence, for each project there are three possible outcoes - the project will fail with probability, with probability 0.5(1- ) the project will yield ( 1 - e) hx ( )( ) q, and with probability 0.5(1- ) the project will yield ( 1 + e) hx ( )( ) q. 14 Our results would be strengthened if banks had to rely on costly legal procedures to induce a delinquent producer to service part of his outstanding debt. 15 See Edwards and Van Wijnbergen (1986) for optial sequencing in a different context.

29 This follows fro the observation that copetitive banks would pass to entrepreneurs the drop in bank's cost of funds, resulting in a drop in lender's interest rate. Equation (6a) iplies that the drop in the lenders interest rate (at a given risk) reduces the arginal cost of risk for banks, inducing banks to tolerate greater risk. 17 This expression is obtained after applying (A6) and (A8) to the resultant first order condition. In deriving (A12) we assue that each bank ignores the ipact of changing the risk on the value of x. The ain results of our analysis continue to hold even if banks internalize this effect.