Evaluating the Financial Performance of Turkish Banking Sector: A Fuzzy MCDM Approach. Aydın Çelen 1

Transcription

1 Journal of Economc Cooperaton and Development, 35, (04), Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Aydın Çelen At the turn of the century Turksh economy was meltdown due to the one of the harshest crses experenced throughout her hstory. Snce then, thanks to very strct measures taken by Turksh government, the economy proceeded to the recovery perod. Ths perod, however, seems to be nterrupted once more agan by global fnancal crss. In ths study, we assess the fnancal performances of the dfferent segments of the Turksh bankng sector for the perod between 00 and 00. In dong ths, a combned methodology of Fuzzy Analytc Herarchy Process (FAHP) and Technque for Order Preference by Smlarty to Ideal Soluton (TOPSIS) methods s utlzed. The results ndcate that the adverse effect of the 00 fnancal crss on the Turksh bankng sector s enormous whle the ongong global crss s not so destructve n the sector.. Introducton All organzatons should measure ther performances to some extent n order to detect organzatonal problems and provde mprovement. The need for a flawless system of performance measurement s even more crucal n today s compettve busness envronment. Bankng sector s not excepton to ths. The measurement of the performance n the bankng sector has been always one of the maor ssues for the banks, governments and academcans. However, t s not easy to evaluate performances of the banks gven that ther products and servces are of an ntangble nature. For ths reason, there s abundant lterature dscussng dfferent research methods appled to performance measurement of banks. Turksh Competton Authorty Ankara Turkey e-mal: acelen@rekabet.gov.tr

2 44 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Performance measurement of banks has been tradtonally acheved by help of several fnancal ratos such as return on assets (ROA), return on equty (ROE) and earnngs per share (EPS). However, the usefulness of tradtonal fnancal performance measurement systems started to be dscussed as the busness envronment changed n the last couple of the decades. As a result, the researchers have attempted to overcome the drawback of the tradtonal measurement system. Ths can be acheved wth two dfferent ways. Frstly, multdmensonal performance measurement systems whch ncorporate both fnancal and nonfnancal measures such as customer satsfacton, management capacty and employee relatons are ntroduced. SMART pyramd (Lynch and Cross, 99), the balanced scorecard (Wu et al. 009), CAMEL (Chrstopoulos et al., 0), DEA (Aydn et al., 009) and TOPSIS (Secme et al., 009) are examples of ths new performance measurement system. Secondly, nstead of usng fnancal ratos ndvdually, sets of these ratos are started to be used. Most of these sets of the ratos are based on ROE and dvde ROE nto ts component ratos. The examples of the sets of ratos are Dupont System of Fnacal Analyss (McGowan et al., 0) and Scherenbeck s Basc ROE Scheme (Badreldn, 009). Snce each of these contemporary performance measurement methods has ts own advantages and dsadvantages, none of them can be seen as perfect. Thus, evaluatng pro and cons of each method, one can select the most sutable method accordng to her study s partcular stuaton. In ths study, we determned our am as evaluatng the overall fnancal performances of the dfferent segments of the Turksh bankng sector. As explaned n the followng secton, there are plenty of lterature on performance measurement of the Turksh banks ndvdually. For ths reason, n ths study, we focused on assessng the overall performances of the dfferent bankng segments nstead of an ndvdual evaluaton at bank level. In addton, although we were aware of the fact that nonfnancal measures show up as mportant performance measurement crtera n the bankng sector, n ths study we delberately focused on only fnancal measures because of several reasons. Frst, t s not an easy task to quantfy non-fnancal measures such as customer satsfacton, qualty and flexblty, management capacty, nnovaton capablty and employee relatons at ndvdual bank level. At the more aggregated segments level, t s more dffcult and even meanngless to try to take nto account these non-fnancal measures. Second, dfferent

3 Journal of Economc Cooperaton and Development 45 from the tradtonal rato analyss whch reles on lmted number of fnancal measures, n ths study, we took nto account a large number of fnancal measures (9 n total) wthn a combned framework of Fuzzy Analytc Herarchy Process (FAHP) and Technque for Order Preference by Smlarty to Ideal Soluton (TOPSIS) methods. Wth FAHP method, the weghts attached to these fnancal performance crtera by expert decson markers are determned. Then, TOPSIS method s used to rank the fnancal performances of the dfferent segments of the Turksh bankng sector. The remanng of the study s as follows: In the next secton, lterature revew on performance measurement n Turksh bankng sector s presented. Ths s followed by a secton explanng the research methodology ncludng FAHP and TOPSIS methods. Then, emprcal results are provded, and n the fnal secton, the results of the study are dscussed.. Performance Measurement n Turksh Bankng Sector In the bankng lterature, three dfferent approaches have been employed n assessment of the effcency of the bankng sector accordng to the actvtes of the banks. These are producton approach, ntermedaton approach and proftablty approach. In the producton approach, banks are assumed to produce deposts and provde loans by usng ther workforce and captal. In the ntermedaton approach, banks are consdered to serve as ntermedares between those who depost and those who wthdraw money. In other words, the man functons of the banks are to convert the deposts whch they collect nto loans. And fnally, accordng to the proftablty approach, the man motvaton of a bank s to ncrease ts proft (Drake, Hall and Smper, 005). The approach used n the assessment of the bankng effcency s crucal gven that the nput/output varables, and probably the results of the applcatons, may vary dependng on the approach used. A revew of the lterature on assessng effcency of the Turksh bankng sector reveals that a great number of studes employng all three types of these approaches seem to be conducted. However, these studes generally utlze the non-parametrc approaches such as Data Envelopment Analyss (DEA) rather than parametrc approaches such as

