Title before and after the East Asian Cri. Citation IDE Discussion Paper. No

Transcription

1 Ttle The Degree of Competton n the Th before and after the East Asan Cr Author(s) Kubo, Koj Ctaton IDE Dscusson Paper. No Issue Date URL Rghts < アジア経済研究所学術研究リポジトリ ARRIDE>

2 INSTITUTE OF DEVELOPING ECONOMIES Dscusson Papers are prelmnary materals crculated to stmulate dscussons and crtcal comments DISCUSSION PAPER No. 56 The Degree of Competton n the Tha Bankng Industry before and after the East Asan Crss Koj Kubo* Abstract Ths paper analyzes the nfluence of the East Asan crss and the subsequent reforms on the olgopolstc nature of the Tha bankng ndustry. Snce the crss, there have been substantal changes n compettve envronment, ncludng a declne n the famly ownershp of banks as well as the arrval of new entrants. How dd these changes affect a bankng ndustry n whch the sx largest local banks accounted for over 70 percent of market share? The estmated Lerner ndex from Bresnahan s [1989] conjectural varaton model ndcates the possblty of a declne n the degree of competton. Keywords: Tha bankng ndustry, degree of competton, Lerner ndex JEL classfcaton: G21, L13 * Research fellow, Internatonal Economcs Studes Group, Development Studes Center, IDE (kubokoj@de.go.jp)

3 The Insttute of Developng Economes (IDE) s a semgovernmental, nonpartsan, nonproft research nsttute, founded n The Insttute merged wth the Japan External Trade Organzaton (JETRO) on July 1, The Insttute conducts basc and comprehensve studes on economc and related affars n all developng countres and regons, ncludng Asa, Mddle East, Afrca, Latn Amerca, Oceana, and Eastern Europe. The vews expressed n ths publcaton are those of the author(s). Publcaton does not mply endorsement by the Insttute of Developng Economes of any of the vews expressed. INSTITUTE OF DEVELOPING ECONOMIES (IDE), JETRO 3-2-2, WAKABA, MIHAMA-KU, CHIBA-SHI CHIBA , JAPAN 2006 by Insttute of Developng Economes, JETRO

4 The Degree of Competton n the Tha Bankng Industry before and after the East Asan Crss 1. Introducton The purpose of ths paper s to nvestgate how the East Asan crss and the subsequent reforms affected the degree of competton n the Tha bankng ndustry. Through the reforms, the ndustry has undergone substantal changes n terms of the ownershp of banks and of regulatons. In many banks, famly ownershp has been replaced by state and foregn ownershp. The bankng sector saw several entrants durng the reforms, a phenomenon that had been absent for more than 20 years before the crss. Despte these developments, a casual observaton of loan market share suggests that the olgopolstc nature of the ndustry remans unchanged 1. Before as well as after the crss, the sx largest banks accounted for 70 percent of total loans of the consoldated bankng sector. The rather stable market share of the large banks, despte the reforms, forms the background of the present analyss on the behavor of leadng Tha banks. To estmate the degree of competton, we apply the method devsed by Bresnahan 1 See Vchyanond [1994] for the structure of the Tha bankng ndustry before the East Asan crss. 1

5 [1989]. Usng estmaton models based on mcroeconomc foundatons, ths method allows us to derve the ndex of the frms market power that s calculated as the devaton of the market prce from the margnal cost. Several studes have appled ths method to the analyss of the bankng ndustry, and these have often found that the bankng ndustres cannot be regarded as beng n perfect competton. One of the earlest studes s that of Shaffer [1993] on the Canadan bankng ndustry. She used tme seres data for the aggregated bankng sector, and estmated an average degree of competton over a long perod of tme 2. Bkker and Haaf [2002] appled the same method to the bankng ndustres of the European countres. Whle the estmaton based on tme seres data yelds only an average degree of competton for a long perod, panel data from the fnancal statements of ndvdual banks allows us to estmate the ndustry-wde average degree of competton and to evaluate ts year-by-year evoluton. The latter method has been appled by Angeln and Cetorell [2003] to the Italan banks, and by Uchda and Tsutsu [2005] to the Japanese banks 3. Our study, also, employs panel data, n our case data relatng to the Tha commercal banks durng the perod By estmatng parameters of a market power ndex for consecutve 2 A smlar study s that of Shaffer and DSalvo [1994], whch deals wth the tme seres data of two Canadan banks. 3 Apart from these, Berg and Km [1994] used the cross-secton data of the Norwegan banks, and estmated a degree of competton for a partcular year. 2

