1 THE DETERMINANTS OF THE TUNISIAN BANKING INDUSTRY PROFITABILITY: PANEL EVIDENCE Samy Ben Naceur ERF Research Fellow Department of Fnance Unversté Lbre de Tuns Avenue Khéreddne Pacha, 002 Tuns Emal : October 2003 ABSTRACT Ths paper nvestgates the mpact of bank s characterstcs, fnancal structure and macroeconomc ndcators on bank s net nterest margns and proftablty n the Tunsan bankng ndustry for the perod. Frst, ndvdual bank characterstcs explan a substantal part of the wthn-country varaton n bank nterest margns and net proftablty. Hgh net nterest margn and proftablty tend to be assocated wth banks that hold a relatvely hgh amount of captal, and wth large overheads. Other mportant nternal determnants of bank s nterest margns bank loans whch have a postve and sgnfcant mpact. The sze has mostly negatve and sgnfcant coeffcents on the net nterest margns. Ths latter result may smply reflect scale neffcences. Second, the paper fnds that the macro-economc ndcators such nflaton and growth rates have no mpact on bank s nterest margns and proftablty. Thrd, turnng to fnancal structure and ts mpact on bank s nterest margn and proftablty, we fnd that concentraton s less benefcal to the Tunsan commercal banks than competton. Stock market development has a postve effect on bank proftablty. Ths reflects the complementartes between bank and stock market growth. We have found that the dsntermedaton of the Tunsan fnancal system s favourable to the bankng sector proftablty. Key words: bank nterest margn, bank proftablty, panel data, Tunsa. I. Introducton Restructurng of the commercal bankng system n Tunsa begun n 987, and was ntended to nstl competton n he bankng sector, moblze savngs and lead to a more effcent allocaton of resources. Reforms were artculated around fve axes: lberalzaton of nterest rates and credt allocaton, ntroducton of new ndrect monetary polcy,
2 strengthenng prudental regulaton, openng the fnancal sector to foregn fnancal nsttutons and promoton of the equty market. All these developments would certanly have mplcatons on the nterest margn and proftablty of the Tunsan bankng ndustry. Ths research paper was ntated by a seres of queston: Why are some commercal banks more successful than others? To what extent are dscrepances n bank s proftablty due to varaton n endogenous factors under the control of bank management and to what extent, do external factors mpact the fnancal performance of these banks? Answers to the questons would be helpful to dentfy the determnants of successful Tunsan commercal banks n order to formulate polces for mproved proftablty of these nsttutons. Ths paper follows n the footsteps of Abreu and Mendes (2002), Demerguç-Kunt and Huzngha (999) and Ben Naceur and Goaed (200) among others. It extends the exstng lterature several ways. Frst, usng bank level data for Tunsa n the perod (Ben Naceur and Goaed, 200 use only the perod), we provde statstcs on sze and decomposton of bank s nterest margn and proftablty. Second, the paper uses regresson analyss (panel data wth random effects) to fnd the underlyng determnants of Tunsan bankng ndustry performance. To ths end, a comprehensve set of nternal characterstcs s ncluded as determnants of bank s net nterest margn and proftablty. These nternal factors nclude equty, overhead, and nterest bearng assets. Thrd, whle studyng the mpact of bank s characterstcs on ther performance, we nclude macroeconomc (nflaton and growth) and fnancal structure ndcators (bank and market sze, and concentraton) to control for the effect of external factors (not ncluded n Ben Naceur and Goaed, 200). The remander of the paper s organzed as follows. A bref revew of relevant lterature s presented n secton II. The emprcal models we employ are descrbed n secton III, along wth a descrpton of the data used n the study. II. The Determnants of Bank Performance: Lterature Revew Studes on the determnants of bank s nterest margn and proftablty have focused whether on a partcular country (Berger, 995; Guru et al., 2002; Barajas et al., 200; Ben Naceur and Goaed, 200) and on a panel of countres (Abreu and Mendes, 2002; Demerguç- Kunt and Huzngha, 999).
