Transcription

1 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: IJBF THE IMPACT OF THE GLOBAL FINANCIAL CRISIS ON AUSTRALIAN BANKING EFFICIENCY Mlnd Sathye Unversy of Canberra Mohamed Arff Sunway Unversy, Malaysa Vvera Unversy of Indonesa, Jakarta. Abstract The theory of fnancal stably postulates that fnancal nstutons n a country experencng fnancal crss would wness productvy losses. Ths study examned whether they experence productvy losses when there s no crss, and whether the fnancal sector s not mmune from global economc events. The Australan fnancal nstuton effcency and productvy durng were examned, that s, after the fnancal system reforms but the test perod ncludes the fnancal crss years. Effcency scores were computed usng Stochastc Fronter Analyss and total factor productvy usng Malmqust ndces. Australan nstutons were found to have experenced productvy declne durng the global fnancal crss. The evdence s just the oppose of the common belef that Australan nstutons remaned nsulated from the crss. Global economc slowdown can also lead to productvy losses n a country not experencng severe fnancal crss because of the reforms taken long before the crss to mprove prudental oversght of the fnancal nstutons n Australa. Keywords: Total factor productvy, Cost effcency, Prof effcency, Global Fnancal Crss, Fnancal frms sze JEL Classfcaton: G2, O3, G11 1. Introducton Theory of fnancal stably suggests that neffcences or nstably n the system are assocated wh welfare costs (the devaton from the nexus between

2 2 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 savngs-nvestments). However, there have been few (f any) studes evaluatng precsely the drect welfare costs assocated wh fnancal nstably and fnancal neffcency (Haldane, Hoggarth, & Saporta 2001). The welfare costs often take the form of output losses. Hoggarth, Res, and Saporta (2001) n ther study of 43 bankng crses found that average output losses were % of annual GDP. What happens, however, when a country s not experencng a severe bankng crss self but s stll not mmune from the fnancal crss that pervaded the global economes? Do fnancal nstutons n such an economy experence output losses or productvy declne? Gong by the theory, as there would be no fnancal crss n Australa from the Global Fnancal Crss, the fnancal nstutons should not experence any output losses or productvy declne, or f at all they do, then these would be only margnal. Fnancal stably theory does not provde gudance, whether there are output losses when a country s not gong through fnancal crss but cannot avod the mpact of a fnancal crss from around the world. Furthermore, theory does not provde a clue what s the magnude of mpact of varous macro-economc varables on productvy loss. Ths study addressed ths gap n the lerature for whch Australa provdes an excellent example. The country dd not experence a notceable bankng crss (of the type wnessed n the US, UK, Japan, and other European countres) though was affected by the global fnancal crss. The productvy and effcency of Australan fnancal nstutons were examned n the pre and post crss years. The study however, went a step further and extended the analyss to nclude fnancal frms other than banks as well. Aganst the above background, our study of natonal and nternatonal mportance had the followng objectves: (a) to measure the effcency and productvy of Australan fnancal frms n a 10-year perod ncludng the years of fnancal crss, (b) to measure the mpact of macro-economc varables on productvy, and (c) also to examne how fnancal crss affects productvy of the fnancal frms, ncludng banks, nsurance companes, nvestment, and fund management frms. The rest of the paper s organsed as follows. Secton 2 provdes a bref revew of the relevant lerature, that s, fnancal stably theory and the theory of effcency and productvy on whch the current research draws upon. Secton 3 contans a descrpton of the data and methodology. In secton 4, the fndngs of the study are provded, whle Secton 5 concludes the paper. 2. What Does the Lerature Suggest? As already stated, the current work drew from two related theores, that s, the theory of fnancal stably, and the lerature on economc effcency and productvy. In hs semnal work on the theory of fnancal stably, Crockett (1997) stated that monetary and fnancal stably are of central mportance to the effectve functonng of a market economy. They provde the bass for ratonal

