Privatization, Foreign Entry, and Bank Risk in an Emerging Banking System: The Case of Argentina,

Transcription

1 Privaizaion, Foreign Enry, and Bank Risk in an Emerging Banking Sysem: The Case of Argenina, Pedro Elosegui Faculad de Ciencias Economicas Universidad Nacional de La Plaa La Plaa 1900 George Pineris * Deparmen of Economics Universiy of Illinois Urbana, IL Tel. (217) gpineri@uiuc.edu Absrac This paper examines he impac of privaizaion and foreign enry on he choices of risk of various ypes of banking insiuions in he Argenine banking sysem during he period Following he empirical mehodology of Shrieves and Dahl (1992) and ohers, we esimae a simulaneous equaions model of a bank s choices of capial and risk and es differences in behavior based on ownership ype. Our resuls do no show any differences in he behavior of various insiuions in erms of capial. However, we do find evidence ha, following privaizaion and foreign enry, boh exising privae banks and privaized banks did increase heir asse porfolio risk. These findings suppor he argumen ha privae banks in Argenina faced increased compeiion from aggressive enry of foreign banks. Finally, we do no find evidence ha foreign banks aemped o increase heir risk and we find only limied evidence ha foreign acquired banks exhibi higher risk. JEL G20, G21 * Corresponding auhor. We would like o hank Charles Kahn, George Pennachi, Waler Sosa-Escudero and seminar paricipans a he Universiy of Illinois. Midwes Finance Associaion 2002 meeings and BCRA seminar, for discussion and commens. All re maining errors are our own.

2 I. Inroducion The Argenine banking sysem wen hrough a period of major resrucuring during he 1990s. Afer an unsuccessful prior experience wih financial liberalizaion in he 1980s, Argenina began a process of srenghening and resrucuring is banking sysem, which accompanied he inroducion of a currency board in This process was characerized by he adopion of sricer regulaory sandards, he privaizaion of several provincial banks, he faciliaion of foreign enry ino he domesic banking sysem, and he inroducion of marke-based approaches for bank discipline. During he period , he srucure of he banking sysem changed subsanially. More han 90 insiuions were closed, including 54 banks and 14 nonbanks. The number of insiuions in he sysem decreased from 212 in 1992 o 119 by mid There were also 18 privaizaions, mainly of provincial banks (Calomiris and Powell, 2000). In addiion, beginning in 1995, several foreign banks enered he domesic marke, primarily hrough acquisiions of domesic insiuions. As a resul, by 1999, around half of he asses in he banking sysem were under foreign conrol. Foreign banks also had minoriy sakes in several oher insiuions. Privaizaion and foreign enry resuled in aggressive compeiion among financial insiuions for marke share. Privaizaion and foreign enry has aken place in a number of emerging banking sysems in recen years as a response o sysemic banking crises. The IMF and he World Bank have offered policy recommendaions emphasizing he benefis of privaizaion and liberalizaion in emerging banking sysems. In a recen sudy, he World Bank (2001) concludes ha Faciliaing he enry of repuable foreign financial firms o he local marke should be welcomed oo: hey bring compeiion, improve efficiency, and lif he qualiy of financial infrasrucure. However, regarding he impac of privaizaion, he World Bank sudy also noes ha privaizaion has o be designed carefully if he 2

3 benefis are o be gained and he risks of an early collapse minimized. In a survey of he role of foreign banks in emerging markes, he IMF (2000) concludes ha The evidence o dae on he effecs of foreign bank enry suggess ha he compeiive pressures creaed by such enry lead o improvemens in banking sysem efficiency, bu i is sill unclear wheher a greaer foreign bank presence conribues o a more sable banking sysem and a less volaile supply of credi. Thus, i is imporan o gain furher undersanding on he impac of privaizaion and foreign enry boh on he overall banking sysem and on he behavior of he privae, domesic and foreign, and public insiuions in emerging economies. A number of recen sudies have provided empirical evidence regarding he effecs of privaizaion and foreign enry on he efficiency and profiabiliy of emerging economies banking sysems (see Claessens e. al. (2001), Clarke e. al. (2000, 2001), Denizer (1999), Claessens and Glaessner (1999)). Furhermore, here have been sudies ha examined he paerns of marke peneraion by foreign insiuions as well as he reacions of domesic insiuions. These sudies have focused on he lending behavior of he various ypes of insiuions in he sysem (Dages e al. (2000), Berger e. al. (2001)). However, here has no been an empirical analysis of he implicaions of privaizaion and foreign enry on he choices of banking insiuions in emerging economies regarding heir argeed levels of asse porfolio risk. In he only relaed sudy, Laeven (1999) examined he behavior of foreign and domesic banks in Asia during and found ha foreign-owned banks ook relaively limied risks compared o oher banks. The purpose of his paper is o conribue o he above lieraure by examining he impac of he process of privaizaion and foreign enry, in he case of he Argenine banking sysem, on he choice of asse porfolio risk of differen ypes of insiuions. The developmens in he Argenine banking sysem during he second half of he 1990s provide a unique case sudy o address he issue of risk behavior of various ypes insiuions in he sysem. The reason is ha, in his case, he process of privaizaion and foreign enry, which could enhance he incenives for risk-aking in some insiuions, was combined wih a serious effor o srenghen he regulaory environmen in he sysem, which would provide he opposie incenives. 3