4 46 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Stochastc Fronter Analyss (SFA). Among the motvatons of these studes, two become promnent: Frst, the studes are motvated to nvestgate the effects of lberalzaton polces on the effcency of Turksh bankng sector. Second, the effects of the ownershp structure of banks on ther performances are frequently examned by usng dfferent tme perods. Some of the studes whch are conducted to nvestgate the relatonshp between the effcency and fnancal lberalzaton n Turkey are as follows: Based on an ntermedaton approach, Zam (995) analyzes the effects of the lberalzaton polces of the 980s on the effcency of the Turksh commercal banks for the perod. Ths study wtnesses substantal mprovements n the effcences of the banks followng to the lberalzaton of captal movements. Mercan and Yolalan (000) also reaches smlar results for the perod: The fnancal performance of the banks ncreases after the fnancal reforms but starts deteroratng wth the 994 crss. However, ths result s not n lne wth Yldırm (999) and Denzer et al. (000). Analyzng the effcency of Turksh commercal banks throughout , Yldrm (999) reaches the concluson that effcency levels n Turksh bankng sector dd not ncrease followng the lberalzaton. Denzer et al. (000), usng both the ntermedaton and the producton approaches smultaneously for the perod , clams that bank effcency decreased followng the lberalzaton n Turkey. Another stream n the lterature s examnng the effects of the ownershp structure of banks on ther fnancal performances: Yolalan (996) compares the effcency performances of foreng, prvate and state-owned banks operaton n Turkey durng the years by usng fnancal ratos as nputs and outputs. Ths study shows that foregn banks n Turkey are more effcent than ther domestc counterparts. Among the domestc banks, prvate banks have hgher effcency scores than publc ones. Ths fndng s also supported by Jackson, Feth, and Inal (998), Cng and Tarm (000) and Mercan and Yolalan (000). Examnng perod by help of a producton approach, Jackson et al. (998) reveals that prvate commercal banks operated more effcency than publc banks durng the 994 crss. Cng and Tarm (000) apples a mxture of the ntermedaton and producton approaches for the years , and concludes that prvate

5 Journal of Economc Cooperaton and Development 47 commercal banks are more effcent compared to state-owned banks. Mercan and Yolalan (000) ndcates that both prvate and foregn banks operate more effcently than state-owned ones durng the before and after 994 crss. Atan and Catalbas (005) also analyzes the mpacts of the dfferentaton n the captal structure of banks and concludes that prvate and foregn banks are more effcent than state-owned banks. Onal and Sevmeser (006) also concludes that foregn banks n Turkey operate more effcently than state-owned and prvate banks although ther growth trends are rather smlar. Isık and Uysal (006) fnd that prvate and foregn banks n Turkey are better at adaptng themselves to changng operatonal envronments and better motvated to utlze emergng opportuntes. Aydn et al. (009) assesses the effcency of the bankng segments for the perod by usng DEA method. Acordngly, the most effcent banks are state-owned, followed by foregn and prvate banks, respectvely. Ths fndng of Aydn et al. (009) does not comply wth fndngs of earler research n lterature. In lterature, several studes am to compare the effcences of the bankng sectors n dfferent countres. For example, employng DEA method on four homogenous groups of countres, Stavárek (006) examnes whether the dfferences among regons and countres n ther stages of European ntegraton and economc development are also vsble n bankng effcency. Results suggest that dfferences n bankng effcency exst among the regons analyzed, and the herarchy corresponds wth the herarchy of regons n terms of economc development and degree of ntegraton. However, the rapd growth of bankng ntermedaton effcency recorded n Central and East European countres could trade off the actual effcency gap compared to tradtonal EU members. Andres (0) studes the determnants of the effcency and productvty of the bankng systems of seven central and east European countres by help of DEA and SFA. The results show that the average effcency of banks n central and east European countres grew durng a fve-year perod, from 004 to 008. The mprovement s attrbuted to the ncreased competton upon EU accesson and the entry of foregn banks, as well as to extensve legslatve changes that led banks to become more effcent. Based on the results, the hghest level of techncal effcency s recorded for the bankng systems of Romana and the Czech Republc, and the lowest s recorded for Slovena.

6 48 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Performance assessment of the bankng sector has been also acheved n a mult-crtera decson makng (MCDM) framework. The am of the MCDM s to obtan the optmum choce that provdes the hghest degree of satsfacton consderng all of the relevant crtera. To the best of our knowledge, Secme et al. (009), Cetn and Cetn (00) and Demrel (00) are the studes whch evaluate the performances of Turksh banks usng a MCDM model. Secme et al. (009) proposes a fuzzy performance evaluaton model wth both fnancal and non- fnancal performance ndcators for fve commercal banks n the Turksh bankng sector by usng FAHP and TOPSIS methods together. Ths study suggests that not only fnancal performance but also non-fnancal performance such as customer satsfacton and servce qualty should be taken nto account n a compettve envronment. Cetn and Cetn (00) evaluates the fnancal performances of the Turksh banks whch are traded n Istanbul Stock Exchange (ISE) by usng Vse Krterumska Optmzaca I Kompromsno Resene (VIKOR) method. Çetn and Çetn (00) concludes that the researchers usng VIKOR method may test ther results wth alternatve crtera s weghts gven that ths method s senstve to these weghts. Demrel (00) analyzes the performances of three state-owned commercal banks. In dong ths, equal weghts are gven to the fnancal crtera whch are used n the study, and then performance scores are obtaned by usng TOPSIS method. Ths study concludes that the state-owned commercal banks are beng effected from local and global fnancal crss, the performance scores contnuously fluctuate, and they cannot show any outstandng performance mprovements over tme. 3. Research Methodology 3.. Fuzzy Analytc Herarchy Process (FAHP) Analytc Herarchy Process (AHP), frstly proposed by Saaty (980), has a wde range of applcatons. AHP uses herarchcal structures to represent a problem and then calculate weghts for alternatves accordng to the udgments of the decson makers n a parwse comparson framework. The conventonal verson of AHP method s often crtczed owng to usng the exact and crsp udgments of the decson makers. On the other hand, decson makers are more confdent about nterval udgments than fxed value udgments. Because of the