3 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: decson makng about the allocaton of real resources through tme and therefore mprove the clmate for savng and nvestment. Schnas (2005), however, stated that no framework exsts now eher for measurng the effcency losses assocated wh market mperfectons n fnance or for assessng the rsks to fnancal stably assocated wh market mperfectons. Goodhart, Sunrand, and Tsomocos (2006) preseneted a model that hghlghts the trade-off between fnancal stably and economc effcency. The model does take nto account the ndvdual behavour of banks, but analyses the trade-off from regulatory and monetary polcy perspectves. Emprcal evdence that hghlghts the trade-off between bank behavour as reflected n the nstutonal economc effcency and fnancal stably was not found n the lerature, save a recent study on Japanese banks. The Japanese bankng sector study by Jones and Tsutsum (2009) examned the fnancal stably and bankng effcency n Japan. These authors found that Japanese bankng sector could whstand the global crss but dd result n sharp contracton n output. These authors recommended mprovng the effcency of bankng sector for greater stably. It s mportant to note that the Japanese government dd nject capal nto banks, whch helped avod the potental crss n the bankng sector n that country. On the background of ths, Australa presents a unque example. There was no fscal ad provded to banks except for a government depos guarantee and wholesale fundng guarantee to ensure smooth flow of cred when the Global crss h Australan fnancal nstutons n September The Australan bankng sector dd not experence dffcultes as n the US, UK, or rest of Europe durng the global fnancal crss. Consequently, a study of bankng sector effcency n Australa pre and post crss perod could add a unque perspectve to extend the lerature on fnancal stably. It s to be noted that central bankers recognse that effcency n the bankng sector s a key contrbutor to macroeconomc stably. It s also a precondon for economc growth and s mportant for the effectveness of monetary polcy (ADBG, 2009, p.1) Effcency and productvy of banks have been a subject matter of study n many countres. Productvy relates the quanty of output produced to one or more nputs used n s producton, rrespectve of the effcency of ther use (OECD, 2004). To compute effcency, two popular fronter analyss methods used are the parametrc stochastc fronter analyss (SFA) and the nonparametrc data envelopment analyss (DEA). Malmqust ndex s the method commonly used n the lerature to compute productvy. Caves, Chstensen and Dewest (1982) appled the Malmqust ndex decomposon for the frst tme n productvy analyss. Ths method defnes an ndex as a rato of two fronter dstance functons, whch represent multple nputs and multple outputs technology whout a need to specfy a frm s behavoural objectve, such as prof maxmsaton or cost mnmsaton. Over the years, many studes examned the effcency of banks, but these were manly confned to the US and Europe. Berger and Humprey (1997) surveyed 130 studes coverng 21 countres and

4 4 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 found that non-parametrc measures of effcency yeld a mean effcency statstc slghtly lower wh a large dsperson: parametrc measures provde a slghtly hgher effcency wh lower dsperson statstcs. For fnancal frms, the averages reported were 0.72 and 0.84 respectvely. Australan studes on fnancal nstuton effcency have been few. Avkran (1999) found evdence of effcency gans post merger n Australan bankng sector. Sathye (2001) studed the x-effcency of Australan banks n 1996 usng data envelopment analyss and found to be low. Strum and Wllams (2004) studed the mpact of foregn bank entry and deregulaton on bank effcency, and found that foregn banks were more effcent than ther domestc counterparts. Neal (2004) found that productvy n Australan banks total factor productvy n the bankng sector was found to have ncreased by an average annual 7.6% between 1995 and Ths author found that effcency does mpact stock returns. The authors of ths present study dd not come across a study that examned the mpact of the fnancal crss on Australan fnancal sector effcency and productvy. Smlarly the mpact of contextual and envronmental varables on bankng effcency s yet to be explored n the Australan context, though a few studes overseas explored the ssue n other countres. Some recent studes ncluded Ladak and Gagans (2009), that examned the relatonshp whch exsts between prof effcency and stock prces of banks n 15 countres over a fve year perod. Arss (2010) reported sgnfcant negatve assocaton betwen market power and cost effcency of 81 banks from 60 developng countres across Afrca, East and South Asa, the Pacfc, and Eastern Europe. Arff and Luc (2009) studed the the pre and postmerger effcency of IMF s restructurng of banks n four East Asan countres. Thangavelu and Fndlay (2010) studed the determnants of bank effcency n South Asan economes. Feng and Serlets (2010) found a decrease n productvy of Amerca s large banks over Berger and De Young (1997), Beard Caudll, and Gropper (1997), and Hunter and Tmme (1995) analysed cost effcency n order for the management and polcy maker to dentfy over or under use of nputs to produce gven output(s). Humphrey and Pulley (1997) found that changes n bankng envronment lead to extra changes n prof effcency. Recently studes ncorporated the lkely mpact of envronmental varables on bank s effcency performance. These may be dvded nto two types of research: nternal usng bank- specfc and external usng macroeconomc factors to dentfy ther relatonshp to effcency. Common macro-economc factors that may have mpact on effcency performance are gross domestc product (GDP), nflaton rate, and nter-bank rate or cash rate. Thoraneenyan and Avkran (2009) examned the mpact of restructurng and country-specfc factors on East-Asan bank effcency. They found that macro-economc condons have more effect on bank s effcency compared to restructurng polcy. In addon, the second study of ths knd to-date by Ramlall (2009) found that bank-specfc varables such as capal and cred rsk affect bank profably sgnfcantly n