4 The policy recommendaions encouraging privaizaion and foreign enry in emerging banking sysems are based on he argumen ha hese changes improve he efficiency and solvency of he sysem. There is ample empirical evidence ha privae banks are more profiable and more efficien han governmen-owned banks. Moreover, i is argued ha he presence of foreign banks in an emerging banking sysem faciliaes he ransfer of improved risk managemen echnology and promoes compeiion. On he oher hand, foreign banks may decide o focus on he low-credi-risk segmen of he marke and arac cusomers away from domesic banks. Faced wih increased compeiion from foreign banks, i is likely ha domesic banks may arge he highcredi-risk segmen of he marke o compensae for heir loss of marke share. Thus, he sraegies ha foreign banks follow o penerae he domesic marke may have significan implicaions for he performance of domesic banks. To provide some addiional empirical evidence on hese effecs, we draw on he recen Argenine experience and aemp o answer he following quesions. How did he privaizaion of domesic insiuions and he enry of foreign insiuions affec he overall level of risk in he sysem? How did he various ypes of insiuions, based on ownership ype, respond o privaizaion and foreign enry in erms of heir choice of a arge level of asse porfolio risk? Finally, did he privaized and foreign-acquired insiuions aler heir behavior in erms of risk, and if so, how? In his sudy, we use a panel daa se of all commercial banks operaing in Argenina during he period To examine any differences in he risk behavior of banks based on ownership, we divide he insiuions in he sysem ino hree caegories: public banks, domesic privae-owned banks, and foreign-owned banks. Moreover, we examine separaely he behavior of privaized and foreign-acquired banks in he sysem. Following previous work by Shrieves and Dahl (1991), Jacques and Nigro (1997) and ohers, we make he assumpion ha he choice of asse porfolio risk by a banking insiuion is made simulaneously wih he choice of is level of capial. Wihin his framework, he argeed risk level of a bank s asse porfolio is affeced by is level of capial. A he same ime, he level of capial is affeced by he level of porfolio risk, as well as he exising regulaory environmen. Thus, our empirical analysis employs a simulaneous equaions framework o examine he effecs of privaizaion and foreign 4

5 enry on boh he level of capial and he level of risk of he various ypes of insiuions in he Argenine banking sysem. Our main findings show ha here was no any significan difference among he hree ypes of insiuions in erms of heir argeed level of capial during he examined period. However, we do find ha privaized insiuions, which were mainly acquired by oher domesic banks, did exhibi a lower arge of capial and a higher arge of risk during he period Bu, our empirical findings are no robus when we divide our sample period ino wo subperiods o accoun for increased foreign bank presence in he sysem. Furhermore, domesic privae banks did exhibi a higher arge level of risk, paricularly during he period The foreign-acquired domesic banks did no exhibi a significanly differen behavior in erms of heir arge level of capial, bu we did find limied evidence ha hey had a higher arge level of risk. Finally, our findings show ha privaized insiuions did aemp o lower heir arge levels of risk afer he privaizaion ook place, paricularly during he period. The paper is organized as follows. Secion II reviews he heoreical argumens on he relaionship beween bank capial and risk. In secion III, we presen he main developmens in he regulaory environmen, as well as he srucural changes ha ook place in he Argenine banking sysem during he 1990s. Secion IV presens he daa and he economeric mehodology, and secion V he economeric resuls. Finally, secion VI presens he conclusions and discusses our main findings. II. The Relaionship Beween Bank Capial and Risk A. Theoreical Argumens The heoreical lieraure on he relaionship beween bank capial and risk emphasizes he relaionship beween changes, raher han levels, in bank capial and changes in risk. In he lieraure, here exis heoreical argumens for boh a posiive and a negaive relaionship. Kim and Sanomero (1988) have saed ha banks wih relaively low risk aversion will selec a low level of capial and a high level of risk. This negaive relaionship beween bank capial and risk has also been emphasized by a number of 5

6 papers ha examined he opion value of deposi insurance (Meron (1977), Kareken and Wallace (1978), Dohan and Williams (1980), Diamond and Dubvig (1986)). Under his argumen, a bank s sockholders may benefi from he deposi insurance subsidy by increasing he bank s leverage and risk. However, maximizaion of he opion value of deposi insurance is resriced by leverage-relaed and risk-relaed coss. If he marginal cos of increased leverage and risk is greaer han he marginal benefi, banks will choose higher levels of capial when hey have higher levels of risk. Banks ha have high levels of leverage and risk also face he coss of complying wih exising regulaory sandards. The hrea of regulaory acion increases he implici coss of higher risk and aemp o change bank behavior. Thus, we should observe a posiive relaionship beween changes in asse risk and changes in bank capial as banks wih higher levels of risk aemp o increase heir capial, bu also banks wih higher levels of capial engage in riskier projecs. Anoher explanaion of he posiive associaion beween bank capial and risk has emphasized he uninended effecs of implemening minimum regulaory capial sandards. Koehn and Sanomero (1980) and Kim and Sanomero (1988) have argued ha consrains on a bank s leverage due o minimum regulaory sandards may cause banks o view leverage and risk as subsiues. A bank ha is forced o lower is leverage due o regulaory sandards will increase is level of risk o achieve a desired level of oal risk. As a resul, we will observe a posiive relaionship beween bank capial and risk for hose banks ha have levels of capial near he minimum regulaory requiremens. Orgler and Taggar (1983) have argued ha here mus be a posiive relaionship beween bank capial and risk coming from he desire of banks o avoid he coss of bankrupcy. Under his argumen, banks ha operae wih high levels of porfolio risk end o increase heir levels of capial due o he fac ha heir probabiliy of bankrupcy is higher. Higher levels of capial reduce he aenion ha a bank receives from regulaors and also send a posiive signal o deposiors regarding is soundness. Finally, here is an argumen for a posiive relaionship beween bank capial and risk based on agency heories of he firm. Agency heories of he firm have emphasized conflics of ineres beween a firm s owners and managers. These divergen ineress could be refleced in banking as differen preferences owards a bank s level of risk. 6

