Segmentation and the Use of Information in Brazilian Credit Markets 1. Armando Castelar Pinheiro 2. Alkimar Moura 3. February 2001

Transcription

1 egmentation and the Ue of Information in Brazilian redit arket 1 Armando atelar Pinheiro 2 Alkimar oura 3 February Introduction Until 1994, when inflation wa brought down from the ky-high level that had prevailed ince the mid - eventie, the financial ytem in Brazil wa almot entirely geared toward maximizing float income, which anwered for cloe to half of the overall earning of commercial bank. In that environment, being efficient in proceing tranaction, uch a the payment of bill and check, wa paramount to bank, wherea credit granting activitie received much le attention, ince very little credit flowed to the private ector anyway. 4 Incentive for inveting in proper credit analyi were further weakened by the fact that mot medium and long term credit to firm and houehold wa provided by tate bank, funded by pecific taxe and government tranfer. And political interference, oft budget and other ditortion common to tate-owned enterprie made the quality of credit analyi in thoe intitution particularly low (ckiney, 1998). Private bank, in turn, not only lent very little, but alo concentrated their loan activitie on hort-term operation, uch a overdraft facilitie and working capital finance, for which borrower cah flow, managed by the ame intitution, worked a collateral. Entry deregulation in the late eightie, although expanding the number of bank, did little to change thi cenario. he dramatic reduction in inflation rate after the eal Plan produced many change in the financial ytem. In particular, it reduced bank float income, cauing everal of them to run into eriou olvency problem. hi wa the cae of mot commercial tate bank, and alo of many mall and ome large private one. ome of the mall bank were liquidated, while mot of the medium and large bank were reolved through purchae-and-aumption tranaction, with the government auming a large hare of their bad loan. 5 In the cae of Banco do Brail, the country econd larget bank, the government had to make a capital infuion of cloe to U$ 8 billion to avoid bankruptcy. Other tate bank were re-capitalized and then privatized. everal of the bank reolved in thi period were acquir ed by foreign intitution. One could expect that the lo of float income, a more table macroeconomic environment and the entry of foreign intitution, which command modern credit analyi technology, would lead to a ubtantial expanion in credit activitie. urpriingly, though, thi did not occur, with the overall amount of credit extended to the private ector actually decreaing a a ratio of GDP (Pinheiro and abral, 1998). he only egment of the credit market that howed a ignificant expanion wa that of conumer and peronal loan, which almot trebled a a proportion of GDP from 1993 to Bank were not prepared for thi, with the 1 hi paper wa written a part of the World Bank reearch project on redit Information in atin America. he author would like to thank argaret iller for comment to a previou verion of the paper. 2 Head of the Economic Department of BNDE, Profeor Economic at the Federal Univerity of io de Janeiro (IE- UFJ) and enior reearch fellow at the Intitute of Economic, ocial and Political tudie of ão Paulo (IDEP). 3 Profeor of Economic at the Getúlio arga Foundation in ão Paulo (FG-P). 4 In , the volume credit to the private ector in Brazil averaged 27 percent of GDP, againt ratio everal time larger in indutrialized and Aian developing countrie (Demirguç-Kunt and akimovic, 1996). 5 ince the eal Plan, 104 bank were reolved by different mean: 42 were liquidated, 7 were incorporated in other intitution, 10 were tranformed into non-financial intitution, 11 changed from univeral bank to pecialized financial intitution, and 34 went through purchae and aumption tranaction. :\Other\catelar-moura redit_ext_10.doc

2 boom in credit upply occurring with eentially no change in credit -granting practice, which continued to rely on old-fahioned method for electing borrower. Other creditor, uch a department tore and mall retailer, were in even wore ituation, ince they lacked even thi limited experience. Not urpriingly, then, default rate increaed dramatically in thi market egment, cauing the bankruptcy of creditor that jut a few month before were poting record ale and interet income. hi wa the cae of two of the country larget department tore, Arapuã and ebla. Bank were not immune either. Public, national private and foreign bank all experienced a urge in default rate in their loan to houehold. Boavita, one of Brazil mot traditional bank, went bankrupt a year after poting the indutry highet profit rate, which wa almot entirely baed on interet income due on loan to conumer. ender were unprepared to ue the available information to elect good borrower, but the quality and nature of that information may alo be blamed for thoe poor reult. redit Informa tion egitrie (I) have exited in Brazil for everal decade, but have traditionally maintained motly black information, obtained from judicial and ecurity regitrie, chamber of commerce and the entral Bank regitry on returned check. ince after the eal Plan many borrower were acceing credit market for the firt time, the information available in thoe I provided little guidance about the likelihood of borrower default. In addition, to ome extent the function of I wa le to inform creditor than to encourage borrower to pay, ince a bad debtor name i eraed from thoe regitrie once payment i done. hat i, the emphai wa on enforcement, rather than on building data bank on borrower payment hitory. It i poible then to ummarize the ituation in Brazil credit market in the immediate aftermath of the eal Plan a one in which bank uperviion and prudential regulation had not been able to prevent the failure of a large number of financial intitution, bank had little expertie in providing credit and I were illequipped to provide the information neceary for adequate credit rik analyi. In um, not an environment conducive to the ort of expanion in private ector credit that could help to accelerate economic growth. 6 Brazil credit market ha changed ince then, with bank inveting in improving their credit analyi, a renewed dynamim in the I indutry and a ubtantial upgrade in the quality of bank regulation. he decline in real interet rate and in reerve requirement ince the 1999 devaluation of the real have alo contributed to timulate credit activitie. But, a we argue in thi paper, one feature of Brazil credit market hould not change in the foreeeable future: it egmentation into different ub-market, with borrower who poe equal rik to bank facing different loan condition, depending on their ize and the nature of their banking relationhip. egmentation, in turn, will limit the cope of I activitie, which hould continue to concentrate on collecting and dieminating black information, largely with enforcement purpoe. Further improvement in bank regulation hould then take into account that a large hare of the information neceary to ae credit rik will remain private to individual bank. aking the Brazilian cae a an example, thi paper analyze the conequence of interbank aymmetry of information for the way credit market operate and, in particular, for the role played by I in dieminating information. We argue that poor accounting and widepread tax evaion caue relationhip banking to be a key ource of information about a wide pectrum of borrower, making much of the relevant information about creditworthine private to individual bank. In thi environment, credit market tend to fragment into egment with different characteritic regarding interet rate and average loan ize. he more pervaive i tax evaion and the more opaque i publicly available information, the larger the hare of credit channeled through non-competitive market egment, where bank exploit their information monopoly to extract rent from borrower, charging interet rate above thoe which would prevail in the preence of 6 For evidence on the poitive impact of financial deepening on economic growth ee King and evine (1993) and Beck, evine and oayza (1999). 2