6 years before and after the crss, we evaluate the year-by-year evoluton of the degree of competton. The present paper contrbutes to the emprcal lterature on the Tha bankng sector. Among exstng studes, n contrast to the abundant lterature that outlnes the fnancal reform measures that followed the East Asan crss 4, there are relatvely few quanttatve analyses of the structural changes that have occurred n the bankng ndustry. Among the few exstng studes, Anuchworawon et al. [2003] and Polsr and Wwattanakantang [2005] descrbe the declne of famly ownershp of banks from the perspectve of corporate governance. In ther study relatng to foregn bank penetraton through the reforms, Okuda and Rungsomboon [2004b] analyze the effects of foregn bank entry on the bankng ndustry, and fnd that an ncreased presence of foregn banks n terms of number of banks s assocated wth a rse n overhead costs, a declne n profts, and an ncrease n the nterest spread of local banks. 5 In contrast to the narrow focus of Okuda and Rungsomboon [2004b] on the effects of foregn bank entres, we consder the effects of the reforms, ncludng changes n the ownershp structure, on the degree of competton. The organzaton of ths paper s as follows. In the next secton, we descrbe 4 For example, Vchyanond [2000]. 5 Apart from ths, Okuda and Rungsomboon [2004a] estmate cost functons of foregn and local commercal banks n Thaland and evaluate changes n cost effcences. 3

7 changes n the Tha bankng ndustry durng the crss and reforms that mght affect the degree of competton. We present our econometrc model n Secton 3, and dscuss the results of estmaton n Secton 4. In Secton 5, we summarze the analyss and offer some concludng remarks. 2. Changes n the compettve envronment of the Tha bankng ndustry 2.1 Structure of the fnancal sector The fnancal sector n Thaland conssts of commercal banks, fnance companes and government-owned specalzed fnancal nsttutons. Commercal banks comprse banks ncorporated n Thaland, and branches of foregn banks. At the end of 2004, there were 3,911 branches of locally ncorporated banks, whle each of 18 foregn banks has been allowed to open one branch wthn Thaland. Fnance companes are nsttutons that are not permtted to accept deposts, and rase funds by ssung debt nstruments nstead. Government-owned specalzed fnancal nsttutons nclude the Government Savngs Bank, the Government Housng Bank, and the Bank for Agrculture and Agrcultural Cooperatves. Apart from the above-lsted fnancal nsttutons, there are savngs cooperatves and agrcultural cooperatves. The average sze of these cooperatves and ther aggregated share n the fnancal sector are small. 4

8 Table 1 summarzes the sze of the respectve categores of fnancal nsttutons. Table 1, here. Table 1 shows that the hgh concentraton of market share by the sx largest local banks remans a salent feature of the Tha fnancal sector. The total loans of these banks account for about a half of the whole fnancal sector before as well as after the East Asan crss. Before the crss, ther share at one tme declned because of a surge n the loans of the fnance companes. The lendng practces of most of these fnance companes were reportedly bold, and 56 fnance companes were closed and lqudated n 1997 due to a severe non-performng loans (NPLs) problem. In the subsequent fnancal restructurng perod, the share of large commercal banks has been relatvely low as ther loan assets declned sharply as a result of NPLs. As the restructurng settled down, ther share pcked up. Ths was partally because some leadng banks merged or succeeded to the vable assets of closed smaller banks and fnance companes. On balance, the bankng ndustry remaned more or less olgopolstc. 2.2 Changes n the compettve envronment 5

9 We can observe two changes n the compettve envronment that mght have affected the degree of competton n the bankng ndustry. Frst, there were sgnfcant changes n the ownershp structure of fnancal nsttutons durng the reforms. The government not only natonalzed and lqudated a number of dstressed banks, but also abolshed the restrcton on the foregn ownershp of commercal banks n order to nvte foregn banks and nvestors to recaptalze the dstressed banks. Htherto, foregn ownershp had been restrcted to less than 25 percent of the equty captal of each fnancal nsttuton, and t was ths regulaton that was lfted for a specfed perod. As a result, foregn banks now own majorty shares n four small and medum banks. As of 1996, foundng famles were the largest shareholders n fve of the eght largest banks. By 2003, foregn nvestors were the largest shareholders n two banks, ncludng the largest one. Two other banks have been ether natonalzed or lqudated, and there remans only one bank n whch the largest shareholder s the foundng famly (Polsr and Wwattanakantang 2005). Table 2, here. Second, there has also been a sgnfcant change n the atttudes of the fnancal 6