7 Saunders, Srock and Travlos (1990) have made he argumen ha a bank s managers have an incenive o arge a lower level of asse risk han a bank s sockholders would prefer. This is because a bank s managers sand o lose much more when he bank becomes insolven, if i is assumed ha hey are compensaed by fixed claims on he bank and have firm and indusry specific human capial. Thus, managers of banks wih higher levels of asse risk are expeced o choose a higher level of capial o reduce he probabiliy of insolvency and vice versa. B. Empirical Evidence Several empirical sudies have esed he above hypoheses on he relaionship beween changes in bank capial and changes in risk. Mos of hose sudies have employed daa from U.S. banks. In an early empirical sudy of he relaionship beween capial and risk, Shrieves and Dahl (1992) emphasized he endogenous deerminaion of a bank s capial and risk. Using a simulaneous equaions framework, hey found ha he majoriy of banks in heir sample miigae he effecs of increases in capial by increasing heir exposure o asse risk. Their resuls also showed ha his behavior was also presen in banks wih levels of capial above he minimum required, which provides suppor for he argumens of he posiive relaionship beween bank capial and risk based on cos avoidance and managerial risk aversion. Jacques and Nigro (1997) used he same framework o examine he effecs of he inroducion of risk-based capial sandards on bank capial and porfolio risk. Examining he impac of he new sandards on he behavior of U.S. banks during he firs year afer hey were implemened, hey found ha he risk-based sandards brough abou significan increases in capial raios and decreases in porfolio risk of banks ha already me he new risk-based sandards. Aggarwal and Jacques (2001) examined he effecs of he Federal Deposi Insurance Corporaion Improvemen Ac (FDICIA) and is mandae for Promp Correcive Acion (PCA) on bank capial and risk. Their empirical resuls showed ha boh undercapialized and adequaely capialized banks responded o PCA by increasing heir leverage raios. A he same ime, banks also reduced heir levels of credi risk in 7

8 response o PCA. Rime (2001) offered similar evidence from he experience of he Swiss banking sysem. He found ha regulaory pressure led Swiss banks o increase heir levels of capial, bu did no affec heir levels of risk. Finally, in a relaed sudy, Saunders, Srock and Travlos (1990) examined he effecs of ownership srucure on bank risk aking. Using a sample of U.S. bank holding companies over he period , hey found ha banks in which managers had a conrolling minoriy share exhibied a lower risk in heir asse porfolios. III. The Resrucuring of he Argenine Banking Sysem A. Developmens in he Regulaory Framework In an effor o achieve macroeconomic sabiliy and figh hyperinflaion, Argenina adoped a currency board wih he implemenaion of he Converibiliy Plan of As par of he Converibiliy Plan, a number of policies were implemened o liberalize he Argenine banking sysem and o srenghen he regulaory framework. During he period , he Banco Cenral de la Republica Argenina (BCRA) imposed inernaional sandards for capial adequacy, accouning and provisioning, and ook seps o improve banking supervision. The new sandards were broadly based on he recommendaions of he Basle Commiee on Banking Supervision as saed in he 1988 Basle Accord and were implemened in a shor period of ime. Oher major changes in he financial sysem included he independence of he BCRA, as well as resricions on he conduc of moneary policy and access o he lender of las resor faciliies. 1 Following he recommendaions of he Basle Commiee, capial adequacy requiremens were deermined by calculaing he implici risk of he differen asses in a bank s porfolio. The capial-o-risk-adjused-asses raios were calculaed by weighing asses by credi, ineres and marke risk exposure. A he end of 1992, a minimum regulaory requiremen of 9.5% was se for he risk-adjused capial raio. This was 1 Alson and Gallo (2001) have demonsraed, by using an insiuional analysis ha incorporaes boh poliical and economic facors, ha he insiuional changes of he Converibiliy Plan and he Cenral Bank independence were he ulimae deerminan of he privaizaion process in Argenina. 8

9 raised o 11.5% in December 1994 during he afermah of he Mexican peso crisis. In addiion, credi risk requiremens were imposed during ha period, which linked credi risk o he ineres rae charged on each loan, and he minimum requiremen was near 14% by he end of In 1996, marke risk requiremens were added o he minimum sandards and a capial requiremen for ineres rae risk was imposed in Furhermore, he Superinendency (a semi-auonomous supervisory uni wihin he BCRA) esablished a CAMEL raing sysem for banking supervision. Banks wih a poor raing faced more pressure o raise heir capial requiremen. In addiion, he BCRA ried o enhance banking supervision hrough marke discipline by requiring banks o issue subordinaed deb equal o 2% of heir deposis each year. However, due o a deerioraing environmen in local and inernaional markes during he second half of he 1990s, no all he banks in he sysem were able o obain financing from he marke. To address his problem, he regulaory auhoriies esablished differen ways o comply wih his requiremen. Banks were allowed o use subordinaed deb ha was also considered as par of Tier 2 capial. In addiion, banks could hold a deposi or loan from a foreign invesor or oher local bank as a way of complying wih he subordinaed deb requiremen. Bu, banks ha failed o comply wih he requiremen were required o increase heir minimum capial requiremen by 5 percenage poins and heir liquidiy requiremen by 1 percenage poin. Moreover, afer he Mexican peso crisis, he BCRA replaced he reserve requiremen ha applied only o deposis wih a liquidiy requiremen ha applied o all liabiliies. The reasoning for his requiremen was ha, during he crisis of , banks wih a greaer fracion of ime deposis had experienced larger losses in deposis compared o banks wih primarily shor-erm deposis because of a low reserve requiremen on ime deposis. Thus he goal of he new requiremen was o inroduce reserves inended for sysemic liquidiy proecion. The requiremen was implemened a raes ha declined depending on he residual mauriy of each liabiliy and was required for all ypes of liabiliies. Furhermore, even hough deposi insurance had been abolished in he early 1990s, Argenina re-inroduced a limied deposi insurance scheme in This deposi insurance sysem was funded by banks hrough risk-based 9