3 II. Sngle country studes As most of the studes on bank performance are conducted n the US and emergng markets, we wll dvde our presentaton n two parts: US evdence and emergng market studes. The emprcal evdence n the US s due to Berger (995), Neeley and Wheelock (997) and Angbazo (997). Berger (995) examnes the relatonshp between the return on equty and the captal asset rato for a sample of US banks for the tme perod. Usng the Granger causalty model, he shows that the return of equty and captal to asset rato tend to be postvely related. Neeley and Wheelock (997) explore the proftablty of a sample of nsured commercal banks n the US for the perod. They fnd that bank performance s postvely related to the annual percentage changes n the state s per capta ncome. Anghazo (997) nvestgates the determnants of bank net nterest margns for a sample of US banks for perod. The results for the pooled sample documents that default rsk, the opportunty cost of non-nterest bearng reserves, leverage and management effcency are all postvely assocated wth bank nterest spread. The man Studes on the determnants of bank s performance n emergng countres were carred out n Colomba (Barajas et al.,999), Brasl (Afanaseff et al., 2002), Malaysa (Guru et al., 2002) and Tunsa (Ben Naceur and Goaed, 200). Barajas et al. (999) document sgnfcant effects of fnancal lberalzaton on bank s nterest margns for the Colomban case. Although the overall spread has not declned after fnancal reform, the relevance of the dfferent factors behnd the bank spreads were affected by such measures. Another change lnked wth the lberalzaton process was the ncrease of the coeffcent of loan qualty after the lberalzaton. Afanaseff et al. (2002) make use of panel data technques to uncover the man determnants of the bank nterest spreads n Brazl. A two-step approach due to Ho and Saunders (98) s used o measure the relatve mpact of the mcro and macro factors. The results suggest that macroeconomc varables are the most relevant elements to explan bank nterest spread n Brazl. Ben Naceur and Goaed (200) nvestgate the determnants of the Tunsan bank s performances durng the perod They ndcates that the best performng banks are those who have struggled to mprove labour and captal productvty, those who have mantaned a hgh level of depost accounts relatve to ther assets and fnally, those who have been able to renforce ther equty. Guru et al. (2002) attempt to dentfy the determnants of successful depost banks n order to provde practcal gudes for mproved proftablty performance of these nsttutons. The study s based on a sample of seventeen
4 Malaysan commercal banks over the perod. The proftablty determnants were dvded n two man categores, namely the nternal determnants (lqudty, captal adequacy and expenses management) and the external determnants (ownershp, frm sze and external economc condtons). The fndngs of ths study revealed that effcent expenses management was one of the most sgnfcant n explanng hgh bank proftablty. Among the macrondcators, hgh nterest rato was assocated wth low bank proftablty and nflaton was found to have a postve effect on bank performance. II.2 Panel country studes The panel country studes were focused on European companes (Molyneux and Thornton, 992; Abreu and Mendes, 2002), MENA countres (Bashr, 2000), and developed and developng countres (Demerguç-Kunt and Huzngha 999, 200). Molyneux and Thornton (992) were the frst to explore thoroughly the determnants of bank proftablty on a set of countres. They use a sample of 8 European countres durng the perod. They fnd a sgnfcant postve assocaton between the return on equty and the level of nterest rates n each country, bank concentraton and government ownershp. Abreu and Mendes (2002) nvestgate the determnants of bank s nterest margns and proftablty for some European countres n the last decade. They report that well captalzedbanks face lower expected bankruptcy costs and ths advantage translate nto better proftablty. Although wth a negatve sgn n all regressons, the unemployment rate s relevant n explanng bank proftablty. The nflaton rate s also relevant. Bashr (2000) examnes the determnants of Islamc bank s performance across eght Mddle Eastern countres for perod. A number of nternal and external factor were used to predct proftablty and effcences. Controllng for macroeconomc envronment, fnancal market stuaton and taxaton, the results show that hgher leverage and large loans to asset ratos, lead to hgher proftablty. The paper also reports that foregn-owned banks are more proftable that the domestc one. There s also evdence that taxaton mpacts negatvely bank proftablty. Fnally, macroeconomc settng and stock market development have a postve mpact on proftablty. In a comprehensve study Demerguç-Kunt and Huzngha (999) examne the determnants of bank nterest margns and proftablty usng a bank level data for 80 countres n the 988-