7 Journal of Economc Cooperaton and Development 49 vagueness and ambguous are nherent of the human udgments and preferences, real lfe stuatons can be modeled more adequately by usng fuzzy values than the exact numercal values. Another handcap of the AHP method s that the preferences n AHP are essentally human udgments based on ther subectve perceptons. Therefore, a fuzzy verson of the AHP method, called Fuzzy Analytc Herarchy Process (FAHP), has been ntroduced n order to take nto consderaton subectve uncertanty of the varables. A fuzzy number s a specal fuzzy set A xr x x A A, where x takes ts values on the real lne : x and s a contnuous mappng from to the closed nterval [0, ]. Trangular fuzzy numbers and trapezodal fuzzy numbers are the most popular fuzzy numbers thanks to ther computatonal smplcty. Trangular fuzzy numbers are preferred for representng the lngustc varables n ths study. ~ and ts A trangular fuzzy number can be donated as M l, m, u membershp functon ~ x: 0, can be gven as M 0 x l or x u ~ l x m () M x x l m l x u m u m x u where l m u and l, m, and u descrbe the smallest possble value, the most promsng value, and the largest possble value of a fuzzy event, respectvely. Membershp functon of a trangular fuzzy number M ~ s llustrated n Fgure (Deng, 999):

8 50 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Fgure : A Trangular Membershp Functon, ~ ( x) ~ M ( x ) M M l(y) M r(y) 0 l m u ~ ~ be two trangular fuzzy numbers, Let M l, m, u, M l, m u, the basc operatons of trangular fuzzy numbers used n ths study are defned as follows: ~ ~ M M l l, m m, u u, ~ ~ M M ll, mm, uu, ~ M l, m u, 0, R, ~ M / u,/ m,/ l, m, u, l, m, u where l 0 () In ths study, crtera weghts of the performance measures are calculated by usng extent analyss of Chang (996). To descrbe the X x x,..., an obect extent analyss of Chang (996), frstly let, x n set, and G g g,..., be a goal set. Accordng to the method,, g n extent analyss for each goal s performed respectvely. Therefore, m extent analyss values for each obect can be obtaned: m M, M,..., M, =,,,n, (3) g g g where all M (=,,,m) are trangular fuzzy numbers. In ths g framework, Chang s extent analyss can be gven as follows (Ertugrul and Karakasoglu, 009):

9 Journal of Economc Cooperaton and Development 5 Step : The value of fuzzy synthetc extent wth respect to the th obect s defned as follows: S m M g n m M g m To obtan M, the fuzzy addton operaton of m extent analyss g values for a partcular matrx s performed as follows: m M g m m m and to obtan m (4) l, m, u (5) n (=,,,m) values s performed as follows: n m M g n M n g, the fuzzy addton operaton of n M g l, m, u (6) and then the nverse of the vector above s computed as follows: n m n n n M g u, m, l (7) Step : When M ( l, m, u) and M ( l, m, u) are two trangular fuzzy numbers, the degree of possblty M l, m, u ) M ( l, m, ) s defned as V ( u ( M M M M ) sup mn ( x), ( ) (8) y yx and can be equvalently stated as: V M M hgt M M, f m m, M ( d) 0, f l u (9) l u otherwse m u m l,

10 5 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Fgure llustrates equaton (9) where d s the ordnate of the hghest ntersecton pont D between M ( x) and (Zhu et al., 999). The M values of both V( M M) and V( M M) are needed to compare M and M. Fgure : Intersecton Between M and M V(M M ) M M D 0 l m l d u m u x Step 3: The degree of possblty for a fuzzy number to be greater than k fuzzy numbers M (,,..., k) can be defned by V M M, M,..., M ) V ( M M ) and ( M M ) and ( M M ) (0) ( k k mn V( M M ),,,..., k () Assume that d ( A ) mn V( S Sk ) for k,,..., n; k. Then the weght vector s gven by ( A T ), d( A ),..., d( An ) W d () where A (,,..., n) are n elements. Step 4: The normalzed weght vectors are obtaned by normalzaton as ( A T ), d( A ),..., d( An ) W d (3)