10 premiums. The sysem iniially covered deposis up o $20,000, bu i was laer expanded o cover deposis up o $30,000. B. Privaizaion and Foreign Enry Even hough here had been some consolidaion aciviy in he Argenine banking sysem in he early par of he 1990s, he process was significanly acceleraed afer he Mexican peso crisis and he negaive spillover effecs on he Argenine banking sysem. The resrucuring process during he second half of he 1990s was characerized by a sronger effor owards privaizaion of several insiuions and by increased foreign enry ino he sysem. The privaizaion effor was mainly concenraed on public provincial banks. During he period , here were 11 privaizaions, followed by an addiional 4 during he period. The privaized insiuions were primarily acquired by oher domesic banks or groups of invesors. As a resul of his privaizaion effor, he number of public-owned banks decreased from 33 in 1994 o 15 in However, here are wo large banks ha remain under public conrol, he Banco de la Nacion Argenina, which is under he conrol of he Federal Governmen, and he Banco de la Provincia de Buenos Aires, which is under he conrol of he provincial governmen of Buenos Aires. These wo banks conrolled approximaely 25% of deposis and 22% of he loan marke a he end of A second major srucural change in he Argenine banking sysem was he significan foreign enry during he second half of he 1990s. Before 1990, here were very few foreign banks presen in Argenina. The removal of resricions on foreign direc invesmen and repariaion of profis ha accompanied he Converibiliy Plan allowed more foreign insiuions o ener he domesic marke. Neverheless, heir share of he domesic sysem was below 20% of oal asses by he end of Afer he Mexican peso crisis, here were effors o encourage he enry of foreign insiuions in an aemp o increase compeiion and efficiency in he domesic banking sysem. Foreign insiuions enered he domesic marke primarily hrough acquisiions of domesic insiuions ha were in good financial condiion and had an esablished nework of branches, raher han rying o acquire privaized provincial banks. This process began 10

11 slowly in , bu acceleraed significanly during As a resul, foreign banks conrolled around half of he asses in he sysem by 1999 (Dages e. al., 2000). Foreign enry resuled in an aggressive compeiion for marke share. Dages e. al. (2000) show ha foreign banks in Argenina peneraed successfully he commercial, governmen, inerbank, and personal loan markes. They also examined he lending behavior of domesic banks, public and privae, as well as foreign banks during he second half of he 1990s and found evidence ha here were differences in lending behavior beween sae banks and domesic privae banks. Bu, domesic privae and foreign banks exhibied comparable lending behavior and had similar composiions of loan porfolios. As Clarke e. al. (2000) also demonsraed, foreign banks compeed aggressively wih domesic privae banks in all segmens of he marke. However, foreign banks did no appear o have a significan effec on he performance of public banks. There is also addiional empirical evidence on he effecs of foreign enry from cross-counry sudies and oher case sudies of banking sysems. Claessens e. al. (2001) examined he effecs of foreign enry in 80 counries during he period Their resuls showed ha foreign banks had higher profis han domesic banks in developing counries. They also found ha an increased presence of foreign banks is associaed wih a reducion in profiabiliy and margins for domesic banks. In a similar sudy, Denizer (2000) examined he effecs of foreign enry on Turkey s banking sysem during he period He found ha foreign enry was associaed wih a reducion on overhead expenses of domesic commercial banks. A he same ime, foreign enry had a srong compeiive effec on he Turkish banking sysem and foreign banks conribued o he developmen of he domesic financial secor. Furhermore, a couple of sudies have examined he impac of he privaizaion and resrucuring process on he degree of marke power and he efficiency of he banking sysem in Argenina. Burdisso e. al. (1998) examined he effecs of privaizaion on he 2 Calomiris and Powell (2000) offer an explanaion for he delayed foreign enry by claiming ha foreign banks waied unil he insiuional changes in he Argenine financial sysem ha ook place in he early 1990s were firs esed by a major exernal shock such as he Mexican peso crisis before hey made he decision o inves in Argenina. 11

12 overall efficiency of he banking sysem during he period They found ha he process of privaizaion and resrucuring ha ook place during ha period had resuled in an improvemen of he performance of banks in he sysem hrough he reducion of heir porfolio risk, hus leading o a more efficien allocaion of credi. In a follow-up sudy, D Amao e. al. (1999) found no evidence of marke power in he reailbanking secor using a sample of he 20 larges banks in Argenina during he period They showed ha he degree of concenraion in more populaed areas where financial markes are more developed was very small. IV. Empirical Mehodology and Descripion of Daa A. Economeric Model Specificaion To examine he effecs of privaizaion and foreign enry on he level of risk in he Argenine banking sysem, we follow he previous lieraure on he relaionship beween bank capial and bank risk (Shrieves and Dahl (1992), Jacques and Nigro (1997)) and assume ha decisions on a bank s level of risk are made simulaneously wih decisions on he level of capial. Thus, following his simulaneous equaions framework, we are also able o examine wheher he various ypes of banking insiuions exhibi differen behavior in erms of heir argeed levels of capial. Following he model of Shrieves and Dahl (1992), we assume ha changes in bank capial and risk are decomposed ino wo componens: a discreionary adjusmen and a change caused by exogenously deermined random shocks. Thus, we wrie (1) d CAP = CAP + E (2) d RISK = RISK + S where CAP j, and RISK j, are he observed changes in capial and risk levels, respecively, for bank j in period. The discreionary adjusmen in bank capial and risk is modeled using he parial adjusmen framework. This recognizes he fac ha banks may no be able o adjus heir desired levels of capial and risk insananeously. Therefore, we wrie 12

