Comparative Performance Evaluation of Small, Medium and Large U.S. Commercial Banks

Transcription

1 Comparatve Performance Evaluaton of Small, Medum and Large U.S. Commercal Bans AUTHORS ARTICLE INFO JOURNAL FOUNDER Chau Chuwuogor-Ndu Jll Wetmore Chau Chuwuogor-Ndu and Jll Wetmore (26). Comparatve Performance Evaluaton of Small, Medum and Large U.S. Commercal Bans. Bans and Ban Systems, (2) "Bans and Ban Systems" LLC Consultng Publshng Company Busness Perspectves NUMBER OF REFERENCES NUMBER OF FIGURES NUMBER OF TABLES The author(s) 28. Ths publcaton s an open access artcle. busnessperspectves.org

2 Bans and Ban Systems / Volume, Issue 2, Abstract COMPARATIVE PERFORMANCE EVALUATION OF SMALL, MEDIUM AND LARGE U.S. COMMERCIAL BANKS Chau Chuwuogor-Ndu, Jll Wetmore Ths paper examnes the comparatve performance of small U.S. commercal bans ($mllon to $3mllon) n assets wth medum sze ($bllon-$5bllon) and large (greater than $5bllon) for the perod of In vew of the banng system deregulaton and ban consoldaton n the Unted States, t s necessary to constantly evaluate the performance of the varous categores of bans to document the possble mpact of these polcy measures. We use proft effcency (PROFEFF), return-on-assets (ROA), nterest ncome, non nterest ncome and loan loss reserve as crtera for ths comparson. We fnd that between 997 and 999, small bans were more proft effcent (PROFEFF) than large bans but less than medum- sze bans. Snce 999, the PRO- FEFF of all szes of bans has been on the declne but the PROFEFF of small bans declned more than that of large and medum-sze bans. The ROA for all the bans under evaluaton declned between 2 and 22. Small bans suffered the largest declne. An examnaton of the trend n net nterest ncome as a percentage of average assets (NII) for the three groups of bans reveals that small bans NII s greater than that of large bans for the entre perod. And n contrast, small bans have the lowest level of non-nterest ncome as a percentage of average assets (NONII). It s apparent that the small bans are vulnerable to ncreased competton offered by deregulaton, technologcal advances, e-commerce and negatve economc stuaton such as the current recesson. These results suggest the survval of small U.S. commercal bans s n jeopardy. Snce these observatons result from the present polcy of consoldaton and ban system deregulaton there s need for the Reserve Bans revst of ths polcy stance. Key words: Proft effcency, Return-on- assets, Commercal banng, Net nterest ncome, Non nterest ncome, non-current loan, loan-loss reserve. JEL classfcaton: G2. Introducton A number of studes comparng the proftablty and safety of peer groups of bans have been wrtten. Most of these studes focus on small bans wth dfferent asset szes than we study. For example, Ahgbe and McNulty (23) study a mnmum sze ban of $5 mllon n assets. DeYoung and Hassan (998) refer to a specfc category of small such as novo bans. Elyasan and Mehdan (995) focus on a rs threshold of small commercal bans wth average assets of less than $5 mllon and wth one branch offce. The perod of s sgnfcant because of the perceved vulnerablty of small commercal bans caused by deregulaton and ncreased competton offered by technologcal advances and e-commerce. The perod also encompasses perods of economc booms and recessons. The ssue of survval of small bans n the present era of ban consoldaton s of tremendous nterest to scholars of fnancal servces and regulators. Regulatory changes contrbutng to the threat of survval of small bans nclude: ntroducton of nterest bearng checng accounts, the removal of regulatory celngs on ban depost rates, relaxaton of branchng laws, ncreased competton from non-ban frms and the emergence of mega bans through mergers and acqustons. Moreover, durng the second half of the 2th century, advances n communcatons technology, (Shaffer, 989) reports n 985, the annual number of falures had rsen over, of whch 77 were smallest bans wth total assets of less than $25 mllon dollars. Some 2 bans n 987, of whch 3 had less than 25 mllon dollars n assets and 67 had less than $5 mllon n assets. Chau Chuwuogor-Ndu, Jll Wetmore, 26