11 Journal of Economc Cooperaton and Development Technque for Order Preference by Smlarty to Ideal Soluton (TOPSIS) TOPSIS s orgnally proposed by Hwang and Yoon (98), and became one of the classcal MCDM methods. Accordng to ths method, alternatves to be evaluated by n attrbutes n a MCDM problem are presented as ponts n an n-dmensonal space. A fundamental assumpton of TOPSIS s that each attrbute has a tendency towards monotoncally ncreasng or decreasng utlty. In ths method, frstly postve deal solutons (PIS) and negatve deal solutons (NIS) are determned. The postve deal soluton s a soluton that maxmzes the beneft crtera and mnmzes the cost crtera, whereas the negatve deal soluton maxmzes the cost crtera and mnmzes the beneft crtera (Wang and Elhag, 006). In short, the postve deal soluton s composed of all best values attanable of crtera, whereas the negatve deal soluton conssts of all worst values attanable of crtera (Wang, 008). TOPSIS method consders the dstances to both the PIS and the NIS smultaneously by defnng relatve closeness to deal soluton. The alternatve whch s the closest to postve deal soluton and farthest from the negatve deal soluton s selected as the best alternatve. To explan the algorthm of TOPSIS, suppose we have m alternatves ( A, A,..., Am ), and n decson crtera ( C, C,..., Cn). Each alternatve s evaluated wth respect to the n crtera. All the ratng scores assgned to the alternatves wth respect to each crteron form a decson matrx denoted by X ( ). Let W w, w,..., w ) be the relatve weght x mxn n ( n vector about crtera, satsfyng w. The algorthm of TOPSIS s as follows (Ertugrul and Karakasoglu, 009): Step : Decson matrx x mxn X ( ) s normalzed by dvdng each performance ratng of the decson matrx wth ts norm, whch s called vector normalzaton procedure. In other words, for beneft attrbutes, the normalzed value r s obtaned by r x m x (4)

12 54 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach whle for cost attrbutes, r s computed as r (/ x ) m (/ x ) (5) where x s the performance ratng of th alternatve for attrbute C. Step : Weghted normalzed decson matrx V ( ) s obtaned by multplyng normalzed matrx wth the weghts of the crtera: v mxn v r w (6) where =,,,m and =,, n. Step 3: Postve deal soluton (PIS) and negatve deal soluton (NIS) are determned: PIS v,..., v max ( v, v vn where ) n NIS v, v,..., v where v mn ( v ) (7) (8) Step 4: The dstance of each alternatve from PIS and NIS are calculated: d n ( v v ) (9) d n where =,,,m. ( v v ) (0) Step 5: The closeness coeffcent of each alternatve ( CC ) s calculated: d CC where =,,,m () d d

13 Journal of Economc Cooperaton and Development 55 Step 6: The rankng of the alternatves are determned accordng to CC values: The bgger CC, the better the relevant alternatve. In other words, the alternatve wth the hghest closeness coeffcent s determned as the best alternatve. 4. Applcaton Followng very severe fnancal crss experenced n 00, rather heavy regulatons regardng fnancal nsttutons were ntroduced n Turkey. In ths study, we analyze the overall fnancal performances of the dfferent segments of the Turksh bankng sector through the perod between 00 and 00. The data used n ths study s obtaned from Banks Assocaton of Turkey (BAT). Turksh fnancal sector has been always domnated by banks: Accordng to the asset szes, as of February 0, 77% of the assets belong to the banks. As can be seen from Table, 44 banks n total operate n Turkey, 3 of them beng depost and 3 development and nvestment banks. The banks whch operate n dfferent segments of the Turksh bankng sector are llustrated n Fgure 3. Amongst depost banks, there are 3 state-owned banks, prvate banks and 6 foregn banks. The Deposts Insurance Fund (DIF) owns bank. As parallel to the growth n the fnancal market n Turkey, the number of branches and employees of banks ncrease contnuously. As of February 0, the number of branches and employees reach to 9,84 and 8,77, respectvely. Total asset of the bankng sector s approxmately 66 bllon US Dollars. Almost all of ths total s owned by the depost banks. Indeed, the depost banks domnate not only bankng sector, but also all fnancal sector.

14 56 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Table : Number of Banks, Branches, Employees and Total Asset n Turksh Bankng Sector Bank Branch Employee Total Asset Turksh Bankng Sector (TBS) 44 9,84 8,77 66,043 Depost banks (DEB) 3 9,798 76,03 605,570 State-owned banks (SDB) 3,894 49,8 85,958 Prvate banks (PDB) 4,996 89,54 334,696 Banks n DIF (DIF) Foregn banks (FDB) 6,907 37,47 84,47 Founded (FFB) 0,88 36,85 80,8 Branched (FBB) ,90 Dev't. and nv. banks (DIB) ,45 0,47 State-owned banks (SDI) 3 3,88,543 Prvate banks (PDI) 6 6,030 7,678 Foregn banks (FDI) ,5 Notes: () The meanngs of the abbrevatons denotng the segments of the bankng sector are as follows: TBS: Turksh Bankng Sector, DEB: Depost Banks, SDB: State-owned Depost Banks, PDB: Prvate Depost Banks, DIF: Banks Under the Depost Insurance Fund, FDB: Foregn Depost Banks, FFB: Foregn Banks Founded n Turkey, FBB: Foregn Banks Havng Branches n Turkey, DIB: Development and Investment Banks, SDI: State-owned Development and Investment Banks, PDI: Prvate Development and Investment Banks, FDI: Foregn Development and Investment Banks. () Source: BAT (3) Total assets are n mllon USD. In analyzng the fnancal performance of dfferent segments of Turksh bankng sector, FAHP and TOPSIS methods are used together. Whle FAHP s used for determnng the weghts of man and sub-crtera n the lght of opnons of an expert group, the TOPSIS method s used for evaluatng the performances of the segments. Meanwhle, the expert group conssts of three decson makers. The frst decson maker s selected from a state-owned depost bank whle the second one s from a prvate depost bank. And the thrd one s an academcan wth a consderable experence on bankng.