10 authortes towards entry to the bankng sector. There had been no new entry and only one ext between 1980 and 1996, whereas fve banks exted through lqudatons and mergers durng the recent reforms, and two state owned banks were establshed to succeed to part of the assets and labltes of the closed banks. Furthermore, a prvate fnance company was upgraded to commercal bank status n 2004 by the ssuance of a brand-new bankng lcense, the frst such ntatve n the last two decades. Such changes n ownershp structure as well as entry and ext n the ndustry may affect the degree of competton. In practcal terms, these changes may make t dffcult to mantan colluson, f any, between banks, and ths n turn mght well lead to an ntensfcaton of competton wthn the ndustry. 2.3 Changes n the performance of banks In order to analyze the degree of competton among banks, we have to take nto account changes n ther performance as well as n the structure of ther revenues and expenses. Frst, there has been a notable fluctuaton n the proftablty of the banks. In Fgure 1, we present a summary of selected ndces concernng the performance of the bankng ndustry usng data compled from the ncome statements of the consoldated commercal banks n Thaland (excludng branches of foregn banks). One of the 7

11 reasons for the fluctuaton n proftablty s the problem of non performng loans (NPLs). The rato of NPLs to total loans rose to around 45 percent n The NPLs problem lowered proftablty n two ways. Frst, the banks ncurred huge expenses n the form of loan loss provsons for the dsposal of NPLs (depcted as the dotted lne n Fgure 1). Moreover, NPLs, ether classfed or non-classfed, reduced nterest ncome from these assets, thus brngng about an abrupt declne n net nterest ncome n 1998 (depcted as the fne contnuous lne n Fgure 1). It was not untl 2001 that the net profts of the banks became postve agan. Fgure 1, here. Changes n nterest rates have also been remarkable. Fgure 2 llustrates lendng and depost nterest rates and the nterest margn between two rates. Whle both the lendng and depost nterest rates fell consderably n the reform process, the spread wdened to more than four percent n the 2000s from around three percent before the crss. Ths wdenng margn partally accounts for the rsng net nterest ncome n recent years (see Fgure 1). 8

12 Fgure 2, here. Changes are also apparent n both revenue and cost structures. Fgure 3 portrays the structure of revenues and expenses as a percentage of total assets for the consoldated commercal banks (excludng branches of foregn banks). So far as the changes n the structure of revenues are concerned, we can see a sharp drop n nterests on loans and deposts 6 n 1999, whch corresponds to the declne n the lendng nterest rate (see also Fgure 2). Another sgnfcant change s the ncrease n other nterests and dvdends revenue after In partcular, there was an ncrease n the nterest revenue from government bonds. As a restructurng scheme, the government njected captal nto some dstressed banks, and these banks receved government bonds. As a result, the portfolo pattern of these banks was modfed, leadng to an ncrease n the nterest revenue from government bonds. As regards changes n the structure of expenses, the effect of the declnng depost nterest rate s reflected n a drop n nterest expenses. In addton, a declne n personnel expenses s apparent. Another mportant change s the rse n the contrbuton to the Fnancal Insttuton Development Fund (FIDF). The FIDF s an 6 Deposts here refer to banks deposts at other fnancal nsttutons. 9

13 nsttuton afflated to the Bank of Thaland (the central bank), and t has been provdng an mplct blanket guarantee for depostors of fnancal nsttutons. The contrbuton of banks to FIDF can be regarded as a premum for depost nsurance. The rate was rased n late 1997 from 0.1 percent of total depost per annum to 0.4 percent, causng a marked ncrease n expenses relatng to ths tem. Fgure 3, here. 3. Analytcal Framework and Data 3.1 Analytcal framework As our analytcal framework, we adopt Bresnahan s [1989] conjectural varaton model of competton. Specfcally, we consder a loan market n whch banks face a demand functon p p( Q, z), wth Q = q = 1, 2,, n, where p s the lendng nterest rate, q s the ndvdual banks loan supply, and z s the vector of exogenous factors affectng the demand for loans. In the exstng emprcal lterature on the bankng ndustry, there has been a controversy over whether deposts should be treated as an nput or output. In lne wth most of the emprcal studes of the degree of competton, we treat both labor and deposts as factor nputs, and loans as banks sole 10

14 output. Thus, Bank s maxmzaton problem s maxπ = p( Q, z) q C( q, ω ), (1) where C( q, ω ) s Bank s cost functon, wth ω beng the vector of the prces of ts factor nputs. The frst order condton s π q = p + q p Q Q q C q = 0. (2) Rearrangng (2) yelds p Q q θ p MC = Q = Q q Q, (3) η where MC s the margnal cost, and η ( Q / p) / Q s the sem-elastcty of loan demand to the lendng nterest rate. θ Q q ) /( Q q ) s Bank s conjectural ( elastcty of total loan of the bankng ndustry wth respect to ts own loans, that s, Bank s expectaton on how other banks react to ts output change. Ths term ndcates the bank s market power.e. the extent to whch the bank can manpulate the loan supply and the lendng nterest rate by colluson wth other banks. When the market s n perfect competton, θ takes the value of zero for all banks. In a monopoly, θ equals one. There are two ways of estmatng the degree of competton usng Equaton (3). One method s to obtan θ by dentfyng t separately from the sem-elastcty of 11