5 995 perod. The set of varables ncludes several factors accountng for bank characterstcs, macroeconomc condtons, taxaton, regulatons, fnancal structure and legal ndcators. They report that a larger rato of bank assets to GDP and a lower market concentraton rato lead to lower margns and profts. Foregn banks have hgher margns and profts than domestc banks on developng countres, whle the opposte preval n developed countres. On an another lnked paper, Demerguç-Kunt and Huzngha (200) present evdence on the mpact of fnancal development and structure on bank proftablty usng bank level data for a large number of developed and developng countres over the perod. The paper fnds that fnancal development has a very mportant mpact on bank performance. Specfcally, the paper reports that hgher bank development s related to lower bank performance (Tougher competton explans the decrease of proftablty). Stock market development on the other hand, leads to ncreased profts and margns for banks especally at lower levels of fnancal development, ndcatng complementartes between bank and stockmarket. III. Emprcal methodology and sample data III. Data sources and varable defnton The data used n the emprcal work were extracted from the Central bank data base. The sample nclude the man depost banks n Tunsa (0 banks) over the perod As all the banks n our sample are observed n the entre perod, we wll use n our emprcal work balanced panel data. The emprcal test s concerned wth the determnants of nterest margn and proftablty of the Tunsan depost banks. We use captal rato, overhead, loan and lqudty ratos as proxes for nternal ndcators. Meanwhle macro-economc measures and fnancal structure ndcators are used as external factors. A lnear equaton relatng the performance measures to a varety of factors s dsplayed n equaton : Per j,t = f (BC j,t + M t + FS t ) () Where: Perf j,t represents two alternatve performance measures for the frm j durng the perod t; BC j,t are bank varables for bank j at tme t; M t are macro-economc varables; FS t are measures of fnancal structure ndcators.
6 Although the prmary focus of ths paper s the relatonshp between net nterest margns and proftablty, and bank s characterstcs ndcators, the ncluson of macro-economc varables and fnancal structure ndcators s ntended to control for cyclcal factors that mght mpact bank proftablty n Tunsa. Two measures of performance are used n the study: the net nterest margn (NIM) and the return of assets (ROA). The NIM varable s defned as the net nterest ncome dvded by total assets. ROA s a rato computed by dvdng the net ncome over total assets. NIM and ROA have been used n most banks performance studes. ROA measures the proft earned per dollar of assets and reflect how well bank management use the bank s real nvestments resources to generate profts whle NIM s focused on the proft earned on nterest actvtes. Fve bank s characterstcs ndcators are used as nternal determnants of performance. They comprse the rato of overhead to total assets (OVERHEAD), the rato of equty captal to total assets (CAP), the rato of bank s loans to total assets (BLOAN), the rato of nonnterest bearng assets to total assets (NIBA) and the log of bank assets (LNSIZE). The rato of overhead to total assets s used to provde nformaton on varaton n bank costs over the bankng system. It reflects employment as well as the total amount of wages and salares. OVERHEAD s expected to have a negatve mpact on performance because effcent banks are expected to operate at lower costs. Bank loans are expected to be the man source of ncome and are expected to have a postve mpact on bank performance. Other thngs constant, the more deposts are transformed nto loans, the hgher the nterest margn and profts. However, f a bank needs to ncrease rsk to have a hgher loan-to-asset rato, then profts may decrease. In addton, as bank loans are the prncpal source of ncome, we expect that non nterest bearng assets mpact negatvely on profts. We also expect that the hgher equty-to-asset rato, the lower the need to external fundng and therefore hgher proftablty. It also a sgh that well captalzed bank face lower costs of gong bankrupt and then cost of fundng s reduced. The sze of the bank s also ncluded as an ndependent varable to account for sze related economes and dseconomes of scale. In most of the fnance lterature, the total assets of the banks are used as a proxy for bank sze. However, snce the other dependent varables n the