15 Journal of Economc Cooperaton and Development 57 Fgure 3: Banks and Segments of Turksh Bankng Sector

16 58 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Although there are many types of fnancal ratos used n the evaluaton of fnancal performance n bankng sector, assessment results can vary accordng to the dfferent ratos. A bank or bankng segment showng a hgh performance accordng to one rato may have a very low performance accordng to another rato (Secme et al., 009). For ths reason, we tred to obtan the evaluatons of the expert group regardng the relatve mportance of all avalable fnancal ratos. Bearng ths n mnd, we determned 9 fnancal ratos whch may be relevant, and grouped them under 6 man crtera. These man crtera are Captal Ratos, Balance-Sheet Ratos, Assets Qualty, Lqudty Ratos, Proftablty Ratos and Income-Expendture Structure. To determne the relatve mportance of two crtera, Saaty -9 scale (Saaty, 980), llustrated n Table, s employed. In ths study, we only explaned how the weghts for man crtera are calculated. The explanatons regardng sub-crtera were not presented here, but may be provded upon request. The abbrevatons denotng fnancal crtera and ther meanngs are presented n Table 3. Ths table also ncludes the calculated weghts for all man and sub-crtera n parentheses. Preferences n lngustc varables Table : Scales for Parwse Comparson Preferences n numerc varables Equal mportance Moderate mportance Strong mportance Very strong mportance Extreme mportance Intermedate values f necessary ,4,6,8

17 Journal of Economc Cooperaton and Development 59 Table 3: Herarchcal Crtera Set Crteron Explanaton and calculated weght CR Captal Ratos (0.0) CR Shareholders Equty / (Amount subect to credt + market + operatonal rsk) (0.0) CR Shareholders' Equty / Total Assets (0.) CR3 (Shareholders' Equty - Permanent Assets) / Total Assets (0.) CR4 Net On Balance Sheet Poston / Total Shareholders' Equty (0.8) CR5 Net On and Off Balance Sheet Poston / Total Shareholders' Equty (0.0) BR Balance-Sheet Ratos (0.06) BR TC Assets / Total Assets (0.6) BR TC Labltes / Total Labltes (0.8) BR3 FC Assets / FC Labltes (0.7) BR4 TC Deposts / Total Deposts (0.5) BR5 Total Deposts / Total Assets (0.7) BR6 Funds Borrowed / Total Assets (0.7) AQ Assets Qualty (0.0) AQ : Fnancal Assets (net) / Total Assets (0.6) AQ : Total Loans and Recevables / Total Assets (0.7) AQ3 : Total Loans and Recevables / Total Deposts (0.7) AQ4 : Loans Under Follow-up (net) / Total Loans and Recevables (0.6) AQ5 : Specfc Provsons / Loans Under Follow-up (0.7) AQ6 : Permanent Assets / Total Assets (0.7) LR Lqudty Ratos (0.9) LR Lqud Assets / Total Assets (0.43) LR Lqud Assets / Short-term Labltes (0.8) LR3 TC Lqud Assets / Total Assets (0.9) PR Proftablty Ratos (0.0) PR Net Proft/Losses / Total Assets (0.8) PR Net Proft/Losses / Total Shareholders' Equty (0.44) Proft/Losses Before Taxes after Contnung Operatons / Total Assets PR3 (0.38) IE Income-Expendture Structure (0.6) IE Net Interest Income After Specfc Provsons / Total Assets (0.9) Net Interest Income After Specfc Provsons / Total Operatng Income IE (0.0) IE3 Non-nterest Income (net) / Total Assets (0.6) IE4 Other Operatng Expenses / Total Assets (0.) IE5 Personnel Expenses / Other Operatng Expenses (0.6) IE6 Non-nterest Income (net) / Other Operatng Expenses (0.7)

18 60 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach In constructng the trangular fuzzy numbers from the decson makers parwse comparson grades, we used respectvely the mnmum and maxmum grades gven by decson makers for the lower (l) and upper (u) bound of the relevant fuzzy number. As for the most promsng value of the fuzzy number, we used the arthmetc mean of the grades gven by decson makers. The parwse comparson matrx ncludng the fuzzy numbers calculated s presented n Table 4. Table 4: Fuzzy Parwse Comparson Matrx CR BR AQ LR PR IE CR (,, ) (7, 8.33, 9) (0.,.7, 7) (0., 5.37, 9) (0.,.4, 7) (7, 8.33, 9) BR (0., 0., 0.4) (,, ) (0., 0.4, ) (0., 0.4, ) (0., 0.4,) (0., 0.4, ) AQ (0.4, 3.38, 9) (, 6.33, 9) (,, ) (0., 4.7, 9) (0., 4.7, 9) (, 5, 9) LR (0., 3.08, 9) (, 6.33, 9) (0., 3., 9) (,, ) (0., 3.37, 9) (0., 3.37, 9) PR (0.4, 6.05, 9) (, 6.33, 9) (0., 3., 9) (0., 3.37, 9) (,, ) (, 6.33, 9) IE (0., 0., 0.4) (, 6.33, 9) (0., 0.44, ) (0., 3.37, 9) (0., 0.4,) (,, ) Then usng the fuzzy numbers n comparson matrx, synthess values respect to man crtera calculated as n equaton (4): S CR S BR S AQ S LR S PR S IE / 06.9,/ 97.48,/ 8.5 (0.0743, 0.888,.4733) ( 5.33, 8.5, 4) / 06.9,/ 97.48,/ 8.5 (0.0075, 0.08, 0.804) (.56,.75, 5.4) / 06.9,/ 97.48,/ 8.5 (0.063, 0.577,.636) ( 3.37, 5., 46) / 06.9,/ 97.48,/ 8.5 (0.08, 0.079,.636) (.44, 0.6, 46) / 06.9,/ 97.48,/ 8.5 (0.063, 0.687,.636) ( 3.37, 6.9, 46) / 06.9,/ 97.48,/ 8.5 (0.08, 0.97, 0.746) (.44,.67,.4) By usng the equaton (9), fuzzy numbers are compared: V( S S ), V( S S ), V( S S ), V( S S ), V( S S ) CR BR CR AQ CR LR V( S S ) 0.9, V( S S ) 0.4, V( S S ) 0.48, V( S S ) 0.4, V( S S ) 0.65 BR CR BR AQ CR LR V( S S ) 0.98, V( S S ), V( S S ), V( S S ) 0.99, V( S S ) AQ CR AQ BR AQ LR CR CR AQ PR PR PR CR CR IE IE AQ IE