15 demand, η. However, nsofar as we assume a homogeneous loan market, we have only one observaton per year for the aggregate loan demand 7. Thus, we cannot obtan annual estmates of η, but an estmate of average sem-elastcty over a long perod of tme, so that we cannot retreve annual estmates of θ. The second method s to estmate θ / η as one parameter (Angeln and Cetorell [2003]). In the latter case, dvdng both sdes of Equaton (3) by p, we obtan a Lerner ndex, L p MC θ /η =. (4) p p Ths Lerner ndex, L [0,1], measures the mark-up of prce over margnal cost, ndcatng the market power of a bank. We estmate a cross-sectonal average of θ /η for each year and derve an ndustry-wde average Lerner ndex to see how t has evolved before and after the crss. Rearrangng Equaton (3) yelds R = q MC θ + q, (3 ) η where R refers to Bank s nterest ncome 8. To calculate the margnal cost, we consder the followng translog cost functon; 7 Uchda and Tsutsu [2005] estmate the demand equaton for each year wth the assumpton that each bank faces heterogeneous loan demand, whch s contradctory to the formulaton of the supply equaton wth loans of banks as a homogeneous product. 8 If we estmate Equaton (3) wthout ths rearrangement, the dependent varable s the lendng nterest rate. Dscrepances n the lendng rates among banks contradct the formulaton of the demand functon that assumes loans of all banks are a homogeneous product. By multplyng both sdes of the equaton wth the quantty of loans, we can transform the lendng nterest rate n the left hand sde to the nterest ncome. Ths allows us to avod ths theoretcal contradcton. 12

16 lnc = β β β1 ln q + (ln q ) + γ k lnωk, + φk ln q lnωk, + γ kk (lnωk, ) 2 k= 1 k= 1 2 k= γ (5) 12 lnω1, lnω2, where ln ω 1, and 2, lnω stand for the depost nterest rate and wage, respectvely. Dfferentatng (5) wth respect to q yelds the followng margnal cost as below: MC 2 C = β 1 + β 2 ln q + φk lnωk,. (6) q k= 1 Substtutng t nto (3 ) completes the supply equaton. As an emprcal strategy, we estmate the bank s supply functon (3 ) smultaneously wth the cost functon (5) on the assumpton that estmatng the supply functon smultaneously wth the cost functon wll mprove the precson of estmaton. We bascally follow Angeln and Cetorell [2003] n the formulaton of the estmaton equatons. One dfference s that we add a control varable, ln( npl ) n the cost equaton, wth npl referrng to the amount of NPLs. Snce the sample span ncludes the perod of the crss, ncludng notably the abrupt rse n non-performng loans (NPLs) and the assocated declne n the net amount of loans, ths term s expected to control such shocks. Thus, the formulaton of the estmaton equatons s as follows; lnc = β β γ k lnωk, + φk ln q lnωk, β1 ln q + (lnq ) + γ kk (lnωk, ) 2 k= R, W k= R, W 2 k= R, W 1 + γ ) + 12 ln ω1, lnω2, + ρ ln( npl ε (7) R = 2 1C + β 2C q + φkc lnωk, + k= R, W m ( θ η) q ν β ln + (8) m 13

17 In Equaton (8), ( θ η) m (m= 1993, 1994,, 2004) stands for the degree of competton for each year, from whch we construct the Lerner ndex. 3.2 Data Our data set covers the sx largest commercal banks for the perod We select these banks accordng to the followng crtera: () banks whch exsted throughout the perod of analyss; () banks that mantaned more than three percent of the bank loan market share throughout the perod of analyss, and () banks whch were not majorty-owned by foregn nvestors. Wth regard to the frst crteron, those banks whch were lqudated or merged wth other banks have been elmnated from the data set snce the cost structure of these banks may dffer from others due to ther bad lendng practce. The thrd crteron s based on the questonnare survey of Okuda [2004] whch showed that the target customers of the foregn-owned banks dffer from those of other locally ncorporated banks, and that there s possble market segmentaton. The data we employ have been derved from the annual reports of ndvdual banks, Statstcal Data on Commercal Banks n Thaland, publshed annually by 9 These are Bangkok Bank, Krung Tha Bank, Kaskornbank (formerly Tha Farmers Bank), The Sam Commercal Bank, TMB Bank (formerly Tha Mltary Bank), and Bank of Ayudhaya. 14