7 models such as ROA were deflated by total assets t would be approprate to log total assets before ncludng t n the models. To solate the effects of bank s characterstcs on performance, t s necessary to control for other factors that have been used as determnants of bank proftablty. Two sets of control varables are expected to nfluence banks performance: the macro-economc and the fnancal structure ndcators. Two macro-economc varables are used: nflaton (INF) and GDP per capta growth (GROWTH). Prevous studes have reported a postve assocaton between nflaton and bank proftablty. Hgh nflaton rates are generally assocated wth hgh loan nterest rates, and therefore, hgh ncomes. However, f nflaton are not antcpated and banks are sluggsh n adjustng ther nterest rates then there s a possblty that bank costs may ncrease faster than bank revenues and hence adversely affect bank proftablty. The GDP per captal growth s expected to have a postve mpact on bank s performance accordng to the well documented lterature on the assocaton between economc growth and fnancal sector performance. We also examne how the performance of the bankng sector s related to the relatve development of the banks and stock markets. Relatve sze (RSIZE) s calculated as the rato of the stock market captalzaton to total assets of depost money banks. In addton, we use stock market captalzaton dvded by GDP (MCAP) as a proxy of fnancal market development and as a measure of the sze of the equty market. The sze of the bankng sector (SBS) s measured by the rato of total assets of the depost banks to GDP and s ntended to measure the mportance of bank fnancng n the economy. MCAP and SBS may also ndcate the complementartes or substtutablty between bank and equty market fnancng. Both varables are expected to nfluence postvely bank performance. Bank concentraton (CONC) equals the fracton of bank assets held by the three largest commercal banks n the country. Most of the evdence on bank structure and performance s devoted to the US bankng ndustry, provdng generally conflctng results. Some evdence ndcates that banks n hghly concentrated local markets charge hgher rates on loans, pay lower rates on deposts, and are slower to reduce rates n response to Federal Reserve decrease n nterest rates than banks n less concentrated markets. Alternatvely, Smrlock (985) fnds that nterest rate spreads are narrower n concentrated bankng ndustry, whle Keeley and Zmmerman (985) fnd more mxed evdence. Berger (995) concludes that the relatonshp between bank concentraton and performance n the US depend crtcally on what other factors are held constant.
8 III.2 Econometrc modelng In ths study, fxed effects as well as random effects models are consdered. The fxed effects model s smpler to conduct and s defned accordng to the followng regresson model: () yt = α + β X t + εt =, L, N ; t =, L, T y t ndcates the dependent varables whle varables. X t determnes the vector of k explanatory α, =, L, N, are constant coeffcents specfc to each country. Ther presence assumes that dfferences across the consdered banks appear by means of dfferences n the constant term. These ndvdual coeffcents are estmated together wth the vector of coeffcentsβ. In order to valdate the fxed effects specfcaton, the queston s to prove, accordng to the emprcal applcaton, that the ndvdual coeffcents α, =, L, N, are not all equal. Ths corresponds to the followng jont null hypothess: (2) H α = L = α = α 0 : N It s rather the acceptaton of the alternatve hypothess whch s nterestng f we want to dfferentate between the stuatons n each bank consdered n the sample and confrm the exstence of sgnfcant heterogenety across banks. The approprate statstc of the test s a Fsher dstrbuted one wth N, hypothess and s defned as follows: N = T N k degrees of freedom under the null (3) SSR 0 SSR F = SSR N = T N k N where SSR 0 and SSR are, respectvely, the sum of squared resduals provded by the estmaton of the constraned model (under the null hypothess that s no ndvdual specfc coeffcents are consdered) and the sum of squared resduals relatve to the fxed effects model (equaton ()).
