Bank Lending Channel in Brazil: Evidence from the Supply of Bank Loans and from the Composition of External Finance of Corporations

Transcription

1 Bank Lending Channel in Brazil: Evidence from he Supply of Bank Loans and from he Composiion of Exernal Finance of Corporaions Fernando Nascimeno de Oliveira 12 (Cenral Bank of Brazil and IBMEC/RJ) 19/09/2007 ABSTRACT This paper analyses he exisence of an acive bank-lending channel in he ransmission of moneary policy in Brazil. We look a he reacions of banks and corporaions o moneary conracions. Our resuls indicae ha small banks and small firms behave very much like he bank-lending channel would predic. We find ha afer a moneary conracion, here is a shif of bank loans financing from small firms o large firms. However, our empirical evidence also indicaes ha in he aggregae he bank-lending ransmission mechanism of moneary policy does no seem o be relevan. Key Words: Bank Lending Channel, Credi Channel, Moneary Transmission. Mechanisms 1 Research Deparmen Rio de Janeiro 2 The ideas expressed in his paper in no way reflec he official opinions of he Cenral Bank of Brazil on hese subjecs

2 1 INTRODUCTION Mishkin (1996) assers ha radiional ransmission mechanisms of Moneary Policy are no fully capable of explaining how economic agens respond o moneary policy. Oher channels are necessary o complemen he explanaions given by he more radiional channels. One of hese oher mechanisms is he bank-lending channel. This heory focus on fricions in he credi marke, caused by problems of asymmeric informaion relaed o he behavior of financial insiuions and corporaions ha have direc impac over he ransmission of moneary policy. Brazil is known for relevan imperfecions in he credi marke. Some of hese imperfecions are relaed o a high cos capial and o an underdeveloped capial marke. Therefore, i seems reasonable o believe ha problems of adverse selecion and moral hazard beween financial insiuions and firms may be imporan o undersand how he moneary ransmission mechanism works in Brazil. The objecive of his paper is o evaluae he exisence and relevance of he bank-lending channel in Brazil. Wih his purpose, we look a he reacion of boh banks and corporaions o moneary conracions. We are paricularly ineresed in comparing he reacions of small banks wih hose of large banks as well as comparing he reacions of small firms o hose of large firms. To classify banks in small or large we use monhly balance shee daa and o classify corporaions in small or large we use quarerly balance shee daa. Our sample period goes from July 1994 o December To idenify he moneary conracions we use wo insrumens: he firs difference in he SELIC ineres rae and he reading of he minues of COPOM. 4 The resuls of he paper indicae ha small banks and small firms reac in accordance wih he bank-lending heory. Small banks respond o a moneary conracion decreasing heir supply of loans and small corporaions also respond decreasing heir financing hrough bank loans. 3 The difference in he frequency of he daa is due o is availabiliy in he case of corporaions and banks 4 SELIC is he basic ineres rae in Brazil. I is he ineres rae of he open marke operaions. COPOM sands for Comiê de Políica Moneária. I is he body of he Cenral Bank of Brazil ha decides he value of SELIC unil he nex meeing 2

3 We also observe ha afer a moneary conracion here is a shif of bank loans financing from small corporaions o large corporaions. However, our empirical evidence indicaes ha in he aggregae he bank-lending mechanism of moneary policy is no effecive. Large banks do no experience a decrease in heir supply of loans and large firms financing decisions also are no affeced. Considering ha large banks in Brazil have a share of more han 85% of he asses of he financial sysem and ha large firms have as much as 70% of oal asses of corporaions, hese resuls cas a doub on he overall significance of he bank-lending heory o explain how moneary shocks affec he real secor of he economy. The bank lending heory is par of a much more exensive heory know as he credi channel heory. This heory poins o oher forms of ransmission mechanisms ha enhance he more radiional ones. I assers herefore ha he radiional Keynesian view does no compleely explain he mechanism of Moneary Policy ransmission. The iming and magniude of he effecs of he Moneary Policy are no saisfacorily explained by he convenional heory. The oher componen of he credi channel is balance shee heory. The balance-shee channel emphasizes he impac of Moneary Policy shocks on he balance shee of he firms. This view affirms ha imperfecions in he credi marke, as asymmeric informaion and moral hazard problems, may deeriorae he balance shee of firms, reducing heir abiliy o obain exernal financing. The reducion in he credi for he firms diminishes heir invesmen level and, consequenly heir oupu ha can lead o a decrease in aggregae demand. The bank-lending channel suggess he exisence of a ransmission mechanism of he Moneary Policy hrough which shifs in he bank loans supply impac he real economy. By his channel, conracioniss moneary shocks are able o reduce demand deposis of banks making hem decrease heir supply of loans o he privae secor. The reducion in bank loans limis he invesmen of corporaions ha are mosly dependen of his form of financing. To be prevalen in an economy, he bank lending channel has o saisfy hree necessary condiions: here are no perfec subsiues for demand deposis; banks decrease heir supply 3