19 Journal of Economc Cooperaton and Development 6 V( S S ) 0.95, V( S S ), V( S S ) 0.97, V( S S ) 0.96, V( S S ) LR CR LR BR LR V( S S ) 0.99, V( S S ), V( S S ), V( S S ), V( S S ) PR CR PR BR PR V( S S ) 0.8, V( S S ), V( S S ) 0.84, V( S S ) 0.89, V( S S ) 0.83 IE CR IE BR Then, accordng to the equaton (), prorty weghts are calculated: d ( CR) mn(,,,,) d ( BR) mn(0.9, 0.4, 0.48, 0.4, 0.65) 0.9 d ( AQ) mn(0.98,,, 0.99,) 0.98 d ( LR) mn(0.95,, 0.97, 0.96,) 0.95 d ( PR) mn(0.99,,,,) 0.99 d ( IE) mn(0.8,, 0.84, 0.89, 0.83) 0.8 IE When we normalze these prorty weghts of man crtera, we obtan the weght of 0.0 for Captal Ratos, 0.06 for Balance-Sheet Ratos, 0.0 for Assets Qualty, 0.9 for Lqudty Ratos, 0.0 for Proftablty Ratos and 0.6 for Income-Expendture Structure. Accordngly, Captal Ratos, Assets Qualty and Proftablty Ratos are seen almost equally as the most mportant crtera whle Balance-Sheet Ratos s evaluated as the least mportant crteron. After determnng the weghts of the all man and sub-crtera (fnancal performance crtera), we proceed to the applcaton of TOPSIS method. Heren we explan only applcaton of TOPSIS method to overall Turksh bankng sector (TBS) to save space. The explanatons about the applcaton of the model to segments may be provded upon request. Fnancal performance crtera are normalzed accordng to equatons (4) and (5). In dong ths, all sub-crtera except BR6, AQ4, AQ6, IE4 and IE5, are consdered as beneft attrbutes rather than cost attrbutes. Meanwhle, the crtera whch are generated from depost (namely BR4, BR5 and AQ3) are not consdered for development and nvestment banks (DIB, SDI, PDI and FDI) gven that these banks do not collect depost. The weghts of these crtera are dstrbuted to other crtera accordng to ther ntal weghts. After gettng the normalzed matrx, we multply each normalzed value of sub-crtera wth ther weghts accordng to equaton (6). Then, these weghted normalzed values of sub-crtera under each man crteron are aggregated, and Table 5 s obtaned. At the end of applcaton, the total values of man AQ AQ AQ IE LR PR LR PR LR IE LR PR PR IE IE

20 6 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach crtera are multpled by the weghts of the man crtera (0.0, 0.06, 0.0, 0.9, 0.0, 0.6), and total weghted values of man crtera (Table 6) are obtaned. Table 5: Total Values of Man Crtera for TBS Years CR BR AQ LR PR IE Table 6: Total Weghted Values of Man Crtera for TBS Years CR BR AQ LR PR IE After calculatng total weghted values of man crtera, postve deal soluton (PIS) and negatve deal soluton (NIS) are determned by takng the maxmum and mnmum values for each crteron accordng to equaton (7) and (8): PIS 0.05, 0.00, 0.074, 0.07, 0.079, maxmum values NIS 0.0, 0.08, 0.039, 0.046, 0.038, mnmum values Then, the dstance of each bank from the postve deal soluton and negatve deal soluton wth respect to each crteron s calculated by usng equaton (9) and (0). Dstances from postve deal soluton and negatve deal soluton are presented n Table 7.

21 Journal of Economc Cooperaton and Development 63 Table 7: Dstances from Postve Ideal Soluton and Negatve Ideal Soluton for TBS Years Dstance Dstance from PIS from NIS Once the dstances from postve deal soluton and negatve deal soluton are determned, the closeness coeffcents of utltes ( CC ) are calculated by equaton (). And fnally, accordng to the closeness coeffcent values, the rankngs of the utltes are determned, as presented n Table 8. Table 8: Rankngs of the Years Accordng to Closeness Coeffcent Values for TBS Rankng Years Closeness Coeffcent By employng the TOPSIS methods to dfferent segments of the Turksh bankng sector, we obtaned the fnancal performances of all segments for the perod Before proceedng to nterpret the fnancal performance scores of the bankng segments, presented n Table 9, one mportant pont should be emphaszed: In ths study, we analyze each bankng segment separately usng TOPSIS method. For ths reason, each bankng segment should be analyzed ndependently. In other words, ths study enables us to evaluate the annual fnancal performances of