9 In the random effects case, the model s defned as follows: (4) yt = β X t + εt =, L, N ; t =, L, T where ε = µ + υ reflect the error component dsturbances. The ndvdual specfc effects t t 2 are random and dstrbuted normally ( IIN ( 0, σµ ) 2 terms υ whch are also dstrbuted normally ( υ IIN( 0 σ ) t µ. They are ndependent of the resdual t, υ. The estmaton of the model s conducted by the feasble generalzed least squares method. Frst, convergent estmates of 2 the varances σ 2 µ and σ υ are needed. They are obtaned by the followng formulae: (5) σˆ 2 υ = N T = t= N = ( υˆ υˆ ) T N k t. 2 N 2 (6) µ υ 2 2 σˆ = y βˆ. b X. σˆ N k = T ˆυ t are the resduals ssued from the estmaton of the fxed effects model (equaton ()) and ˆυ. are ndvdual means of these resduals over each tme perod relatve to each bank. Next, the frst term n equaton (6) ndcates the resduals ssued from the estmaton of the unt means regresson where ˆβ b are called the between estmators. The second stage conssts n the estmaton by ordnary least squares of the followng transformed regresson model: (7) yt + θˆ y ˆ ˆ. = β X t X + θ. + ε t + θ ε. wth: 2 (8) ˆ σˆ υ θ = =, L, N 2 2 σˆ + T σˆ υ µ Fnally, a Hausman specfcaton test s conducted n order to compare the two categores of specfcatons. It s proven that, under the null hypothess, the two estmates
10 (equatons () and (7)) could not dffer systematcally snce they are both consstent. So, the test can be based on the dfference. Under the null hypothess, the Hausman statstc s asymptotcally dstrbuted as ch-square wth k degrees of freedom and s wrtten down as follows: (9) ( ˆ ˆ ) ( ( ˆ H = β β Vˆ β ) Vˆ ( βˆ ) ( βˆ βˆ ) GLS F F GLS GLS F where β ˆ and βˆ are, respectvely, the estmates of the fxed effects and random effects F GLS models. Vˆ (.) are the correspondng varance-covarance matrces of these estmated coeffcents. IV. Emprcal fndngs Ths secton provdes emprcal evdence on the determnants of bank nterest margns and proftablty n the Tunsan Bankng ndustry. A broad descrpton of the characterstcs of the varables used n the study s gven n table whch reports ther statstcal means and standard devaton. Next, we report the results of regresson of the net nterest margn and return on asset varables, respectvely. The tables nclude several specfcatons, wth the basc specfcaton ncludng a set of bank characterstc varables. Subsequently, we add the macroeconomc varables and the fnancal structure varables. The estmaton technque s the balanced panel data regressons. <INSERT TB HERE> The frst bank-level varable s the equty varable (CAP). Buser et al. (98) argue n theory that banks generally have an optmal captalzaton rato and need to reman well-captalzed when they have a hgh franchse value. Berger (995) and Dermerguç-Kunt and Huzngua (999) fnd a postve relatonshp between bank performance and captalzaton. Consstent wth the prevous evdence, we confrm the postve relatonshp whether we use nterest margn or return on assets as a dependant varable and n all specfcatons. Ths may ndcate that well-captalzed banks support lower expected bankruptcy costs for themselves and ther costumers, whch reduce ther cost of captal.
11 Next, there s a postve and sgnfcant coeffcent on the overhead to assets rato varable (OVERHEAD) n the net nterest margn and return on assets equatons. The overhead varable has an estmated coeffcent of n the net nterest equaton, whch suggest that 87.8% of a bank s overhead costs s passed on ts depostors and lenders (n terms of lower depost rates and/or hgher lendng rates). In all net nterest margn equaton specfcatons, we see that the coeffcent on bank loans (BLOAN) s postve and sgnfcant. Ths notably reflects that bank loans are nterest-payng contrary to the cash, thereby ncreasng net nterest margn. Conversely, non-nterest bearng assets (NIBA) has no sgnfcant mpact on net nterest margn and return on assets, provng that bank proftablty stems manly from nterest bearng assets. Many researchers fnd that lttle cost savng can be acheved by ncreasng the sze of the bankng frm (Berger et al., 987) and others report sgnfcant scale economes for banks whose asset sze extends well nto the bllon range (Shaffer, 985 and many others). In table 2, the sze varable (LNSIZE) has mostly negatve and sgnfcant coeffcents on the net nterest margns equatons. Ths suggests that larger banks tend to lower margns and s consstent wth models that emphasze the negatve role of sze arsng from scale neffcences. The macroeconomc ndcators (.e. nflaton and economc growth) ncluded n column 2, 3 and 4 of table 2 and 3 are nsgnfcant n both spread and proft regressons. Ths may suggest that banks tend to not proft n nflatonary envronment. In addton, economc growth does not reflect any aspects of bankng regulatons and technology advance n the bankng sector omtted from the regressons. In table 2 and 3 we nclude two sets of fnancal market or structure varables. The frst set, nclude the market concentraton rato and the second, fnancal structure varables n the sense that they measure the mportance of bank and stock market fnance and the fnancal development. These varables among other thngs may reflect any complementary or substtutablty between bank and stock markets. Turnng to market concentraton, we see that the concentraton rato has a negatve and sgnfcant mpact only on net nterest margn. Ths result means that concentraton s less benefcal n terms of proftablty to the Tunsan commercal banks than competton.