5 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: Tawan. Drake, Maxmlan, and Smper, (2006) found n the context of Hong Kong, that bank techncal effcency s sgnfcantly affected by bank sze. Followng from the above, the current study was motvated to seek answer the followng questons. (a) What s the mpact of the Global fnancal crss on the productvy and effcency (usng both cost and prof effcency measures) of Australan fnancal nstutons? (b) What s the magnude and mpact of bank specfc and macro-economc factors on Australan bankng effcency? (c) How do Australan banks compare wh other fnancal frms n terms of effcency and productvy? 3. Data and Method Ths study combned a tme seres and cross-sectonal data of 44 publcly-lsted fnancal frms. Of the 44 frms n the fnal sample, there were nne banks (all banks), nne nsurance frms, and 26 other fnancal frms. The complete data on all the varables n the model were not avalable for all frms for all years. Accordngly, the researchers could obtan and used 10 years of data for banks ( ), fve years for nsurance ( ), and four years for other frms ( ). The data after 2009 were left out snce the crss was offcally over n November 2008, and the test perod was stopped at The data were collected from varous sources: databases such as DatAnalyss, FnAnalyss, Dunn and Bradstreet, DataStream and fnancal nformaton avalable at frm s webse. 1 The study used the followng procedure to answer the queston whether there was a sgnfcant dfference n the mean effcency and mean productvy of fnancal nstutons durng the study perod, that s, durng crss and noncrss perods. To compute effcency and productvy scores, followng the steps used n pror studes and the commonly adopted ntermedaton approach (Sealey & Lndley, 1977), ths study used the value of net loans (y 1 ), and other earnng assets (y 2 ) as output varables whle nput varables were purchased funds (x 1 ), labour cost (x 2 ), and capal nput (x 3 ). To estmate the cost and prof effcences, ths study used addonal use of the prce data of nputs and outputs. Malmqust ndces were computed to measure total factor productvy (TFP), economc change (EC), techncal effcency (TE), overall effcency (OE), and allocatve effcency (AE). Parametrc SFA method was used to compute the cost effcency (CE) and prof effcency (PE) measures. To check the statstcal sgnfcance, ths study appled the Mann-Whney and Kruskal Walls statstcal test. The panel regresson was performed usng suable econometrc software (E-vews) wh accountng for correctons for seral correlaton and heteroscedastcy. 1 Ths project arose from one of the authors beng granted the Australan Prme Mnster s Fellowshp to study ths ssue through an nternatonal study award. The project was completed n 2009, so the data collecton ended n 2009 wh the fundng for that project. Hence, the study could not be extended beyond the test perod stated n ths paper.

6 6 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: Malmqust Productvy and Scale Economes Followng Fare and Grosskopf (1994), the Malmqust productvy change ndex can be wrten as: 1/2 d t1 0 ( y 1, x ) 1 d t 0 ( y 1, x 1 ) d t 0 ( y, x ) t t m 0 ( y, x, y 1, x 1 ) t t t t t t t t t d (, ) t 1 t 0 y x d t t (, ) d 0 ( y, x ) 0 y t 1 x t1 t t (2) where y and x are outputs and nputs across tme t to t+1 to ndcate Malmqust ndces whch are computed relatve to the prevous perod. The techncal effcency change measures the change n effcency between perod t and t+1, whle techncal change captures the shft n the technology appled over tme as banks adopt newer methods of producton. A value greater than one n both cases ndcates growth n productvy, that s, posve factor values suggest that banks are functonng wh gans n productvy. 3.2 Cost Effcency Cost effcency descrbes how the cost of a frm compares wh the best practce n the fted fronter, to produce the same output under the same envronmental condons (Berger & Mester, (1997). In ths, cost effcency s measured usng SFA, whch allows for ncorporaton of both allocatve and techncal effcences from nputs. It measures the change n a frm s varable cost adjusted for random error relatve to the estmated cost needed to produce ouput(s) as effcently as the best practce frm n the sample. Ths study used the translog stochastc functon by Battese and Coell (1995) to estmate frms relatve effcency. The general form of cost effcency can be wrten as: C q, p, ) ( U V ) = 1,2,...,N; t=1,2,...,t ( Where C represents the costs (expenses) of the -th frm n the t-th perod, q s vector prces of varable nputs, p s vector of quanty of varable outputs, V s random error, whle U s neffcency, and β s unknown vector parameter. The specfc form of cost functon used n ths study can be wrten as follows: (3) LnTC 0 ln Z ln W j 1 j ln Z Q j ln Q 1 2 ln Z j j ln W ln W W jt j ( V jt U ) 1 2 j ln Q Q j jt ln W ln Q j j jt (4)

7 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: where, TC s defned as the total costs, W s the vector of nput prces, Q s a vector of varable outputs, and Z s a vector of fxed netputs. Ths model s estmated usng maxmum lkelhood estmaton. 3.3 Prof Effcency Model Accordng to Humphrey and Pulley (1997), prof effcency refers to the closeness of the frm to generate maxmum possble outputs gven a partcular level of nput and output prces. It s the rato of predcted maxmum prof, whch could be earned f a frm was as effcent as the best practce frms after adjustng for random error. The value s bounded between 0 and 1. The hgher the measured value, the more prof effcent the banks are. A score of 1 means the frm has acheved optmal prof effcency. The general form of prof functon can be wrten as: P f ( q, p ; ) ( V U ) 1,2,..., N; t 1,2,..., T Where P s the prof before tax of the -th bank n the t-th perod, q represents the vector of output quantes of the -th bank n the t-th perod, p s vector of quantes output varable of the -th frm n the t-th perod, and β s unknown vector parameter. To estmate prof effcency, ths study used the translog fronter prof functon developed by Altunbas and Chakravarty (2001), as n the followng: Ln 0 ln Z ln W ln Q j ln WW jt 2 j 2 ln Z Q ln Z ln W ( V U ) where p s defned as pre-tax prof; to avod negatve value for prof, the mnmum absolute value of prof plus one to the prof values was added; W s the vector of nput prces; Q s a vector of varable outputs; and Z s a vector of fxed netputs. Ths model s estmated usng maxmum lkelhood estmaton. 3.4 Panel Regresson j j j j jt The effcency measures (TFP, AE, CE, and PE) were modelled as beng determned/assocated wh two sets of varables. To dentfy the lkely relatons, Panel Data Regresson method was used, whch corrects the mpact of changng varances n the tme-seres and cross-secton data, thus yeldng robust results. Hence, two sets of regressons (one for macroeconomc and another for bankspecfc factors) were run usng the panel data avalable for N number of banks over t to T perods of annual observatons. The dependent varables are the four measures of effcency from fronter methods as n the equatons above. Followng from pror studes, the ndependent varables were three macroeconomc factors (GDP growth, nflaton, and nterest j ln Q Q j jt ln W ln Q j j jt (5) (6)