18 Bangkok Bank, and Thaland s Commercal Banks, Nov. 25, 1997 by Tha Investment and Securtes Publc Company Lmted. We provde the defnton of varables and ther descrptve statstcs n Table 3. Wth regard to output, we use net earnng assets, whch comprse loans as well as nvestments. As we have seen n the last secton, after the crss, banks shfted ther portfolos towards securtes. Usng net loans as the volume of output leads to an underratng of bankng operatons n recent years. Second, nterest rates are ex post values, that s, nterest ncomes (expenses) dvded by net earnng assets (total labltes). Therefore, the lendng nterest rate gets lower than the nomnal ex ante lendng rate when the net earnng assets comprses unclassfed NPLs, ths beng partcularly the case n Thrd, we defne the costs and the depost nterest rate n two ways. In the basc defnton of the costs ( C ), we count only the nterest expendture and personnel expenses because these tems are two man varable components of ordnary expendture. In the alternatve defnton, gven the marked ncrease n contrbuton to FIDF n recent years, we nclude ths expendture tem n the costs, ( C 2 ). Accordngly, the depost nterest rate, ω R 2,, s calculated by dvdng the sum of the nterest costs and the contrbuton to FIDF wth total labltes. 15

19 Table 3, here. 4. Estmaton Results 4.1 Results We estmate Equatons (7) and (8) smultaneously by three-stage least squares (3SLS) regressons, usng the pooled data of the sx largest commercal banks for As for nstrumental varables, n addton to the exogenous factor prces and the amount of NPLs, we employ the lagged varables of the terms that nclude q and C. We frst estmate Equaton (7) and (8). The estmated parameters of the cost functon do not fulfll the condton of monotoncty (not reported). Consderng the small sample sze, we mght need a more parsmonous specfcaton of the cost equaton. In fact, we test the null hypothess that the cost equaton takes the form based on the Cobb-Douglas producton functon, and cannot reject the null hypothess. Accordngly, we estmate the system wth the restrctve specfcaton of the cost functon as below The data set covers the perod As we use some lagged varables for nstruments, the sample s shortened to The wrong sgns of the parameters of the translog cost equaton are also observable n prevous studes ncludng Angeln and Cetorell [2003] and Uchda and Tutsu [2005]. Instead of a problematcal cost equaton, we opt for a cost functon, albet 16

20 ln C β 0 + β1 ln q + γ 1 lnωr, + γ 2 lnωw, + ρ ln( npl ) + δ k bankk + ε (7 ) = R = C + m ( θ η) q ν β 1 + (8 ) m Also, we nclude ntercept dummes for ndvdual banks, bank k (k=2, 3, 4, 5, 6), to account for possble dfferences n cost effcences. Table 4 reports the results of estmaton. In general, the ft of the cost functon s satsfactory. All the estmated coeffcents have the expected sgns. β < 1 suggests 1 economes of scale n the bankng ndustry. Wth regard to the dfferences n cost effcences among banks, a Wald test rejects the null hypothess of no dfferences. We also perform a Hausman specfcaton test, whch also s n favor of the model that ncludes the bank dummes. On the other hand, whether or not the contrbuton to FIDF s ncluded n the costs does not much affect the results of the estmaton. Thus, we put forward the analyss based on the estmaton result of model (3) n Table 4. Table 4, here. To construct the Lerner ndex, we dvde ( θ η) m for each year by an average of the ex post lendng nterest rates of sx banks for the correspondng perod. We depct restrctve, wth approprate propertes. 17

21 the derved ndex n Fgure 4. As can be seen, the Lerner ndex swung down consderably n However, we do not consder ths s due to ntensfed competton among banks. On the contrary, the low score of the Lerner ndex s related to non-fled NPLs n the earnng assets. As the loan classfcaton was stll loose n the late 1990s, banks had a consderable amount of non-fled NPLs on ther balance sheets. Non-fled NPLs lowered the ex post lendng nterest rate, whch resulted n the low score of the Lerner ndex. On the other hand, a more nterestng fndng from Fgure 4 s the rse n the Lerner ndex n the post-crss perod. How can we nterpret ths change? We devote the rest of the analyss to dscusson of ths queston. Fgure 4, here. 4.2 Dscusson The rse n the Lerner ndex n recent years mples that, despte changes n the compettve envronment of the bankng ndustry n terms of ownershp structures and regulatons, competton n the ndustry mght have declned durng the 2000s. In ths regard, we cannot perform a statstcal analyss to regress the derved Lerner ndex on some relevant state varables because of the shortness of the sample perod. Instead, we 18