12 The second set of fnancal structure varables has a more sgnfcant mpact on bank proft as opposed to bank margns. Accordng to Dermercuc-Kunt and Huzngua (999), ths may ndcate that these varables have a smaller ncdence on banks loan and depost costumers compared to the other clents. The stock market captalzaton to GDP rato enters the return on assets equaton postvely, whch suggest that a larger equty market per se gves banks the opportunty to ncrease ther proftablty. Ths may be due to the complementarty s effect between equty and debt fundng. As stock markets enlarge, mproved nformaton avalablty ncrease the potental number of customers to banks by easng the dentfcaton and montorng of borrowers. The ncrease of bank actvty contrbutes to enhance proftablty. In addton, the stock market captalzaton to bankng assets rato (RSIZE) enters the return on equty equaton postvely, whch suggest that a larger stock market relatve to the bankng sector ncrease bank profts and confrm the complementarty s effect. All the above results on fnancal structure mean that the move of the Tunsan fnancal system towards a more market based fnancal structure s proftable to the bankng ndustry. <INSERT TB 2 AND 3 HERE> V. Concluson Ths paper nvestgates the mpact of bank s characterstcs, fnancal structure and macroeconomc ndcators on bank s net nterest margns and proftablty n the Tunsan bankng ndustry for the perod. Frst, ndvdual bank characterstcs explan a substantal part of the wthn-country varaton n bank nterest margns and net proftablty. Hgh net nterest margn and proftablty tend to be assocated wth banks that hold a relatvely hgh amount of captal, and wth large overheads. Other mportant nternal determnants of bank s nterest margns bank loans whch have a postve and sgnfcant mpact. The sze has mostly negatve and sgnfcant coeffcents on the net nterest margns. Ths latter result may smply reflect scale neffcences. Second, the paper fnds that the macro-economc ndcators such nflaton and growth rates have no mpact on bank s nterest margns and proftablty.
13 Thrd, turnng to fnancal structure and ts mpact on bank s nterest margn and proftablty, we fnd that concentraton s less benefcal to the Tunsan commercal banks than competton. Stock market development has a postve effect on bank proftablty. Ths reflects the complementartes between bank and stock market growth. We have found that the dsntermedaton of the Tunsan fnancal system s favourable to the bankng sector proftablty. As a matter of polcy mplcatons, we need to draw several proposals at the bank and naton levels: - At the bank level, the mprovement of the proftablty of Tunsan commercal banks need to be conducted by a renforcement of the captalzaton of banks through natonal regulaton programs, by reducng the proporton of non-nterest bearng assets to the beneft of bank loans and by reducng the sze of large banks to optmal levels. - At the naton level, we need to reduce concentraton and spur competton, and to boost the development of the equty market n order to mprove bank s proftablty as bank and stock market was found to be complementary. References Abreu M. and V. Mendes Commercal bank nterest margns and proftablty: evdence from E.U countres, Porto Workng paper seres. Afanaseff T., P.Lhacer and M. Nakane. (2002). The determnants of bank nterest spreads n Brazl, Banco Central d Brazl Workng Papers. Angbazo, L Commercal bank net nterest margns, default rsk, nterest-rate rsk, and off-balance sheet bankng, Journal of Bankng and Fnance, Vol.2: Barajas, A., R. Stener, and N. Salazar Interest spreads n bankng n Colomba , IMF Staff Papers, Vol. 46: Bashr A Assessng the Performance of Islamc Banks: Some Evdence from the Mddle East, Paper presented at the ERF 8 th meetng n Jordan. Ben Naceur S. and M. Goaed The determnants of the Tunsan depost banks performance, Appled Fnancal Economcs, Vol.:37-9. Berger A The relatonshp between captal and earnngs n bankng, Journal of Money, Credt and Bankng, Vol.27:404-3.