3 24 Bans and Ban Systems / Volume, Issue 2, 26 fnancal marets and banng producton technques have contrbuted to the eroson of the fundamental advantages of communty bans namely: the personal nteracton between baners, borrowers and depostors, especally small, unsophstcated borrowers and depostors. These developments have led to the margnalzaton of some small bans that flourshed when the regulatons and envronment favored small-szed bans. Research results concernng the danger to small bans are mxed. The lterature fnds a dramatc ncrease n the proporton of falures occurrng among the small bans and they cte performance data as evdence corroboratng ths vew. (See, for example, Shaffer, 998; Kutter, 99 and Fx, 988). The survval of small bans s mportant because f they do not survve, credt to small busnesses may dmnsh wth the resultng negatve effect on job creaton. However, other researchers fnd that small bans may have both an nherent nformaton advantage over large bans (Naumura, 993; and Mester, Naumura and Renoult, 998). Berger, Alan, Mller, Peterson, Rjan and Sten (22) suggest that small bans may have a comparatve advantage n developng and usng the soft nformaton often assocated wth small busness lendng. Dfferences n ban asset and lablty composton, expenses, non-nterest ncome, captal rato, competton and access to credt nformaton, all emanatng from the dfferences n ther asset szes pose problems for scholars wshng to conduct comparatve evaluaton of ban performance. To mtgate ths dffculty n evaluatng small bans n comparson to other bans, we use the proft effcency (PROFEFF) analyss as one of our nstruments of analyss n ths paper. We estmate the PROFEFF measure for each peer group of U.S. bans based on asset szes but we do not regress t on varables reflectng dfferences n asset and lablty composton, competton, locaton, organzatonal structure and other factors (Ahgbe and McNulty, 23 and DeYoung and Hassan, 998). Rather we determne the PROFEFF, absolute proftablty usng (ROA), the operatonal rs threshold of each category of bans. Ths analyss provdes a comprehensve pcture of the dfferences n proft effcency, absolute proftablty and the rs profle between small bans and other categores of bans for the perod of Lterature Revew There are sgnfcant emprcal fndngs n support of theores advanced to explan why small bans fnancal performance may dffer from that of other bans. The lterature suggests ceters parbus, small bans n small communtes can charge hgher rates on loans and pay lower rates on deposts than other bans because there s less competton n small banng marets. (See, for example, Glbert, 984; Hannan, 99a, b; Berger, Hanwec and Humphrey, 987; and Gllgan and Smrloc, 984). Moreover, many researchers fnd that lttle cost savng can be acheved by ncreasng the sze of the banng frms (Berger, Hanwec and Humphrey, 987 and Gllgan and Smrloc, 984). Other research suggests the presence of sgnfcant scale economcs for bans whose asset sze extends well nto the multbllon dollar range. (See, for example, Shaffer, 985; Hunter and Tmm, 986; Evanoff, Isralevch and Merrs, 99; Noulas, Ray and Mller, 99 and Shaffer and Davd, 999). Naumura (993) and Mester et al. (998) fnd that small bans have access to better credt nformaton than large bans, such as daly data on frm cash flows, whch s avalable through montorng checng accounts. Several authors fnd managers of bans n less compettve marets may dsspate part of ther advantage by enjoyng perqustes such as hgher salares, more assstants, lavsh offce quarters, etc. (See, for example, Arnold, 985; Berger and Hannan, 998; Hannan and Mavnga, 98; Purroy and Salas, 2; and Rhoades, 98). Rhoades and Rutz (982) and Clar (986) fnd that ban managers n smaller, less compettve marets may also shft the ban s asset composton to less rsy loans and securtes out of a desre to enjoy a quet lfe. Many studes of ban performance report that small bans have hgher ROA (but not necessarly equty) than large ones. Boyd and Runle (993) study bans wth assets of more than $ bllon and fnd an nverse relatonshp between ban sze and ROA, whch they attrbute to monopoly rents. Berger and Mester (997) report greater PROFEFF at small bans than at large bans. Elyasan and Mehdan (995) suggest that because of deregulaton, the future survval of small bans s n serous queston. McNulty et. al. (2) fnd no consstent evdence of superor loan qualty at small bans. However ths analyss s