3 ymmetric information. We how that in uch a egmented credit market the role of private I i limited eentially to upplying black information about borrower, the main objective of which i to foter debt repayment. Although the analyi i centered on the Brazilian cae, we believe that our reult are alo relevant for other developing countrie. We proceed in three tep. In ection 2 we focu on the ue of credit information by bank in their loan operation, howing how interbank aymmetry of information reulting from relationhip banking caue market egmentation. In ection 3 we argue that market egmentation help to explain why I in Brazil tend to collect and upply motly black information, and explain why thi ituation i unlikely to change even if better mechanim to creen borrower type become available. In ection 4 we look at how market egmentation affect uperviion and prudential regulation of the banking ector. A final ection um up our main concluion and the policy implication of our analyi. 2 Bank lending to private partie in Brazil 2.1. he redit Deciion Proce Up to 1994, private Brazilian bank were not very active in lending and therefore were not careful in implementing credit deciion policie and procee. In the high inflation pe riod, from 1974 to June 1994, with full indexation of wage, rent, contract, foreign exchange and financial aet, monetary policy wa generally aimed at controlling the nominal interet rate, therefore providing liquidity to utain increaing level of aggregate demand. Default ratio by both firm and individual were low, change in loan-lo reerve were a mall hare of bank total expenditure, and credit income anwered for an equally mall fraction of their overall revenue. Under thoe circumtance, credit policie were almot non-exitent, being limited to maintaining internal cutomer file (cadatro) to tore particularly negative information. Bank exchanged information about their cutomer with other lender (both bank and non-bank) through a completely informal network of informer (the o-called informante ) whoe ole function wa to crocheck retrictive data about the bank borrower. It wa only after price tability in 1994 that financial intitution became keen on expanding their lending operation, particularly in financing the ale of durable conumer good. In fact, there wa an incipient credit bubble beginning with the tabilization plan in July 1994 and lating up to arch Expene and income aociated with credit activitie began to account for a ignificant proportion of total bank expenditure and revenue. However, both bank and borrower were not prepared to operate in the new environment of eay acce to credit. he default ratio on bank loan increaed fater than total performing loan during thi period, which i, among other factor, an indication of the generally poor quality of credit management then prevailing in the country. In January 1995, on average, for each real of performing loan bank poted $ 0.08 of non-performing loan; in January 1997, the correponding figure wa $ 0.18, that i, a 125% increae. hi unucceful experience prompted many bank to retructure their credit area, trying to introduce new policie and procedure to cope with credit rik. ix year after the burt of thi credit bubble, there are till ubtantial cro-bank difference with repect to the tage of their organizational development, a far a the formulation, implementation, monitoring, controlling and evaluation of credit policie, procedure and practice. On the one hand, there are ome bank with a relatively trong credit culture, which tend to make intenive ue of internally generated information a well a external data (that i to ay, mainly data provided by the variou I) a input to their credit deciion proce. On the other hand, bank with a looe or ill-formulated credit policy in mot cae do not make ue of formal criteria to allocate credit (other than the traditional method of etablihing fixed credit limit to cutomer) and therefore ue information le intenively, including I data, to decide 3