14 Berger A., D Hanweck and D. Humphrey Compettve vablty n bankng: Scale, scope and product mx economes, Journal of Monetary Economcs, Vol.20: Buser, S., A. Chen, and E. Kane. 98. Federal depost nsurance, regulatory polcy, and optmal bank captal, Journal of Fnance, Vol. 35: Demerguç-Kunt A. and H. Huznga Determnants of commercal bank nterest margns and proftablty: Some nternatonal evdence, World Bank Economc Revew, Vol.3: Demerguç-Kunt A and H. Huznga. 200 Fnancal Structure and Bank Proftablty n Fnancal Structure and Economc Growth: A Cross-Country Comparson of Banks, Markets, and Development, Eds. Asl Demrguc-Kunt and Ross Levne. Cambrdge, MA: MIT Press, 200. Guru B., J. Staunton and Balashanmugam Determnants of commercal bank proftablty n Malaysa, Unversty Multmeda workng papers. Keeley M. and G. Zmmerman Competton for money market depost accounts, Federal Reserve Bank of San Francsco Economc Revew, Vol. :5-27. Molyneux P. and J. Thornton The determnants of European bank proftablty, Journal of Bankng and Fnance, Vol. 6: Shaffer S Competton, economes of scale, and dversty of frm szes, Appled Economcs, Vol.7: Smrlock M Evdence of the (Non)Relatonshp Between Concentraton and Proftablty n Bankng, Journal of Money, Credt, and Bankng, Vol.7:
15 Table. Descrptve statstcs of varables Varable name Mean Mnmum Maxmum Standard devaton NIM ROA CAP BLOAN NIBA OVERHEAD GROWTH INF SBS MCAP RSIZE CONC
16 Table 2. Determnants of Tunsan depost banks Net Interest Margns Constant CAP BLOAN NIBA OVERHEAD LNSIZE INF GROWTH CONC RSIZE SBS MCAP Regressons () (2) (3) (4) Nb. observatons Adj R² Haussman test : Random vs Fxed effects χ² P-values (0.797) 0.080*** (4.059) 0.02*** (3.342) (-0.267) 0.878*** (8.40) - (-.253) (0.045) 0.075*** (3.792) 0.02*** (3.396) - (-0.096) 0.848*** (7.557) -* (-.742) - (-.087) (0.70) T-Student are n parentheses. *, ** and *** ndcate sgnfcance levels of 0, 5 and percent respectvely *** (4.725) 0.033* (.647) 0.09*** (3.095) (0.46) 0.743*** (6.792) -0.03*** (-3.096) (.087) (0.279) -0.2*** (-5.446) (0.598) (0.089) *** (4.884) 0.034* (.7) 0.02*** (3.399) (0.660) 0.728*** (6.7) *** (-3.224) (.37) (0.466) -0.3*** (-6.05) (0.253)
17 Table 3. Determnants of Tunsan depost banks Return on Assets Constant CAP BLOAN NIBA OVERHEAD LNSIZE INF GROWTH CONC RSIZE SBS MCAP Regressons () (2) (3) (4) Nb. observatons Adj R² Haussman test : Random vs Fxed effects χ² P-values * (-.836) 0.055*** (4.596) (0.506) - (-0.55) 0.224*** (3.383) ** (2.335) (-0.4) 0.049*** (4.78) (0.72) (0.093) 0.94*** (2.84) (0.507) -* (-.706) (.397) T-Student are n parentheses. *, ** and *** ndcate sgnfcance levels of 0, 5 and percent respectvely. 0.07* (.745) 0.03** (2.373) (0.045) (-0.493) 0.36* (.927) - (-.6) - (-0.865) (.49) (-.593) (0.624) 0.04*** (2.782) (.467) 0.033** (2.54) (0.228) - (-0.237) 0.28* (.822) - (-0.972) - (-0.780) (.289) (-.227) 0.0*** (2.764)