8 8 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 rates) n the frst four regressons. Sx bank-specfc varables were regressed wh effcency as the dependent varables. The general model s: DV jt 0 N 1 ( IV ) j jt jt (7) 2 where, DV j : eher TFP j : AE j : CE j : PE j, as defned earler. Thus, panel data regressons were run respectvely ncludng (a) three for macroeconomc, and (b) sx for bank-specfc varables. : : m 0 measured ntercept and coeffcents of respectve ndependent varables n each regressons; IV j : the ndependent varables, macroeconomc and bank-specfc varables n respectve regressons; and ej: the error term n each regressons. As s the procedure n panel data regresson, random effect or fxed effect model was teasted for. Test statstcs suggested that the fxed effect model was approprate for the data set n ths study. The test results were then obtaned from eght regressons. The ndependent varable are measured as below: CAR=Capal Adequacy Rato, NPL=rato of non-performng loans to total loans, cost rato=cost to ncome rato, NIM=net nterest margn percentage, Core earnng assets=rato of core earnng assets to total assets, and sze=total assets. Descrptve statstcs and varables 4. Fndngs and Dscusson Tables 1 and 2 present a summary of descrptve statstcs and the theorypredcted relatonshp of measures wh factors. Table 2 s a summary of expected sgns of theoretcal relatonshp between factors and productvy measures. Economc growth s favourable to banks, and so productvy s ganed durng economc growth. Productvy declnes as growth declnes. Hence a posve sgn for ths factor was predcted. Increases 2 The symbols are defned as: j ndcates the banks; I ndcates varables, and t ndcates tme from t=2000 to 2009.

9 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: n nflaton and nterest rates retard real sector growth, so ther effects are negatve on bank effcency. There are sx bank-specfc varables n the panel regresson (we could also nclude them as factors n the SFA, but we dd not do so because we wanted to use the panel regresson because of s known robustness). Increases n factors such as NPL (non-performng loans ratos) reduce bankng effcency, so the sgns were expected to be negatve. The other factors were lkely to ncrease effcency except n the case of capal adequacy rato (CAR) where takes a posve or negatve value. Table 1: Summary Statstcs of the Input and Output Varables Adjusted for Inflaton ($ mllon) Varables Descrpton Mean Standard Devaton Outputs: LOAN Total loans-loan loss reserve OEA Placement wh other banks, secures, and nvestments Inputs: Purchased funds Deposs, money market fundng, and others Labcost Salary and wages Capal Net value of fxed assets Dependent varables Y1 Total Costs Y1 Pre-tax Prof Input Prces w1 Prce of Purchased Funds w2 Prce of Labour w3 Prce of Physcal Capal Note: All values are n $ mllons, except for nput prces. Input prces were derved usng the procedure n Altunbas and Chakravarty (2001).

10 10 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 Table 2: Predcted Relatonshp between Effcency and Factors Determnant Varables Dependent Varables Productvy Allocatve Cost Prof Growth Effcency Effcency Effcency Macro-economc varables GDP Inflaton rate Cash rate Bank - specfc varables CAR ± ± ± ± NPL Cost rato NIM Core Earnng Assets SIZE CAR (cumulatve abnormal returns) can affect effcency n eher way. Hence ths study ndcated the effect as not predetermned. For example, reducng CAR may act as a sgn of rsk, so effcency may declne (-). It can also be argued that banks may mnmse CAR n order to decrease capal nput, so the profs are rased (+). Ths study determned the actual effect by notng the sgns obtaned. 4.1 Total factor productvy of banks and other fnancal frms Table 3 provdes some new evdence on the performance of the fnancal frms studed n ths case. The mean values of the TFP, EC, and TE are summarsed n Table 3 for the 10-year perod. Results showed that the mean TFP was 1.072, that s, the average productvy gan over the study perod was 7.2 ndex ponts ( ) suggestng a gan equal to about 7.2% per year. Ths suggests that the banks are hghly productve, experencng ncreasng returns to scale n each year of the test perod wh very good ncrease of effcency: exceptons are durng 9/11 and global crses years. As can be seen from above table, for the banks, the total factor productvy ncrease of 7.2% s manly drven by a 6.8% shft n the fted fronter, that s, from adopton of newer ways of dong busness, such as technologcal change. There was a slght ncrease n catchng-up to the fronter due to manageral effcency. Ths suggests that the fronter self had shfted (gven effcent use nputs) by an annual average of 6.8%, whle annual catchng-up to the fronter contrbuted just 0.4%. Thus, management ncreased output effcency whle technology/newer methods contrbuted a bulk of the ncreases. Panel B shows the estmates for nsurance companes. These numbers ndcated a decrease of

11 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: Table 3: Malmqust Productvy Indces Panel A: Banks: Annual means of total factor productvy change and s components Year Malmqust dex of Effcency change Technologcal total factor (catch-up) change productvy change (fronter shft) mean Panel B: Insurers: Annual means of total factor productvy change and s components mean Panel C: Other fnancal frms: Annual means of total factor productvy change and s components mean average productvy of 6.1%. In contrast to the case of banks, the TFP decrease was prmarly due to negatve effcency change (6.3%). However, the average technologcal change of these frms ncreased by 0.2%. Ths ndcates that the TFP declne as beng entrely due to negatve techncal change, although there are small mprovements n catchng-up to the average fronter. The effcency estmaton for other fnancal frms ndcated an ncrease of total factor productvy by 7.7% as n Panel C. Smlar to the cases of nsurance frms, the TFP growth was due most lkely to shfts n the fronter (18%) rather

12 12 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 than mprovement effcency relatve to the sample frms fronter. These results were smlar to pror studes. Importantly, despe the challenges of the 9/11 n 2001 and the Global Fnancal Crss n , there were overall gans n productvy of Australan banks. 4.2 Global Fnancal Crss and Effcency performance On testng the mean dfferences across the three groups (banks, nsurance companes, and other frms), nterestng results were revealed. The effcency declned substantally by 3.6% (banks) and 21.8% (others). Although the fnancal sector dd not collapse as was the case n Europe and the US, the results dentfed a large drop n producton effcency n the Australan fnancal sector. Tabel 4.1 and 4.2 respectvely present a summary of test results of mean effcency and productvy dfferences of fnancal frms durng the crss and the pre-crss years. An examnaton of the results showed that there were no sgnfcant dfferences of mean productvy growth of banks and nsurance frms between the pre-crss and crss perod, although there was a substantal declne n productvy n the case of other frms. The dfference was statstcally sgnfcant for other fnancal frms at a 0.05 probably level. In addon, the results showed that there were no sgnfcant dfferences of mean effcency of banks and other fnancal frms between the pre-crss and crss perod, whle there was also a substantal declne n effcency of nsurance frms at 0.05 probably level. The declne n bank effcency was 1.7% durng the crss perod (n ), for nsurance 35.4% for others (nvestment and fund management frms) 7.4%. These fndngs have sgnfcant mplcatons for the theory of fnancal stably. In the context of Australa where there was no bankng crss, the macro-economc suaton faced dffcultes due to the mpact of global fnancal crss engulfng the rest of the world. The banks and nsurance frms dd not experence statstcally sgnfcant productvy loss but other fnancal frms dd. The lesson for theory of fnancal stably s healthy bankng and nsurance sectors can whstand nfluence of macro-economc crss, however, the weakest nstutons are affected frst whch n the Australan context were found to be other fnancal frms (other than banks and nsurance companes). It s often sad that n a herd under attack from predators s the weakest n the herd that falls prey frst. Table 4: Test of Mean Effcency Dfference: Crss versus Pre-Crss Perods Descrpton Mean Standard Devaton Test for Sgn Mann-Whney U Test Panel A: Banks Crss Perod ( ) (contnued)

13 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: Descrpton Mean Standard Devaton Test for Sgn Mann-Whney U Test Pre-Crss Perods ( ) (0.398) Panel B: Insurers Crss Perod ( ) Pre-Crss Perods ( ) (0.047)* Panel C:Other Fnancal Frms Crss Perod ( ) Pre-Crss Perods (2006) (0.419) *Statstcally sgnfcant at 0.05 sgnfcance level. Insurance frms had sgnfcant dfference. Table 5: Test of Mean Productvy Dfference: Crss versus Pre-Crss Perods Descrpton Mean Standard Devaton Test for Sgn Mann-Whney U Test Panel A: Banks Crss perod ( ) 780 Pre-crss perod (0.388) ( ) Panel B: Insurers Crss perod ( ) 78 Pre-crss perod (0.450) ( ) Panel C: Other Fnancal Frms Crss perod ( ) 494 Pre-crss perod (0.027)* (2006) *Statstcally sgnfcant at 0.05 sgnfcance level. Other frms had sgnfcant dfference.

14 14 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 Evdence reported n Table 5 suggests that both banks and nsurance frms were not as sgnfcantly affected by the crss as the other fnancal frms (nvestment and fund management frms). Ths makes sense as well. The other frms were capal market frms, whch experenced the worst mpact from fnancal crss occurrng across the world. The other frms n our study were nvestment and fund management frms, whch generate revenues from nvestng n secures that were severely affected by the crss wh stock and bond market prce declnes. The other fnancal frms such as penson funds, stockbrokers, and nvestment frms all had experanced sgnfcant mpact losng by a factor of two-thrds of the average effcency durng non-crss perod. 4.3 Frms s Sze and Effcency Performance Ths study examned the mpact of frm sze on effcency. The null hypothess was: the mean effcency growth of large frms was greater than that of nonlarge (small frms). The test used was the Mann-Whney U-test. Test statstcs and sgnfcance values are ncluded n Table 6. The table shows that there was no sgnfcant dfference n the effcency of small and large banks. Ths fndng s dfferent from the fndngs by Kwan (2006) for Hong Kong where small banks were found to be more effcent than large banks. Walker (1998), usng Australan data, found that smaller banks were more effcent than the larger ones. Both studes dd not provde statstcal tests, so s dffcult to say f the dfferences were sgnfcant. Our test result suggested that, small banks tend to be more cautous n playng ther role as ntermedary nstutons compared to large banks. Perhaps ths s symptomatc of the moral hazard problem that has been hghlghted n the debate across the world on the Global Fnancal Crss about the too-bg-to-fal argument. The banks n our tests account for about 97% of the total assets of the bankng sector. These results ponted to the need to contnue wh the four pllar polcy that bans merger n bg four banks. Table 6: Summary Test of Mean Dfference: Bg versus Non-bg Frms Frms Mean Standard Test for Sgn Devaton Mann-Whney U Test Panel A: Banks Bg Non-bg (0.387) Panel B: Insurers Bg Non-bg (0.431) (contnued)

15 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: Frms Mean Standard Devaton Test for Sgn Mann-Whney U Test Panel C: Other Fnancal Frms Bg Non-bg (0.395) Note: These test statstcs show that the dfferences are not sgnfcant. 4.4 Macroeconomc and Bank-specfc Determnants Do non-parametrc effcency measures have dentfable relatonshps wh macroeconomc and frm-specfc factors? The macroeconomc factor effects are summarsed n Panel A whle the bank-specfc factor effects are presented n Panel B (Table 7). The dependent varables were the values of total factor productvy, and allocatve effcency, whch are regressed aganst three macroeconomc factors, GDP growth, nflaton, and nterest rates. The coeffcents were obtaned after correctons for seral correlaton and heteroscedastcy run as panel regressons, so the coeffcents are robust. The results are presented n Table 7. Table 7: Bank Effcency Measures and ther Macro- and Mcro-Determnants Explanatory Varables Total Factor Productvy (TFP) Panel A: Macro Factors GDP (2.162)* (0.339) INFL (-8.879)*** (0.619) INTR (-1.028) (2.040)** Durbn-Watson stat R-squared F-statstc (Probaly) *** 3.381*** Allocatve Effcency (AE) (contnued)

16 16 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 Explanatory Varables Total Factor Productvy (TFP) Allocatve Effcency (AE) Panel B: Bank-Specfc Factors CAR (-0.110) (2.439)** NPL (0.970) (0.054) CR (2.616)*** (-0.055) NIM (-4.794)*** (-1.790)* CEA (-7.894)*** (-8.777)*** SIZE (-1.099) (1.967)* Durbn-Watson stat R-squared F-statstc (Probaly) *** 8.424*** Note: *** ndcates sgnfcance at.01; ** at 0.05; and * Regresson results summarsed n Panel A of Table 7 show F-statstcs of (sgnfcant at lower than probably) and adjusted R-square of 33.98%. Ths ndcated that macroeconomc factors have a sgnfcant mpact on effcency. The GDP growth has posve effect on effcency wh a coeffcent of (sgnfcant at 0.05 levels). The other two were also sgnfcant wh a negatve mpact on effcency snce hgher nflaton and hgher nterest rates reduce effcency gven ther negatve mpact on frm s operatons. The nput varables purchased funds and labour costs would be drectly affected by nflaton and nterest rates. Meanwhle n Panel B, the F-value and R-squares are respectvely and 50.20% for TFP: the correspondng numbers for AE are and 54.10%. Of the bank-specfc varables, sgnfcant assocaton was observed for the ndependent varables: CR (cost rato), NIM (net nterest margn), and CEA (core earnng assets). Gven that the banks dd not have non-performng loans durng the test perod, the varable NPL dd not have any effect. However, the capal adequacy rato and sze were not found to have a sgnfcant assocaton, although these had negatve sgns consstent wh predctons. Hgher levels of CAR and sze were found to have negatve assocaton to TFP. In the case of allocatve effcency, the fndngs were smlar except for the fndngs relatng to NPL, Sze, and CR. Larger banks have easy access to

17 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: capal, and so s mpact s sgnfcantly posve, although judged by the sze of the coeffcent, ths mpact s but margnal. Strangely NPL had a posve mpact, cost rato was found to have a sgnfcant mpact. Overall, as n prevous studes, effcency gans were reported for banks although there were declnng treads n bank effcency durng the crss perod. Compared to banks and nsurance frms, the other fnancal frms had sgnfcant loss of effcency the loss n effcency was two-thrds durng the crss years. Unlke prevous reports, sze appears not to provde a dsadvantage for effcency n ths test perod compared to prevous reports for other countres, for example, Hong Kong banks. Fnally, banks domnated gans n effcency over nsurance and other fnancal frms. In the lerature, how macro-economc factors affected Australan bankng effcency s an unanswered queston. Ths study found that macroeconomc varables sgnfcantly mpact bankng effcency, as well as strong correlatons of four bank-specfc factors wh these non-parametrc effcency measures. Together, these results add mportant nsghts about the effcency of fnancal frms n the last decade, the mpact of fnancal crss on effcency and the mpact of macro-economc and frm specfc varables on effcency. 4.5 Parametrc Effcency and s Determnants In ths sub-secton, the results are presented on cost and prof effcency as these relate to determnants specfed n the test model where ths study compared ths wh a pror study by Worthngton (2000). Cost and prof effcency Mean values of cost and prof effcency estmates are presented n Table 8. It shows that mean bank effcency was (70.1%), whle prof effcency was (31%). Bankng effcency was hgh as compared to mean effcency reported for Australan cred unons scores range between 63 and 67% (Worthngton, 2000). Table 8: Cost and Prof Effcency of Banks Mnmum Maxmum Mean S.D Banks Cost Effcency Prof Effcency Macroeconomc and Bank-specfc Determnants Table 9 provdes a summary of test results. As descrbed n the prevous subsecton, these statstcs were obtaned from two panel regressons wh correctons

18 18 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 for seral correlatons and heteroscedastcy, as was deemed approprate. The dependent varables are two parametrc effcency measures namely CE or cost effcency, and PE or prof effcency. The macro and the bank-specfc varables are lsted n the table. The overall f of models are good as can be judged by the explaned varaton and the sgnfcant R-squares. For CE, these numbers were: 39.3% and 6.247, sgnfcant at levels. For PE, the numbers were 54.40% and , agan sgnfcant. The three macroeconomc factors affect these effcency measures as predcted by theores. GDP growth affects effcency posvely: the coeffcent on cost effcency has 0.59 and sgnfcant whle the for PE ndcated a strong sgnfcant effect. Table 9: Bankng Effcency Measures and ther Macro-and Mcro- Determnants Explanatory Varables Cost Effcency Prof Effcency Panel A: Macro Factors GDP (6.816)*** (6.894)*** INFL (-0.043) (-0.058) INTR (-2.491)** (5.300)*** Durbn-Watson stat R-squared F-statstc (Probaly) 6.247*** *** Panel A: Frm-Specfc Factors CAR (1.981)* (5.764)*** NPL (1.150) (2.216)** CR (3.187)*** (2.101)** NIM ( )*** (-4.791)*** CEA ( )*** (-6.929)*** SIZE (contnued)

19 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: Explanatory Varables Cost Effcency Prof Effcency ( ) (-0.703) Durbn-Watson stat R-squared F-statstc (Probaly) *** *** Note: *** ndcates sgnfcance at.01; ** at 0.05; and * Whle nflaton does not have a sgnfcant mpact (although the sgns are as predcted), nterest rate ncreases have negatve and sgnfcant effect on effcency. The coeffcents were and for CE, and and for PE. As for bank-specfc factors, all factors except sze were sgnfcantly correlated as predcted by theory. In the case of CE, the NPL effect was not sgnfcant. All other coeffcents were strongly correlated wh cost and prof effcency measures. 5. Concluson The objectve of ths study was to measure the effcency and productvy of Australan banks and non-bank fnancal frms over a 11 year perod, In so dong, ths study also examned the mpact of fnancal crss on effcency and productvy to extend the theory of fnancal stably to a unque economy that dd not experence bankng crss and yet was not mmune from global fnancal crss effects on some parts of the fnancal nstutons. It s a specal case for whch theory dd not provde an answer. It was found that whle banks and nsurance frms dd not record a sgnfcant declne n productvy and effcency as contemplated by the theory, other fnancal frms wnessed a sharp and sgnfcant declne n effcency. The study also demonstrated that several frm-specfc and macro-economc varables do mpact cost and prof effcency an ssue that escaped the attenton of researchers so far. As a summary of fndngs of ths study, s noted that Australan fnancal frms effcency s on trend, and dd not suffer a declne over the recent tested years, but, the global crss had a knock-on effect on the trend gans for banks, nsurance frms and other fnancal frms n the last two years of the data seres. So, the common belef that Australan fnancal ntermedares are nsulated from the global fnancal crss s not borne out by emprcal evdence n ths study. Though banks were found to have hgher effcency than other fnancal frms, sze dd not provde any ncremental effcency gans to large Australan banks. Fnally, ths study dentfed three mportant macro-economc and four key frmspecfc factors as the key sources determnng producton effcency. Further cross country research on smlar lnes may help refne the theory of fnancal stably for other countres affected by the global fnancal crss.

20 20 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 References Afrcan Development Bank Group (ADBG), (2009). Confrontng the global fnancal crss: Bank effcency n Afrca. Retreved: org/en/news-events/artcle/confrontng-the-global-fnancal-crss-bankeffcency-n-afrca-5340/ Altunbas,Y., & Chakravarty, S. P. (2001). Fronter cost functons and bank effcency. Economc Letters, 72(5), Arff, M., Luc, (2009). IMF bank-restructurng effcency outcomes: Evdence from East Asa. Journal Fnancal Servces Research, 35, Arss, R. (2010). On the mplcatons of market power n bankng: Evdence from developng countres. Journal of Bankng and Fnance, 34, Avkran, N. (1999). The evdence of effcency gans: The role of mergers and the benefs to the publc. Journal of Bankng and Fnance, 23, Battese, G. and T. Coell. (1995). A model for techncal neffcency effects n a stochastc fronter producton functon for panel data. Emprcal Economcs, 20, Berger, A. and D. Humprey. (1997). Effcency of fnancal nstutons: Internatonal survey and drecton for further research. Journal of Bankng and Fnance, 98, Berger, A. and L. Mester. (1997). Insde the black box: What explans dfferences n effcences of fnancal nstutons? Journal of Bankng and Fnance, 21, Beards, T., Caudll, S., and D. Gropper. (1997). The dffusson of producton processess n the U.S bankng ndustry: A fne mxture approach. Journal of Bankng and Fnance, 21, Berger, A. and R. DeYoung. (1997). Problem loans and cost effcency n commercal banks. Journal of Bankng and Fnance, 21, Caves, D., Chrstensen, L., and W. Dewert The economc theory of ndex numbers and the measurement of nput, output and productvy. Econometrca, 50, Crockett, A., (1997). Why s fnancal stably a goal of publc polcy? In: Mantanng Fnancal Stably n a Global Economy. Proceedng of a symposum of the federal bank of Kansas Cy, Jackson Hole, Wyomng, August: Drake, L., Maxmlan J., and R. Smper. (2006). The mpact of macroeconomc and regulatory factors on banks effcency: A non-parametrc analyss of Hong Kong s bankng system. Journal of Bankng & Fnance, 30, Fare, R. and S. Grosskopf. (1994). Measurng Productvy: A Comment. Internatonal Journal of Operaton & Producton Management, 14(9): Feng, G. and A. Serlets. (2010). Effcency, techncal change, and returns to scale n large U.S banks: Panel data evdence from output dstance functon

21 The Impact of the Global Fnancal Crss on Australan Bankng Effcency: satsfyng theoretcal regulary. Journal of Bankng and Fnance, 34, Goodhart, C., Sunrand, P. and D. Tsomocos, (2006). A model to analyse fnancal fragly. Economc Theory, 27, Haldane, A., Hoggarth, G., and V. Saporta. (2001). Assessng fnancal stably, effcency and structure at the Bank of England. Publshed n Marryng the Macro Prudental Dmensons of Fnancal Stably, BIS Papers No 1. Hoggarth, G., Rcardo, R. and V. Saporta. (2002). Costs of bankng system nstably. Journal of Bankng and Fnance, 26(5), Humphrey, D., and L. Pulley., (1997). Banks responses to deregulaton: Profs, technology, and effcency. Journal of Money, Cred and Bankng, 28(4), February. Hunter, W., and Tmme, S., (1995). Core deposs and physcal capal: A reexamnaton of bank scale economes and effcency wh quas-fxed nputs. Journal of Money, Cred and Bankng, 27(1), Jones, R. and M. Tsutsum, (2009). Fnancal stably: Overcomng the crss and mprovng the effcency of the bankng sector, OECD Economcs Department Workng Papers, No.738, OECDPublshng.Retreved: lbrary.org/docserver/download/fulltext/5ks5ljtzrw0x.pdf?e xpres= &d=0000&accname=guest&checksum=8af5cb C465C08EC2FE2415D030 Kwan, S. (2006). The X-effcency of commercal banks n Hong Kong. Journal of Bankng and Fnance, 30(4), Ladak, A. and C. Gagans. (2009). Effcency and stock performance of EU banks: Is there a relatonshp? OMEGA, do: / j.omega Omega. Neal, P. (2004). X-Effcency and prodyctvy change n Australan bankng. Australan Economc Paper, 43(1), OECD, (2004). The 8 th OECD NBS Workshop on Natonal Accounts, 6-10 December OECD Headquarters, Pars. Retreved: ttp:// Ramlall, I. (2009). Bank-Specfc, Industry-Specfc and macroeconomc determnants of profably n tawanese bankng system: Under panel data estmaton. Internatonal Research Journal of Fnance and Economcs, 34, Sathye, M. (2001). X-Effcency n Australan bankng: An emprcal nvestgaton. Journal of Bankng and Fnance, 25, Schnas, G. (2005). Preservng fnancal stably. Economc Issues, 36. Internatonal Monetary Fund. Sealey, C., and J. Lndley. (1977). Inputs, Outputs and a Theory of Producton and Cost at Deposory Fnancal Instutons. The Journal of Fnance, 32: Strum, J., and B. Wllams. (2004). Foregn bank entry, deregulaton and bank effcency: Lessons from the Australan experence, Journal of Bankng and Fnance, 28(7),

22 22 The Internatonal Journal of Bankng and Fnance, Vol. 12, No. 2, 2016: 1-22 Thangavelu, S. and C. Fndlay. (2010). Bank effcency, regulaton and response to crss of fnancal nstutons n selected Asan countres. Retreved: 22 August Thoraneenyan, N., and Avkran, N. (2009). Measurng the mpact of restructurng and country-specfc factors on the effcency of post-crss East Asan bankng systems: Integratng DEA wh SFA. Soco-Economc Plannng Scences, 43, Walker, G. (1998). Economes of Scale n Australan Banks, Australan Economc Paper 37: Worthngton, A. (2000). Cost effcency n Australan non-bank fnancal nstutons: A non-parametrc approach. Accountng and Fnance, 40(1),