22 lst some ponts that mght be related to the rse n the Lerner ndex. One factor that mght account for the rsng trend of the Lerner ndex s the busness cycle. In the condtons of slowdown that have been evdent n the economy snce the fnancal crss, both lendng and depost nterest rates have declned markedly n recent years. Moreover, the moderate declne of the lendng nterest rate n relaton to the depost nterest rate, and the wdenng nterest margn mght be ndcatve of a declne n the credtworthness of borrowers. However, a quanttatve evaluaton of these changes s beyond the scope of ths present paper. Second, we look at loan demand elastcty. The Lerner ndex s composed of the sem-elastcty of loan demand to the lendng nterest rate, and the conjectural elastcty. Thus, the ncrease n the Lerner ndex n recent years mght be due to the declne n the sem-elastcty of loan demand. Temporarly suspendng the assumpton that the loans of all banks are homogeneous products, we estmate, as a tentatve exercse, the followng loan demand functon wth fxed effects regresson. ln q, t = α 0 + α1 lnyt + α 2 lnyt 1 + ηt p, t + ( υ + ε, t ), (9) q, t n ths estmaton refers to the loans of Bank, net of allowance for doubtful t accounts. and Y stand for the ex post lendng nterest rate for Bank and the p, t t Gross Domestc Product (GDP), respectvely. η t s the ndustry-wde average 19

23 sem-elastcty of loan demand. We summarze the result of estmaton n Table 5. We have to note that ths s possbly a based estmaton because of mssng explanatory varables. Takng ths nto account, we compute the conjectural elastcty, θ, and depct t n Fgure 5. In ths fgure, though tentatve, we can confrm the trend of rsng θ. Thus, the results from the prelmnary estmaton of demand functon do not deny the possblty of a declne n the bankng competton snce the fnancal crss. Table 5, here. Fgure 5, here. If the degree of competton really s declnng, what are the possble reasons for such a behavoral change among the banks? One possble explanaton s that the banks need to ncrease net nterest ncome to make up for the losses from the dsposal of NPLs. As can be seen n Fgure 1 n Secton 2, despte the rsng net nterest ncome, the return on assets stll remans much lower than the pre-crss level. Such a condton mght be related to a possble declne n bankng competton. 5. Concludng remarks In ths artcle, we have estmated the degree of competton n the Tha bankng 20

24 ndustry n order to evaluate how the East Asan crss and the subsequent reforms affected the structure of the bankng ndustry. In partcular, applyng the method of Bresnahan s [1989] conjectural varaton model, we have derved a Lerner ndex whch measures the mark-up of prce over margnal cost. In spte of changes n the compettve envronment of the bankng ndustry n terms of ownershp structures and regulatons, the Lerner ndex showed a rsng trend n the post-crss perod. Durng the post-crss reforms, famly ownershps n a number of banks were replaced wth state and foregn ownershps. There were also new entrants to the ndustry, a phenomenon that had been absent for a long perod. These changes may make t dffcult to mantan colluson, f any, between banks. However, prelmnary estmaton results reveal the possblty of a declne n competton. The present analyss s lmted n the sense that we cannot perform statstcal analyss on the causes of a change n the degree of competton due to the shortness of the sample perod. Nonetheless, t s hoped that ths paper can be regarded as a useful contrbuton to the emprcal lterature on the Tha bankng sector untl such tme as an accumulaton of quanttatve studes makes t possble to undertake more n-depth polcy analyses. There reman several outstandng ssues arsng from our analyss. Frst, the 21

25 sample perod s short, whch not only affect the robustness of estmatons, but also makes t dffcult to judge the effects of the fnancal reforms on the degree of competton as the effect may have not yet have fully worked ts way through. Second, the assumpton of the quantty competton among banks n a statc framework may be restrctve, especally for the crss and reform perod. We leave the resoluton of these problems as a task for future research. 22

26 References Angeln, P., and Cetorell, N. [2003] Bank Competton and Regulatory Reform: The Case of the Italan Bankng Industry. Journal of Money, Credt and Bankng, Vol. 35(5), pp Anuchworawong, C., Souma, T., and Wwattanakantang, Y. [2004] Famly Ownershp of Fnancal Insttutons: Evdence from Thaland, n Hanazak, M., and Teransh, J. (eds.) An Economc Analyss of Corporate Governance: Japan and East Asa n Changng Envronment. Unversty of Tokyo Press: Tokyo, pp (n Japanese) Berg, S.A, and Km, M. [1994] Olgopolstc Interdependence and the Structure of Producton n Bankng: An Emprcal Evaluaton. Journal of Money, Credt and Bankng, Vol. 26(2), pp Bkker, J.A., and Haaf, K. [2002] Competton, concentraton and ther relatonshp: an emprcal analyss of the bankng ndustry, Journal of Bankng and Fnance, Vol. 26(11), pp Bresnahan, T.F. [1989] Emprcal studes of ndustres wth market power, n Schmalensee, R., and Wllg, R., (eds.), Handbook of Industral Organzaton.Vol.2, North-Holland: Amsterdam, pp