22 64 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach bankng segments over the perod To be more precse, the fnancal performance scores presented n Table 9 should be nterpreted horzontally, not vertcally. As an alternatve to our study, consderng each year ndvdually, one may compare the fnancal performances of the segments wth each others n a gven year. We dd not follow ths delberately: It seems to be more meanngful and nterestng to analyze each bankng segment over the years rather than to compare the dfferent segments whch have rather dfferent structures and prortes n a gven year. However, admttedly, a cross evaluaton of the fnancal performances of the bankng segments n a gven year may be carred out, whch can be topc of another study. As seen from the frst two row of the Table 9, Turksh bankng sector n general and ts depost bankng segment has followed smlar fnancal performances durng perod: The worst fnancal performance was exhbted n 00, whch s the year ust after the 00 fnancal crss, whle the best performance was shown n 007, whch s the year ust before the ongong fnancal crss. Indeed, the smlarty between performance trends of the bankng sector and ts depost bankng segment s not surprsng gven that Turksh bankng sector s domnated by depost banks. As for the remanng segments of the bankng sector, the results are as follows: State-owned depost banks, state-owned development and nvestment banks, foregn depost banks n general and foregn depost banks founded n Turkey seem to be affected most negatvely by the ongong crss, thus take the lowest performance scores n 008, the begnnng year of the crss. On the other hand, banks under the Depost Insurance Fund, foregn development and nvestment banks, prvate depost banks and development and nvestment banks n general exhbted ther worst fnancal performances n years 003, 004, 005 and 007 respectvely, whle foregn banks havng branches n Turkey and prvate development and nvestment banks performed poorly n 006. As for the year n whch the best fnancal performances were shown, the results are somewhat mxed: State-owned depost banks overperformed n 004 whle prvate depost banks showed ther best performance n 008. On the other hand, 003 was the year when foregn depost banks n general, foregn depost banks havng branches n Turkey, development and nvestment banks n general, prvate and foregn depost and nvestment banks exhbted ther best fnancal performances. Foregn

23 Journal of Economc Cooperaton and Development 65 depost banks founded n Turkey and state-owned development and nvestment banks took ther hghest performance scores n 005, depost banks n general delvered an outstandng performance n 007. Table 9: Fnancal Performance Scores of Dfferent Segments of Turksh Bankng Sector Segment TBS DEB SDB PDB DIF FDB FF B FB B DIB SDI PDI FDI Note: The meanngs of the abbrevatons denotng the segments of the bankng sector are as follows: TBS: Turksh Bankng Sector, DEB: Depost Banks, SDB: Stateowned Depost Banks, PDB: Prvate Depost Banks, DIF: Banks Under the Depost Insurance Fund, FDB: Foregn Depost Banks, FFB: Foregn Banks Founded n Turkey, FBB: Foregn Banks Havng Branches n Turkey, DIB: Development and Investment Banks, SDI: State-owned Development and Investment Banks, PDI: Prvate Development and Investment Banks, FDI: Foregn Development and Investment Banks. 5. Concluson The rapd advancements n nformaton technologes and ncreasng competton n natonal and nternatonal markets have started to challenge all banks to take tmely measures to survve. Undoubtedly, the frst and crucal step n determnng the necessary management decsons s measurng the performances of the banks adequately. However, t s not easy to fnd the perfect method for the effectve evaluaton of the performance n bankng sectors supplyng ntangble products and servces. For ths reason, n lterature, a wde range of methods have been appled for performance evaluaton of bankng sectors. In ths

24 66 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach study, we am to assess the overall fnancal performances of the dfferent segments of the Turksh bankng sector. To be more precse, the effects of two fnancal crses, whch are experenced n 00 and n 008 and contnung, on dfferent bankng segments are assessed. In dong ths, dfferent from other studes n the lterature, a large number of fnancal measures (9 n total) are consdered wthn a combned framework of FAHP and TOPSIS methods. FAHP method s used to determne the weghts attached to these fnancal performance crtera by expert decson markers. Then, the fnancal performances of the dfferent segments of the Turksh bankng sector over perod are ranked accordng to the TOPSIS method. The results of FAHP method ndcates that Captal Ratos, Assets Qualty and Proftablty Ratos are almost equally the most mportant crtera whle Balance-Sheet Ratos s the least mportant crteron. Accordng to the TOPSIS method usng the fuzzy weghts of the crtera calculated by FAHP, all segments of Turksh bankng sector were adversely affected by 00 fnancal crss n whch almost half of Turksh banks were faled. After exhbtng rather bad fnancal performance n 00, a recovery perod started thanks to the restructurng measures taken n the frst stage of the fnancal collapse. Turksh bankng sector wth ts all segments has been somewhat affected by the current fnancal crss startng n 008. However, dfferent from 00 crss, the adverse effects of the current crss s not destructve. One may safely clam that the mmunty of Turksh bankng sector from the global fnancal crss s agan the result of the favorable restructurng reforms started n perod. Surrounded wth heavy regulatons and rules, the sector seems to acheve ts stablty durng the current fnancal crses. Especally, the lmtaton of exposure to the toxc assets by regulatons helps Turksh banks to save ther asset qualty. As the global fnancal crss turned out to be more serous than what was expected, Turksh government took more measures aganst t at the end of 00, amng mostly contractng the loan supply. As long as the poltcal stablty s preserved, the Turksh bankng sector seems to have capablty of overcomng the current fnancal crses. Ths study may be extended n several drectons by carryng out a number of consstency analyses. Frst of all, the performance of the Turksh bankng sector can also be assessed wth other MCDM models

25 Journal of Economc Cooperaton and Development 67 such as VIKOR and ELECTRE. Another topc of a consstency study may be analyzng the robustness of the fnancal performance estmatons by usng alternatve weght values attached to the performance crtera. The result of a gven MCDM method wll be more relable when ts decson outcome does not vary sgnfcantly dependng on the normalzaton procedure used. Thus, t would be of great nterest to repeat ths TOPSIS applcaton wth alternatve lnear normalzaton procedures n addton to the vector normalzaton procedure.