13 ( CAPj, CAPj, 1) ( RISK RISK ) d CAP = α (3) d RISK = β (4) 1 where CAP, and j RISK, are he arge levels of capial and risk for bank j, j respecively. Wihin he parial adjusmen framework, he discreionary changes are proporional o he difference beween he arge level and he level exising in period 1. Now, subsiuing equaions (3) and (4) ino equaions (1) and (2), we obain ( CAPj, CAPj, 1 ) E j ( RISK ) RISK 1 S j CAP = +, α (5) RISK = +, β (6) The above wo equaions show ha he observed changes in capial and risk in period are a funcion of he arge levels of capial and risk, he previous period s levels of capial and risk, and any exogenous shocks. In his modeling approach, arge levels of capial and risk are assumed no o be observed direcly. However, we assume ha hese levels depend upon a se of observable variables. The exogenous shocks could capure, for example, a change in he overall macroeconomic environmen or an unexpeced change in a bank s financial condiion. In our model, we use a number of variables ha could influence he arge levels of capial and risk. Several of hese variables have also been used in previous empirical sudies of he relaionship beween bank capial and risk. 3 We use a bank s size ( SIZE ), measured by he naural log of a bank s oal asses, o accoun for he fac ha larger banks may be able o access he capial markes more easily or may be able o have a more diversified porfolio of asses. As Aggarwal and Jacques (2001) poin ou, larger banks may be willing o hold less capial due o heir abiliy o raise addiional funds from he capial marke. In his case, we would expec a negaive relaionship beween a bank s size and is arge level of capial. We, hus, hypohesize ha here is a negaive relaionship beween a bank s size and is arge level of capial, as well as a negaive relaionship beween a bank s size and is arge level of risk due o porfolio diversificaion. also include a bank s profiabiliy (ROA ) o accoun for he fac ha more profiable We 3 See he sudies by Shrieves and Dahl (1992), Jacques and Nigro (1997), Aggarwal and Jacques (2001), Rime (2001), and Aggarwal, Jacques and Rice (2000). 13

14 banking insiuions may be able o increase heir level of capial hrough reained earnings. Based on previous empirical findings, we expec a posiive relaionship beween changes in capial and profiabiliy. Furhermore, Argenine banks are characerized by holding a buffer of excess capial. Due o he fac ha raising capial hrough he markes is a cosly process in Argenina, we expec ha banks would aemp o reain profis as a means of raising heir capial raios. We also include a proxy of a bank s curren losses. This variable, approximaed by he raio of curren loan loss provisions o poenial bad loans ( LLOSS ), accouns for he fac ha loan losses will affec a bank s risk-adjused asses o oal asses (Rime, 2001). We include his variable in he risk equaion and assume ha banks wih higher loan losses will exhibi lower fuure levels of risk-adjused asses. 4 We also assume ha banks wih more expeced losses would aemp o raise heir levels of capial o comply wih regulaory requiremens. In addiion, we include he raio of governmen securiies o oal asses ( BONDS ) o conrol for he fac ha Argenine banks held a significan porion of heir asses in erms of governmen securiies. We expec ha banks wih high levels of governmen securiies in heir porfolios will be able o have higher capial raios hrough sales of securiies, paricularly in an environmen of falling ineres raes. Alernaively, as Aggarwal and Jacques (2001) poin ou, if banks wih large holdings of governmen securiies reained, raher han sold, hese securiies during a falling rae environmen, hen hey may be required o have lower levels of capial o comply wih exising regulaions. Moreover, banks wih high raios of governmen securiies in heir asse porfolios will be exhibiing lower levels of risk. To capure he effec of he liquidiy requiremen imposed on Argenine banks afer 1995, we include he raio of liquid reserves o oal asses ( LIQUIDITY ). Banks wih higher liquidiy raios are faced wih less risk and, hus, need o hold less capial. A he same ime, banks wih higher liquidiy may be more willing o increase heir levels of risk. Thus, we expec o find a negaive relaionship beween he raio of liquid reserves o oal asses and he level of a bank s capial and a posiive relaionship beween his raio and he level of a bank s porfolio risk. 4 Banks may also be faced wih unexpeced losses due o evens no under heir conrol. Our model capures hese unexpeced losses hrough he error erm in he capial and risk equaions. 14

15 Finally, we use dummy variables in he esimaed equaions of capial and risk o accoun for differences in he behavior of banking insiuions based on ownership ype. We firs include wo dummy variables, ( PRIVATE ) and ( ) FOREIGN, o accoun for differences in he arge levels of capial and risk beween he domesic privae banks and foreign banks, on he one hand, and public banks, on he oher. Based on previous empirical findings (see La Pora e. al. (2000) and World Bank (2001)), we assume ha privaely owned banks, eiher domesic or foreign, will be employing beer echniques o measure and manage risks, and, hus, will have a lower arge level of risk compared o public banks. Furher suppor of our hypohesis comes from evidence by Barh, Caprio and Levine (2001) and Caprio and Marinez-Peria (2000), which shows ha saeownership of banks has a posiive effec on he probabiliy of a banking crisis and is associaed wih less compeiion in he banking sysem. Furhermore, evidence from case sudies of banking sysems in disress shows ha sae ownership of banks ends o reduce he incenives ha bankers have for acquisiion of informaion, which is essenial for he efficien use of he banking sysem (see Goldsein and Turner (1996) and Goldsein (1997)). Second, we use wo dummy variables, ( PRIVATIZED ) and ( FOREIGN 1) separae hose banks ha were privaized afer 1994, as well as hose ha were acquired by foreign banks during he same period, from he res of he banks in he sysem. The firs group includes mainly former provincial banks ha were acquired by oher domesic banks or groups of domesic invesors. The second group includes several domesic banks ha were acquired by foreign banks, as well as domesic banks in which foreign banks have a minoriy shareholding. Wih he excepion of one bank, he Banco Frances del Rio de la Plaa, which is lised as having a minoriy shareholding of 32% by BBVA in 1999, he remaining banks in his group have a foreign majoriy shareholding. We expec ha hese banks will have a lower arge of risk and higher arge of capial afer hey were privaized or acquired by foreign banks. This will be due o he applicaion of beer managemen echniques, he removal of bad asses from heir porfolios, and he incenives of he new managemen o esablish a repuaion and improve he solvency of he new insiuion., o 15

16 Furhermore, here is evidence from Dages e. al. (2000) and Clarke e. al. (2000) ha foreign banks in Argenina were compeing primarily wih domesic privae banks for he same porion of he marke during our examined period. Dages e. al. (2000) also show ha he composiion of he loan porfolios of domesic privaely owned and foreign banks were very similar beween 1994 and 1999, despie he fac ha he share of foreign banks in he Argenine lending marke rose from 18% in 1994 o 48.1% in Thus, we could expec ha, given he rapid increase in he share of foreign banks in he Argenine banking sysem, here may have been a negaive effec on domesic privae banks based on he implicaions of he cherry pick hypohesis of foreign enry. According o his hypohesis, foreign banks cherry pick he mos profiable cusomers in he marke, forcing domesic banks o cover for heir los marke share by increasing heir lending o higher-risk cusomers and, hus, increase he overall level of risk of heir porfolios. We would, herefore, expec o find a higher arge level of risk among domesic privae banks, as well as privaized banks acquired by oher domesic banks or groups of invesors, compared o he res of he banks in he sysem, especially during he period afer 1997 when several foreign banks enered he Argenine banking sysem. Finally, wih respec o he foreign-acquired banks, we would expec o find ha hey exhibi a lower arge of risk and a higher arge of capial. This would be explained by he applicaion of beer managemen echniques and heir abiliy o arge he low credi risk segmen of he lending marke. B. Definiions of Capial and Risk In our analysis, we employ variables o measure bank capial and risk similar o hose used in several oher sudies. For he definiion of capial, we use boh he capialo-risk-adjused-asses raio ( CAPRWA ) and he raio of oal capial o oal asses ( CAPTA ). Boh variables have been used in previous sudies on bank capial and risk. 5 In Argenina, he BCRA follows he recommendaions of he Basle Commiee o calculae minimum risk-adjused capial raios. The minimum capial requiremen is calculaed by weighing asses based on heir credi, marke and ineres rae risks. The 5 For example, see Jacques and Nigro (1997) and Aggarwal and Jacques (2001). 16

17 BCRA requires ha banks mainain a 11.5% minimum requiremen. This is he raio of oal (Tier I plus Tier II) capial over risk-adjused asses. The oal capial used in he calculaion of his raio is defined wih he erm Inegraion in he source of bank informaion published by he BCRA ha we use in his paper. Thus, in our model, we use boh he raio of Inegraion over risk-adjused asses and he corresponding raio over oal asses. As for our definiion of a bank s asse porfolio risk, we follow previous sudies by Avery and Berger (1991), Berger and Udell (1993) and Berger (1995) and use he raio of risk-adjused asses o oal asses as our measure of risk ( RISK ). Avery and Berger (1991) have shown ha his raio is highly correlaed wih a bank s risky behavior. In addiion, mos of he banks in our sample do no have raded sock. Therefore, we could no use a capial marke measure of risk, such as he variance of sock reurns. CAP β 5 Thus, our esimaed model is = β 0 LIQUIDITY + β SIZE β + β RISK 2 ROA + β CAP 7 + β 3 1 LLOSS BONDS + β OWNERSHIP + ε 8 + β 4 + (7) RISK δ 4 = δ 0 LIQUIDITY + δ SIZE 1 + δ 5 CAP + δ 2 LLOSS + δ 6 RISK + δ 1 3 BONDS + δ OWNERSHIP u (8) In our esimaion approach, we accoun for he panel naure of our daa and we assume ha he error erm in he above equaions is wrien as µ +, i.e., we assume a j v j, one-way error componen model. In his case, µ j denoes he unobservable bankspecific effec and v, denoes he remainder disurbance. We follow a singe-equaion j esimaion approach, such as 2SLS, o esimae our simulaneous equaion sysem. A full-informaion esimaion approach, such as 3SLS, would resul in more efficien esimaes, bu i also depends grealy on a correc specificaion of he equaions in our sysem. To avoid any esimaion problems due o misspecificaion of he model, we used he 2SLS mehod. Given our assumpion abou he error erm, he generalized 2SLS esimaes are consisen, bu are no efficien. To obain efficien esimaes, we employ Balagi s (1981) error-componen 2SLS (EC2SLS), which akes ino accoun he one-way 17

18 error componen specificaion in our model. Thus, we esimae each of he equaions in he sysem wih EC2SLS. C. Descripion of Daa In his sudy, we use daa from commercial banks in Argenina during he period The daa source is he annual publicaion Informacion de Enidades Financieras published by he BCRA. This publicaion conains deailed balance shee and income saemen daa, as well as a number of indicaors on efficiency, performance, capial adequacy and risk profile for all he commercial banks operaing in Argenina. Due o he exisence of a one period lag in our esimaed economeric model, we lose one annual observaion from As a resul, our sample is a panel of 330 bankyear observaions on Argenine banks during he period The examined period coincides wih he major changes ha ook place in he Argenine banking sysem during he second half of he 1990s. To consruc our final sample, we excluded all he cooperaive banks in Argenina. This reduced our sample by a oal of 20 observaions over he four-year period. These banks have a very small share of he banking sysem and hey also have a differen ownership srucure compared o he res of he banks in our sample. Moreover, we omied 12 bank-year observaions for which here was missing informaion on he variables used in our model. Table 1 summarizes he main characerisics of he banks used in our sample. The number of banks in each year varies from 74 o 93. Banks in Argenina hold on average abou half of heir asses in he form of loans and around 12 percen of heir asses are in he form of governmen securiies. During our examined period, banks are on average well capialized wih a capial-asse raio around 16 percen and a capial-o-riskadjused-asses raio beween 22 and 27 percen. The Argenine banks also exhibi a decline in heir raio of risk-adjused asses o oal asses, as well as he share of problemaic loans o oal asses. This indicaes ha here was an effor o reduce he overall level of risk exposure of banks in he sysem. Finally, Argenine banks exhibied on average an improvemen in profiabiliy during , followed by a decline in 1999 due o a deerioraing macroeconomic environmen. 18

19 Examining he disribuion of capial and risk by ownership ype during our examined period, as presened in Table 1, we firs observe ha all insiuions in he sysem are well capialized. Only he former provincial banks ha were ransferred under privae conrol have lower capial adequacy raios compared o he res of he banks in he sysem, bu hese raios also increase by In erms of leverage raios, public and privae banks show an increase during our examined period, while foreign banks exhibi a considerable decrease and foreign acquired banks do no exhibi any significan change. Bu, in erms of risk-adjused capial raios, all banks in he sysem, excep foreign banks, have increased hese raios by he end of he 1990s. The foreign banks show a U-shaped paern wih a fall unil 1998 and a subsequen increase in Finally, in erms of risk, all banks in he sysem, excep for he privaized banks, show a decrease in risks during he examined period. As a preliminary analysis of our hypoheses, we presen in Table 2 he resuls of nonparameric ess of differences in means for bank risk by various ypes of ownership and by year. We conduc wo ypes of ess. Firs, we es for differences in he mean values of risk-adjused asses o oal asses by ype of ownership holding he year consan. Then, we es differences in mean values across years holding he ype of ownership consan. To examine he effecs of he process of privaizaion and foreign enry on he risk of he various insiuions in he Argenine banking sysem, we chose o compare banks in 1996 wih banks in The resuls of Table 2 show ha for 1996 here is a saisically significan difference in risk only beween privae and foreign banks. In his case, privae banks have a saisically significan higher mean level of risk compared o foreign banks. However, we do no find any saisically significan difference in he mean levels of risk eiher beween privae and public banks or beween foreign and public banks for ha year. Performing he same ess for he year 1999, we do find a saisically significan difference in means beween boh privae and public banks and beween privae and foreign banks. In boh cases, privae banks have a higher average level of asse porfolio risk compared o oher banks in he sysem in This suppors our hypohesis ha privae banks aemped o cover heir losses of marke share due o foreign enry by argeing he higher risk segmen of he marke. 19

20 We also performed -ess of differences in means beween 1996 and 1999 holding consan he ownership ype of he insiuion. The resuls of hese ess did no show any saisically significan difference in he mean levels of risk for he various ypes of insiuions beween he wo years. However, we mus poin ou ha he mean levels of risk are lower in 1999 compared o 1996 for all he ypes of insiuions. Finally, we performed a -es for difference in he mean levels of risk beween 1996 and 1999 for he overall banking sysem. The resuls of his es showed a saisically significan difference and, more specifically, a decline in he average level of risk in he overall sysem beween 1996 and These resuls provide supporing evidence for he argumen ha foreign enry improves he sabiliy of he banking sysem in emerging economies. V. Economeric Resuls A. Main Resuls The main resuls from he esimaion of our simulaneous equaions model wih EC2SLS are presened in Tables 4 and 5. In Table 4, he variable ( CAP ) is defined as he raio of oal capial (Tier I plus Tier II) o risk-adjused asses. In Table 5, we use he raio of oal capial o oal asses in our esimaion. In boh ables, he resuls presened are for he whole period, , and he variable ( RISK ) is given by he raio of risk-adjused asses o oal asses in all esimaed equaions. Our resuls show a negaive relaionship beween changes in capial ( CAP) and changes in risk ( RISK ) in he Argenine banking sysem during he second half of he 1990s. This relaionship is significan in mos of he equaions and, paricularly, when we use he raio of oal capial o risk-adjused asses as a measure of capial. Previous sudies by Shrieves and Dahl (1992), Aggarwal and Jacques (2001) using daa from U.S. banks and by Rime using daa from Swiss banks have found a posiive relaionship beween changes in capial and changes in risk. However, Jacques and Nigro (1997) found an inverse relaionship beween hese wo variables wih daa from U.S. banks 20

21 during 1991, he period afer he implemenaion of risk-based capial adequacy sandards. Our resuls seem o show suppor for he deposi insurance subsidy argumen, which suggess ha a bank will maximize he opion value of deposi insurance by increasing is leverage and risk. This explanaion is consisen wih he fac ha Argenina re-inroduced a limied, fully funded, deposi insurance sysem in 1995, which covered deposis up o $20,000 iniially, and was subsequenly raised o $30,000. Bu, he above argumen implies a fla-rae deposi insurance pricing formula, while he deposi insurance scheme in Argenina was funded hrough premia on banks calculaed using a risk-based pricing formula (Calomiris and Powell, (2000)). A possible explanaion of he negaive relaionship beween bank capial and risk shown in our resuls is ha banks in Argenina were sill operaing under false incenives. Even hough several insiuions were allowed o fail afer he Mexican peso crisis in 1995 and, in some cases, deposiors suffered significan losses, he governmen-assised privaizaions and he re-inroducion of deposi insurance may have sen he signal o he banks ha he marginal benefis from increased asse risk were higher. Given he hisorical experience of he banking sysem in Argenina, several banks may have aemped o exploi he re-inroducion of deposi insurance by increasing heir risk and leverage. Alernaively, as Marcus (1984) and Keeley (1990) have poined ou, increased compeiion and enry in he banking sysem could cause bank charer values o decline and lead banks o swich from a posiive capial-risk relaionship o a negaive one. This argumen is also consisen wih negaive relaionship beween capial and risk given by our esimaed model and he experience of he Argenine banking sysem wih privaizaion, foreign enry and increased compeiion during he second half of he 1990s. The oher explanaory variables of he arge levels of capial and risk are overall saisically significan and have he expeced signs. A bank s size ( SIZE ) has a negaive and saisically significan effec on is arge level of capial. This negaive effec appears mainly when capial is defined as he raio of capial o oal asses. We also find a negaive and saisically significan effec of a bank s size on is arge level of risk. This would imply ha larger banks have a greaer abiliy o manage he risk in heir asse porfolio or can arge cusomers wih lower credi risk. We find a posiive, bu no 21

22 saisically significan, effec of a bank s profiabiliy ( ROA ) on is level of capial. The posiive coefficien is consisen wih resuls from previous sudies, which show ha banks wih higher earnings can raise heir levels of capial hrough inernal sources. Banks wih higher loan loss provisions ( LLOSS ) have a higher arge level of capial. This relaionship is significan in all of he esimaed equaions. This is consisen wih he argumen ha banks wih higher expeced losses aemp o increase heir levels of capial o comply wih exising regulaory sandards and cover any poenial losses. We also find a saisically significan posiive relaionship beween loan loss provisions and he level of bank risk. This confirms he argumen ha banks wih lower asse qualiy, meaning higher loan loss provisions, have higher risk. Banks wih more governmen securiies ( BONDS ) in heir porfolios have a lower level of risk. This relaionship is saisically significan in all of he esimaed equaions. This resul is explained by he lower risk involved wih invesmen in governmen securiies and also reflecs he mehodology of he Basle Commiee s recommendaions of assigning a lower risk weigh o governmen securiies. We also find ha he share of governmen securiies in a bank s asse porfolio had a saisically significan posiive effec on he level of capial when capial is measured in erms of risk-adjused asses. The raio of liquid asses o oal asses ( LIQUIDITY ) has a negaive, bu no saisically significan, effec on he level of capial and, in general, a posiive and no saisically significan effec on he level of risk in mos of he esimaed equaions. This finding is consisen wih he hypohesis ha banks wih more liquid asses are exposed o lower risks and, hus, are in need of less capial. Finally, in all of he esimaed equaions, he lagged values of capial and risk are saisically significan and all appear wih a negaive sign. The esimaed coefficiens for lagged capial range from o The corresponding esimaed coefficiens for lagged risk range from o 0.949, which imply a much faser adjusmen of risk owards he desired levels. Overall, hese resuls imply a relaively fas adjusmen of capial and risk owards he desired levels among Argenine banks during he examined period. This could be explained by he sric regulaory environmen, which led banks o ake seps o ensure compliance wih exising regulaions. 22

23 B. The Effec of Ownership on Bank Capial and Risk The impac of he various ypes of ownership on a bank s arge level of capial and, paricularly, on is arge level of risk is examined by including dummy variables in he esimaed equaions presened in Tables 4 and 5. Firs, in equaions (1) and (2) in Table 4, we include dummy variables for domesic privae banks ( PRIVATE ) and foreign banks ( FOREIGN ). Our resuls show ha here was no saisically significan difference in eiher he levels of capial or he levels of risk beween he domesic privae banks or he foreign banks and he public banks. This finding mos likely reflecs he sric regulaory environmen in Argenina during he second half of he 1990s. Regulaions regarding capial and risk were sricly enforced by he BCRA and here was a srong incenive for all he banks in he sysem o comply wih exising regulaions in order o avoid any penalies. In addiion, banks in Argenina did hold excess amouns of capial during he lae 1990s, which reflecs a response o he negaive effecs of he Mexican peso crisis on he sysem and he deerioraion in world capial markes during he period of he urmoil in Eas Asia and Russia. Neverheless, o examine our hypohesis furher, and o capure any addiional effecs, paricularly on he arge levels of risk of domesic privae banks as a resul of he rapid growh in marke share of foreign banks during he second half of he 1990s, we decided o divide our sample in wo subperiods and re-esimae our model. Hence, we reesimaed our model using daa for he period and daa for he period separaely. The reasoning for his cuoff is ha many foreign banks enered he Argenine banking sysem aggressively during 1997 and afer. Thus, we expec ha he increased compeiion from foreign banks could have had a negaive effec on he credi qualiy of he asse porfolios of domesic privae banks afer 1997, as he domesic privae banks aemped o cover for heir los clienele by expanding heir businesses o higher-risk areas. Tables 6 and 7 presen he resuls from he esimaion of our model for he periods and , respecively. Even hough we do no find any saisically significan difference among he behavior of privae or foreign banks and public banks for he period , we do find ha he domesic privae banks exhibied a higher, 23