Analysis of Technical Efficiency of Indian Banking Sector: An Application of Data Envelopment Analysis

Transcription

1 Internatonal Journal of Fnance & Bankng Studes IJFBS Vol.3 No.1, 2014 SSN: avalable onlne at Analyss of Techncal Effcency of Indan Bankng Sector: An Applcaton of Data Envelopment Analyss Dpayan Roy a a Indan Insttute of Management, Indore, Inda Abstract The proactvely evolved bankng regulatons n the Indan Bankng sector under the authoratve drectve of the Reserve bank of Inda (RBI) has often brought about a change n the busness strategy, captal structure and operatons of the banks n the Indan bankng sector. Durng these events of contnuous change and adopton of Basel norms, we analyse the effcency of the Indan bankng sector wth usng Data Envelopment Analyss across three economc eras andacross the dfferent ownershp structures. The determnants of effcency are selected on the bass of ntermedaton approach. We also attempt to dentfy whether the neffcency arses from manageral ncompetence or mproper sze and resource allocaton. From our analyss, we dentfy the man cause of neffcency n the Indan Bankng sector to be arsng out of mproper sze allocaton. Keywords: Effcency, Basel norms, Data Envelopment Analyss, Intermedaton Approach JEL classfcaton: M Publshed by SSBFNET 1. Introducton The need to measure the effcency of the bankng sector s spread across varous stakeholders- regulators, bank management, share holders of banks, customers of the banks. From a regulators pont of vew, t s essental to measure the effcency of the banks to assess the probablty of default snce banks are sources of funds n the economy and hence shapes the productvty of an economy. The bank management needs to estmate the effcency to devse strateges for survval n the compettve envronment. The shareholders are concerned about the returns generated by the banks where as the customers are concerned about the servce qualty and prce of the products traded by the banks. Earler studes on effcency of the banks mostly dealt wth the study of fnancal ratos and the analyss of scale and scope effcences. However researchers later noted that there were more aspects of effcency such as techncal effcency and allocatve effcency whch needed to be taken nto account when studyng effcency (Molyneux, Altunbas& Gardener, 1996). These two factors were ntroduced by Farell (1957) and Lebensten (1966) wth the concept of X-effcency whch encompassed both these measures. Although the advantage of X-effcency over scale and scope effcences were known for a long tme, t s just a decade and a half snce researchers has turned ther attenton to X-effcency. The measurement of X-effcency nvolves constructon of the effcency fronter from the a Correspondng author.dpayan Roy, Tel.:

2 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: sample data by usng any parametrc or non-parametrc approach (Farel, 1957). Varous parametrc and non parametrc approaches are used for estmatng the techncal effcency. The Indan bankng sector can be broadly classfed on the bass of four ownershp structures- natonalsed banks, prvate sector banks, foregn banks and state bank and ts assocates. These banks are regulated by the Reserve Bank of Inda (RBI). Although, most of the regulatons are smlar across the dfferent ownershps, the RBI mposes some regulatons whch vary across the ownershps. 2. Lterature Revew A comprehensve lterature revew of the studes on effcency of banks s out of the scope of study. In ths secton, we shall manly focus on the studes pertanng tothe analyss of effcency of banks n the Indan bankng sector.the man objectve of fnancal lberalzaton was to make the fnancal nsttutons more effcent and competent. Indan fnancal system contnues to be a bank based fnancal system and the bankng sector plays an mportant role as a resource moblser. It remans the prncpal source of resources for many households, small and medum enterprses and also caters the large ndustres. Studes by Bhattacharya et. al(1997), Sathye (2003), Das &Ghosh (2006), Ray & Das (2009), Koeva(2003) have tred to measure the effcency of the Indan bankng sector. Bhattacharya et. al (1997), n ther study examned the productvty effcency of 70 Indan commercal banks between 1986 to 1991 usng non-parametrc DEA to estmate the cost and proft effcency of the Indan bankng sector n the post reform perod. Ray and Das (2009) found that the publc sector banks are more effcent than the prvate banks. The latest study n ths regard s that of Das and Drne (2011). The study revealed that the publc sector banks and the prvate banks have the mean effcency whch s hgher than the all banks mean effcency. The foregn banks are found to be least effcent among the bank groups. The publc sector banks are the most effcent, followed by the domestc prvate banks.another set of lterature comes from the works of Das and Ghosh (2006) and Ghosh (2009). In the frst study, t was found that there was an asymmetry n the techncal effcency of the banks over the years. It was also found that there was a decreasng trend n the techncal effcency score durng the sample perod ( ). The next study ndcates hgh levels of effcency n costs and lower levels n profts, reflectng the mportance of neffcences on the revenue sde of bankng actvty, the proxmate determnants of proft effcency appear to suggest that bg state-owned banks performed reasonably well and are more lkely to operate at hgher levels of proft effcency. Kumar and Gulat (2008) analysed the pure techncal effcency and scale effcency of the natonalsed banks and found that the techncal effcency of the Indan bankng sector s not affected by the asset qualty. Sarkar et al. (1998) had suggested a prvatzaton of the bankng sector n the emergng economes lke Inda to ncrease the effcency n the sector. However Ray and Das (2009) also argue that prvatzaton has not ncreased the effcency of the Indan bankng sector. The reason for ths may be due to the fact that the stock market and other forms of fnancal markets have not yet grown n Inda.In another recent study, Das (2010) had compared the effcency scores of dfferent sectors of the banks and found that the domestc prvate sector banks were becomng more effcent compared to that of the publc sector banks and the foregn banks. It was also found that unlke other countres where the bankng sector reforms were able to ncrease the cost effcency, there has been no sgnfcant change n the Indan publc sector banks n the post reform perod. Although ths area s wdely researched, there exsts no lterature whch compares the effcency of the Banks across the eras of three dfferent economc reforms.e. Bankng Sector reforms, Basel I and Basel II. In ths secton, we try to compare the effcency of the Indan bankng sector across the three economc era separately. 151

3 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: Research and Methodology 3.1 Research Objectve Analyss of effcency of the bankng sector has been of great nterest to the researchers all over the world. The exstng set of lterature also contans a rch volume of research pertanng to the Indan Bankng sector. However, these researches comprse of the study of the effcency of the bankng sector n the post lberalzaton era. Even after the fnancal lberalzaton of The Indan Bankng sector has undergone a seres of reforms n order to comply wth the Basel Norms. A comparatve analyss of effcency of the Indan Bankng sectors across the dfferent eras of economc reforms remans unexplored. In ths paper we dvde the tme perod of our study nto three economc eras: pre Basel, Basel I and Basel II (detaled n the later secton) and analyse the overall techncal effcency of the Indan Banks across these three eras. Further we also analyze the causes of neffcency by decomposng the overall Techncal effcency nto pure techncal effcency and scale effcency and construct a set of effcent leaders of the Indan Bankng sector. The most wdely used non-parametrc approach s the Data Envelopment Analyss (DEA) whle Stochastc Fronter Approach (SFA) and Dstrbuton Free Approach (DFA) are the most frequently used parametrc approaches (Kosak&Zajc, 2006). However there s no general consensus about the choce of methodology. The dfferent approaches for measurement of effcency s expressed n fgure 1as documented by Kumar and Gulat (2008). Fronter Approaches for measurement of Bank Effcency Parametrc Approach Non Parametrc Approach Stochastc Fronter Approach (SFA) Data Envelopment Analyss (DEA) Thck Fronter Approach (TFA) Free Dsposal Hull Dstrbuton Free Approach (DFA) Fgure 1:Fronter Approaches for measurement of Bank Effcency Before proceedng any further, t s essental to elaborate certan concepts of techncal, pure and scale effcency. Techncal Effcency related to the productvty of nputs (Sathye, 2003) and s the measure of the bank s ablty of transforms multple resources nto multple fnancal servces (Bhattacharya et al., 1997; Kumar and Gulat, 2008). A bank s sad to be techncally effcent f t operates along the fronter. The measure of techncal effcency under constant returns to scale s termed as Overall techncal effcency (OTE) whch can be further decomposed nto two components - pure techncal effcency (PTE) and scale effcency (SE). The PTE s the measure of techncal effcency under varable returns of scale and reflects the manageral competence for proper allocaton of resources. The SE s a measure of the rato of OTE to PTE and reflects the ablty of the management to choose the optmum sze of resources or n other words to choose the scale of producton to attan the expected producton level Choce of Methodology In ths study we shall try to measure effcency of Indan Bankng Sector wth DEA. We prefer ths methodology over the parametrc approaches for the followng reasons: 152

4 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: DEA easly accommodates multple-nputs and multple-outputs of the banks.. v. It allows the estmaton of overall techncal effcency (OTE) and decomposes t nto two mutually exclusve and non-addtve components, namely, pure techncal effcency (PTE) and scale effcency (SE)(Kumar and Gulat, 2008). There s no need to select a pror functonal form relatng to nputs and outputs lke Cobb-Douglas and Translog producton/cost functons (Banker, 1984) DEA provdes a scalar measure of relatve effcency, and the areas for potental addton n outputs and reducton n nputs (Kumar and Gulat,2008) v. It s not necessary to provde values for weghts assocated wth nput and output factors, although the user may exert nfluence n the selecton of weght values v. DEA works partcularly well wth small samples (Evanroff and Isralevch, 1991) However one major shortcomng of DEA s that t assumes the data to be free of measurement error and may gve unrelable results f the ntegrty of the data source s not assured. 3.3 Data Envelopment Analyss (DEA) DEA s a lnear programmng technque whch converts multple nputs and outputs (measured n any unt) of each Decson Makng Unt nto a scalar measure of effcency by assgnng weghts to the nputs and outputs of a DMU that gve t the best possble effcency (Ray,2004).The mathematcal expresson for DEA can be expressed as : Maxmze Y / X for each value of k from 1 to n, subject n each case to k k o Y j / X j <= 1, j= 1 to n, where o Y k means *Y k, = 1 to s, X k means *X k, = 1 to m Y j means *Y j, = 1 to s and j = 1 to n X j means *X j, = 1 to m and j = 1 to n The soluton s the set of maxmum values for Y / X and the assocated values for and k k DEA computes the scores by comparng the performance of a Decson Makng Unt (DMU) wth respect to ts peers. Hence the effcency scores reflect the comparson of a DMU wth respect to the best performance n the ndustry. The orentaton of DEA s on the bass of two approaches- Input orentaton and Output orentaton. In Input orentaton, we explore the possblty of reducng nputs to produce a gven level of output level. It s the measure of the rato of the actual output to the maxmum possble output. In the output orented approach we explore the possblty of a possble expanson of outputs for a gven set of nput varables measured by the rato of mnmum possble nput to actual nput. The two approaches comprse of the prmal and dual soluton of the optmzaton problem. 3.4 Choce of Varables The choce of the nput and output varables play an mportant role n the studes of effcency. Although there s no general consensus n the lterature about the varables whch consttute the nput and output varables (Casu and 153

5 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: Grardone, 2002; Sathye, 2003), there exsts two approaches for the choce of varables n the studes of effcency of the bankng sector Producton Approach and Intermedaton Approach (Bauer et al, 1998). In the producton approach, banks are vewed as servce provders to customers (Benston, 1965) where as n the ntermedaton approach banks are vewed as channels between depostors and credtors (Gupta et al, 2008). In the producton approach emphass s gven to operatng cost by completely gnorng nterest expenses. The varables that are commonly used n ths approach are proxes of servces provded to the customers and are best measured by the number and type of transactons, documents processed or specalzed servces provded over a gven tme perod and costs related to labour, technology, materal space etc. The wdely set of varables n ths approach comprses of data on the number of deposts and loan accounts, as a surrogate for the level of servces provded as the output varables (Kumar and Gulat, 2008).In the producton approach, banks produce ntermedaton servces through the collecton of deposts and other labltes and ther applcaton n nterest-earnng assets, such as loans, securtes, and other nvestments and consder both operatng cost and nterest cost. Thus Deposts are consdered to nput varable compared to the producton approach where deposts are taken as output. In the real scenaro, banks functon as provders of servces as well as fnancal ntermedares (Berger & Humphrey, 1991). However ntermedaton approach s sutable for analysng Bank level effcency whereas producton effcency s sutable for analysng branch level effcency (Kumar and Gulat,2008; Berger & Humphrey, 1991). In ths study our choce of varables s on the bass of ntermedaton approach. Although varous factors have been used by researchers to measure the effcency as nputs and outputs, the performance of a bank s determned by monetzed value of a set of nputs and outputs. In ths study the output varables are () Net Interest Income () Non nterest ncome and the set of nput varables are () Operatng Expenses () Number of Employees () Physcal Captal- measured as the sum of fxed assets and other assets (v) Loanable Funds- measured as the sum of deposts and borrowngs. DEA results are nfluenced by the number of nputs and output varables. Our choce of the nput and output varables s n agreement wth the rule of thumb lad down by Cooper et al. (2007). The rule of thumb can be mathematcally expressed as: n >max{m x s ; 3(m+ s)}; where : n= Number of Banks; m= Number of Input varables ; s= Number of Output varables 3.5. Data Set and Tme perod The data set of our study comprses of an unbalanced panel of 62 scheduled commercal banks dvded across the four ownershp structures: Natonalsed banks(20), Prvate sector banks(21), Foregn banks(13) and the State bank of Inda (SBI) and ts accocates(8). We have dvded the tme perod nto three eras: pre Basel era ( ) whch marks the era n between the bankng sector reforms and the mplementaton of Basel I norms wth the market rsk amendment; Basel I era ( ) and Basel II era ( ). The dataset s not a balanced panel owng to the mergers and entry of new banks n the study perod. 4. Results and Fndngs The average OTE of the Prvate sector banks was found to be the hghest n all the eras and that of the Natonalsed banks was found to be the least. However n the Basel II era, there was an ncrease n the average OTE of the foregn banks and t matched up wth that of the prvate sector banks. There were maxmum number of foregn banks whch were sgnfcant throughout our study perod although the number of foregn banks n our sample was the least. In some years, there was not a sngle natonalsed bank or SBI and ts assocates whch was effcent Although there has been an ncrease n the effcency of the natonalsed banks after the adopton of the Basel accords durng the Basel I era, t reduced to a great extent n the Basel II era. There was not much sgnfcant change n the effcency scores of SBI and ts assocates and the prvate sector banks n the Basel I era compared to that of the pre - Basel era but reduced sgnfcantly n the Basel II era. The effcency scores of the foregn banks were found to ncrease over the three eras. Although there has been an ncrease n the effcency of the natonalsed banks after the adopton of the Basel accords durng the Basel I era, t reduced to a great extent n the Basel II era. There was not 154

6 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: much sgnfcant change n the effcency scores of SBI and ts assocates and the prvate sector banks n the Basel I era compared to that of the pre -Basel era but reduced sgnfcantly n the Basel II era. The effcency scores of the foregn banks were found to ncrease over the three eras. Insert Table 1 Here The next step n the analyss of techncal effcency s to analyse the causes of neffcency. We therefore decompose the overall techncal effcency nto pure techncal effcency (PTE)and scale effcency (SE). The pure techncal effcency s calculated usng varable returns to scale. The summary of the results are llustrated n Table 2 below. We have further calculated the scale effcency, a measure of the rato of PTE to OTE. The summary s detaled n Table 3. The mean PTE of the natonalsed banks ncreased from pre-basel era to Basel I era but decreased n the Basel II era. There was no varaton n the mean PTE from pre-basel era to Basel I era but t reduced n the Basel II era for the SBI group. There was a gradual ncrease n the mean PTE of the foregn banks across the three eras and margnal varaton n case of the domestc prvate sector banks. Insert Table 2 Here In the pre-basel era, the mean PTE of the foregn banks were observed to the hghest whereas that of the natonalsed banks were observed to the least. However n the Basel I era, the domestc prvate sector banks were observed to the least effcent whereas the foregn banks mantaned ther hghest poston n terms of average PTE. However, n case of scale effcency analyss, we observed that the scale effcency of the domestc prvate sector banks were hghest n the pre-basel era and Basel I era and that of the natonalsed banks were the least. In the Basel II era, the SE of the foregn banks were observed to the hghest and the natonalsed banks were observed to the least. There was a huge decrease n the mean SE scores of the natonalsed banks and the SBI group over the years. Insert Table 3 Here On the bass of the number of effcent banks across the dfferent ownershps, we can dentfy the cause of neffcency. In all cases, we noted a consderable amount of neffcency arsng from both manageral ncompetence and mproper sze. However the number of banks whch were found to be neffcent on the bass of mproper sze was much more compared to the banks that were found to be neffcent on the bass of manageral ncompetence. 5. Concluson and Implcatons Our analyss suggests that the effcency of the foregn banks have ncreased through manfolds over the the three eras and were consstently effcent n terms of pure techncal effcency and scale effcency across the three eras. The prvate sector banks showed margnal varaton across the three eras n case of all the effcences. However, n case of the SBI group and the natonalsed banks, there was a sgnfcant decrease n the OTE scores wth the major cause of the neffcency beng mproper sze allocaton. The problem of mproper sze and resource allocaton remans an area of concern for the banks across all the four ownershp structures. The followng study only meausres the effcency scores and compares t across the dfferent economc eras and dfferent ownershps. However, dscrmnaton of the effcent banks on the bass of the analyss of slacks remans to be explored for gettng a clear pcture about the cause of neffcency. Ths process would also allow us to buld a reference set of the effcenct banks whch could be vewed as a example for the neffcent banks n order to devse 155

7 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: ther busness strateges to maxmse ther effcency. Further, ths study refers to the Indan bankng sector and hence the results cannot be genralsed wthout further analyss, although the framework of the study can be used for analyss for any study on effcency of the banks. References Banker, R,. D,. (1984). Estmatng most productve scale sze usng data envelopment analyss. European Journal of Operatonal Researc., 17, 1, Bauer, P.W.; Berger, A.N.; Ferrer, G.D. and Humphrey, D.B. (1998). Consstency condtons for regulatory analyss of fnancal nsttutons: a comparson of fronter effcency methods. Journal of Economcs and Busness. 50, Benston, G,,J.(1965). Economes of Scale and Margnal Costs n Bankng Operatons.Natonal Bankng Revew2, Berger, A.,N. and Humphrey, D.,B. (1991). The domnance of neffcences over scale and product mx economes n bankng.journal of Monetary Economcs.28, Bhattacharyya, A., Lovell, C., &Sahay, P. (1997).The Impact of Lberalzaton on the Productve Effcency of Indan Commercal Banks.European Journal of Operatonal Research, 98(2): Casu, B,.andGrardone,C. (2004). An Analyss of the Relevance of Off-Balance Sheet Items n Explanng Productvty Change n European Bankng. Money Macro and Fnance (MMF) Research Group Conference , Money Macro and Fnance Research Group. Cooper, W.W., Seford, L.M. & Tone, K. (2007).Data Envelopment Analyss: A Comprehensve Text wth Models, Applcatons, References and DEA-Solver Software (Second Edton). New York: Sprnger Scence + Busness Meda. Das, A and Ghosh, S. (2006). Fnancal Deregulaton and Effcency: An Emprcal Analyss of Indan Banks Durng the Post Reform Perod. Revew of Fnancal Economcs, 15(3), Das, S.K. and Drne, I. (2011). Fnancal Lberalzaton and Bankng Sector Effcency n Inda: A fourer Flexble Functonal Form and Stochastc Fronter Approach. Internatonal Busness and Management, 2(1):42-58 Evanoff, D. D., and. Isralevch, P.R (1991). Productve effcency n bankng, Economc Perspectves. Federal Reserve Bank of Chcago, Farrell, M. J. (1957) The Measurement of Productve Effcency. Journal of the Royal Statstcal Socety, Seres A, 120 (3): Ghosh, S. (2009). Fnancal Deregulaton and Proft Effcency: A Non-parametrc Analyss of Indan Banks. Journal of Economcs and Busness, 61, 6, Gupta,O.K., Dosht,Y. and Chnnubha, A.(2008) Dynamcs of Productve effcency of Indan Banks. Internatonal Journal of Operatons Research 5, 2, Koeva, P.(2003). The Performance of Indan Banks durng Lberalzaton. IMF Workng Paper no WP/03/150 Kosak, M. and Zajc, P. (2006).Determnants of bank effcency dfferences n the new EU member countres, Fnancal Stablty Report, Expert Papers, May Ljubljana: Bank of Slovena, Kumar, S., and Gulat, R., (2008), Evaluaton of Techncal Effcency and Rankng of Publc Sector Banks n Inda: An Analyss from Cross-sectonal Perspectve, Internatonal Journal of Productvty and Performance Management, 57,7, Lebensten, H,.(1966), Allocatve Effcency vs X-Effcency.Amercan Economc Revew 156

8 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: Molyneux,P., Altunbas, Y., and Gardener, E,P,.M.(1996)."Effcency n European Bankng", Wley Ray, S. C. (2004). Data envelopment analyss: Theory and technques for economcs and operatons research. Cambrdge, UK: Cambrdge Unversty Press. Ray, S. C and Das, A. (2009), Dstrbuton of Cost and Proft Effcency: Evdence from the Indan Bankng, European Journal of Operatonal Research Sarkar,J., Sarkar,S., &Bhaumk, S.,K. (1998). Does Ownershp Always Matter? Evdence from the Indan Bankng Industry. Journal of Comparatve Economcs, 26, Sathye M (2003). Effcency of Banks n a Developng Economy: the Case of Inda. European Journal of Operaton Research, 148(3),

9 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: Table 1: Summary of OTE across dfferent ownershps for dfferent years Pre- Basel Basel I Basel II Natonalsed Banks of Era No of Banks No. of effcent Banks Pvt Sector Banks of Era No of Banks No. of effcent Banks Foregn Banks of Era No of Banks No. of effcent Banks SBI and ts Assocates of Era No of Banks No. of effcent Banks

10 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: Table 2: Summary of PTE across dfferent ownershps for dfferent years of Era No. of effcent Banks of Era No. of effcent Banks of Era No. of effcent Banks of Era No. of effcent Banks Pre- Basel Basel I Basel II Natonalsed Banks Pvt Sector Banks Foregn Banks SBI and ts Assocates

11 Roy /Internatonal Journal of Fnance & Bankng StudesVol3, No 1, 2014 ISSN: Table 3: Summary of SE across dfferent ownershps for dfferent years of Era No. of effcent Banks of Era No. of effcent Banks of Era No. of effcent Banks of Era No. of effcent Banks Pre- Basel Basel I Basel II Natonalsed Banks Pvt Sector Banks Foregn Banks SBI and ts Assocates