4 on lending operation. In between thee two type, and thi might be the cae of mot bank, there are many intitution trying to introduce formal policie, procedure and practice of credit management, including purchae of foreign method and model of credit analyi and coring. Procedure adopted in the credit deciion proce differ according to the type of bank and the characteritic of the loanborrower. For loan to conumer and to mall buine, the general trend i toward the introduction of a highly decentralized proce of credit management. According to thi, all loan requet are treated automatically by tatitical method (credit coring, for intance), baed on information upplied by the client andor available from public record, with a deciion being rapidly reached at the branch level. aking into conideration the borrower characteritic, the tatitical model aign himher a quantity of point and the correponding automatic credit limit. Exception are dealt with at higher level of the credit bureaucracy, generally by credit committee. hi aet allocation proce i motly ued in lending operation uch a overdraft facilitie, conumer intallment credit, leaing, credit card loan and ecured or unecured peronal loan. hat eem to be the mot efficient way to guarantee peedy deciion in large retail bank, which can receive a much a 2000 loan application per day. 7 For loan to mall buine, typically for working capital need (baed on dicounting of predated check and duplicata ), the deciion proce i very imilar, with branche having their own credit limit for ecured operation. 8 hi mean that a large hare of all loan, a much a 80% in ome bank, i decided at the branch level, baed on automatic credit evaluation method relying on tatitical analyi. For loan other than to the o-called retail market i.e., conumer and mall buinee -- the traditional method of credit management i to etablih credit limit per cutomer in order to retrain the lender expoure to a particular obligor. A recent trend in the banking indutry ha been to tranform the credit deciion into a group deciion, that i to ay, a deciion made by formal credit committee which are generally organized according to criteria uch a the value of the loan, the exitence and kind of collateral, and the type of operation. Each application i treated on a cae-by-cae bai by the correponding credit committee, taking into conideration variable uch a the client file ( cadatro ), it economic and financial ituation, it relationhip with the bank, it buine tradition and it indutry propect. In ome large bank, branche do not extend buine loan (except to the mall buinee mentioned earlier), with loan application being decided by thee committee or by the bank credit department. ome mall wholeale bank with relatively large loan value per buine cutomer have rather formalized rule for the credit committee organization, including variable uch a it compoition, ize of expoure, maximum and minimum loan maturity, type of collateral, rule for functioning and alo for voting on loan requet. Our interview with bank manager revealed ubtantial difference with repect to the intenity with which financial intitution reort to I data -- both black and white (ee Annex A). A a general rule, it can be aid that all bank ue negative information provided by I a a firt filter in the credit deciion-making proce, that i to ay, in order to decide whether or not to continue with the analyi of the credit application. herefore, that type of information i the relevant barrier to dicriminate between potential borrower and applicant with no acce to credit market. 9 In the retail market, where a large number of 7 In fact, in thi type of retail operation bank compete with each other with repect to the peed at which they can decide on loan application, with peed in thi cae being meaured in number of econd. 8 Given the precariou quality of accounting and other financial information on mall buinee, where the firm and the owner banking account often mix together, lender tend to conider them a a ingle entity for credit granting purpoe. 9 Of coure, the larger the geographical coverage of the I databae the better the quality of the information. But bank alo value two other apect of the information upplied by I. Firt, the timeline and accuracy of the information, that i to ay, the time lag between any event affecting borrower behavior and it tranmiion to the creditor file. he horter thi time interval the more rapidly the credit regitry trace the change in the borrower economic and financial condition. econd, the degree of completene of the information in term of it market 4

5 low-value loan to mall buinee and individual take place, the dicriminating variable i the borrower credit record (cadatro), which i heavily biaed toward weighing the importance of retrictive information. In thi cae, black information provided by I i probably the mot relevant and poibly the only data ued in the credit deciion proce. 10 Once a credit relationhip ha been etablihed, information provided by I become ueful alo for monitoring the borrower financial ituation, i.e., to learn about the occurrence of event that might lead to default. In thi way, both upgrade and downgrade in the borrower creditworthine can be anticipated by monitoring change in hiher economic fortune a recorded in thoe regitrie. hi i not a trivial conideration, given that in the eightie and ninetie the Brazilian economy ha been ubjected to evere macroeconomic hock, which deeply affected borrower ability to pay. High volatility in interet and exchange rate, varying retriction on term and condition for lending, and trade liberalization are example of macroeconomic outcome that have made bank face large wing in market a well a credit rik. he importance of negative information in credit analyi decreae a the ize and complexity of the loan operation increae. It role i therefore le important in the o-called middle-market, which eem to be the mot profitable buine lending activity in Brazil. Bank ue two type of information to make loan deciion in thi credit egment: firtly, black and white informatio n provided by I and by other lender and, econdly, data collected by the bank itelf through balance-heet analyi and on-ite viit to firm. In mot cae, lender ue the I information either to check or to complement their own private information and analyi. here are financial intitution that even maintain their own in-houe credit rating facility. ome more aggreive bank in thi market egment almot diregard the uual publihed balance-heet data, on account of their mirepreentation of the actual economic and financial ituation of companie. Intead, they replace the formal accounting information with an internally created managerial information ytem to trace the actual change in the company financial condition. One important part of uch a ytem i to monitor the liquidity of the borrower receivable (mainly duplicata ) ince the latter are the mot commonly accepted collateral for buine loan in Brazil. he data and analyi of borrower creditworthine gathered by internally developed management information ytem remain private to the bank and are not hared with credit bureau. In the cae of loan to large firm (private corporation, Brazilian and multinational, and ome tate-owned firm), information prov ided by I ha a very limited role in the credit analyi proce, in comparion to the reearch and analyi conducted by the lender itelf, coupled with whatever private information ha been previouly gathered by the financial intermediary. Audited ba lance heet and other financial tatement are alo valuable in uch cae becaue they are more reliable than for maller firm. In particular, many of thee borrower, being public corporation (with hare quoted in dometic or U tock market, or having raied fund abroad through iuance of eurobond or through debt intrument in Brazilian market, uch a debenture and commercial paper) have to provide invetor with a regular flow of information on their economic and financial condition. he credit proce take longer and i obviouly more cotly, relatively to other lending operation. ending to the o-called corporate ector account for a large proportion of the total coverage, meaning by that the I capacity to provide informa tion on the borrower behavior in other egment of the credit market, uch a trade finance, conumer credit, real e tate market, capital market, and o on. 10 Indeed, ome of the large commercial bank have replaced their own buine record by imilar information gathered and proceed by EAA, Brazil larget I. 5