4 of bank loans afer a moneary conracion; and finally corporaions observe a decrease in heir volume of loans ha affec heir privae invesmen and herefore he aggregae invesmen of he economy. We show in his paper ha hese condiions are me for small banks and small corporaions bu he same does no happen for large banks and large corporaions. One possible explanaion for he case of large banks is ha he level of concenraion of he banking secor makes easier for hem o find good subsiues for he loss of demand deposis. In he case of large corporaions, possible explanaions are ha hey have more access o he exernal credi marke and also more access o boh shor and long erm financing from Banco Nacional de Desenvolvimeno Econômico e Social BNDES - an official developmen bank. The srucure of he res of his paper is as follow. In secion 2 we survey he lieraure. In secion 3, we describe he daa we laer use in our empirical analysis, boh of banks and corporaions. In secion 4, we perform he empirical analysis relaed o reacion of banks o moneary conracions. In secion 5, we presen he empirical analysis relaed o he reacion of corporaions o moneary conracions. In secion 6, we presen he main conclusions. 2 SURVEY The lieraure on he ransmission mechanisms of he Moneary Policy developed very much in he las years. During more han fify years, i was believed ha he Moneary Policy was ransmied hrough he long run real ineres rae. Tha moneary ransmission mechanism, known as ineres rae channel, is based in Keynes and can be represened by he IS-LM exbook model. By he ineres rae channel, an increase in shor run nominal ineres rae implicaes, hrough he price rigidiy and agens expecaions, an increase in he long run real ineres rae. Tha causes an increase in he capial cos and he consequen reducion in he consumpion and invesmen levels of he economy. Recenly, Taylor (1995) idenified a channel ha reaffirms he imporance of he ineres rae mechanism, known as exchange rae channel. This channel assers ha he ineres rae affecs 4

5 he real economy no only hrough he capial cos, bu also hrough variaions in he exchange rae. Tha occurs because an increase in he real ineres rae aracs foreign invesmens, ha appreciaes he domesic currency. Tha appreciaion of he exchange rae causes negaive pressures in he rade balance, reducing he growh rae of he economy. Mishkin (1996) idenifies anoher approach o he moneary ransmission mechanism, he monearis one. The moneariss argue ha, insead of jus analyzing he ineres rae, he prices of oher asses of he economy should be noiced. In his way, moneariss sugges a ransmission mechanism by which he relaive prices of oher asses and he real wealh of individuals ransmi moneary effecs o he real economy. We highligh wo channels of Moneary Policy ransmission idenified by he moneariss: Tobin s q Theory of Invesmen and wealh effecs on consumpion. The predominance of he ineres rae channel for so many years was due, parly, o he Modigliani-Miller Theorem (1958). Modigliani and Miller showed ha he invesmen decisions of firms do no depend on heir capial srucure. As consequence, sudies on he role of credi in economy were discouraged. In ha way, for several years, banks were only considered imporan in he Moneary Policy ransmission for heir power o creae fiduciary currency. In he sevenies, however, informaion economiss highlighed he imporance of he fricions in he credi marke. By hose auhors, problems of asymmeric informaion and ransacion coss generae agency problems in he finance markes. As a consequence, he validiy of Modigliani-Miller Theorem (1958) was quesioned and he special role of he bank loans as financing form was acceped. In ha conex, he credi channel heory emerged as an alernaive mean of undersanding he moneary ransmission. The credi channel heory, as defined by Bernanke and Gerler (1995), foresees he exisence of channels of Moneary Policy ransmission ha amplify and spread he effecs of he ineres rae on he economy. In conras o ha heory, he credi channel esablishes ha he real ineres rae does no compleely explain he ransmission of ha policy. Facors as he iming and he magniude of he effecs of he Moneary Policy are no saisfacorily explained by he radiional heory. 5

6 The credi view saes ha marke fricions creae agency problems in he financial markes. As consequence, here is a difference beween he cos of he funds obained exernally (hrough he issuance of sock and bank loans) and he opporuniy cos of he funds generaed inernally (profis reenion). Tha difference is known as exernal finance premium and shows how he credi marke imperfecions creae a discrepancy beween he expeced reurn of he lenders and he coss incurred by he borrowers (Bernanke and Gerler, 1995). By he credi channel heory, a moneary shock by changing he ineres rae, affecs he exernal finance premium, amplifying he effecs of moneary shocks. This complemenary movemen of he exernal finance premium helps o explain he magniude and he iming of he Moneary Policy effecs in addiion wih he variaion of he real ineres rae. The credi channel can be divided in wo disinc heories: he balance-shee channel and he bank lending channel. 2.1 The Balance-Shee Channel The balance-shee channel is based on he hypohesis ha he exernal finance premium of a firm depends on is financial posiion. In his way, as he financial posiion of he firm affecs is credi erms, a decrease in he qualiy of is balance shee impac heir invesmen decisions as in Oliner and Rudenbusch (1996). If he balance-shee channel is acive, moneary policy shocks affec no only he ineres rae of he marke, bu also he financial posiion of he firms. As Bernanke and Gerler (1995) explain a moneary conracion ( M ) deerioraes he balance shee of he firms in a leas wo ways. Firs, an increase in he ineres rae ( i ) increases he expenses of he companies wih credi ineres, reducing is cash flow and weakening is financial posiion. Tha reduces he firm s capaciy of inernal financing. Besides, increases in he ineres rae end o decrease he marke value of he asses of he firm, ha reduces is collaerals. The deerioraed balance shee of he firm reduces is condiions of offering collaerals o poenial lenders, increasing he adverse selecion problem. A smaller equiy posiion also resuls in moral hazard problems, as he managers of he firms have incenives o engage in 6