4 Bans and Ban Systems / Volume, Issue 2, restrcted to one large state (Florda) n the U.S. The result s evdence on the vablty of small bans s mxed and the queston of vablty deserves further study. 3. Data and Methodology The sample studed ncludes all U.S. bans wth asset szes $ mllon-$3 mllon for the small peer group, $bllon-$5 bllon for the medum-sze peer group, and wth assets greater than $5 bllon for the large peer group. These bans are found n the Report of Condton and Income (call report) database avalable at the Federal Reserve Ban of Kansas, whch contan data for dfferent ban peers from 997 to 22 and the Federal Reserve Ban of Chcago s web page for whch at least one year of data are avalable, ncludng newly chartered bans. The number of observatons s 2579 n 997, 265 n 998, 2655 n 999, 2693 n 2, 2724 n 2 and 2728 n 22. Ths paper adopts several technques n evaluatng dfferent asset categores of U.S commercal bans. We estmate the proft effcency (PROFEFF) for the small, medum and large bans n order to determne ther operatonal effcency durng the perod. We calculate the return on assets (ROA) for small, medum and large bans. We compare the annual mean PROFEFF and ROA of the varous sze bans and we apply the t-stats at, 2, and 5 percent levels to determne the sgnfcance. The PROFEFF test and ROA analyss gve an ndcaton of proftablty and degree of management effectveness n the utlzaton of ban assets. We examne the two man sources of ban ncome, the net nterest ncome and the non nterest ncome. We use the Man-Whtney (U) test, non parametrc varance analyss test two sample test, to test the sgnfcance of the dfferences n net nterest ncome and non nterest ncome as percentages of average assets for the small, medum and large bans for the perod of We compare the operatonal rs for the varous sze bans as ndcated by the level of non-current loan, loan-loss reserve and net actual loan loss each as a percentage of total loans. 3.. Proft Effcency (PROFEFF) Analyss Vrtually all proft effcency studes use a lnear functon to analyze the correlates of the proft effcency functon. PROFEFF s a sophstcated fnancal performance statstc, measurng how actual fnancal performance compares to a theoretcal best practce fronter. For a ban under evaluaton, t s measured as a percentage of the PROFEFF of the best practce ban. The fronter s estmated separately for each year and each ban s PROFFEF s also estmated usng the followng non-standard, Fourer-flexble 2 : PREROA j 3 Z m j j v, Y 2 2 W Z m cos j m X X snx X cos Z Z 9 Y Y cos X W W YW sn X X X X snx X X l l l m j j j j j j mn m j m m n 3 3 j Y Z () Non-current loans are loans that are past due for 9 days or more. 2 Berger and Mester (997), Altunbas, Evans, and Molyneux (2), Ahggbe and McNulty (23), DeYoung and Nolle (996).

5 26 Bans and Ban Systems / Volume, Issue 2, 26 where: PREROA = operatng profts (earnngs before taxes, extraordnary tems, and loan losses) measured as a percentage of total assets. level: Y represents a vector of three outputs defned for each ban as: total loans (the sum of consumer, commercal/ndustral and real estate loans) retal deposts (the sum of demand deposts and tme deposts) and non-nterest ncome (representng fee-based fnancal servces). W represents a vector of three maret prces for ban nputs, measured at the country the wage rate for labor the average nterest rate for borrowed funds a prce for physcal captal. Z vector contans three varables: equty captal (defned separately for each ban) to control for the potental ncreased cost of funds due to fnancal rs, a Hrschman-Herfndahl Index (HHI, defned at the country level) to control for dfferences n maret structure among countres, and the average non-performng loan rato (defned at the country level) to control for dfferences n economc condtons across marets. X represents a set of nne varables that transform the output (Y) varables to place them on an nterval from to 2 2. We assume that profts depend on nput prces and output quanttes. Ths s a reasonable assumpton for loans, deposts and fee-based servces. The Fourer functon has been used n a large number of recent cost and proft effcency studes. (See, for example, Ahgbe and McNult, 23; Berger and Mester, 997, 2; DeYoung and Hassan, 998; DeYoung and Nolle, 996; McAllster and McMamus, 993; and Mtchell and Onvurall, 996). For bans n whch Y, W, and Z dffer maredly from the sample mean, the Fourer form provdes a better ft than other functons, such as the translog functons. The non-standard Fourer form assumes that bans have some control over output prces (DeYoung and Hasan, 998 and Humphrey and Pulley, 997). Profts are assumed to depend on nput prces and output quanttes. Snce output prces are not exogenous under these assumptons, Equaton () s very smlar to the functon used by Ahgbe and McNulty (23) and DeYoung and Hassan (998). Ths functon avods the dffculty n measurng output prces. Output quanttes, rather than output prces explan a larger porton of varaton n proftablty. We apply the stochastc fronter approach suggested by Jondrow et al. (982) and used by Ahgbe and McNulty, (22) and DeYoung and Hassan, (998) to capture the ban s dvergence from the best practce fronter. The stochastc fronter approach assumes that devatons from the fronter nclude neffcences (proft neffcences n our case) and random errors. Ineffcences are assumed to follow an asymmetrc, half normal dstrbuton, and the random errors follow a symmetrc normal dstrbuton. We estmate the neffcency term as the expected value of proft neffcency, condtonal on the resduals from each year s proft functon. Equaton () reflects the non-standard Fourer hybrd form snce t contans both a quadratc proft functon and a seres of trgonometrc (Fourer) terms. Because of software lmtatons and lmtatons on the number of observatons, we estmate a slghtly modfed verson of ths functon. Our functon contans 8 trgonometrc terms and 54 other terms for a total of 72 ndependent varables. Lmtng the number of terms (especally the thrd-order terms) s consstent wth other re- The wage rate for labor equals total salares and benefts dvded by the number of full-tme employees. The prce of captal equals expenses of premses and equpment dvded by premses and fxed assets. The prce of deposts and purchased funds total nterest expense dvded by total deposts and purchased funds. 2 See Berger and Mester (997, p. 97 n) for the methodology for performng these transformatons.