6 credit portfolio of Brazilian retail bank, though it client-bae i very mall. pread are alo rather narrow in thi type of credit operation. 11 An important feature of the Brazilian credit market that come through rather clearly from the above decription i it diviion in three egment, which differ with repect to typical loan ize and the nature and amount of information on borrower ued by creditor. In the retail market the number of loan application i very large, loan ize i mall, interet rate are high and the credit-deciion proce i decentralized and automated, relying motly on outourced, black information. In the middle market, bank tend to bae their deciion on internally collected information, which i often obtained from a continued banking relationhip with borrower. hi information remain private to the bank. he very poor quality of the information contained in thoe borrower balance heet, largely a reult of pervaive tax evaion and poor accounting practice, make information on the borrower cah flow extremely valuable to ae her actual creditworthine. A third market egment comprie foreign and large national corporation, which for variou reaon keep much better accounting, which i largely public information. In thi market egment there are fewer borrower, but loan tend to be larger and interet rate lower than in the other egment. ome of thee feature of the Brazilian credit market are evident in able 2.1, which how the ditribution of borrower with total debt of $ 20,000 or more in any ingle financial intitution, according to loan ize and number of intitution with which they contracted thoe loan. 12 On the whole, there were 1.1 million individual borrower in thi group on June 30, 2000, with outtanding debt of 65.0 billion reai, each of them owing on average $ 58,878. For the thouand firm in the ame category, total debt added to billion reai, correponding to an average loan of $ 907,489 per firm. able 2.1: Ditribution of borrower according to value of total loan (l) extended by the financial ytem and the number of intitution that have extended thoe loan (June 30, 2000)(*) um of loan extended to 1 Intitution 2 Intitution 3 Intitution 4 Intitution 5 or more Intitution otal each borrower (l) in 1000 $ Indivi dual Firm Indivi dual Firm Indivi dual Firm Indivi dual Firm Indivi dual Firm Indivi dual Firm 20 l < l < l < l < l < l < l < l > otal ource: entral Bank. (*) Include only borrower with debt of $ 20 thouand or more in any ingle financial intitution. On June 30, 2000, 1 UD = $ hey can vary from 0.5% to 5.0% p.a., with the lending rate following cloely the change in the baic dometic interet rate (Gazeta ercantil, eptember 22nd, 1999). hee contrat with much higher average pread on commercial loan, which in eptember 1999 reached 36.9% (entral Bank). 12 Not hown in able 2.1 are the loan to borrower with debt of le than 20 thouand reai in any ingle financial intitution. Overall, thee account for 30 percent of the total credit extended to individual and firm in Brazil ($ billion on June 30, 2000), with $ 53.0 billion lent to individual and $ 44.8 billion to firm. 6

7 Brazilian borrower tend in general to prefer ingle -bank relationhip. In the cae of individual borrower, only one in every twenty owe money to more than one intitution. hi pattern i heavily influenced by the behavior of mall debtor, that i, thoe with total loan between $ 20,000 and $ 50,000, who account for 66% of all individual with debt above 20 thouand reai, and who only exceptionally (0.5% likelihood) borrow from more than one intitution. Among mid -ized individual borrower ($ 50,000 < l $ 200,000) ingle banking i alo prevalent, with only 11.4% of them having debt with more than one intitution. Even among large borrower (l > $ 200,000) ingle banking i common, although 52% of them have debt with two or more intitution. ingle-banking i alo dominant among firm, which however tend to diverify their ource of loan more than individual. onidering all firm with debt of $ 20,000 or more, we have that jut 36.1% of them borrow from more than one bank. In the cae of mall debtor ingle banking i more prevalent, with only one in every ten firm owing money to more than one intitution. id-ize commercial debtor tend to operate with more bank than mall one, but yet 49.3% of them have debt with only one intitution and 35.0% with jut two intitution. For large commercial debtor, however, multiple-bank relationhip are the norm, with only 23.8% of them operating with a ingle intitution. Below we preent a imple model with interbank aymmetry of information that produce the ort of market egmentation decribed above. egmentation arie becaue good borrower are informationally captured by the intitution with which they bank, and a a conequence end up paying higher interet rate tha n they would in the tandard ymmetric information cae. he aumption that in the context of interbank aymmetry of information inide bank are able to extract rent from afe borrower i tandard in model that feature relationhip banking. 13 Differently from thoe model, however, we aume that, one, even for the inide bank there i a cot to ae borrower type; and, two, that a borrower can evade thi information trap by making it type known to outide bank, even if at a cot, what contrain the monopoly power of the inide bank. 14 We then ue the model to conduct ome comparative tatitic exercie to ee how thi market tructure change with bank borrowing rate and the cot of information he odel he model ha 2 type of borrower, afe and riky, with probabilitie of default 1- q and 1- ( q q ) q <, repectively. When borrower default the bank receive nothing. Borrower differ alo with repect to how much the loan i worth to them (v) and to the ize of the loan (l) they eek, which i aumed to be independent of v. For conumer v may be interpreted a the rate of time preference and for firm a the rate of return they expect to obtain from inveting the money they borrowed. We aume that v i uniformly ditributed in [ 0, ], while l ha an exponential ditribution with mean. 15 For all interval x, 0, 0, there i a proportion p of afe borrower. Borrower rate of return or time [ ] [ ] [ ] ( ) 0, x preference and the ize of the loan they eek are private information, but their ditribution are common knowledge. All bank and borrower are aumed to be rik neutral. r r 13 ee, for intance, harpe (1990), Beanko and akor (1993) and Padilla and Pagano (2000). 14 here i alo a difference of degree: the information advantage that the inide bank ha vi a vi outide bank from being able to infer a borrower hadow account i likely to be much larger than the one it would be able to derive from a normal banking relationhip. 15 In Annex B we preent the empirical ditribution of loan ize for individual and firm on June 30, Becaue thee reflect the ditribution of loan actually extended, they do not necearily have to be exponentially ditributed, even if the ex-ante ditribution i. omparing the empirical ditribution with the exponential, logonormal, uniform and Weibull ditribution with the ame mean we find that the exponential ditribution provide the bet fit in the cae of individual borrower, but not for firm, for which the lognormal give a cloer approximation. 7