7 riskier projecs. Amplifying he problems of adverse selecion and moral hazard deerioraes he balance shee of he firm, wha cause an increase in he exernal financing coss incurred by i ( EFP ). Concluding, he firm needs o appeal o exernal financing when i is more expensive or when is exernal finance premium is elevaed. Tha reduces he invesmen level of he economy ( I ) and, consequenly, is oupu ( Y ). Tha effec is responsible for amplifying and spreading he effecs of an increase in he ineres rae. Tha is illusraed below: M i cash flow & collaeral adverse selecion & moral hazard EFP I Y I is imporan o highligh ha he balance shee channel may or no exis for some agens in he economy. Tha happens because he exernal finance premium depends on he financial posiion of each firm individually. Firms wih smaller access o alernaive financing forms, as he capial marke, should be more affeced by Moneary Policy shocks (Hubbard, 1995). 2.2 The Bank Lending Channel The bank-lending channel assers he exisence of a Moneary Policy ransmission mechanism by which changes in he bank lending level affec he economy. Several auhors quesion he asymmeric reamen given by he radiional view of moneary ransmission o currency and credi 5. While he currency receives a special saus, he bank loans, bonds and oher financing forms are concenraed a a "bonds marke", ruled jus by an ineres rae (Bernanke and Blinder, 1988). Nowadays, he lieraure on he Moneary Policy ransmission has been emphasizing he financial inermediary's imporance in he provision of he credi. This view assers ha banks and oher credi insiuions presen a special role in he economy. This occurs because hese insiuions specialize in collecing informaion and monioring he performance of he borrowers. Besides, hey have he capaciy o esablish a long run relaionship wih heir 5 See Alfaro e al (2004), Bernanke and Blinder (1988), Bernanke and Gerler (1995), Kashyap and Sein (1994, 1995). 7

8 cusomers. These characerisics guaranee o banks comparaive advanages when working wih asymmeric informaion problems. Consequenly, he bank's abiliy o finance aciviies ha would no be financed a he capial marke gives he banks loans a special saus in he credi marke (Kashyap and Sein, 1994). In his conex, he bank lending channel arose as a new channel of Moneary Policy ransmission. By he bank-lending channel, he economy can be modeled as being composed by hree asses - currency, bonds and bank loans - which differ considerably amongs hem and should be analyzed individually when observing he impac of Moneary Policy shocks. In his conex, banks can be special in he moneary ransmission in wo ways: in addiion o he creaion of fiduciary currency as indicaed by he radiional view, he banks lend. Thus, in conras wih he ineres rae channel, he Moneary Policy canno only be ransmied by he impac in he real ineres rae, bu also hrough an independen impac in he bank lending level. For insance, even if a moneary conracion has reduced effec on he real ineres rae, i can affec he spread beween loans and bonds significanly. This affecs he invesmen decisions of he firms ha depend on bank loans, which impac he oupu of he economy (Kashyap and Sein, 1995). The bank-lending channel indicaes ha he Moneary Policy ransmission works hrough he following mechanism. In he case of a conracionis moneary shock ( M ), he reducion in he economy's reserves level diminishes he supply of deposis (insured financing form) o banks. These suffer a cu in heir financing sources and have o reduce he supply of loans. The reducion in he supply of bank credi increases he exernal finance premium of he firms dependen o ha financing form. Consequenly, he invesmen level of hose companies is reduced ( I ), wha affecs negaively he growh of he economy ( Y ) (Kashyap and Sein, 1994). M deposis bank lending I Y 8

9 Bernanke and Blinder (1988) idenify wo necessary condiions o he exisence of he bank lending channel 6. Firs, here mus be firms dependen o bank loans, which are unable o subsiue hem by oher financing forms wihou coss. Besides, he Cenral Bank has o be capable o modify he supply of bank loans hrough he Moneary Policy. In agreemen wih Kashyap and Sein (1994), he validiy of he firs condiion requess ha firms do no face bank loans and oher financing forms as perfec subsiues. In oher words, he Modigliani-Miller Theorem (1958) has o fail a leas for some firms, so ha hose are unable o compensae a reducion in he bank loans supply for oher financing form (issuing promissory noes, debenures ec). Thus, a reducion in he bank lending level leads some firms o incur in coss associaed o he esablishmen of a new credi relaionship. In oher words, here is lock-in in he credi marke. To es he firs condiion, we can analyze he balance shee of he firms of an area and he percenage of heir hird paries financing ha comes from bank loans. Bonomo and Graminho (2002) made ha es for Brazil in he year of They used daa of 224 companies exraced from Economaica. The auhors found ha, on average, he oal financial deb of he companies was, in 2000, 41% of ouside financing and 22% of he oal financing. Besides, 20% of he analyzed companies presened more han 63% of ouside financing in bank loans. These resuls sugges ha, in Brazil, a considerable number of companies are dependen of bank lending. The condiion ha he Cenral Bank has o be able o modify he bank credi supply hrough he Moneary Policy is more conroversial. This is because when promoing a conracionis moneary shock, he Cenral Bank compromises he bank abiliy o generae insured financing forms (as demand deposis). However, i does no resric he use of uninsured financing forms, as he issue of Cerificaes of Deposi (CD's). Thus, for he second condiion o be valid, he banks canno consider insured deposis and oher financing forms as perfec subsiues, since ha would allow hem o compensae he deposis reducion compleely. 6 A hird condiion is, like in every non-neuraliy of money model, ha here mus be some ype of price rigidiy (Kashyap and Sein, 1994). 9

10 Several auhors argue ha he proposiion above is heoreically dubious. Romer and Romer (1990), for insance, believe ha he banks can always be financed, wihou addiional coss, hrough uninsured financing forms. Indeed, even if a conracionis Moneary Policy shock diminishes he amoun of available deposis o he banks, hey are able o compensae ha reducion hrough he emission of CD's, promissory noes or oher securiies. Tha argumen is an applicaion of he Modigliani-Miller Theorem (1958) for he banks. For he bank-lending channel o be acive, one mus prove ha marke fricions, relaed asymmeric informaion and agency problems, do no allow ha a reducion in he banks deposis o be compleely compensaed by oher financing forms. In oher words, he heorem of Modigliani-Miller (1958) canno be valid for he bank firm. Concluding, he bank-lending channel depends on he flexibiliy degree of he financial marke. The more flexible (guaraneeing financing forms similar o insured deposis), less relevan he bank lending channel is in an economy. Kashyap and Sein (1994) es he validiy of Modigliani-Miller Theorem (1958) for banks. In oher words, hey es if here are fricions in he marke of uninsured exernal financing. If we ake as example he issue of CD's, which are no insured by he governmen, we noiced ha he qualiy of he issuing bank is imporan in he invesor's decisions. This occurs because here is some degree of asymmeric informaion beween he issuing bank and he invesor so ha problems of adverse selecion are relevan. These consideraions allow cross-secional differences in he cos of uninsured exernal financing beween banks of differen characerisics. Kashyap and Sein (1994) conclude ha, as in he case of he no-bank firms, small banks incur in larger coss when subsiuing beween differen financing forms han he big ones. This resul suggess ha he Modigliani-Miller Theorem (1958) is no valid for he bank firm. The banks can relax he impac he reducion in he insured deposis sock on is loan supply also by selling liquid asses (Kashyap and Sein, 1995). This occurs because he banks can compensae he decrease in heir liabiliy side hrough he reducion of he share of heir asse ha is more liquid, proecing heir loan porfolio. This sraegy is jusified by he rade-off beween liquidiy and expeced reurn of he bank. While loans are less liquid and have larger reurn, bonds and securiies are more liquid and have less reurn. Though, ha sraegy is only 10

11 possible for banks ha are more liquid (grea proporion of ( Bonds + Securiies) Toal Asse ), since a very low sock of liquid asses can represen risks o he bank. Empirical sudies ha ry o es he bank lending channel use, in general, wo mehodologies: he sudy of aggregae daa or micro daa. An example ha uses aggregae daa is Bernanke and Blinder (1992). These auhors find evidences ha a conracionis moneary shock, represened by an increase in Federal Funds Rae 7, is followed by he reducion in he aggregae level of bank loans. Such a resul is consisen wih he exisence of he bank lending channel, bu i also admis anoher inerpreaion. The reducion in bank lending level can be caused by a diminuion in he demand 8 for loans, and no in he supply of bank credi. Trying o solve he idenificaion problem found by Bernanke and Blinder (1992), Kashyap e al. (1993) observe how an imporan subsiue o he bank loans, he promissory noes, behave during a moneary conracion. The auhors jusify ha analysis by he following inuiion. Suppose ha he Moneary Policy operaes only hrough he ineres rae channel and ha he reducion in he bank lending observed during a moneary conracion is consequence jus of he reducion in he demand for credi. If ha is rue, he demand for nobank financing forms should also decrease, wha would cause a reducion in he level of promissory noes issued. However, if he credi channel is acive, a moneary conracion should reduce he supply of bank loans. In ha case, promissory noes issue should increase. This indicaes ha he companies have some abiliy in subsiuing heir financing sources and ha he demand for credi does no explain he decrease in loans level. Using aggregae daa, Kashyap e al. (1993) find ha a moneary conracion changes he composiion of he exernal financing of he firms. There is an increase in he promissory noes issuance and a reducion in he amoun of bank loans. This resul implies ha here was a reducion in he supply of bank loans, no in he demand for hose, wha is consisen wih he exisence of he bank-lending channel. Besides, he auhors observed ha he changes in 7 For a deailed discussion abou Moneary Policy insrumens, see Bernanke and Mihov (1998). 8 The demand effec on he bank loans represens he naural reducion in he demand for hose resuling from an increase in he ineres rae afer a moneary conracion. Tha effec is foreseen by he radiional Moneary Policy ransmission channel. The effec ha he bank lending channel auhors are ineresed is an independen shock of Moneary Policy over he supply of bank loans. 11

12 he exernal financing composiion of he firms affeced he invesmen level. Tha resul implicaes ha he bank loans and oher forms of exernal financing are no perfec subsiues. Oher auhors, in conras, argue ha he resuls found by Kashyap e al. (1993) can be explained by he fac ha, in a recession, big companies presen beer performance han he small ones, wha would make he bigges o demand more credi. Besides, hose companies are he ones which issue more promissory noes, wha explains he increase in ha finance form wihou proving ha here was a reducion in he supply of loans. Tha idenificaion problem occurs in sudies wih aggregae daa because hey do no consider individual characerisics of he agens, wha made ha pracice broadly criicized 9. The use of micro daa permis he analysis of he disribuive effecs of he Moneary Policy. In oher words, during a moneary shock, i is possible o observe differences of behavior among banks wih disinc characerisics. This analysis allows us o solve he idenificaion problem beween supply and demand for bank loans found by sudies wih aggregae daa. Indeed, several researches have been using ha mehodology o es he bank lending channel in disinc counries. The main conclusion of Kashyap and Sein (2000) is ha, for he small banks, Moneary Policy shocks affec more he loan level of he less liquid banks. This occurs because a moneary conracion increases he liquidiy consrains of he smaller banks, which, for hypohesis, are no indifferen o insured and uninsured financing sources. Therefore, he banks wih more liquid balance shees are able o proec heir loan porfolio, by selling heir liquid asses sock. This resul confirms he main hypohesis of Kashyap and Sein and he exisence of he bank lending channel in he Unied Saes during he sudied period. For he Brazilian economy here are some empirical papers in he lieraure ha also use micro daa. Graminho and Bonomo (2002) and Takeda and Nakane (2005) analyze he exisence and relevance of he bank lending channel in Brazil. Takeda and Nakane (2005) use monhly micro daa of Brazilian commercial banks balance shee from December 1994 o December These auhors consider as Moneary Policy insrumens he shor run ineres rae and 9 See Bernanke and Gerler (1995), Hubbard (1995), Kashyap and Sein (2000). 12

13 he reserve requiremens on bank deposis. The resuls found by Takeda and Nakane (2005) sugges ha he impac of he reserve requiremens is relevan and sronger on big banks' loans. Tha can be explained by he progressive naure of he reserve requiremens, which affecs more srongly banks wih larger levels of deposis. Shocks in he ineres rae, in conras, seem no o affec he bank loans level. Concluding, Takeda and Nakane (2005) find evidences ha he bank lending channel was acive in Brazil during he sudied period. Graminho and Bonomo (2002) es he exisence of he bank-lending channel in Brazil based in he mehodology of Kashyap and Sein (2000), explained above. These auhors conclude ha, unlike he expeced, posiive shocks in he ineres rae relax he banks liquidiy consrain. By Graminho and Bonomo (2002), his resul suggess ha increases in he ineres rae can be beneficial o he Brazilian banks. This occurs because an increase in he banks profis, caused by he increase of he ineres rae, raises he inernal financing capaciy of hose banks, reducing heir dependence of deposis as source of financing. Tha resul does no corroborae he exisence of a bank lending channel in Brazil. 3. Daa We divide our descripion of he daa in hree pars. In he firs par, we idenify he moneary shocks. In he second par, we describe he daa of banks showing how we classify hem in small and large. Finally, in he hird par, we presen he daa of he corporaions and how we classify hem in small and large. 3.1 Measures of Moneary Conracions In he firs place, we explain how we define a moneary conracion. A prerequisie for all our ess is a good indicaor of moneary policy. However as Bernanke and Mihov (1998) poin ou here is no consensus in he lieraure as o he bes indicaor of moneary sance. We decide o use wo measures of moneary policy 10. Our firs measure is he SELIC rae. Our second measure is he Boshen-Mills (1995). 10 Bernanke and Mihov (1998) propose anoher form of idenifying moneary shocks, in paricular moneary conracions. They build a flexible VAR model ha ness previous VARs based on more specific assumpions 13

14 Bernanke and Blinder (1993) advocae ha he ineres rae se by he Cenral Bank in is open marke operaions is a good indicaor of moneary policy excep in periods where he ineres is very volaile, which was no he case in Brazil in our sample period (ha goes from he hird quarer of 1999 o he hird quarer of 2005). We use he quarerly series of he effecive SELIC rae. SELIC rae is a nominal ineres rae ha he Cenral Bank of Brazil ses as is arge in open marke operaions. We consider his series more relevan o characerize moneary conracions han he real SELIC rae because he laer depends on expecaions on inflaion. Only recenly, have expecaions of inflaion of he privae agens became public. If we use his series herefore his would hamper our capaciy o perform empirical ess. We define a moneary conracion by looking a he firs difference of SELIC. A moneary conracion occurs in he quarer in which we observe ha he modulus of he firs difference of he SELIC is greaer han he mean of he series plus one sandard deviaion. Using his crieria, we observe 4 moneary conracions, as Table 1 Panel A shows. They occur in he following quarers: firs quarer of 1995 (March), fourh quarer of 1997 (November), hird quarer of 1998 (Sepember) and firs quarer of 1999 (March). Our second mehodology of idenifying moneary is relaed o he Boshen-Mills (1995) index. Boshen and Mills read he FOMC documens and classify moneary conracions in five caegories: srongly expansionary, mildly expansionary, neural, mildly conracionary, and srongly conracionary. The classificaion is based on relaive weighs hey perceived he FED pu on he shor erm radeoff beween inflaion agains unemploymen. To build Boshen-Mills (1995) index for Brazil we read all COPOM documens since is creaion and for each documen classified moneary policy in one of he five caegories menioned above. Panel B of Table 1 deails he resuls of our classificaion. We idenify 3 COPOM meeing ha can be caegorized as srongly conracionary. These meeings were in abou FED s moneary policy, such as funds rae arge, and non-borrowed reserves arge. The mehodology is useful for calculaing high frequency moneary shocks or as indicaor of he overall sance of moneary policy. 14

15 he fourh quarer of 1998 (December), he firs quarer of 1999 (March) and in he fourh quarer of 2002 (Ocober). 3.2 Banks The banks were divided in hree size classes, based on he average monhly asses in he sudied period. The banks ha are locaed above he percenile 95 of he sample (15 banks) were considered large; he ones ha are locaed among he perceniles 75 and 95 (58 banks), medium; and he ohers were considered small (220 banks). This classificaion was based in he grea concenraion of he Brazilian banks marke, as can be observed in Table 2. The observaion of he main accouns of he banks balance shee allows us o draw some conclusions on how he Brazilian banks behavior since he Real Plan. The main purpose of his analysis is o observe how he banks have been financing heir credi operaions in our sample. This is imporan because if here is an acive bank lending channel, he Cenral Bank can affec he supply of bank credi by inducing a conracionis moneary shock. However, for his o be possible, he banks canno easily subsiue he demand deposis as financing form of heir loans. In his secion, he daa are analyzed in four caegories. The complee sample and each size caegory of banks are observed separaely. The daa are sudied in level and in firs difference. We observed he following balance shee indicaors: in he liabiliy side, demand deposis and "oher financing forms" 11 ; in he asses, freely allocaed bank credi 12 and a balance shee liquidiy indicaor 13. The use of he freely allocaed bank credi, insead of he oal credi, is jusified for he compulsory allocaion of some credi modaliies in Brazil, as housing and rural loans. This makes hese credi modaliies less sensiive o Moneary Policy (Cenral Bank of Brazil, 2002). 11 The balance "oher financing forms" resuls from he sum of he following accouns of COSIF: ( ) "Saving Deposis", ( ) "Inerfinance Deposis", ( ) Time Deposis and ( ) "Obligaions for Compromised Operaions" excluded he accoun ( ) Third Pary Porfolios. 12 "Freely Allocaed Bank Credi" corresponds o he oal credi supply excluded from he compulsory allocaed credi modaliies in Brazil. 13 The variable "Liquidiy" was creaed in he following way: [( ) Inerfinance Liquid Applicaions + ( ) Tiles and Valores Mobiliários]/[( ) Toal Asses - ( ) Compensaion]. 15

16 Figure 1 presens some ime series evoluion of balance shee characerisics of he complee sample. As can be observed, since he beginning of he Real Plan, he freely allocaed bank credi was always higher han he demand deposis. This indicaes ha he Brazilian banks consider he use of oher financing forms o finance heir loans. Besides, he growh of oher forms of financing was larger han he growh of demand deposis in he las years of he sample 14. Anoher relevan fac is he higher increasing rae of he freely allocaed bank credi since middles of Tha increase was no followed by he demand deposis, wha indicaes ha he banks have been using oher financing forms more acively. In relaion o he banks asses, i is imporan o observe he balance shee liquidiy. The evoluion of ha variable since he beginning of he Real Plan can be analyzed in Figure 1. As can be observed, he banks liquidiy suffered considerable increase from July 1994 o December Tha guaranees o he banks larger margin of proecion o heir loan porfolio in moneary conracion periods. Figure 2 presens balance shee characerisics of he banks considered large. As can be observed, he balance shee of hese banks is similar o he one of he complee sample. On he liabiliy side, here was considerable increase in oher financing forms, wha was no followed by he demand deposis. In he asses, he liquidiy of he big banks also suffered noable increase, overcoming he average of he sample. The evoluion of he balance shee characerisics of small banks is presened in Figure 3. As we can observe, he demand deposis represen a small share of he financing alernaives of hose banks. In relaion o liquidiy, hose banks presen he smalles proporion of ne asses among he hree caegories. Table 3 presens he correlaion marix among freely allocaed bank credi, demand deposis and oher financing forms for each size caegory. The correlaion beween oher financing forms and freely allocaed bank credi is larger in all caegories if compared o demand deposis and freely allocaed bank credi. Besides, in he case of he small banks, he 14 The accoun ( ) "Demand Deposis" added o he balance of "Oher Deposis" represened, in December of 2005, more han 94% of he oal of deposis of he Brazilian financial sysem, wha shows ha hose are he more relevan modaliies of deposis. 16

17 difference among hose correlaions is paricularly high. This suggess ha he small banks are more dependen of oher financing forms han of he demand deposis o finance heir loans. Anoher ineresing characerisic is ha, unlike expeced, he correlaion beween demand deposis and oher financing forms for he small banks is low. Tha resul suggess ha hose banks use oher financing forms as subsiues o demand deposis. The analysis of he daa in level indicaes ha he Brazilian banks, including he small ones, are no dependen of he demand deposis as a financing form of heir loans. Tha iniial resul does no confirm he exisence of a bank-lending channel in Brazil. This occurs because he Moneary Policy affecs he real economy by reducing he bank loans supply, which resuls from he falling in he demand deposis availabiliy o banks afer a moneary conracion. Tha resul, however, i is no conclusive. Oher analyses are necessary o es he exisence of an acive bank-lending channel in Brazil. 3.3 Classifying Firms in Large or Small We ake size, measured by oal asses, as our classificaion crieria following Gerler and Gilchris (1994). We observe ha size is highly correlaed wih oher financial variables ha indicae he capaciy firms have o access he financial markes. We classify corporaions in small and large. We will show ha our small corporaions have relaively less access o he financial markes han large corporaions. Our ineres in separaing firms in large and small ones is ha, as Gerler and Gilchris (1994) poin ou, is ha by doing his we can infer he level of access o he financial markes of he corporaions. In heory, small firms will depend much more on bank loans han large firms. The laer will also issue much more shor and long erm commercial paper and have much more access o capial markes, issuing more ordinary and preferred socks. Our classificaion scheme of small and large firms is he following. Our sample period sars in he hird quarer of 1994 and ends up in he fourh quarer of In he firs place, we exclude from our sample of public corporaions in Brazil financial insiuions. We keep rack of he firms whose financial saemens are no available in all periods, because here were no public firms ye or because hey closed heir capial, or because here was a akeover or fusion 17

18 or even because hey wen bankrup during our sample period. Our daabase is composed of 291 corporaions. Laer on, in our empirical analysis we will sudy he reacion of hree variables o moneary conracions. The variables are growh raes of invenories, growh raes of ne operaional revenues and growh raes of shor-erm deb. These variables are as Gerler and Gilchris (1994) sress he mos imporan variables o idenify he balance shee channel. We are assuming ha size of firms, which is he crieria we use o selec our sample, is independen of hese growh raes. This assumpion guaranees ha our seleced sample is unbiased. We consider a possible candidae for being small, a firm whose logarihm of oal asses is less or equal o he percenile 30 of he disribuion of oal asses in all quarers. In a similar fashion, we consider a possible candidae for being a large firm, one whose logarihm of oal asses is greaer or equal o he percenile 70 in all quarers. To choose he small firms, we consider hose ha we consider o be small in all quarers. By doing his we obain 72 small firms and 55 large firms. We look a every quarer a he skewness of he disribuion of small and large. We could have problems in our sample selecion if he disribuion of small firms were skewed o he righ or if he disribuion of large firms were skewed o he lef. This could indicae ha our cu-off for small and large is no a good one. The average of skewness (considering all periods) we observe for small firms was 0.80 and for large firms was 1.5. These resuls indicae ha our classificaion scheme is no a bad one as far as he cu-off is size concerned. Panel A of Table 4 shows he small and large firms separaed by he secor of he economy hey belong o. As one would imagine, large firms (40%) come from he concessionaries followed by he elecommunicaions secor (15%) while small firms come mosly from services secor (20%) followed by he exile secor (16%). Panel B of Table 4 liss mean values of some financial characerisics of small and large firms for he whole sample relaive o is asses. As we can easily verify, large firms have greaer long and shor-erm deb in average han small firms. Large firms also issue much more longerm commercial paper, ordinary and preferred socks han small firms. 18

19 Panel C of Table 4 shows some mean ess for hese characerisics considering he financial saemens of he las quarers of he years 1999, 2002 and As one can see all p-values of he differences of characerisics means beween large and small are close 0. Therefore, i seems ha small firms in our sample differ from large firms as far as access o he financial marke is concerned. They have less access o he financial markes. Finally, Panel D of Table 4 shows correlaions of several balance shee characerisics of he firms ha we will us as conrol variables in our regressions. As i is clear, he size of he company (measured by asses) is posiively relaed o he oher financial characerisics relaed o access o he financial markes. 4. Empirical Analysis of Banks 4.1 Descripive Analysis of he Reacion of Banks In his secion we analyze how he balance shee accouns freely allocaed bank credi, demand deposis and oher financing forms of he banks of he hree size caegories reac o he idenified moneary shocks. The daa are analyzed in firs difference, ha allows us o visualize he variaion of hose accouns afer a moneary conracion. The ables presened in his secion demonsrae how he small, medium and big banks reaced o he idenified moneary conracions. Tables 5, 6 presen he reacion of he Brazilian banks o he idenified moneary shocks, SELIC rae and Boshen-Mills index (1995). We repor he average monhly variaions of he analyzed accoun in six periods before and afer each shock. By he credi channel heory, i is expeced ha a moneary conracion would diminish he availabiliy of demand deposis and, consequenly, he bank lending level. Table 5 presens he reacion of he banks o shocks idenified by he firs difference of he SELIC ineres rae. As can be observed, he demand deposis suffered reducion afer he shocks of SELIC in mos of he cases, mainly for he small banks. This fac can be used o 19

20 jusify he decreasing endency of he freely allocaed bank credi afer he moneary conracions. I is ineresing o noice ha in all shocks here was increase in oher financing forms by he small banks. This indicaes ha hose banks seek alernaives o he demand deposis as form of financing heir credi operaions. The analysis of he reacion of he banks o increases of he SELIC ax suggess ha he Cenral Bank was able o affec he bank credi supply hrough he Moneary Policy in he sudied period. This resul corroboraes he exisence of he bank lending channel in Brazil. The analysis of he reacion of he banks o he shocks idenified hrough he reading of he COPOM documens is represened in Table 6. As can be observed, here are indicaions of reducion in he bank loans supply in periods afer he shocks. In relaion o he financing forms, he small banks presened a decline endency in he demand deposis availabiliy, bu here was an increase in he recepion of oher financing forms. This indicaes ha, a leas for he small banks, he moneary conracions diminished heir financing sources and, consequenly, heir credi supply. This resul suggess ha he small banks do no consider oher financing forms as perfec subsiues o he demand deposis, wha corroboraes he exisence of he bank lending channel for small banks. 4.2 The Two-Sep Model Based in Kashyap and Sein (2000), his paper uses a wo-sep esimaion mehod o evaluae he relevance of he bank-lending channel in Brazil. The model of Kashyap and Sein ess he hypohesis ha he Cenral Bank is capable o affec he bank lending by inducing Moneary Policy shocks. In oher words, i ess if he bank-lending channel is acive as a ransmission mechanism of he Moneary Policy. By he bank lending channel heory, he impac of he Moneary Policy on he credi supply should be larger for banks wih difficulies in subsiuing heir financing sources. Besides, he impac of moneary shocks on he credi supply should be larger for banks wih less liquid balance shees, since he more liquid banks are able o proec heir loans porfolio. In his conex, he economeric models used in his sudy are based in he following hypoheses: he liquidiy consrains o loans are inensified during moneary conracions; he sensibiliy of he 20

21 volume of loans in relaion o he Moneary Policy is larger for banks wih less liquid balance shee; and he small banks are less capable of subsiuing heir financing source The firs sep of he model consiss in a cross-secional esimaion ha aims in esimaing he sensibiliy of he volume of loans in relaion o an index ha represens he liquidiy of he balance shee of he bank. Wih his purpose, we regress he variaion of he bank loans [ ln( L i ) ] agains an index ha represens he liquidiy of he bank ( B i ). In addiion, we 1 propose he addiion of hree lags of he variaion of he bank loans as explanaory variables o conrol for any inerial influences ha hose can have. The esimaed equaion for each bank class for each period is he following 15 : 3 () 1 ln( Li ) = α ij ln( Li j ) + β Bi + ε 1 i, j= 1 where L i represens he volume of loans 16 of each bank and B i is he liquidiy variable. The key iems of ha regression are he esimaed coefficiens for B i, represened by β 1. Those coefficiens represen he sensiiviy of he volume of loans o he srucure of he balance shee (or he liquidiy consrain dimension) of each bank class. Thus, as resul of he firs sep esimaion, i will be found a value of β in every monh for each bank class, wha allows he creaion of ime series of β for each one of hose classes. The firs sep is based on he hypohesis ha banks wih more liquid balance shees (larger value of B i ) have greaer capaciy o sofen shocks in heir financing sources on he credi supply. Tha is because hose banks can proec heir loans porfolio by selling heir sock of ne asses. The firs sep regression was esimaed using he Ordinary Leas Squares - OLS - mehod and, in he cases ha presened heeroskedasiciy problem, he Newey-Wes covariance marix 15 We do no inser he inercep in he firs sep equaion because he expeced value for he loans of a bank when here is no liquidiy in he previous period is zero. Tha specificaion is based on Kashyap and Sein (2000). 16 I was used he freely allocaed bank credi. 21

22 was used. As a resul of he firs sep esimaions, here were found β ime series for each size class of bank. These series, ha represen he sensibiliy of he bank loans supply o he liquidiy index of he bank, are used as dependen variable in he second sep of he esimaion. Figure 4 illusraes he graphs of he β series. Table 7 shows some descripive characerisics of β series. As can be observed, he volailiies of hese series presen significan differences. The observaion of he sandard deviaion and of he maximum and minimum values of each series shows ha he mos volaile series is he one of he big banks. I indicaes ha he loans supplies of hose banks are more sensiive o he liquidiy of heir balance shee. The small banks, oherwise, seem o presen he credi supply less sensiive o heir liquidiy level. Anoher imporan characerisic is he form how he β values are disribued. The resul of he normaliy es, represened by he Jarque-Bera values, indicaes ha he hree analyzed series are no disribued accordingly o a normal disribuion. The second sep of he esimaion analyses he impac of Moneary Policy shocks on he sensibiliy represened by β. Wih his purpose, he β series of each bank class should be regressed agains a Moneary Policy indicaor ( M k ). The addiion of lags of ha explanaory variable is jusified for he fac ha Moneary Policy shocks are no insanly responded and heir effecs las long over he economy. Thus, he following ime series regression is esimaed: n β = η + φk M + χr + µ k = 0 ( 2) k, where M represens he seleced Moneary Policy indicaor and R is a dummy variable ha disinguishes he Real Plan sages. The inserion of his las explanaory variable is imporan because, in he analyzed period, Brazil crossed wo Moneary and exchange Policy regimes 17 : 17 The dummy variable ha disinguishes among he Real Plan sages was esed isolaed and ineracing wih he moneary shock. However, we oped o inser i isolaed in he second sep of he model. Tha is because he effec ha we inend o observe is a possible quaniaive difference beween he liquidiy consrains of he banks before and afer January of

23 he Crawling Band regime unil December of 1998 and he inflaion argeing and floaion exchange saring from January of We also esimae anoher version of he previously exposed regression, which includes a variable of economic growh (PIB ) and he inflaion rae of he period ( π ) 18. There were also added hree lags of each one of hose explanaory variables, because heir effecs persis over he economy 19. The objecive of esimaing ha muliple version is o evaluae he exisence of oher facors, as capial or macroeconomic shocks, which can affec he bank lending supply. In his case here are no oher facors affecing he bank loans, γ j and ρ j coefficiens should no be saisically significan. The regression is he following: n 3 3 = η + φk M k + γ j GDP j + ρ j π j + χr µ ( 2 ). k = 0 j= 0 j= 0 β + The lags of M insered in he equaions (2) and (2 ) is represened by n because i depends on he Moneary Policy indicaor used. The number of lags was deermined by he use of a VAR and he observaion of he Akaike and Schwarz lag lengh crieria. I was defined he use of en lags for he SELIC shock, zero for he reserve requiremens and hree for he COPOM Documens. To analyze he exisence of he bank-lending channel hrough he proposed model we consider he following. In he case of a conracionis moneary shock, he elasiciy of he volume of loans in relaion o he liquidiy of he bank should increase ( β ). In addiion, when here is a moneary conracion, he banks end o suffer a reducion in he value of heir ne asses (elevaions of he ineres rae depreciae he asses). Therefore, if β increase in a period when he banks suffer fall in heir financing sources and in is sock of ne asses, he 18 For he variable of economical growh, we used he lagged indusrial produc and for inflaion, he Price Index o he Wide Consumer (IPCA), boh made available by he Brazilian Insiue of Geography and Saisics IBGE. 19 Like he mehodology adoped o define he moneary shocks number of lags, for he macroeconomic variables a VAR was also used in he definiion of he number of lags. The observaion of Akaike and Schwarz lag lengh crieria suggess wo lags for he indusrial produc variable and zero for inflaion. However, i is believed ha he use of hree lags of hose variables is more appropriae since heir effecs las long in he economy and oher empiric works use similar mehodologies [Bonomo and Graminho (2002), Kashyap and Sein (2000) and Takeda and Nakane (2005)]. 23

24 volume of loans ends o decrease more inensely. In his conex, he bank lending will diminish if a moneary conracion elevaes he liquidiy consrains of banks ha canno subsiue heir financing sources. Tha scenery characerizes an acive bank lending channel. The main resul o be analyzed in he proposed model is he sum of he coefficiens of M, k φ k. I is expeced ha, a leas in he case of he small banks, a conracionis moneary shock increases he liquidiy consrain of he banks. Thus, he sum of φ k should be saisically significan and posiive, wha would confirm he exisence of a bank-lending channel in Brazil. We expec ha he resul is significan for small banks because, by hypohesis, hose banks face larger difficuly in raising oher financing forms. In case he larger banks can easily subsiue heir financing sources, he sum of heir φ k coefficiens should no be saisically significan. In he second sep regression, we esed he addiion of up o six lags of β on he righ side of he equaion. However, hose variables were no saisically significan in any case. Tha es is imporan when working wih ime series, since i is usual o observe inerial effecs in ha kind of sample. The resuls of he second sep regression can be observed in Table 8 Panel A. These ables presen he Wald es of he sum of he coefficiens of moneary conracions. Our null hypohesis is ha his sum is posiive, which would indicae he exisence of a bank-lending channel. The main resul is ha he sum of φ k coefficiens (moneary conracions) is significan a 5% for he small banks class. I is no significan for he large banks class, hough. This sum represens he marginal effec of he Moneary Policy shocks on β and will be he key iem o define he relevance of he bank-lending channel in Brazil. In he case of small banks, he sum of he coefficiens indicaing moneary conracions relaed o SELIC of he more simple specificaion is 0.41 (p-value 0.03) and sum of he same coefficiens in he case of he more complee specificaion is 0.31 (p-value 0.0). In he case of 24