27 Okuda, H. [2004] The Impact of Foregn Penetraton on the Tha Bankng Industry, Kahatsu-knyu-kenkyujo-ho Vol. 19, JBIC Insttute, Japan Bank for Internatonal Cooperaton (n Japanese) ---, and Rungsomboon, S. [2004a] Comparatve Cost Study of Foregn and Tha Domestc Banks : Estmatng Cost Functons of the Tha Bankng Industry, CEI Workng Paper Seres, No , Htotsubash Unversty ---, and ---. [2004b] The Effects of Foregn Bank Entry on the Tha Bankng Market: Emprcal Analyss from 1990 to 2002, CEI Workng Paper Seres, No , Htotsubash Unversty Polsr, P., and Wwattanakantang, Y. [2005] Corporate Governance of Banks n Thaland. Paper presented at Corporate Governance of Banks n Asa-Second Semnar, Asan Development Bank Insttute, Tokyo, January 21, 2005 Shaffer, S. [1993] A Test of Competton n Canadan Bankng, Journal of Money, Credt, and Bankng, Vol. 25(1), pp , and DSalvo, J. [1994] Conduct n a bankng duopoly, Journal of Bankng and Fnance, Vol. 18(6), pp Uchda, H., and Tsutsu, Y. [2005] Has competton n the Japanese bankng sector mproved? Journal of Bankng and Fnance, Vol.29 (2), pp

28 Vchyanond, P. [2000] Fnancal Reforms n Thaland, Thaland Development Research Insttute: Bangkok ---, [1994] Thaland s Fnancal System: Structure and Lberalzaton, Thaland Development Research Insttute Research Monograph No. 11, Thaland Development Research Insttute: Bangkok 25

29 Table 1: Credt Extended by Fnancal Insttutons (FIs) and Number of FIs Credt extended by Fnancal Insttutons mllons of baht share(%) mllons of baht share(%) mllons of baht share(%) mllons of baht share(%) Commercal banks 2,161, ,825, ,585, ,081, of whch 6 largest locally ncorporated banks 1,659, ,086, ,597, ,607, of whch foregn bank branches 111, , , , Fnance companes 547, ,488, , , Credt-foncer (mortgage loan) companes 5, , , , Lfe nsurance companes 21, , , , Government-owned fnancal nsttutons 170, , , ,189, Total 2,906, ,911, ,962, ,598, Number of Fnancal Insttutons Commercal banks local banks foregn banks Fnance companes Credt-foncer (mortgage loan) companes Government-owned fnancal nsttutons Source: Economc and Fnancal Statstcs (varous ssues), Bank of Thaland Statstcal Data on Commercal Banks n Thaland (varous ssues), Bangkok Bank 26

30 Table 2: Reorganzaton of the Tha Bankng Industry Bangkok Bank Bangkok Bank Tha Farmers Bank Kaskornbank (Tha Farmers Bank) Sam Commercal Bank Sam Commercal Bank Bank of Ayudhaya Bank of Ayudhaya Tha Mltary Bank TMB Bank (Tha Mltary Bank) new bankng lcense granted to a fnance company n 2001 Thanachart Bank Krung Tha Bank (state owned bank) Krung Tha Bank natonalzed n 1998 Sam Cty Bank Sam Cty Bank Bangkok Metropoltan Bank natonalzed n 1998 and ntegrated wth Sam Cty Bank n 2002 Frst Bangkok Cty Bank natonalzed and closed n 1998 loans/ labltes transferred to Krung Tha Bank Bangkok Bank of Commerce natonalzed n 1996 and closed n 1998 loans/ labltes transferred to Krung Tha Bank natonalzed n 1998, ntegrated wth a fnance company, transformed nto a new bank n 1998 BankTha Unon Bank of Bangkok BankTha BankTha natonalzed n 1999, acqured by ABN AMRO Bank n 1999 Bank of Asa Bank of Asa acqured by Development Bank of Sngapore n1998, merged wth TMB Bank Tha Danu Bank acqured by Standard Chartered Bank Nokornthon Bank Standard Chartered Nakornthon Bank natonalzed and newly establshed as Radanasn Bank n1997, acqured by Unted Overseas Bank n1998 Laem Thong Bank UOB Radanasn Bank Source: Compled from data n Table 2:2 of Polsr and Wwattanakantang [2005] 27 State owned banks Foregn owned banks

31 Table 3: Defnton of varables and descrptve statstcs Table 3. (A) Defnton Varable defnton Total earnng assets, net of allowance for doubtful accounts q p R npl C C 2 Interest ncome/ Total earnng assets Interest ncome Non performng loans The sum of nterest expenses and personnel expenses The sum of nterest expenses, personnel expenses, and contrbuton to FIDF ω R, Interest expenses/ Total labltes ω R 2, [Interest expenses+ contrbuton to FIDF]/ Total labltes Source: Author ω w, Personnel expenses/ Number of employees Note: FIDF = Fnancal Insttuton Development Fund (see text for explanaton) Table 3. (B) Descrptve statstcs q p R npl C C(2) ω(r) ω(r2) ω(w) ml. baht percent ml. baht ml. baht ml. baht ml. baht percent percent baht Average 539, ,488 79,780 34,379 36, ,115 Maxmum 1,129, , , , , ,427 Mnmum 136, ,754 1,297 8,913 10, ,471 Standard Devaton 251, , ,807 21,170 21, ,997 Source: Author s calculaton 28

32 Table 4: Results of estmaton (cost and supply equatons) (1) (2) (3) (4) wthout FIDF contrbuton wthout FIDF contrbuton wth FIDF contrbuton wth FIDF contrbuton wth bank dummes wthout bank dummes wth bank dummes wothout bank dummes coeffcent p-value coeffcent p-value coeffcent p-value coeffcent p-value β β γ γ δ bk bk bk bk bk (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) (θ/η) No. of observatons R-squared Cost equaton Supply equaton Wald test Hausman test Source: Author s calculaton Note: The system of equaton estmated s ln C 0 + β1 ln q + β 2 lnωr, + β3 lnωw, = β + δ ln( npl ) + ε R = 1 C + m ( θ η) q ν β + m Followng varables are used as nstrumental varables. (1) Exogenous varables: ln( npl ) q C, and. lnω r,, w, lnω, and dummy varables for each year. (2) Lagged varables of endogenous varables:, ln q 29

33 Table 5: Results of estmaton (demand equaton) coeffcent p-value α α α η η η η η η η η η η η η No. of observatons 72 R-squared wthn between overall Source: Author s calculaton 30

34 Fgure 1: Selected ndces of performance of commercal banks, Mllon baht 400, , percent , , , , , ,000 ROA (%, rght scale) Net nterest ncome Net profts Provson Source: Statstcal Data on Commercal Banks n Thaland, varous ssues. Note: ROA = return on assets 31

35 Fgure 2: Interest rates and nterest margn, Lendng Rate Depost Rate Interest margn M1 1989M9 1990M5 1991M1 1991M9 1992M5 1993M1 1993M9 1994M5 1995M1 1995M9 1996M5 1997M1 1997M9 1998M5 1999M1 1999M9 2000M5 2001M1 2001M9 2002M5 2003M1 2003M9 2004M5 Source: Internatonal Fnancal Statstcs CD-ROM, Internatonal Monetary Fund. 32

36 Fgure 3: Structure of expenses of commercal banks, (A) Interest and non-nterest ncome year Interests on loans and deposts Other nterests and dvdends revenue Non-nterest ncome % 2% 4% 6% 8% 10% 12% 14% percentage of total assets (B) Interest expenses and provson for possble loan losses year Interest expenses Provson for possble loan losses % 2% 4% 6% 8% 10% 12% 14% 16% percentage of total assets (C) Non-nterest expenses year Personnel expenses Other operatng expenses Contrbuton to FIDF Loss on sales of nvestment and securtes Other expenses 0% 1% 2% 3% 4% 5% percentage of total assets Source: Statstcal Data on Commercal Banks n Thaland, varous ssues. Note: Other operatng expenses consst of premses and equpment, taxes and dutes, and drectors remuneraton. Untl 1997, Contrbuton to FIDF and Loss on sales of nvestment and securtes are counted n Other expenses. 33

37 Fgure 4: Lerner ndex Source: Author s calculaton Note: The Lerner ndex s calculated usng the followng formula: L ( θ / η) / p m =, m=1993, 1994, m m, θ s obtaned from the estmaton result (4) n Table 4. p m s an average of the ex ( / η) m post lendng nterest rates of the sx banks for the correspondng year. 34

38 Fgure 5: Conjectural elastcty of loan supply Lerner ndex Conjectural elastcty Source: Author s calculaton 35