8 At the beginning of each period, each borrower ha a banking relationhip with a ingle bank. 16 It i eaier for thi incumbent bank than for all other bank to acertain thi borrower type, but thi i not without cot. 17 o make her type known to outide bank, entering what we call the corporate market, a borrower ha to pend. Alternatively, he may decide to reveal her type to the incumbent bank at a cot (< ). 18 We aume and to be public information. Finally, a borrower may opt not to reveal her type and borrow in the retail market. here i no fixed cot in acceing the retail market. 19 he credit market i then divided in three egment: corporate, middle and retail. In the firt, there i perfect ymmetry of information and all bank know whether the potential borrower i afe or riky. In the retail market no bank know a borrower type, and complete aymmetry of information prevail between bank and borrower. We aume that the number of bank i ufficiently large for loan in the corporate and retail market to be priced competitively. he middle -market i characterized by a cloe banking relationhip, o that only one bank know whether the potential borrower i afe. hi ort of interbank information aymmetry exit in all credit market, but what i peculiar in Brazil and poibly in other developing countrie i the magnifying effect of tax evaion and poor accounting practice, with only one bank being able to oberve the actual cah flow, and indirectly the creditworthine, of the borrower. We aume that each bank act a a monopolit in it middle market egment. 20 he game i played a follow. Initially, bank et a menu of interet rate for the three market egment. Borrower then decide whether or not to tap the market, and, if they decide to borrow, in which market egment to do o. A Nah equilibrium may then be derived by fixing the menu of interet rate o that bank maximize their expected profit conditional on borrower optimal reaction. A afe borrower i will chooe the corporate market iff ( vi ) li and < li ( ) q q > v +, that i, i > (2.1) and liq 16 It i not neceary to aume that borrower do not bank with other intitution, but imply that only one of thee ha ufficient information on the borrower financial data to ecurely acertain at a low cot whether he i afe or riky. 17 In practice, becaue it i neceary to proce the relevant information and due to high market volatility, which caue borrower type to change frequently. 18 One way to interpret i a being the cot to hire a rating agency. But in the Brazilian context a more relevant factor tend to be the cot of conducting buine in a formal way, with proper accounting. In particular, becaue thi make tax evaion more difficult, increaing borrower tax expenditure. o we attribute mot of the difference between and to the difficulty of borrower who tap the corporate market to evade their taxe. 19 he reult do not change if we aume intead that it i the bank that initially incur the cot or and later charge it to the borrower, a long a (i) the latter till incur ome upfront cot to apply for credit and (ii) the mechanim to creen borrower type alway correctly identifie her type and he know that. he model would not change either if we aumed that bank pay for creening borrower and afterward charge thi cot to them by way of interet rate pread o that (l) = jl, where j = or for the middle and corporate market, repectively. 20 he realization that a continuou financial contact between a bank and it client generate valuable information date back to Kane and alkiel (1965) and Black (1975). Fama (1985) noted that thi information i important for both creening and monitoring purpoe. ummer and connell (1989) preent empirical evidence that the information generated by cloe bank relationhip i well valued by market participant. Beanko and hakor (1993) analyze the importance of relationhip banking for both lender and borrower and argue that the rent earned by bank from that relationhip may be an important incentive to avoid the moral hazard problem created by rik-inenitive depoit inurance. Boot (forthcoming) review the recent literature on relationhip banking and the empirical evidence on it value for borrower and lender. 8

9 ( ) l i > = (2.2), q where ( ) and are the interet rate charged to afe borrower in the corporate and middle market, repectively. ompetition in the corporate market enure that ( 1 + ) q 1 =, where i the bank borrowing rate. ondition (2.1) tate that a afe borrower will be intereted in acceing the corporate market only if the expected net benefit of the loan exceed the cot of informing the market about it type. ondition (2.2) ay that a afe borrower will go to the corporate market if the aving from paying a lower rate of interet rate in thi market egment more than compenate the additional cot of revealing her type to outide bank. hee imply eentially that the corporate market i the bet option for large afe borrower. afe borrower will chooe the middle market iff ( vi ) li, > li ( ) q, and li ( ) q q > v + <, that i i > (2.3), and liq q i ( ) = < l < (2.4), where = 1+ p q + ( 1 p ) q r 1 i the interet rate charged from borrower in the retail market, with p being the proportion of credit extended to afe borrower in that market egment. 21 and are, repectively, the lower and upper bound of the middle market and, becaue they depend on the menu of interet rate, are determined endogenouly. Finally, a afe borrower will operate in the retail market iff v i > (2.5) and l i < (2.6), In equilibrium, all riky borrower, independently of loan ize and expected return, will be in the retail market, ince it i alway the cae that 21 Note that becaue the average loan ize granted to riky borrower in the retail market will be larger than the average loan ize extended to afe borrower, p will be maller than the proportion of afe borrower in the retail market. 9

10 r 1+ = 1 q r r with if i being trictly lower than if p > hen, a riky firm or conumer will be willing to borrow v > r. herefore, the hare of loan extended to afe borrower in the retail market, p, i: p p(1 (1 + ) e ) = (2.7) 1 p(1 + ) e Under thee aumption, and conidering an univere of N potential borrower, the volume of credit extended in each of the three market egment i given by: orporate arket: iddle arket: ( ) = Npe q = Np ( ) ( + ) e ( + ) [ ( + ) ] [ ] [ ] e e e ) q etail arket: ( ) [ p( + ) e ] N = In equilibrium, bank make zero expected profit in the retail and corporate market, which are actuarially balanced, and derive all their economic profit from the middle market, where they earn (1 + ) q 1 for each dollar lent. hu, total expected profit i given by E ( Π ) = ((1+ ) q (1 + )) (2.8) he only deciion variable in thi model i the interet rate charged by bank to their (afe) client in the middle market. When fixing thi rate, bank have to take two effect into account. On the one hand, the higher, the more they earn from their middle -market client. On the other hand, the more they raie thi rate, the higher and the lower, o that le credit i extended through the middle market. aximizing (2.8) with repect to doe not yield a cloed form olution, but we can do ome exercie on comparative tatic uing numerical maximization. In able 2.2 to 2.5 we conduct uch an exercie, examining how the market behave when we change the bank borrowing cot (), the cot of acceing the corporate and middle market ( and ) and borrower diverity (q-qr). In all cae, unle otherwie tated, we fix = 15%, =$1, N=1,000, =0.8, =$0.1, =$0.01, p=0.6, q=0.98 and qr= Becaue only afe borrower acce the corporate and middle market, from now on we drop the upercript from interet rate in thee two market egment. 10

11 o ae the impact of egmentation we compare the reult of the model with two benchmark in which interbank aymmetry of information i not preent. he firt i the cae in which borrower type i public * extended to afe and riky borrower and the correponding information, o that the volume of credit ( ) market interet rate ( *) are, repectively, ( ) * j = p jn j (2.9), and * (1 + ) q 1 (2.10), j = j where j = and r tand for afe and riky borrower, repectively, p = p and p r = 1 p. With perfect information the average loan ize would be for both afe and riky borrower. he other benchmark i given by a ituation in which there i no way to determine borrower type i.e., there i interbank ymmetry of information, but aymmetry between borrower and bank. In thi cae the market interet rate and the volume of credit to afe and riky borrower would be ** * = ( 1+ ) q 1; (2.11), and ** ( ) ** j = p jn (2.12), ** where q = pq + (1 p) q. It i immediate to verify that r * q = q = q > q > q > q r, * ** * < < < r =, and that < <. It follow that if borrower type i not known by any bank, afe (riky) borrower are wore (better) off than if borrower type wa known to all bank. he intermediate ituation of a egmented credit market, a modeled above, produce loer and winner. arge borrower will tend to gain, particularly if and are mall, but riky and mall afe borrower will loe. Yet, even though default and interet rate are higher in the retail market than in the ret of the credit market, riky borrower will be better off with market egmentation than if complete ymmetry of information prevailed, and in thi ene interbank aymmetry of information create a cro ubidy from afe to riky borrower in the retail market. hi ubidy will increae with borrower diverity ( ) q r q and decline with and omparative tatic ompared to a ituation of perfect ymmetry of information, market egmentation penalize afe borrower in all three market egment (able 2.2). hoe in the corporate market pay the ymmetric-information interet rate for afe borrower, but have to pend to reveal their type, which i equivalent to an additional 2.8% pread (conidering the parameter value ued in thi exercie). Interet rate to borrower in the middle market are 4.6 percentage point above the perfect information rate of interet, with borrower incurring an information cot that i equivalent to a 1.4% pread. afe borrower in the retail market pay an interet rate mark-up that range between 24.6% and 28.6% depending on bank borrowing rate. iky borrower pay an interet rate only marginally lower than they would if perfect information prevailed. hee additional cot reduce the volume of credit extended to afe borrower between 7% and 9%, compared to 23 What i le intuitive i that afe borrower in the retail egment alo ubidize borrower in the middle market, by reducing and in thi way contraining the ability of the incumbent bank to extract rent. 11