6 Bans and Ban Systems / Volume, Issue 2, cent PROFEFF studes. (See Ahgbe and McNulty, 22; DeYoung and Hassan, 998; DeYoung, Spong and Sullvan, 2; and Berger and Mester, 997, 2). POTENTIAL PREROA s defned as the estmated proftablty of the ban f t s operated on the best-practce fronter. Snce effcency cannot be negatve, as n other PROFEFF studes we defne: PROFEFF = (ACTUAL REROA/POTENTIAL PREROA), f PREROA > PROFEFF = f PREROA <. (2) PROFEFF s an effcency measure whch ranges from zero for bans experencng losses to one for bans operatng on the best practce fronter. We estmate a separate PROFEFF functon (fronter) for each year. Ths approach allows the regresson coeffcents and the effcency measures to vary over tme, thereby allowng flexblty n the estmaton procedure Return on Assets Return on Asset s the best rato for comparng proftablty performance of companes across ndustres. Whereas the ROA ndcates the overall proftablty of a company, t can be dstorted by the occurrence of nonrecurrng gans and losses, changes n the company s leverage and the ncdence of restructurng and acqustons. The ROA s used n ths paper subject to these lmtatons. ROA = Net Income /Total Assets (3) We examned the trends n the two man sources of ban ncome, net nterest and non nterest ncome. We used the Mann-Whtney U Test, a nonparametrc varance analyss test, to test the equalty of the small bans mean net nterest ncome and non net nterest ncome wth frst that of large bans and second wth that of medum bans for the perod of 997 to 22. U n nn2 R, n 2 where: n = number of observatons for small bans; n 2 = number of observatons for large bans; R = sum of the rans of observatons for small bans; R 2 = sum of the rans of observatons for large bans. We test the hypothess: H o : = 2 null hypothess: There s no dfference between the net nterest ncome of small and large bans, n partcular, both have the same mean. H o : 2 alternatve hypothess: There s no dfference between the net nterest ncome of small and large bans: n partcular, they have dfferent means. =.5 level of sgnfcance for testng these hypotheses We repeat ths test for small and medum bans. We also tested the hypothess on the non nterest ncome of small, medum and large bans for the perod of Rs Analyss We evaluate the major commercal ban rs factor, credt rs by examnng the loan loss reserve as a percentage of total loans and non-current loan as a percentage of total loans. Snce bans hold lttle owners captal relatve to aggregate value of ther assets, only a relatve small percentage of total loans need to turn bad n order to push any ban to the brn of falure (Rose, 999). The loan-loss reserve ndcates the extent to whch a ban s preparng for loan losses through annual charges aganst current ncome. The non-current loans are loans that are past due for 9 days or more. Fnally we analyze the actual charge-off by examnng the net loan-losses as a percentage of average loans.

7 28 Bans and Ban Systems / Volume, Issue 2, Emprcal Results Table contans the summary statstcs data for our estmated PROFEFF for three classes of bans for the perod Panel A presents the results when a sngle PROFEFF fronter s estmated for small and large bans. Panel B contans the results when a sngle PROFEFF fronter s estmated for small and medum sze bans Between 997 and 999, the small bans, wth asset sze between $ mllon and $3 mllon were more proft effcent than the large bans but less than medum-sze bans. Snce 999, even though the PROFEFF of all the asset sze has been on the declne, the PROFEFF of small bans declned more than that of large bans and medum-sze bans. Medum-sze bans, wth asset sze of between $bllon and $5bllon acheved the hghest PROFEFF durng the perod of Table Summary Statstcs for Proft Effcency of a Sample of U.S. Bans, Year Small bans Medum-sze bans (Asset sze $m-$3m) (Asset sze$b -$5b) Panel A Small versus Medum bans usng a sngle fronter for all bans Dfference N Mean Std N Mean Std Mean t-stat *** *** *** *** Year Small bans Large bans Dfference Panel B Asset sze $m-$3m Asset sze greater than $5b Mean t-stat Small versus Large bans usng a sngle fronter for all bans N Mean Std N Mean Std *** *** *** *** *** Ths table presents our PROFEFF estmates for the three classes of bans for the perod of Panel A presents the results when a sngle PROFEFF fronter s estmated for small and large bans. Panel B presents the results when a sngle PROFEFF fronter s estmated for small and medum sze bans. *** Sgnfcant at the percent level. Fgure depcts the comparatve PROFEFF performance of the small, medum and large bans.

8 Bans and Ban Systems / Volume, Issue 2, Estmated mean PROFEFF Years Estmated mean PROFEFF of small bans Estmated mean PROFEFF of medum bans Estmated PROFEFF of large bans Fg.. Comparatve PROFEFF, All categores of bans under evaluaton acheved approxmately the same return on asset n 2. The calculated ROA s.6% for small bans,.5% for medum- sze bans and.5% for large bans. The ROA for all the bans under evaluaton declned between 2 and 22. The small bans suffered the hghest declne. Between 997 and 2, the small bans were more proftable than the large bans. Agan as was the case wth the PROFEFF analyss, the medum sze bans acheved the hghest proftablty durng the perod of Table 2 contans the ROA summary statstcs data and the comparatve ROA of the three classes of the bans under evaluaton s contaned n Fgure 2 below. Summary of ROA Statstcs for ROA for each sample of US bans, Table 2 Small bans Medum bans Asset sze $m-$3m Asset sze$b -$ b Panel A. Small bans versus Medum bans Year N Mean Std N Mean Std Dfference Mean t-stat

9 3 Bans and Ban Systems / Volume, Issue 2, 26 Table 2 (contnuous) Small bans Large bans Asset sze $m-$3m Asset sze greater than $5b Panel A. Small bans versus Large bans Year N Mean Std N Mean Std Dfference Mean t-stat Return on assets (ROA).6%.4%.2%.%.8%.6%.4%.2%.% Years Peer Group US Bans asset sze $m-$3m Peer Group US Bans asset sze $b-$5b Peer Group US bans asset sze greater than $5b Fg. 2. Comparatve Return on assets (ROA) The two man sources of commercal ban ncome are nterest on loans and non nterest ncome such as fees and commssons. The net nterest ncome, sometmes referred to as nterest margn s a ey determnant of ban proftablty. An examnaton of the trend n net nterest ncome as a percentage of average assets for the three peers of bans reveals that small bans net nterest ncome as a percentage of average assets s greater than that of large bans for the whole perod, greater than that of medum bans between 997 and 999 but less than that of medum bans for the perod from 2 to 22. The large bans had the lowest level of net nterest ncome (Fgure 3). Net nterest Income as a percent of average assets Years Peer Group US Bans asset sze $m-$3m Peer Group US Ban asset sze $b-$5b Peer Group US ban asset sze greater than $5b Fg. 3. Comparatve Net Interest Income as a percent of average assets, It s beleved that new communcatons technology enables large bans to erode the doman of communty bans and compete n local marets through the use of networs of ATMs, Internet oss, and transactonal Internet webstes. The credt-worthness of local communty ban loan

10 Bans and Ban Systems / Volume, Issue 2, 26 3 customers can now be evaluated by fnancal nsttutons located outsde the maret usng new lendng and fnancal technologes. Use of nternet fnancal models, questonnares for credt scorng and on-lne credt applcatons permts large bans to buld up a huge data base of the credt rs of small customers at low cost. The large bans also beneft from economes of scale by combnng a hgh volume of loans wth the use of asset securtzaton rs management technques. It appears that despte these advantages, the small bans wth asset sze $ mllon -$3 mllon relatve to ther asset sze are stll dervng more ncome from loans than large bans. Table 3 below shows the results of the normalty on the Net nterest ncome for small, medum and large bans for the perod of The W Test results ndcate the non normalty of the dstrbuton. Results of W Test, Test for Normalty Table 3 Year R P value StDev R P value StDev R P value StDev The results of the Mann-Whtney test on the equalty of the Net nterest ncome for small and large bans, small and medum bans for the perod of are shown n Table 4. The results ndcate that for both comparsons the dfferences were not sgnfcant at 5% sgnfcant level. So we accept the Null hypotheses that there s no dfference between the net nterest ncome of small, medum and large bans between 997 and 22. However, the Mann-Whtney (U) test shows that the dfferences for some years were sgnfcant at very low levels. For example n 999, the dfference between the net nterest ncome as a percent of average assets for small and large bans was sgnfcant at.24% level. In 999 and 2, the dfference between the net nterest ncome as a percent of average assets for small and medum bans was sgnfcant at.24% level, n 998 t was sgnfcant at.77% and n 2 at.2%. Results of the Mann-Whtney tests Year Small bans/large bans Small bans / Medum bans W C W C *.4/ *.5/ ***.259/ *.3937/ *.6652/ *.7378/.3243 * Sgnfcant at zero percent ** Sgnfcant at.2 *** Sgnfcant at.24 **** Sgnfcant at **** -.62/ *** -.998/ *** -.688/ ** / * / Table 4

11 32 Bans and Ban Systems / Volume, Issue 2, 26 Small bans have the lowest level of non-nterest ncome as a percentage of average assets. The large bans have the hghest level of non-nterest ncome earnngs (Fgure 4). It must be noted that n recent tmes the relatve mportance of loan revenue versus non nterest revenue sources (fee ncome for example) has been changng rapdly as fee ncome grows much faster than nterest ncome on loans. Ths s because baners are worng hard towards developng fee-based servces. Non Interest Income as a percent of average assets Years Peer Group US Bans asset sze $m- $3m Peer Group US Bans asset sze $b-$5b Peer Group US Bans asset sze greater than $5b Fg. 4. Comparatve Non Interest Income as a percent of average assets Accordng to the results of the U test n Table 5, we also accept the Null hypotheses that equalty of small, medum and larges non nterest ncome as a percent of average assets for the perod of at 5% sgnfcant level. Some of the dfferences tested sgnfcant at very low levels of between.2% and.7%. Results of the Mann-Whtney tests Table 5 Year Small bans Large bans Small bans = Medum bans W C W C * -6.23/ / * -6.44/ 498********* -.569/ * -.38/ 435******** / * -.2/ 3994******* -.548/ * -.293/ 444****** / * -.69/ 433***** -.597/ * Sgnfcant at percent ****** Sgnfcant at.97 ** Sgnfcant at.2 ******* Sgnfcant at.7 *** Sgnfcant at.24 ******** Sgnfcant at.4 **** Sgnfcant at.77 ********* Sgnfcant at.34 ***** Sgnfcant at.58 ********** Sgnfcant at.24

12 Bans and Ban Systems / Volume, Issue 2, The small bans have the lowest loan loss reserve provsons. Ths ndcates better credt management and greater stablty n generatng ncome from loans. The large bans have the hghest provson for loan loss durng the perod. Wth the lowest net nterest ncome as earler observed, large bans seem to be experencng greater rs n ther loan management operatons (Fgure 5). Loan loss reserve as a percentage of total loans Year Peer Group US Ban asset sze $m-$3m Peer Group US Ban asset sze $b-$5b Peer Group US Ban asset sze greater than $5b Fg. 5. Comparatve Loan loss reserve provsons, Small bans consstently mantaned the lowest level of net actual loan losses. Medum bans experenced the hghest level of loan loss n 997 and 998. From 999 to 22 large bans suffered the hghest level of loan losses (Fgure 6). Net loan losses as a percentage of average loans Years Peer Group US Bans asset sze $m-$3m Peer Group US Ban asset sze $b- $5b Peer Group US Bans asset sze greater than $5b Fg. 6. Comparatve net actual loan losses, Small bans mantaned the lowest level of non-current loan as a percentage of total loans. From 998 to 22, the large bans mantaned the hghest level of non current loan. The medum sze bans mantaned the hghest n 997 and 998 (Fgure 7).

13 34 Bans and Ban Systems / Volume, Issue 2, Non-current loan as a percentage of total loans Years Peer Group US Bans asset sze $m-$3m Peer Group US Bans asset sze greater $b-$5n Peer Goup US bans asset sze greater than $5b Fg. 7. Comparatve actual Non-current loan Conclusons Small bans were more proft effcent than the large bans but less than medum sze bans. Snce 999, even though the PROFEFF of all the asset sze has been on the declne, the PROFEFF of small bans declned more than that of large bans and medum bans. Medum sze bans, wth asset sze of between $bllon and $5bllon acheved the hghest PROFEFF durng the perod. The dfferences n the PROFEFF for small, medum and large bans tested sgnfcant only at %, T test sgnfcance level. Wth respect to ROA, all categores of bans under evaluaton acheved approxmately the same return on asset n 2. The ROA for all the bans under evaluaton declned between 2 and 22. The small bans suffered the hghest declne. Between 997 and 2, the small bans were more proftable than the large bans. Agan as was the case wth the PROFEFF analyss, the medum sze acheved the hghest proftablty durng the perod of The dfferences n the ROA for small, medum and large bans tested sgnfcant only at %, T test sgnfcance level. An examnaton of the trend n net nterest ncome as a percentage of average assets for the three peers of bans reveals that small bans net nterest ncome as a percentage of average assets s greater than that of large bans for the whole perod, greater than that of medum bans between 997 and 999 but less than for the perod from 2 to 22. The result of the Mann-Whtney equalty of means on net nterest ncome for all the bans tested sgnfcant only at very low sgnfcant levels for some years durng the perod. In contrast, small bans have the lowest level of non nterest ncome as a percentage of average assets. The large bans have the hghest level of non nterest ncome earnngs. Agan the result of the Mann-Whtney equalty of means on non nterest ncome for all the bans tested sgnfcant only at very low sgnfcant levels for some years durng the perod. The small bans have the lowest loan loss reserve provsons. Ths ndcates better credt management and greater stablty n generatng ncome from loans. The large bans have the hghest provson for loan loss durng the perod. Wth the lowest net nterest ncome as earler observed, large bans seem to be experencng greater rs n ther loan management operatons. Small bans mantaned the lowest level of non-current loan as a percentage of total loans. From 998 to 22, the large bans mantaned the hghest level of non current loan. The medum sze bans mantaned the hghest level n 997 and 998.

14 Bans and Ban Systems / Volume, Issue 2, Small bans consstently mantaned the lowest level of net actual loan losses. Medum bans experenced the hghest level of loan loss n 997 and 998. From 999 to 22 suffered the hghest level of loan losses. It s apparent that the small bans are vulnerable to ncreased competton offered by deregulaton, technologcal advances, e-commerce and negatve economc stuaton such as the current recesson. These results suggest the survval of small U.S. commercal bans s n jeopardy. Hstorcally, the regulatory protecton afforded small bans could have been consdered part of a broader publc polcy desgned to preserve smaller, more rural communtes. Whether the declne n small bans adversely mpacts the economes of smaller communtes s an open ssue as s whether the publc sector should ntervene to support small communtes. In any case, the evdence suggests the contnued consoldaton of the U.S. banng ndustry and the contnued declne of small bans. The general lower levels of PROFEFF, ROA and non nterest ncome especally durng recessonary perods underscore the contnued vulnerablty of small US bans n the present era of ban consoldaton and banng system deregulaton. References. Ahgbe A., J.E. McNulty. The Proft Effcency of Small U.S. Commercal Bans// Journal of Banng and Fnance, pp Arnold R., J. Agency. Costs n Banng Frms: An Analyss of Expense Preference Behavor // Journal of Economcs and Busness pp Berger A.N., L.J. Mester. Insde the Blac Box: What Explans Dfferences n the Effcences of Fnancal Insttutons?// Journal of Banng and Fnance pp Berger, A.N., N.H. Mller., M.A. Peterson., R.G. Rajan and J. Sten. Does Functon Follow Organzatonal Form? Evdence From the Lendng Practces of Large and Small Bans, Paper presented at the Annual Conference on Ban Structure and Composton, Federal Reserve Ban of Chcago, May Berger A.N., G.A. Hanwec, D.B. Humphrey. Compettve Vablty n Banng Scale, Scope, and Product Mx Economcs // Journal of Monetary Economcs pp Berger A.N., T. Hannan. The Effcency Cost of Maret Power n the Banng Industry: A Test of the Quet Lfe and Related Hypotheses// Revew of Economcs and Statstcs pp Boyd J.H., D.E. Runle. Sze and Performance of Banng Frms, Testng the Predctons of Theory// Journal of Monetary Economcs pp Clar J.A. Maret Structure, Rs and Proftablty: The Quet-Lfe Hypothess Revsted, Quarterly Revew of Economcs and Busness pp De Young R. and I. Hassan. The Performance of De Novo Commercal Bans: A Proft Effcency Approach, Journal of Banng and Fnance pp DeYoung R., W. Hunter. Deregulaton, the Internet and Compettve Vablty of Large Bans and Communty Bans, The Future of Banng, Benton Gup(ed.) West-port, CT: Quorum Boos, 22.. DeYoung R., D.E. Nolle. Foregn-Owned Bans n the Unted States: Earnng Maret Share or Buyng It? // Journal of Money, Credt and Banng pp Elyasan E.S. Mehdan. The Comparatve Effcency Performance of Small and Large U.S. Commercal Bans n the Pre- and Post-Deregulaton Era. // Journal of Appled Economcs pp Evanoff D.D., P.R. Isralevch., R. Merrs. Relatve Prce Effcences, Techncal Change, and Scale Economes for Large Commercal Bans // Journal of Regulatory Economcs pp Evanoff D.D., O Evren. Banng Industry Consoldaton and Productve Effcency Proceedngs of a Conference on Ban Structure and Competton, Federal Reserve Ban of Chcago. 2. pp Federal Depost Insurance Corporaton webste,

15 36 Bans and Ban Systems / Volume, Issue 2, Federal Reserve Ban of Chcago webste, 7. Fx J. Bg bans Lve-The Small to De, Why the FDIC has a Dual Polcy, Phladelpha Inqurer , March. 8. Glbert R.A. Ban Maret, Structure and Competton: A Survey. //Journal of Money, Credt and Banng pp Gllgan T., M. Smrloc. An Emprcal Study of Jont Producton and Scale Economcs n Commercal Banng // Journal of Banng and Fnance March, Hannan T. H. Ban Commercal Loan Marets and the Role of Maret Structure: Evdence From Surveys of Commercal Lendng // Journal of Banng and Fnance. 99a. 5, pp Hannan T.H. Foundatons of the Structure-Conduct-Performance Paradgm n Banng // Journal of Money, Credt, and Banng. 99b 23. pp Hanna T.H., F. Mavnga. Expense Preference and Manageral Control: The Case of the Banng Frm // Bell Journal of Economcs. 98. pp Hughes J.P., L.J. Mester., C. Moon. Are Scale Economcs n Banng Elusve or Illusve? Evdence Obtaned by Incorporatng Captal Structure and Rs-Tang nto Models of Ban Producton // Journal of Banng and Fnance pp Hunter W.C., S.G. Tmm. Techncal Change, Organzaton Form, and the Structure of Ban Producton // Journal of Money, Credt and Banng pp Jondrow J.C., C.A. Knox Lovell., I. Materov., P. Schmdt. On the Estmaton of Techncal Ineffcency n the Stochastc Fronter Producton Functon Model // Journal of Econometrcs pp Kutter J. Forecast for Year 2: 24% Fewer Bans, Amercan Baner, October, 99,. 27. McAllster P.H., D. McMamus. Resolvng the Scale Effcency Puzzle n Banng// Journal of Banng and Fnance pp McNulty J.E., A. Ahgbe.,.J.A. Verbrugge. Small Ban Loan Qualty n a Deregulated Envronment: The Informaton Advantage Hypothess //Journal of Economcs and Busness pp Mester L.J., L.I. Naamura., M. Renault. Checng Accounts and Ban Montorng, Federal Reserve Ban of Phladelpha, Worng Paper WP Naumura L. Recent Research n Commercal Banng: Informaton and Lendng. Fnancal Marets, Insttutons and Instruments pp Noulas A.G., S.G. Ray., S.M. Mller. Return to Scale and Input Substtuton for Large U.S. Bans // Journal of Money, Credt, and Banng pp Purroy P. and V. Salas. Strategc Competton n Retal Banng Under Expense Preference Behavor // Journal of Banng and Fnance pp Rhoades S.A. Monopoly and Expense Preference Behavor: An Emprcal Test of a Behavoralst Hypothess // Southern Economc Journal pp Rhoades S.A., R.D. Rutz. Maret Power and Frm Rs: A Test of the Quet-Lfe Hypothess// Journal of Monetary Economcs pp Rose P. Commercal Ban Management, 4th edton, McGraw Hll, Shaffer S. Competton, Economes of Scale, and Dversty of Frm Szes Appled Economcs pp Shaffer S. A Revenue Restrcted Cost Study of Bans. Unpublshed worng paper, Federal Reserve Ban of New Yor, Shaffer S., E. Davd. Economcs of Superscale n Commercal Banng// Appled Economcs, pp Small Busness Admnstraton. Small Busness Lendng n the Dstrct of Columba// 997. Washngton, DC. U.S. 4. U.S. Baner webste,