26 68 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach References Andres, A.M. 0. The Determnants of Bank Effcency and Productvty Growth n the Central and Eastern European Bankng Systems. Eastern European Economcs 49, no. 6: Aydn, N., A. Yalama, and M. Saym Bankng Effcency n. Developng Economy: Emprcal Evdence from Turkey. Journal of Money, Investment and Bankng 8: Atan, M., and G. Catalbas Bankacılıkta Etknlk Ve Sermaye Yapısının Bankaların Etknlğne Etks. İşletme ve Fnans Dergs 37: Badreldn, A.M Measurng the Performance of Islamc Banks by Adaptng Conventonal Ratos. Workng Paper No. 6, German Unversty n Caro. Cetn, M.K., and E.I. Çetn. 00. Mult-crtera Analyss of Banks Performances. Internatonal Journal of Economcs and Fnance Studes, no.: Chang, D.Y Applcatons of the Extent Analyss Method on Fuzzy AHP. European Journal of Operatonal Research 95: Chrstopoulos, A.G., J. Mylonaks, and P. Dktapands. 0. Could Lehman Brothers Collapse Be Antcpated? An Examnaton Usng CAMELS Ratng System. Internatonal Busness Research 4, no. : -9 Cng, S., and A. Tarm Turk Banka Sstemnde Performans Olcumu: DEA-Malmqust TFP Endex Uygulamas. No , Turkye Bankalar Brlg Arastrma Teblgler Sers. Demrel, E. 00. TOPSIS Multcrtera Decson Makng Method: An Examnaton on State Owned Commercıal Banks n Turkey. Journal of Entrepreneurshp and Development 5, no.: 0-. Deng, H Multcrtera analyss wth fuzzy par-wse comparson. Internatonal Journal of Approxmate Reasonng : 5 3. Denzer, C., M. Dnc, and M. Tarmclar Measurng Bankng Effcency n the Pre- and Post-Lberalzaton Envronment: Evdence from the

27 Journal of Economc Cooperaton and Development 69 Turksh Bankng System. Workng Paper No. 476, World Bank Polcy Research. Ertugrul, I., and N. Karakaşoğlu Performance Evaluaton of Turksh Cement Frms wth Fuzzy Analytc Herarchy Process and TOPSIS Methods. Expert Systems wth Applcatons 36, no.: Hwang, C.L., and K. Yoon. 98. Multple attrbutes decson makng methods and applcatons. Berln: Sprnger. Isk I., and D. Uysal The Sources of Productvty Growth n Turksh Banks by Ownershp: The Intal Responses of Publc, Prvate, and Foregn Banks to Lberalzaton. Turksh Studes 7, no.3: Jackson, P.M., M.D. Feth, and G. Inal Effcency and Productvty Growth n Turksh Commercal Bankng Sector: A Non-Parametrc Approach. Presented at the European Symposum on: Data Envelopment Analyss - Recent Developments and Applcatons, Werngerode, Germany, 6-8 October, 998. Lynch, R L and Cross, K F (99). Measure Up!, Yardstcks for Contnuous Improvement. Cambrdge: Blackwell Publshers. McGowan, C.B., R. Andrew, and Z. Sulong. 0. Fnancal Analyss Of Bank Al Blad. Internatonal Busness & Economcs Research Journal 0, no.3: 9-6. Mercan, M., and R. Yolalan Turk Bankaclk Sstemnde Olcek ve Mulkyet Yaplar le Fnansal Performansn Ilsks, No:000-0, Yap Kred Tartsma Teblgler Sers. Onal Y.B., and N.C. Sevmeser The Effects of Foregn Banks Entrance n Turksh Bankng System: The Effcency Analyss of Domestc and Foregn Banks. Journal of Çukurova Unversty Insttute of Socal Scences 5, no. : Saaty, T.L The analytc herarchy process. New York: McGraw-Hll. Secme, N.Y., A. Bayrakdaroglu, and C. Kahraman Fuzzy performance evaluaton n Turksh bankng sector usng Analytc Herarchy Process and TOPSIS, Experts Systems wth Applcatons 36, no.9:

28 70 Evaluatng the Fnancal Performance of Turksh Bankng Sector: A Fuzzy MCDM Approach Stavárek, D Bankng Effcency n the Context of European Integraton. Eastern European Economcs 44, no. 4: 5-3. Wang, Y.M., and T.M.S. Elhag Fuzzy TOPSIS Method Based on Alpha Level Sets wth an Applcaton to Brdge Rsk Assessment. Expert Systems wth Applcatons 3, no.: Wang, Y.J.008. Applyng FMCDM to Evaluate Fnancal Performance of Domestc Arlnes n Tawan. Expert Systems wth Applcatons 34, no.3: Wu, H.Y., G.H. Tzeng, and Y.H. Chen A fuzzy MCDM Approach for Evaluatng Bankng Performance Based on Balanced Scorecard. Expert Systems wth Applcatons 36, no.6: Yldrm, C Lberalsaton Before Stablsaton: Polcy and Performance n Turksh Bankng. Unpublshed PhD Thess. Unversty of Lancaster. Yolalan, R Turk Bankaclk Sektoru cn Gorel Mal Performans Olcumu. Turkye Bankalar Brlg Bankaclar Dergs 9: Zam, O The Effect of Fnancal Lberalzaton on the Effcency of Turksh Commercal Banks Appled Fnancal Economcs 5: Zhu, K., Y. Jng, and D. Chang A Dscusson on Extent Analyss Method and Applcatons of Fuzzy AHP. European Journal of Operatonal Research 6: