João André Marques Pereira* 1 Richard Saito**

Transcription

1 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank (Coordenação enre capal buffer e perfl de rsco sob supervsão do Banco Cenral) João André Marques Perera* Rchard Sao** Absrac Ths work nvesgaes how banks reac o he capal consrans mposed by he Cenral Bank Usng models ha ncorporae he smulaney of capal decsons and rsk decsons our fndngs confrm he capal buffer heory whch predcs ha adjusmens o capal and adjusmens o rsk are posvely relaed Moreover we fnd ha regulaory pressures nduce banks o ncrease her rsk levels n response o capal adjusmens bu no vce versa Keywords: regulaory capal asse rsk prudenal regulaon JEL Classfcaon: G2 G28 Resumo Ese argo nvesga como bancos reagem a resrções de capal mposas pelo Banco Cenral Usando modelos que ncorporam smulanedade enre decsões de capal e decsões de rsco nossos resulados confrmam a eora de capal buffer que preconza que os ajuses a capal e os ajuses ao rsco são posvamene relaconados Além dsso enconramos que a pressão regulaóra exercda pelo Banco Cenral nduz bancos a aumenar o nível de rsco em resposa ao ajuse de capal mas não vce-versa Palavras-chave: capal regulaóro rsco dos avos e regulação prudencal Submedo em 26 de junho de 204 Reformulado em 09 de março de 205 Aceo em 8 de mao de 205 Publcado on-lne em 5 de novembro de 205 O argo fo avalado segundo o processo de duplo anonmao além de ser avalado pelo edor Edor responsável: Rcardo Leal *Banco Cenral do Brasl Brasíla DF Brazl E-mal: joaoandre@bcbgovbr **EAESP/FGV São Paulo SP Brazl E-mal: rsao@sanfordalumnorg Rev Bras Fnanças (Onlne) Ro de Janero Vol 3 No January 205 pp 73 0 ISSN ISSN onlne Socedade Braslera de Fnanças under a Creave Commons Arbuon 30 lcense - hp://creavecommonsorg/lcenses/by/30

2 Perera J Sao R Inroducon The Basel Accord of 988 (BCBS 988) ses a new sandard for mnmum capal requremens whch was adoped by Brazl cenral bank n 994 The Accord ses a mnmum capal adequacy rao (CAR) of 8% whch was ncreased by Brazl cenral bank o % n 999 Snce he adopon of he mnmum capal requremens banks have adoped a hgher sandard han ha se by he fnancal auhores Ths can be parly explaned by he cos of adjusmen for he opmal level of he CAR for he bank For nsance banks opmal decsons may be nfluenced by marke dscplne mechansms by agens nvesmen sraeges or even by ndrec schemes of regulaory pressure relaed o oher aspecs of rsk profles capured by boh on- and off-se supervsory measures In aempng o undersand and evaluae he effec of regulaory nervenons on bank solvency a new lne of research on capal buffer heory has shown ha a ypcal bank s capal cushon may be drven by he explc and mplc coss of prudenal regulaon Ths paper conrbues o he prudenal regulaon on shor-erm smulaneous decsons abou capal and rsk n he Brazlan bankng sysem To our knowledge hs s he frs work o address how prudenal regulaon has been conduced n Brazl The resuls show ha banks closer o he regulaory mnmum seem relavely more rsk-averse n ha hey make larger posve adjusmens o her capal levels and smaller adjusmens o her porfolo rsk levels Low-capalzed banks also manage her solvency raos more acvely coordnang her capal and rsk adjusmens n he same drecon The supervsory monorng effec also appears lkely o suppor capal resrcon ncreasng rsk averson as he evaluaon auhory srenghens The res of he paper s dvded as follows Secon wo provdes a bref heorecal revew of he role of prudenal regulaon n banks behavor relaed o adjusmens n leverage and porfolo rsk followed by some emprcal evdence Secon hree deals he smulaneous paral adjusmen model used as a reference n he emprcal analyss and presens he esng hypohess Secon four descrbes he daabase Secon fve presens he esmaon mehods and nerpres he resuls Secon sx concludes 74 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

3 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank 2 Regulaon and capal-rsk dynamcs n he bankng leraure The leraure on mnmum capal requremens was frs approached n he lae 970s o look a how capal rules may correc perverse ncenves generaed by he radonal rsk-nsensve srucure of depos nsurance (Sharpe 978) Generally moral hazard auhors assume ha capal s defned a he mnmum regulaory lm and focus on analyzng banks possble porfolo-rsk decsons Kahane (977) and Koehn and Sanomero (980) use effcen froner models n whch banks maxmze asse reurns subjec o porfolo rsk consrans They demonsrae ha mposed lms on leverage can ncrease an nsuon s rsk as banks end o reallocae her porfolos among rsker asses lookng for hgher expeced reurns (asse subsuon moral hazard) Km and Sanomero (988) show ha he perverse ncenve may be mgaed by rsk-based capal requremens unless he defned rskweghs do no correcly reflec he porfolos poenal losses Furlong and Keeley (989) and Keeley and Furlong (990) ncorporae he value of he depos nsurance opon no he model and show ha regardless of rsk weghng leverage resrcons creae ncenves for nsuons o manan proper rsk levels The heory of capal bufferng suppors he hypohess ha banks manan a capal surplus (buffer) o reduce nerference by supervsory auhores and o mgae he evenual regulaory coss assocaed wh volang he mnmum capal rule Unlke radonal moral hazard lnes he models of he capal buffer heory (Furfne 200; Mlne & Whalley 200; Peura & Keppo 2006) ake capal as an endogenous response o regulaon and add an neremporal perspecve o he bank s recapalzaon process As hey approach he mnmum banks end o ncrease her capal and/or reduce her rsk exposure o resore her solvency raos and avod regulaory coss Subsequenly hey agan coordnae her behavor o mee her combned capal and rsk arges Mlne and Whalley (200) exend he behavoral model of he buffer heory addng o he capal regulaon resrcons he aud funcon of he supervsor agen Those auhors fnd ha bank s rsk averson s a posve funcon of supervsory monorng Therefore banks end o enlarge capal cushons and o choose less rsky porfolos n reacon o hgher levels of bank supervson The recenly formalzed capal buffer heory however proposes relaonshps among capal rsk and regulaon ha seem more algned o he evdence n he emprcal leraure In general s observed ha Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

4 Perera J Sao R less capalzed banks engage n major posve shor-erm adjusmens for capal and lower adjusmens for rsk The precursory work of Shreves and Dahl (992) analyzes he smulaenous relaonshp beween capal and rsk n he bankng secor These auhors run ess on he US marke n he 980s a perod when he capal requremen was no rsk-adjused and sugges ha banks under regulaory pressure (below he mnmum capal requremen) offse ncreases n capal by ncreasng rsks Afer he deploymen of rskbased capal rules n he Uned Saes n 99 Jacques and Ngro (997) replcaed he mehodology and observe ncreases n capal levels and reducons n rsk levels suggesng ha he Basel Accord has played an mporan role n changng banks opporunsc behavors Followng he enacmen of Federal Depos Insurance Corporaon Improvemen Ac of 99 (FDICIA) Aggarwal and Jacques (998) nvesgaed how FDICIA may have helped n supervsory correcve acons for banks ha are no properly capalzed (Promp Correcve Acon) Ther resuls ndcae ha low-capalzed banks have shown ncreases n capal levels and reducons n rsk levels Rme (200) observes ha Swss banks c ha less-capalzed banks esablsh hgher capal levels whou promong correspondng adjusmens o her rsk levels Solz (2007) performs a smlar sudy of German savngs banks argung ha hese banks comprse he mos exensve and homogeneous group of nsuons n he counry She fnds evdence ha banks adjus capal faser han rsk and banks wh smaller buffers promoe faser adjusmens han do well-capalzed banks Wh respec o he coordnaon beween capal decsons and rsk adjusmens she observes ha he wo are negavely correlaed for banks wh smaller buffers bu posvely correlaed for banks wh larger buffers The auhor nerpres he resuls as ndcave ha regulaory pressure ncreases banks rsk averson One crcsm of he emprcal research s ha may no properly denfy and separae he drec effecs of regulaory resrcons from hose ndrecly caused by eher he pressure of supervson or dscplnary forces n he marke Furfne (200) dsngushes greaer srcness n he monorng crera of supervsory auhores as one of he major facors ha led o he cred crunch n he Uned Saes durng he 990s The auhor noes ha gher supervsory rules have had a greaer nfluence on banks balance shee decsons han have he mposon of mnmum capal lms 76 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

5 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank Berger e al (200) analyze he solvency rangs gven by he supervsors of US commercal banks (CAMEL Capal Adequacy; Asse Qualy; Managemen; Earnngs; Lqudy) and noe ha supervsory rgor has an mpac on he cred supply bu only o a moderae degree De Young e al (200) explore he nformaonal value of he supervsory CAMEL rangs and fnd evdence ha CAMEL scores reflec he rsks aken by fnancal nsuons and may generae ncenves o engage n rsk managemen ha s more effcen Our work hopes o fll he gap by ncludng propreary cenral bank as a proxy for supervsory pressure and s mpac on he coordnaon beween capal and rsk 3 Model of smulaneous paral adjusmens To es he decsons of Brazlan banks capal and rsk adjusmens we follow he paral adjusmens approach proposed by Shreves and Dahl (992) TObserved change n he capal and rsk levels of he nsuon n he perod beween - and ( e ) are smulaneously defned and may be decomposed no a d dscreonary poron ( d e ) ha s endogenously deermned and anoher par composed of exogenous shocks ( u e w ): d u () d w (2) Our assumpon s ha he exogenous shocks are formed by wo orhogonal componens whch are ndependen and dencally dsrbued: a frm-specfc effec and whe nose In urn he dscreonary varaons are modeled on a paral adjusmen approach assumng ha nsuons canno perform mmedae adjusmens due o some ype of rgdy and ransacon coss Thus he opmum levels of capal and rsk * and * are followed based on he adjusmen speeds and respecvely Consderng he smulaney n capal and rsk decsons he model s descrbed as follows: Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

6 Perera J Sao R * ( ) u (3) * ( ) w (4) The paral adjusmen n equaons (3) and (4) suggess ha banks se her capal and rsk arges Alhough hese varables are no drecly observable he emprcal leraure ndcaes ha hey depend on varables relaed o boh he specfc characerscs of he nsuon and he economc envronmen whch are represened by he vecors Y and Z The capal and rsk arges should also be nfluenced by regulaory consrans and monorng by fnancal auhores Thus o capure he respecve effecs of regulaory and supervsory pressures he bnary varable ndcaes less-capalzed banks and he connuous varable SUPERV denoes he assessmens of he bank conduced by he supervsory auhory accordng o whch he hgher he score s he worse he perceved condon of solvency s By ncludng hese new varables he smulaneous equaons model (Specfcaon I) s defned as follows: SUPERV a Y u (5) SUPERV b Z w (6) By allowng dfferen nerceps for banks under dfferen levels of regulaory pressure hrough he explanaory dummy Solz (2007) suggess ncludng varables consruced by neracng he varable wh he varables Δ and Δ whch helps verfy wheher he paern of coordnaon beween capal and rsk s mananed along capalzaon levels Addonally o es wheher banks under greaer regulaory pressure adjus her capal and rsk more quckly han do Gropp and Heder (2008) fnd evdence for publcly raded banks n he US and Europe ha he varables commonly used as capal srucure deermnans for non-fnancal companes such as sze profably he marke-o-book rao and angbly are also deermnng facors n explanng banks leverage Çağlayan and Şak (200) show smlar resuls for he Turksh bankng sysem dsngushng he peckng order heory as he prmary drver of bank behavor 78 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

7 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January oher banks she suggess ncludng he neracon of he varable wh he varables - and - Defned hs way he sysem assumes he form of Specalzaon II n equaons (7) and (8) u Y a SUPERV 0 0 (7) w Z b SUPERV 0 0 (8) u Y a SUPERV 0 0 (7) w Z b SUPERV 0 0 (8) Fnally n Specfcaon III he neracon beween he supervson varable SUPERV and he regulaon varable s ncorporaed no boh he capal and rsk-adjusmen equaons wh he am of esng he nfluence of he supervson crera on he capal rules u Y a SUPERV SUPERV (9) w Z b SUPERV SUPERV (0) 3 Tesng hypoheses Under he null hypohess of no effecs of capal and monorng rules on banks behavor he smulaneous equaons model presened n Specfcaon I allows us o confron hree aspecs of banks heorecal and observed managemen conduc Thus he hypoheses are defned o

8 Perera J Sao R conduc he followng ess: frs he drec mpac of regulaon and supervson on capal and rsk arges; second he coordnaon beween adjusmens n capal and rsk; and hrd he speeds of hose adjusmens Wh respec o he drec mpacs of prudenal regulaon and supervson Hypoheses H and H2 consder he nfluence of mnmum capal requremens and supervsory evaluaons on changes n banks opmum levels of capal and rsk As nended by fnancal auhores and conssen wh he capal buffer heory banks are expeced n he shor erm o experence greaer regulaory pressure o adjus her capal levels upwards and o ake less rsk n her porfolos (Hypohess H esed by Specfcaons II) I s also expeced ha percepons of supervson are lkely o nfluence banks decsons n he same drecon as shown n Hypohess H2 (esed by Specfcaon I) Moreover he jon effec of regulaon and monorng descrbed n Hypohess H2 Hypohess H Capal regulaon on capal and rsk adjusmens: Banks closer o he mnmum regulaory capal requremen are lkely feel ncreasng regulaory pressure; herefore he mpac of hs regulaory pressure on banks decsons mgh be posve for capal adjusmens ( 0 ) and negave for porfolo rskadjusmens ( 0 ) Hypohess H2 Bankng supervson on capal and rsk adjusmens: Banks negavely evaluaed by a supervsory auhory mgh reac by ncreasng capal ( 0 ) and reducng porfolo rsks ( 0 ) Addonally he effec of hose evaluaons should be sronger for less-capalzed banks ( 0 and 0 ) Wh respec o he shor-run nerdependence beween capal and rsk decsons he expeced effec of regulaon may also depend on a bank s capalzaon level Prudenal regulaon should exer a mnor nfluence on well-capalzed banks; neverheless a posve relaonshp beween he adjusmen of capal and rsk should be jusfed by agency conflcs or by nernal solvency arges se by a bank s admnsraon In he case of low-capalzed banks he cos assocaed wh regulaory penales should make banks even more sensve o changes n eher capal or rsk; herefore more srongly coordnaed behavor beween 80 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

9 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank capal and rsk adjusmens should be expeced o avod he volaon of a regulaon as saed n Hypohess H3 Hypohess H3 Capal-rsk coordnaon: The cos assocaed wh he volaon of capal regulaon should enhance he ncenves for low-capalzed banks o posvely coordnae capal and rsk adjusmens Thus an ncrease (decrease) n rsk may lead o a compensaory ncrease (decrease) n capal and a reducon (ncrease) n capal may be compensaed by a reducon (ncrease) n rsk ( 0 and 0 ) Fnally s expeced ha banks under regulaory and supervsory pressures wll seek o replensh her capal and rsk arges faser han well-capalzed nsuons Hypohess H4 Velocy of capal and rsk adjusmens: Banks closer o he mnmum regulaory capal requremen mgh adjus her capal and rsk levels more quckly o escape regulaory coss ( 0 and 0 ) 32 Defnons of capal and rsk The capal and rsk measures used o compose he varables and respecvely follow he provsons of local prudenal regulaon whch n urn follows he proposals of he Basel Accord 2 The oal capal n he varable s defned a wo dfferen levels dependng on s nsrumen loss-absorpon capacy Ter I capal s prmarly composed of equy capal and dsclosed reserves Ter II capal whch has a lower capacy o absorb losses s prmarly composed of subordnaed deb nsrumens and hybrd nsrumens of capal and deb 2 Prudenal rsk-based capal regulaon was mplemened n Brazl by Resoluon 2099 of Augus In 997 Brazlan capal regulaon became more rgd by elevang he facor of RWA from 8% o % In 999 Brazl began o mplemen capal rules o cover he marke rsk caused by exposure o foregn currences Recenly Resoluon 3444 of February amended he defnon of regulaory capal (Parmôno de Referênca - PR) In parallel Resoluon 3490 of Augus whch ook effec n June 2008 has provded new models for calculang he mnmum capal requremen (Parmôno de Referênca Exgdo PRE) Ths norm nroduces capal o cover operaonal rsk and changes he form of calculaon for marke rsk and cred rsk All models are based on he smplfed or sandardzed mehodologes proposed by he Basel Accord Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205 8

10 Perera J Sao R An nsuon s rsk exposure whch s conaned n he varable s deermned by s rsk-weghed asses (RWA) Rsk weghs are based on mehods adaped from sandardzed models of he Basel Accords and encompass hree prmary rsk sources: cred rsk marke rsk and operaonal rsk RWA n urn are obaned by mulplyng he sum of he capal requremen porons o cover marke and operaonal rsks by 909 and by addng he poron of he rsk-weghed exposures relaed o cred rsk A bank s CAR s hen calculaed by dvdng he regulaory capal by RWA I s worh nong ha he formula as descrbed reflecs n erms of he CAR he mnmum capal requred by Brazlan regulaon whch s % and herefore more conservave han he 8% proposed by he Basel Accord 33 Measuremen of regulaory and supervsory pressures Regulaory pressure can be measured eher by an evaluaon of specfc changes n regulaon or hrough varables ha capure ha pressure over me In economerc models ha arbue s usually represened by a dchoomous varable whch denfes he lesscapalzed banks ha are also he mos suscepble o regulaory acon The man ssue n hs consrucon however les n he defnon of he capal level ha characerzes a bank under regulaory pressure An obvous choce would be he regulaory capal lm; however a few banks fall below ha hreshold Wha happens n pracce s ha regulaory nerference s rggered whle a bank s capal remans whn posve levels of capal buffers despe he fac ha hese levels are no formally defned Accordngly researchers have aemped o capure hs feaure n dfferen ways Par of he leraure uses fxed capal buffer hresholds Rme (200) for nsance fxes a lm of wo percenage pons above he mnmum CAR Solz (2007) alernavely proposes a sascal creron n whch he hreshold for regulaory pressure s defned by he enh percenle of he se of sandardzed capal buffers (e a capal buffer over s sandard devaon) of a specfc perod In hs sudy hree dfferen mehods based on dfferen parameers for banks capal buffers (acual CAR mnus %) have been esed o defne he regulaory pressure hreshold: () absolue approaches n whch he fxed absolue capal buffer hresholds are % and 2%; () a sascal creron n whch he hreshold s he enh percenle of he sandardzed capal buffers; and () relave approaches n whch he 82 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

11 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank hresholds are he ffh enh and ffeenh percenles of he se of absolue capal buffers n each quarer Due o beer resuls n he esmaed regressons we adoped he hrd opon usng he enh percenle of absolue buffers as a cuoff for he varable Fgure shows he evoluon of some capal buffers percenles over he sample perod Fgure Percenles of absolue capal buffers The graph shows he evoluon over me of he ffh enh and ffeenh percenles of he capal buffers of he banks n he sample The buffers are calculaed n percenage pons as he excess of he curren capal adequacy rao (CAR) over he mnmum regulaory CAR n Brazl (%) The effecs of regulaory pressure are nrnscally lnked o he performance of supervsors The Cenral Bank should ensure ha nsuons are complan wh he lms esablshed by he regulaons; accordngly effecve supervsory acons brng credbly o he regulaory consrans Moreover supervsors have ools and nformaon no avalable o he marke ha may resul n percepons abou an nsuon s solvency condons ha go beyond s capal rao Therefore he ndrec pressure of supervson may devae from he drec pressure of regulaon To conrol for he nfluence of supervson n he model he varable SUPERV consss of he scores gven o he nsuon by he Cenral Bank The assessmen crera are confdenal bu nvolve no only managemen feaures bu also nformaon abou accounng and prudenal conrols n he CAMEL syle Therefore he fnal score represens he supervsor s perspecve on economc and fnancal Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

12 Perera J Sao R condons rsk profle and nsuonal effcency and s an mporan ool among ohers wh whch o decde wheher o nensfy focus on a specfc frm Thus a poorly raed nsuon even f s solvency rao does no presen problems s more lkely o suffer from drec supervsory acons I s mporan o hghlgh ha an nsuon s no always nformed of s score Even n hs case he argumen abou supervsory pressure should reman vald because f he percepon of he rsk relaed o an nsuon s hgh s supervsor mgh ncrease prevenve acons ha consequenly may nfluence he nsuon s behavor 34 Varables ha affec he arge levels of capal and rsk Followng he emprcal leraure on bank capal srucure decsons he opmum levels of capal and rsk depend drecly on frmspecfc and economc envronmen facors Because reaned earnngs correspond o a sgnfcan source of fnancng for Brazlan banks reurn on asses (ROA) s used n he capal equaons wh posve expeced sgns Expenses wh loan losses (LLOSS) n urn reduce he value of exposures subjec o rsk weghng and as a resul reduce he defned rsk measure For hs reason hs facor s ncluded as an explanaory varable n he rsk equaons wh negave expeced sgns Sze (SIZE) as an ndcaor of exernal capal access may have a negave effec on he level of capal I may also have a negave mpac on rsk level reflecng larger banks greaer nvesmen opporunes and beer porfolo dversfcaon (Acharya e al 2006) However larger banks ha are perceved as oo bg o fal can ake advanage of her safey ne by ncreasng boh her leverage and her porfolo rsk Thus he expeced sgn of he varable n he capal equaon remans negave bu becomes ambguous n he rsk equaon The capal buffer heory suggess ha banks wh more lqud asses (LIQUID) need less nsurance agans breaches of capal requremens The buffer can be rapdly replenshed by changng lqud asses wh hgher rsk weghs (for example socks) for ohers lower RWA Accordngly he expeced mpacs of he asses lqudy should be negave on he capal arge and posve on he rsk arge 84 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

13 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank To capure he effec of economc flucuaons on banks decsons boh equaons nclude he varable of GDP growh (GPDG) One of he regulaon s curren concerns refers o he amplfcaon effec on he busness cycle ha orgnaed n he prudenal rules and n he managemen conduc of some banks In an economc expanson phase ncreases n leverage and porfolo rsk may presen evdence of harmful pro-cyclcal behavor by he bankng sysem In hs case he sgns observed would be negave n he capal equaon and posve n he rsk equaon Table I presens a summary descrpon of he varables consdered n he models of capal and rsk arges and her expeced sgns Indeed me dummes are also ncluded o capure possble seasonaly and specfces for each year n he sample along wh a dummy for conrollng changes n he regulaory sandardzed models for calculang capal requremens n July 2008 Table I Explanaory varables of he capal and rsk arges Descrpons and he expeced sgns of he varables ha explan he opmal levels of capal and rsk n he proposed model of he paral adjusmen of smulaneous equaons Varable Defnon Hypohess SUPERV ROA LLOSS SIZE LIQUID GDPG 4 Daabase Dummy for low-capalzed banks CAMEL rangs Reurn on asses Provson losses Bank sze Lqud asses Real GDP growh Effec on capal Effec on rsk Proxy for regulaory pressure + - Proxy for supervsory pressure + - Reaned earnngs represenng a sgnfcan source of recapalzaon Expenses wh loan provsons drecly affec RWA by reducng he rsk-asse base - Access o fnancal markes - Invesmen opporunes and porfolo dversfcaon - The oo-bg-o-fal problem - Lqudy as nsurance agans volaons of capal requremens Busness cycle nfluence Shorsghed capal managemen wh pro-cyclcal effecs / The daase consss of quarerly nformaon for he 2 commercal banks and bankng holdng companes wh cred porfolos ha Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

14 Perera J Sao R comprse he Brazlan bankng sysem 4 The perod of analyss covers 36 quarers from he frs quarer of 200 unl he fourh quarer of 2009 Insuons subjec o governmen nervenon or lqudaon processes and hose wh less han fve observaons n he perod were removed from he sample Developmen banks as well as hose whose man acves are nvesmen bankng or reasury operaons were also excluded Therefore he fnal unbalanced panel ncludes 3846 bankquarer observaons 5 Due o mergers and acqusons he number of nsuons oscllaes by approxmaely 00 durng all quarers Table II shows he number of banks n December of each year A he end of 2009 he sample comprsed more han 70% of he oal number of nsuons n he Brazlan bankng sysem and represened approxmaely 97% of he sysem s oal asses Along wh he hgh concenraon n he bankng ndusry he en larges banks n he sample hold approxmaely 88% of he oal asses n he sample for he same dae Table II Number of banks Number of nsuons n he sample a he end of each year Year # Banks Year # Banks The varable represens he rao beween he nsuon s regulaory capal and s oal asses Analogously he asses rsk of an nsuon s deermned by dvdng s RWA by s oal asses The rao beween he wo varables represens s CAR In urn s 4 The bankng sysem s defned as he se of nsuons ha form Bankng Consoldaed I and II accordng o he Cenral Bank of Brazl Tha group ncludes commercal banks unversal banks savng banks nvesmen banks and fnancal conglomeraes composed of a leas one of hese nsuons 5 The mehodology appled o separae he fnancal nsuons accordng o her operaonal profles s descrbed by Capelleo (2006) Ths procedure s used by he Cenral Bank of Brazl for supervsory purposes 86 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

15 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank capal buffer (BUF) s calculaed by subracng % from he CAR The dummy whch ndcaes nsuons under greaer regulaory pressure assumes a un value f he nsuon presens a buffer below he sample s enh percenle for a specfc quarer as dscussed n he prevous secon The supervsory evaluaon varable SUPERV s consruced from he average of he CAMEL scores gven o he nsuon The lower he score s he beer he evaluaon s The profably defned by he varable ROA represens a bank s quarerly ne prof n relaon o s average oal asses Provson expenses are compued n he varable LLOSS as a proporon of ousandng cred operaons The varable SIZE s calculaed by he naural logarhm of he nsuon s oal asses Imporanly he oal asse values do no nclude a bank s fnancal nermedaon accoun The lqudy cushon LIQUID s defned by he rao of lqud asses o oal asses Lqud asses are composed of cash nerbank operaons governmen bonds oher lqud bonds socks quoes from nvesmen funds and Cenral Bank reserves Fnally he economc cycle componen (GDPG) s se based on he seasonally adjused quarerly change n GDP deflaed by he Brazlan nflaon ndex IPCA The descrpve sascs for he man varables n he sample are presened n Table III excep for he varable SUPERV whch has been suppressed for confdenaly reasons I s observed ha he levels of capal and rsk boh presen wde varaons n erms of sandard devaon as he mnmum and maxmum values Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

16 Perera J Sao R Table III - Descrpve sascs Summary of basc sascs for he varables n he sample on a quarerly bass Varable Un Mean Sd Dev Mn Max Δ (%) (%) TIER - (%) Δ (%) (%) ROA (%) LLOSS (%) SIZE (Mllons R$) LIQUID (%) GDPG (%) BUF - (%) Table IV shows he means and sandard devaons for he varables separang he sample no banks wh lower levels of capal ( = ) and hose wh hgher capal raos ( = 0) I s noed ha on average less-capalzed banks have posve capal adjusmens (059%) and negave rsk adjusmens (-29%) whereas for oher banks he average adjusmens are made n he oppose drecons (capal reducons of 026% and rsk ncreases of 038%) The dfferences n profably and lqudy beween he wo groups are also evden The beer-capalzed banks have hgher average ROA (058% versus 036%) and larger lqud asse cushons (269% versus 57%) 88 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

17 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank Table IV - Descrpve sascs by level of capalzaon Means and sandard devaons of he varables n he sample on a quarerly bass and separaed no a group of less capalzed banks ( = ) and ohers ( = 0) = = 0 Varable Un Mean Sd Dev Mean Sd Dev Δ (%) (%) TIER - (%) Δ (%) (%) ROA (%) LLOSS (%) SIZE (Mllons R$) LIQUID (%) GDPG (%) BUF - (%) Table V presens he correlaon marx for he man varables n he sample There s a hgh posve correlaon beween Ter capal (TIER) and oal regulaory capal (); moreover he means and sandard devaons of he varables n Table III are very smlar ndcang ha on average hgher-qualy capal s he prmary form of he nsuons capalzaon The correlaons beween measures of rsk and capal are also posve boh n levels and n frs dfferences alerng us o coordnaed acon by he banks n her decsons relaed o capal and rsk adjusmens Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

18 Perera J Sao R Table V - Correlaon marx Correlaons among he varables n he sample on a quarerly bass The ndex * represens a sgnfcance level of 5% Δ Δ - TIER - Δ - ROA LLOSS SIZE LIQUID GDPG BUF * TIER - -06* 099* Δ 022* * 030* 029* -022* ROA 00* 007* 008* 003* 008* LLOSS 003* 02* 0* -004* 007* -07* SIZE * -063* * -005* -008* LIQUID 004* 038* 039* * * -0* GDPG -005* * -004* BUF - -02* 053* 053* 007* -09* * -030* 040* 00 5 Emprcal analyses In general he relaed emprcal leraure uses pooled daa approaches o esmae he smulaneous equaons model (Shreves & Dahl 992; Jacques & Ngro 997; Aggarwal & Jacques 200; Rme 200) Gven he endogeney beween he and varables he ordnary leas squares (OLS) esmaor should be dsmssed because provdes based and nconssen coeffcen esmaes A common sraegy n hese cases s he leas squares esmaors n wo or hree sages (2SLS/3SLS) In sum hose approaches use he exogenous and predeermned varables as nsrumens for he endogenous varables Thus by srucurng he combnaon of all avalable exogenous varables boh esmaors provde conssen and effcen esmaes 6 The gap n he mehodology however les n s omsson of he evenual unobserved banks heerogeney (fxed effecs) whch can lead o based esmaons Indeed even approaches ha specfcally address he fxed effec ssue should be sources of bas n he case of dynamc panels because he whn-group ransformaon gnores he correlaon beween he lagged dependen varable and he regresson error erm (Nckell 98) 6 Because he 3SLS esmaor ncorporaes he cross-correlaons among he equaons presens coeffcen esmaes asympocally effcenr 90 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

19 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank Arellano and Bond (99) propose a generalzed mehod of momens esmaor (GMM) o correc he bas n he dynamc panels Known as dfference GMM he esmaon procedure nally elmnaes he unobserved heerogeney effec usually by dfferenaon and subsequenly apples he GMM usng he lagged varables n level form as nsrumens for he ransformed explanaory varables The problem wh hs mehod s ha n he case of hghly perssen seres he lagged varables are merely weak nsrumens n he ransformed equaon generang a bas n fne samples as demonsraed by Blundell and Bond (998) Formulang addonal hypoheses Arellano and Bover (995) and Blundell and Bond (998) ncorporae he frs dfferences of he nsrumenal varables used n he dfference GMM mehod as nsrumens for he orgnal equaon Ths mehod generaes a sysem of wo equaons he orgnal (n level) and he ransformed (n dfferences) equaons and s herefore called sysem GMM The equaon n level form s nsrumened by lags of he explanaory varables frs dfferences and analogously he equaon n dfferences s nsrumened by lags of he varables n level form In addon o reducng he dfference GMM bas for fne samples he sysem GMM approach allows a wder use of nsrumens and can subsanally ncrease he sascal effcency of he coeffcen esmaor Consderng he characerscs of he emprcal consrucon presened he esmaons n hs sudy have been performed usng he descrbed GMM wh correced sandard errors (sysem GMM) and wh some adjusmens o accoun for he smulaney beween he dependen varables Because he second-sage esmaor may produce nconssenly smaller sandard errors especally n cases of small samples and large numbers of nsrumens he Wndmejer (2005) mehod has been used o correc varances and he covarance marx The number of nsrumens has been conrolled nally by reducng o egh he maxmum number of lags of he varables and hen by combnng (collapsng) he nsrumens no smaller ses 7 In addon orhogonal devaons have been used raher han frs dfferences o remove he unobserved dosyncrac effecs because he 7 The prolferaon of nsrumens may generae he overdenfcaon of he endogenous varables hnderng he proper reamen of endogeney and hus resulng n based esmaes (Roodman 2009) Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205 9

20 Perera J Sao R frs dfferences ransformaon may ncrease he gaps n unbalanced panels Fnally wh he purpose of ncorporang smulaney no capal and rsk decsons we have adoped he adjusmen proposed by Solz (2007) and Jokp and Mlne (200) n whch Δ s modeled as an endogenous varable n he capal equaon ncludng lags of he varable n he relaed se of nsrumenal varables and n urn Δ s modeled as endogenous n he rsk equaon usng lags of as s nsrumens 5 Emprcal resuls Table VI presens he sysem GMM regresson resuls for he sysem of smulaneous equaons of Specfcaons I II and III In he dagnosc analyss of all equaons he auocorrelaon ess of he frs and second orders proposed by Arellano and Bond (99) ndcae he valdy of he hypohess of GMM denfcaon and Hansen s J ess sugges he correc model specfcaon 7 7 Furher he auocorrelaon does no mply a bas bu hs underesmae he sandard errors One way s o cluser a he bank level However cluserng a bank level wll ncrease he sandard erros as you decrease he degree of freedom from N x T o N As we use dynamc GMM panel daa we have esed and we have no found any major changes on he resuls 92 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

21 Coordnaon of capal buffer and rsk profle under supervson of Cenral Bank Table VI - Esmaons of he smulaneous equaons model of capal and rsk adjusmens Panel A presens Specfcaons I o III regardng he equaons ha explan he capal adjusmens Δ Panel B presens Specfcaons I o III regardng he equaons of rsk adjusmens Δ The coeffcens are esmaed hrough he sysem-generalzed mehod of momens (Sysem GMM) K s he nercep and DModel s a dummy for he 2008 crss The model ncludes me dummes whose coeffcens were omed Indexes * ** *** represen sgnfcance levels of 0% 5% and % respecvely and he -sascs are repored n parenheses The Hansen es refers o he es of overdenfcaon resrcons and ess AR () and AR (2) refer o frs- and second-order auocorrelaon ess The p-values are repored for hose ess Panel A Equaons of capal adjusmens Δ Panel B Equaons of rsk adjusmens Δ Spec I Spec II Spec III Spec I Spec II Spec III Eq (5) Eq (7) Eq (9) Eq (6) Eq (8) Eq (0) * ** -08 *** (80) (00) (036) (-202) (-424) (-02) SUPERV * 0009 * 00 * SUPERV ** ** -002 (80) (67) (68) (-234) (-22) (-25) - SUPERV SUPERV * (-06) (-79) ROA 0396 ** 0380 ** 0374 ** LLOSS -3 * (237) (2) (204) (-66) (-074) (-080) SIZE ** ** *** SIZE -003 *** -003 *** -004 *** (-223) (-229) (-306) (-32) (-390) (-436) LIQUID 07 *** 026 *** 03 *** LIQUID *** *** *** (309) (36) (367) (-437) (-473) (-454) GDPG *** ** -008 ** GDPG (-258) (-244) (-22) (-050) (-036) (-035) Δ 0062 *** 0053 *** 0053 *** Δ 08 *** 074 *** 0748 *** (444) (346) (369) (524) (52) (522) - Δ 003 ** 0096 * - Δ 2548 *** 260 *** (200) (78) (537) (57) *** -059 *** -078 *** *** *** *** (-296) (-339) (-42) (-342) (-45) (-44) *** 025 *** (29) (56) (282) (3) DModel 0009 *** 002 *** 000 *** DModel * (3) (359) (327) (-72) (-26) (-27) K 009 * 0099 * 04 ** K 0583 *** 064 *** 062 *** (68) (65) (22) (426) (495) (496) AR() AR() AR(2) AR(2) Hansen Hansen Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

22 Perera J Sao R Resuls regardng he mpac of regulaon and supervson on he arges of capal and rsk The coeffcens of he varable n he capal adjusmens equaons are posve ( 0 ) bu hey are sgnfcan a he 0% level n only one of hree equaons (equaon 5) In he rsk adjusmen equaons shows negave coeffcens ha are sgnfcan n wo of hree specfcaons a he 5% and % levels (equaons 6 and 8) Accordngly here s evdence ha banks wh lower capal surpluses ncrease her capal a hgher amouns and parcularly ncrease her rsk by lower amouns compared o oher banks Ths resul s n lne wh he expeced effec descrbed n Hypohess H abou he nfluences of regulaory pressure Smlarly he coeffcens of he varable SUPERV whch represens he supervsory auhor s rangs of banks economc and fnancal condons are posve and sgnfcan n he capal equaons ( 0 ) a he 0% level and negave and sgnfcan n he rsk equaons ( 0 ) Tha s poorly evaluaed banks end o pursue shor-erm adjusmens o mprove her solvency raos by eher ncreasng capal or decreasng rsk Moreover he neracon beween he varables SUPERV and n Specfcaon III shows ha he effec of he scores may be hgher for banks close o he capal lms parcularly wh respec o rsk adjusmens In equaon (0) he coeffcen of he neraced varable s negave and sgnfcan ( 0 ) a he 0% level showng ha nensy n he rsk adjusmens s hgher for lower-raed banks near he regulaory capal lm However he observed resul s no mananed for he capal adjusmens The coeffcen of he combned varable SUPERV n equaon (9) s no sgnfcan ( 0 ) suggesng ha wh respec o capal behavor none of he dfferences beween low-capalzed banks and ohers can be explaned by he scores receved Overall he resuls are also algned wh he expeced effecs of Hypohess H2 relaed o he nfluence of supervsory pressure and s jon effec wh prudenal rules Resuls regardng he coordnaon beween capal and rsk adjusmens Wh regard o coordnaed decsons abou capal and rsk he coeffcens of he varables of capal and rsk adjusmens n 94 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

23 equaons (5) and (6) are boh posve and sgnfcan a he % level ndcang ha banks ncrease capal when rsk ncreases and vce versa The resul suggess ha banks may have an opmal level of solvency (CAR) and may coordnae her levels of rsk and capal o acheve ha arge ( 0 and 0 ) Includng he neracon of he varable wh he varable Δ (equaons 8 and 0) and wh he varable Δ (equaons 7 and 9) he coeffcens of he neraced varables reman posve and sgnfcan ( 0 and 0 ) ndcang ha banks under regulaory pressure are more sensve o changes n her levels of capal and rsk The resuls suppor Hypohess H3 and sugges ha less-capalzed Brazlan banks acvely manage her capal raos wh he probable nenon of avodng he regulaory coss of a breach of mnmum capal requremens Resuls regardng he veloces of capal and rsk adjusmens The esmaed adjusmen speeds of capal ( abou -06) and rsk ( of abou - -09) are relavely smlar and sugges ha capal and rsk arges are fully acheved afer sx (as sx mes equals one) and fve quarers ( of abou -020 me fve equals one) respecvely Compared o he resuls of relaed works he resuls of hs analyss mply ha Brazlan banks seem o adjus boh capal and rsk levels much more quckly han banks abroad For nsance Rme (200) shows ha Swss banks may ake approxmaely en and weny years o adjus capal and rsk levels respecvely Solz (2007) esmaes ha German savngs banks may requre more han hry years o reach her arges The dfferences can be jusfed by he hgher capal rao levels n he Brazlan bankng sysem so he banks may be closer o her opmal levels Regardng he dfferences n speed adjusmens beween he low-capalzed banks and ohers only he coeffcen of lagged rsk s sgnfcan and posve ( 0 ) a he % level as shown n equaons (8) and (0) Ths resul ndcaes ha he beercapalzed banks may adjus rsk faser han banks under Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

24 Perera J Sao R regulaory pressure Ths resul s a odds wh he predcons of he buffer capal heory and herefore does no suppor Hypohess H4 One explanaon may be relaed o he effec of capal ncreases and rsk reducons capured by he dummy In hs raonale less-capalzed banks ake longer o reach her arges because hey se hgher capal arges and lower rsk arges n he shor run Addonal resuls As expeced he esmaed coeffcen of ROA s posve and sgnfcan a he 5% level n all of he capal equaons ndcang ha nsuons may rely on reaned earnngs as an mporan source of capal ncreases In urn he loan loss provsons (LLOSS) have coeffcens wh negave sgns due o he expeced negave mpac of provsons on ousandng rsk exposures; however he esmaed values are no que sgnfcan wh only equaon (6) presenng a sgnfcance level of 0% The resuls sugges a mnor nfluence of hs varable on RWA Sze of bank asses (SIZE) has a negave nfluence on changes n capal because he coeffcens of ha varable are negave and show sgnfcance levels of 5% n equaons (5) and (7) and % n equaon (9) Ths resul s n lne wh oher emprcal works suggesng ha larger banks promoe lower adjusmens n her capal srucures han do smaller banks The reason may be relaed o larger banks wder access o oher sources of fnancng and greaer economes of scale n her cred managemen acves Ths resul may also provde evdence of he presence of moral hazard n he nsuons consdered oo bg o fal However a negave effec of sze on he level of rsk s observed whch conradcs he resuls usually found n he leraure The negave coeffcens whch are sgnfcan a he % level n all rsk equaons show ha large banks may lower her rsks n greaer proporons han do small banks Ths resul conrbues o undermnng he hypohess of he asse subsuon moral hazard relaed o large banks mplc safey ne hus 96 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

25 supporng he argumen ha hose banks have greaer porfolo dversfcaon The effecs of he lqudy cushon (LIQUID) n capal and rsk adjusmens also show nrgung resuls The esmaed coeffcens for he varable are sgnfcan a he % level bu have oppose sgns from hose expeced Ineresngly a posve mpac on he capal adjusmens and a negave mpac on rsk adjusmens are observed These resuls conradc he proposons of he capal buffer heory n whch he lqudy cushon can replace capal as nsurance agans volaons of he mnmum capal requremen One explanaon for he fac ha capal and lqudy reserves go n he same drecon may be relaed o longerm nvesmen sraeges pursuan o whch banks may hold capal and lqud asses whle wang for beer nvesmen opporunes Anoher mporan oucome concerns he effec of he economc cycle (GDPG) on he dependen varables The negave and sgnfcan (a % and 5% levels) coeffcens n he capal equaons may be nerpreed as pro-cyclcal behavor by banks whch end o ncrease her leverage durng perods of economc expanson Moreover s observed ha he effec of hs varable on RWA s no sgnfcan whch can be explaned by he fac ha new operaons may belong o he same rsk-ype buckes as curren cred porfolos and ha sandardzed weghs for cred rsk n he regulaory models are nvaran over he economc cycle Despe no havng he reference of he rsk-sensve Basel II models he resuls sugges ha sandardzed models of capal may be less procyclcal han would be expeced 6 Concluson The capal requremen rules have become predomnan nsrumens n he conex of prudenal bankng regulaon Ths sudy examnes he effecs of regulaory resrcons on he shorerm dynamcs of capal and rsk n he Brazlan fnancal sysem beween 200 and 2009 One novely s our analyss of he Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

26 Perera J Sao R nfluence of he supervsory auhory ha monors hs process Usng he confdenal nformaon of banks CAMEL rangs he emprcal model allows an assessmen of he ndrec conrbuon of supervsory percepon n banks solvency decsons In lne wh he capal buffer heory our resuls ndcae ha regulaory coss may pressure banks o manan capal cushons I s observed ha when a bank approaches regulaory consrans makes hgher adjusmens o capal and lower adjusmens o rsk The supervsory evaluaon seems o nfluence hese banks decsons n he same drecon nducng banks o recompose her capal raos upon recevng poor rangs Addonally here s evdence of a posve relaonshp beween he levels of capal and rsk Accordngly rsk levels are defned n erms of adjusmens o he level of capal and smulaneously capal levels are adjused based on changes n rsk levels; all are n he same drecon Ths coordnaed behavor s shown o be even more nense for less-capalzed banks suggesng ha nsuons under regulaory pressure may acvely manage her regulaory capal raos o avod he regulaory coss of a breach of mnmum capal requremens Among he addonal resuls s worh emphaszng he nfluence of he economc cycle on banks capal formaon In phases of economc expanson Brazlan banks show sgnfcan ncreases n leverage whch may evdence of pro-cyclcal behavor References Acharya Vral Hasan Ifekhar & Saunders Anhony (2006) Should banks be dversfed? Evdence from ndvdual bank loan porfolos Journal of Busness 79: Aggarwal Raj & Jacques Kevn (200) The mpac of FDICIA and promp correcve acon on bank capal and rsk: Esmaes usng a smulaneous equaons model Journal of Bankng and Fnance 25: Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

27 Arellano Manual & Bond Sephen (99) Some ess of specfcaon for panel daa: Mone Carlo evdence and an applcaon o employmen equaons Revew of Economc Sudes 58: Arellano Manuel & Bover Olympa (995) Anoher look a he nsrumenal varables esmaon of error componens models Journal of Economercs 68:29-5 Basel Commee on Bankng Supervson (BCBS) (988) Basel Commee Inernaonal Convergence of Capal Measuremen and Capal Sandards BCBS Basel Swzerland Berger Allen Kyle Margare & Scalse Joseph (200) Dd US bank supervson ge ougher durng he cred crunch? Dd hey ge easer durng he bankng boom? Dd maer o bank lendng? In Mshkn Frederc edor Prudenal Supervson: Wha Works and Wha Doesn? Unversy of Chcago Press Chcago s edon pp Blundell Rchard & Bond Sephen (998) Inal condons and momen resrcons n dynamc panel daa models Journal of Economercs 87:5-43 Çağlayan Ebru & Şak Nazan (200) The deermnans of capal srucure: evdence from he Turksh banks Journal of Money Invesmen and Bankng 5:57-65 Capelleo Luco (2006) Classfcação das nsuções fnanceras pela avdade operaconal UnB Conábl 9:6-88 DeYoung Rober Hughes Joseph & Moon Choon-Geol (200) Effcen rsk-akng and regulaory covenan enforcemen n a deregulaed bankng ndusry Journal of Economcs and Busness 53: Furfne Crag (200) Bank porfolo allocaon: The mpac of capal requremens regulaory monorng and economc condons Journal of Fnancal Servces Research 20:33-56 Furlong Frederck & Keeley Mchael (989) Capal regulaon and bank rsk akng: A noe Journal of Bankng and Fnance 3: Gropp Ren & Heder Floran (2009) The deermnans of bank capal srucure Workng paper Seres n 096 European Cenral Bank Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January

28 Perera J Sao R Jacques Kevn & Ngro Peer (997) Rsk-based capal porfolo rsk and bank capal: A smulaneous equaons approach Journal of Economcs and Busness 49: Jokp Terh & Mlne Alsar (20) Bank capal buffer and rsk adjusmen decsons Journal of Fnancal Sably 7:65-78 Kahane Yehuda (977) Capal adequacy and he regulaon of fnancal nermedares: A porfolo approach Journal of Bankng and Fnance 4: Keeley Mchael & Furlong Frederck (990) A re-examnaon of he mean-varance analyss of bank capal regulaon Journal of Bankng and Fnance 4:69-84 Km Daesk & Sanomero Anhony (988) Rsk n bankng and capal regulaon Journal of Fnance 43: Koehn Mchael & Sanomero Anhony (980) Regulaon of bank capal and porfolo rsk Journal of Fnance 35: Mlne Alsar & Whalley A Elzabeh (200) Bank capal and ncenves for rsk-akng Avalable a hp://ssrncom/absrac=30376 Accessed on Nckell Sephen (98) Bases n dynamc models wh fxed effecs Economerca 49: Orgler Yar & Taggar Rober (983) Implcaons of corporae capal srucure heory for bankng nsuons: Noe Journal of Money Cred and Bankng 5:22-22 Peura Samu & Keppo Juss (2006) Opmal bank capal wh cosly recapalzaon Journal of Busness 79: Rme Berrand (200) Capal requremens and bank behavour: Emprcal evdence for Swzerland Journal of Bankng and Fnance 25:37-78 Roodman Davd (2009) A noe on he heme of oo many nsrumens Oxford Bullen of Economcs and Sascs 7:35-58 Sanos Joao (200) Bank capal regulaon n conemporary bankng heory: A revew of he leraure Fnancal Markes Insuons & Insrumens 0: Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205

29 Saunders Anhony Srock Elzabeh & Travlos Ncholaos (990) Ownershp srucure deregulaon and bank rsk akng Journal of Fnance 45: Sharpe Wllam (978) Bank capal adequacy depos nsurance and secury values Journal of Fnancal and Quanave Analyss 3:70-78 Shreves Ronald & Dahl Drew (992) The relaonshp beween rsk and capal n commercal banks Journal of Bankng and Fnance 6: Solz Sephane (2007) Capal and rsk adjusmens afer an ncrease n capal requremens In Solz Sephane edor Bank Capal and Rsk-Takng: The Impac of Capal Regulaon Charer Value and he Busness Cycle Sprnger New York s edon pp Wndmejer Frank (2005) A fne sample correcon for he varance of lnear effcen wo-sep GMM esmaors Journal of Economercs 26:25-5 Revsa Braslera de Fnanças (Onlne) Ro de Janero Vol 3 N January 205 0