12 the amount that would have been extended in the cae of perfect information ymmetry. A a conequence, default rate are alo higher than when borrower type i known to all bank. able 2.2: redit arket eaction to a eduction in the ot of Fund to Bank ariable Bank cot of capital () 25,0% 20,0% 15,0% 10,0% 7,5% With complete ymmetry of information Interet rate afe borrower 27,6% 22,4% 17,3% 12,2% 9,7% iky borrower 56,3% 50,0% 43,8% 37,5% 34,4% olume of credit ($) (1) 512,2 581,6 651,2 720,7 755,4 afe borrower 393,4 431,6 469,9 508,2 527,3 iky borrower 118,8 150,0 181,3 212,5 228,1 arket default rate 6,2% 6,6% 7,0% 7,3% 7,4% Without interbank aymmetry of information Interet rate (q ** = 9.2% in all cae) 37,7% 32,2% 26,7% 21,1% 18,4% olume of credit ($) (1) 529,2 598,0 666,9 735,7 770,1 afe borrower 317,5 358,8 400,1 441,4 462,1 iky borrower 211,7 239,2 266,7 294,3 308,0 With interbank aymmetry of information Interet rate orporate () 27,6% 22,4% 17,3% 12,2% 9,7% iddle () 32,0% 27,0% 21,9% 16,8% 14,3% etail () 56,2% 50,0% 43,7% 37,4% 34,3% iddle-market limit ($) Upper bound () 2,062 2,039 2,019 2,003 1,996 ower bound () 0,042 0,044 0,047 0,049 0,051 olume of redit ($) All borrower 475,7 545,0 614,3 683,7 718,4 afe borrower 356,7 394,8 432,8 470,9 490,0 orporate () 143,5 160,8 178,2 195,6 204,3 iddle () 213,1 233,7 254,4 275,0 285,3 etail () 0,2 0,2 0,3 0,4 0,4 iky borrower (in retail market) 119,0 150,2 181,5 212,8 228,4 Number of borrower afe borrower orporate iddle etail Avg loan ize (afe borrower, $) (2) orporate 3,654 3,561 3,486 3,425 3,399 iddle 0,736 0,733 0,731 0,729 0,729 etail 0,021 0,022 0,023 0,025 0,025 arket default rate (3) 6,50% 6,96% 7,32% 7,60% 7,72% Bank Overall Profit ($) 9,30 10,32 11,34 12,36 12,87 (1) Without interbank aymmetry of information, the average loan ize to both afe and riky borrower i equal to $1, o that the number of borrower in each cae i equal to the volume of credit extended. (2) Average loan ize for riky borrower in retail market i $1 in all cae. (3) he default rate in the retail market decline very lightly, from 19.98% when = 25.0% to 19.97% when = 7.5%. 12

13 ontrated to a ituation in which bank have equal ignorance about borrower type, market egmentation benefit borrower in the corporate and middle market, who are able to borrow at a lower cot, information cot included. iky borrower, in turn, are penalized with much higher rate of interet, a are mall afe borrower, for whom it i not worth to pay to let the market know it type. A a conequence, although with market egmentation more credit flow to afe borrower and le to riky one than when borrower type may not be inferred by bank, and a conequence the market default rate i lower, credit become concentrated on large borrower. A decline in the cot of capital to bank lower interet rate to both type of borrower in all ituation, but to a larger degree in the cae of riky borrower, when either market egmentation or perfect information prevail. A a conequence, a lower proportion of total credit goe to afe borrower, raiing in the market default rate, which however tay below the rate oberved when bank are equally ignorant about borrower type. 24 With market egmentation, all player benefit from a reduction in, with a rie in borrower urplu and bank profit. 25 An important concluion that follow from thee reult i that market egmentation doe not ubtantially affect the overall volume and the quality of credit. By treating debtor who do not want to pay to credibly reveal their type eentially a riky borrower, bank limit the volume of bad credit and keep the market relatively afe. he main negative conequence of market egmentation i then the high interet rate impoed on mall afe borrower. With the parameter value ued in able 2.2, approximately 30% of all credit i granted to borrower in the corporate market, who however repreent le than 10% of the total number of borrower (and between 4% and 6% of the univere of potential borrower). ike corporate borrower, all participant in the middle market are afe. But interet rate in thi market egment are coniderably highe r, with giving the rent extracted by bank from middle -market borrower. A remarkable reult i that thi rent actually rie when the cot of fund to bank decline, with pread dropping le in the middle than in the corporate, alo contract market. A a reult, the range of loan value granted through the middle market ( ) when come down, cauing the expanion in the volume of credit to be particularly pronounced in the retail market. he increae in lead to a rie in the average loan extended to afe borrower in the retail market, but ince mot of the expanion in credit goe to riky borrower, o doe the cro ubidy provided by afe borrower. he average loan ize extended to afe borrower i very different in the three market egment. A remarkable reult i that the average loan extended to riky borrower ( = $1), all in the retail market, i actually higher than that given to afe borrower in the middle market, while 40 to 50 time bigger than that granted to afe borrower in the retail egment. o ome extent, therefore, the difference between client in the retail and middle-market i more one of rik than of ize. Alo important i that, to diguie themelve, large riky borrower will tend to pread their loan requet through variou bank. hi may help to explain why, a hown in able 2.1, many mid-ized individual and commercial borrower operate with more than one bank Note that thi exercie abtract from the likely increae in q and qr a a reult of lower interet rate, which would contribute to reduce default rate. 25 Note, though, that ince credit volume rie more than profit the rate of profit per dollar lent come down. 26 hi may alo explain why, a hown by ole (1998), borrower who concentrate their financial tranaction on a ingle bank are more likely to obtain credit than firm with multiple ource of financial ervice. 13

14 An important parameter in our model i the cot of informing outide bank about one type ( ). he higher the value of, the more bank are able to extract rent from middle-market borrower, and a a conequence the higher and bank profit (able 2.3). An increae in and in, in turn, move both and upward, reducing the volume of credit extended through the corporate market, but actually increaing lending in the middle and retail market, even if the net effect i a decline in the total volume of lending. Becaue acce to the corporate and middle market become more expenive, proportionately more credit i extended to afe borrower in the retail market, lowering and attracting more riky borrower, in the end raiing the market default rate. herefore, although beneficial to bank, an increae in harm almot all borrower: (i) it move borrower in the lower end of the corporate market into the middle market, where they pay higher rate of interet; (ii) it allow bank to extract more rent from borrower in the middle market; and (iii) it move borrower in the lower fringe of the mid dle market into the retail market, where interet rate are the highet. till, afe borrower already in the retail market and riky borrower in general obtain a mall gain from an increae in, due to the marginal decline in. It follow, converely, that meaure that contribute to reduce the cot for outide bank to infer borrower type, uch a the adoption of better accounting practice, hould help to reduce default rate, expand the volume of credit and benefit mot borrower. till, inamuch a thi cot i due more to tax evaion than to poor accounting, the capacity of public policy to reduce the information capture of middle -market borrower i in a ene rather limited. able 2.3: eaction to hange in ot of Acce to orporate arket (with interbank aymmetric information) ariable ot of acce to corporate market (, 0,015 0,040 0,080 0,100 0,150 in $) Interet rate in iddle arket () 17,6% 19,0% 21,0% 21,9% 24,1% iddle-market limit ($) Upper bound ( ) 1,841 1,893 1,977 2,019 2,125 ower bound ( ) 0,039 0,041 0,045 0,047 0,052 olume of redit ($) All borrower 642,1 633,5 620,6 614,3 599,6 afe borrower 460,6 452,1 439,1 432,8 418,0 orporate () 210,0 200,1 185,2 178,2 161,6 iddle () 250,5 251,8 253,6 254,4 256,1 etail () 0,2 0,2 0,3 0,3 0,4 iky borrower in retail market 181,4 181,5 181,5 181,5 181,6 Number of afe borrower orporate iddle etail Avg loan ize (afe borrower, $) (1) orporate 2,896 3,051 3,329 3,486 3,943 iddle 0,686 0,699 0,720 0,731 0,756 etail 0,019 0,020 0,022 0,023 0,026 arket default rate (2) 7,09% 7,16% 7,26% 7,32% 7,45% Bank Overall Profit ($) 0,68 3,99 8,98 11,34 16,87 (1) Average loan ize for riky borrower in retail market i $1 in all cae. (2) he default rate in the retail market decline very lightly, from 19.98% when = $0.015 to 19.96% when = $

15 A reduction in the value of, the cot of informing the inide bank of one type, attract afe borrower in the upper fringe of the retail market to the middle market, hifting upward the demand for middle-market loan, and conequently allowing for a rie in and in profit (able 2.4). Depite facing higher interet rate, mot borrower in the middle market experience a net gain from thi proce, with the only loer being thoe in the upper end of thi market egment, including thoe that a a conequence move to the corporate market. With fewer and only the very mall afe borrower remaining in the retail market, goe up, lowering the demand for credit of riky borrower, which in turn caue the market default rate to decline. With a low, the retail market will eentially erve only riky borrower. able 2.4: eaction to hange in ot of Acce to iddle arket (aymmetric information) ariable ot of Acce to iddle arket () 0,020 0,010 0,005 0,001 Interet rate iddle ( ) 21,3% 21,9% 22,2% 22,4% etail ( ) 43,6% 43,7% 43,7% 43,7% iddle-market limit ($) Upper bound ( ) 2,041 2,019 2,009 2,002 ower bound ( ) 0,092 0,047 0,024 0,005 olume of redit ($) All borrower 611,4 614,3 616,1 617,7 afe borrower 429,2 432,8 434,8 436,4 orporate () 175,7 178,2 179,4 180,2 iddle () 252,4 254,4 255,4 256,2 etail () 1,1 0,3 0,1 0,0 iky borrower in retail market 182,2 181,5 181,3 181,3 Number of afe borrower orporate iddle etail Avg loan ize (afe borrower, $) (1) orporate 3,512 3,486 3,475 3,466 iddle 0,774 0,731 0,709 0,691 etail 0,045 0,023 0,012 0,002 arket default rate (2) 7,36% 7,32% 7,30% 7,28% Overall redit arket Bank Overall Profit ($) 9,89 11,34 12,07 12,67 (1) Average loan ize for riky borrower in retail market i $1 in all cae. (2) he default rate in the retail market rie very lightly, from 19.89% when = $0.02 to 20.00% when = $ In able 2.5 we ee that borrower diverity ( q q r ) i good for bank. If riky borrower become le o, a in the econd column of able 2.5, bank have to bring pread down in the middle market, and although thi contribute to raie lending volume, profit and epecially profit rate come down. Interet rate drop very ubtantially in the retail market, cauing the number of borrower and volume of credit to boom. If the decline in borrower diverity arie from an increae in the default rate of afe borrower (column 3), profit fall even more, a a reult of lower pread and volume of credit in the middle market. he retail market alo become relatively more attractive in thi cae, with an increae in and a decline in. 15