13
.0/msohtmlclip1/01/clip_image001.gif”>
AN ANALYSIS OF PRODUCTIVITY CHANGE: ARE UAE BANKS OPERATING EFFICIENTLY WHEN COMPARED TO GCC BANKS?
Hossein Attar Kashani, University of Dubai
Lamia Abdelaziz Obay, University of Abu Dhabi
ABSTRACT
This study, for the first time, takes on the issue of productivity changes of UAE banks in comparison to those in the rest of the Gulf Cooperation Council (GCC) countries over the period 2000-2005. Given the growing importance of Islamic banking, this study also examines the difference in efficiency between Islamic and Conventional banks. Based on the non-parametric approaches of data envelopment analysis, DEA, and Malmquist productivity index, MPI, the results show that the banks in the GCC countries show relatively similar levels of efficiency. While Kuwait and Qatar had higher efficiency scores than the UAE and the UAE higher than the rest of the remaining GCC countries, these were not found to be statistically significant. Commercial Bank of Abu Dhabi and National Bank of Dubai appeared constantly on the best practice frontier. The statistical tests also show no significant differences between the performance of the Islamic banks and their conventional counterparts. Over time, UAE banks were able to show gains in efficiency (4%) when other GCC banks were actually recording losses of efficiency of the same magnitude. Both Islamic and Conventional banks in the UAE recorded gains in technical and pure efficiency and losses in scale efficiency.
INTRODUCTION
Efficiency analysis of commercial banks is particularly needed in a region dominated by bank-intermediated finance. A well developed banking sector was found to bear a positive relationship to economic growth (Levine and Zervos, 1998). It was further suggested that the legal environment within which banks operate can significantly affect economic growth through its effect on bank behavior (Levine, 1998).
The economies of the Gulf Cooperation Council (GCC) have gone through periods of peaks and troughs, mainly because of their dependence on oil revenues and the volatility of oil prices. After a period of economic slump due to declining oil prices during most of the 1990s, GCC countries have been witnessing breathtaking growth thanks to skyrocketing oil prices since 2000. The surge in oil revenues has led to a lifting of the region’s economy and the accumulation of large
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image002.gif”>
14
.0/msohtmlclip1/01/clip_image001.gif”>
amount of liquidity. This has stimulated an unprecedented investment boom, especially in the property market, and increasing demand for bank services. The question becomes whether banks are well poised to meet the challenges and take advantage of the new opportunities presented to them.
Banks in the region are generally small in size, preventing them from participating fully in the financing of energy-related and other local infrastructure projects. They have long benefited from the protectionist policies of the local governments. Entry and branching restrictions have limited the ability of foreign banks to capture a material share of the local loan and deposit market, even when outnumbering the local ones.
Bank regulators, however, have expressed intentions to, or starts of, changes, which, once completely implemented, would confront the banks specially the local ones with serious challenges. To start with, all the GCC member countries are signatories of the WTO agreement, which should result in further liberalization of their economies, in general, and the financial sector in particular. Indeed, most of these six countries have already revised, or are in the process of revising, their company and investment laws to allow for higher level of foreign ownership of banks and presence of foreign investors in the local stock markets. It is these changes that made Kuwait, Oman and Saudi Arabia extend new licenses to foreign banks in 2004. The UAE recently indicated its intention to adopt a reciprocal treatment to foreign banking presence when issuing new banking licenses.
Further, the GCC countries are envisioning a monetary union by the year 2010. This not only assumes the removal of all barriers towards the flow of capital between the member countries. It may also lead, through mergers and acquisition, to a wave of consolidation that would unveil hidden inefficiencies that were only made viable because of government protectionist policies.
One salient feature common to all GCC countries is the emergence of Islamic banking and finance. It all started with the establishment of Dubai Islamic Bank which was able to bring about a marriage of faith and finance that many thought could not coexist in modern times. While the Islamic finance sector has been growing at a faster rate than its conventional counterpart, it still is far from capturing a sizable market share of the banking industry. During 2001-2005, the UAE Islamic Banks outperformed their GCC counterparts in terms of growth of their asset base (20.5%), loan portfolio (21.6%), and deposit mix (21.2%) compared to GCC’s growth rates of 7%, 6% and 4% respectively. Islamic banks in Bahrain lead in terms of market share: viz., 38% of the countrys banking sector real assets, 28% of real loans, and 36% of real deposits in 2005 compared with 12.4%, 16.1% and 13.6% respectively in the UAE.
With these scenarios, the efficiency and/or over-the-time efficiency gain of each bank in the system would play a major role in their competitiveness and survival as an independent entity, for the less efficient firms have traditionally been the prime target of well-functioning competitors. On a cross-country basis, on the other hand, the possibility of a monetary union between the member countries would suggest homogenous banking system in each and every GCC countries. Without this homogeneity cross-country mergers and acquisitions are possibilities that could not be ignored.
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image003.gif”>
15
.0/msohtmlclip1/01/clip_image001.gif”>
Once again the efficiency of banks and their over-the-time efficiency gain play major role in provoking a take-over initiation.
This study is, therefore, important for bankers as a higher efficiency means higher profit and increased chance of survival in an increasingly deregulated and competitive market. Higher efficiency could also lead to a higher customer satisfaction as efficient banks are better positioned to offer quality and new services at competitive prices. The study is also important for the policy makers. An awareness of efficiency features is important to help them formulate policies that affect the banking industry as a whole and the local banks in particular.
This study aims to examine the overall cost efficiency of the banks in all six GCC member countries of the GCC over the period of 2000-2005 and the comparative efficiency between Islamic banks (IBs) and conventional banks (CBs). We measure the efficiency of each bank and the average efficiency of the banks in each country and for each specialization (IBs vs CBs) in each year of the operation within the time period in question. We proceed by investigating the efficiency gain/loss of each bank during the mentioned period to shed some lights on overall performance of the banking industry in the GCC countries.
We also test whether there has been a significant improvement in their efficiency over time. Given the protectionist environment in which they operate, the high level of government ownership in the sector, the low level of financial deepening of the GCC economies and the most recent oil bonanza, banks may have had little incentive to strive towards improving their productive efficiency, and accordingly, achieve little efficiency gains.
The remainder of this study is organized as follows. The next section looks at some of the existing works on the issue. This will be followed by discussing the methodologies used in this study. We will see how data envelopment analysis can be used to measure the relative efficiency of the banks and how we can apply the Malmquist productivity index (MPI) technique to break down the efficiency changes in various components and how it could be used to measure the efficiency changes of the banks over time. To measure the efficiency of the banks, we need the input and output data. Section four discusses the variables we use in this study. We then proceed to introduce the results. The study comes to its end with a summary and some concluding remarks.
LITERATURE
The literature focusing on the efficiency of the financial sectors of various countries, in general, and the banking sector, in particular, is vast. To mention only a few of the literature during the last 10 years we can name: Sufian and Abdul Majid (2007) discusses the relationship between X-efficiency and share prices in the Singaporean banking sector; Sufian (2007) evaluates the efficiency of domestic and foreign Islamic banks active in Malaysia; Barros and Garcia (2006) use DEA to evaluate the performance of Portuguese pension funds from 1994 to 2003; Lozano-Vivas and Pastor (2006) relate macro-economic efficiency of fifteen OECD countries over a period of eighteen years to the financial efficiency of the countries; Grigorian and Manole (2006) use DEA
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image002.gif”>
16
.0/msohtmlclip1/01/clip_image001.gif”>
to estimate indicators of commercial bank efficiency to bank-level data from a wide range of transition countries; Brown and Skully (2006) evaluate the cost efficiency of banks in the Asia-Pacific region and test whether the operating performance of banks in poorer economies improves with the inclusion of environmental proxies; Kirkwood and Nahm (2006) investigate the relationship between the Australian banks’ efficiency to their stock returns; Lo and Lu (2006) discuss the profitability and marketability benchmark of financial holding companies in Taiwan; Samjeev (2006) evaluates the efficiency of the public sector banks in India to investigate whether there exists any relationship between the efficiency and size of the banks; Camanho and Dyson (2005) use DEA to measure the cost efficiency of a British bank branches in phase 1 to be applied in the analysis of branch network and the production and value-added approaches to have a more comprehensive assessment of bank branch efficiency; Weill (2004) investigates the consistency of efficiency frontier models on some European (France, Germany, Italy, Spain, and Switzerland) banking samples; Krishnasamy, Hanuum Ridzwa, and Perumal (2003) apply Malmquist Productivity Index to discuss the efficiency of the Malaysian banks’ post-merger productivity; Sathye (2002) measures the productivity changes in Australian banking sector using DEA and Malmquist Productivity Index; Athanassopoulos and Giokas (2000) use DEA to discuss the efficiency of the banking sector in Greece; Chen and Yeh (2000) measure the bank efficiency and productivity changes in Taiwan banking sector and investigate the impact of ownership on the resulted efficiency scores; Zenios et. al. (1999) use DEA to develop a benchmark on the relative efficiency of the Cyprus banks branches, provide guidelines for improvement to management, and isolate the effects of the environment on branch efficiency; Camanho and Dyson (1999) too use DEA to assess the performance of Portuguese bank branches to complement the profitability measure used by the bank; Ayadi, Adebayo, and Omolehinwa (1998) measure the bank performance of Nigerian banks and conclude that the seeming inefficiency is attributed to the banks’ poor management;
Closer to the region under this study, Rao (2005) looks at 35 banks operating in the UAE for the years 1998 and 2000, and concludes that these banks suffer substantial cost, X- (managerial), and scope inefficiency. Saif and Yaseen (2005) look at Scope and Scale efficiency of 100 banks operating in the Middle East and North Africa (MENA). They concluded that banks in the MENA region exhibited reasonable levels of efficiency and that ownership structures (foreign vs. domestic) had no significant effect on efficiency. Their study, however, uses the intermediation approach and makes no cross-country comparison of efficiency.
Of particular interest to our research are comparative efficiency studies between Islamic banks and conventional banks. Al Jarrah and Molyneux (2003) investigate the efficiency of 82 banks in selected MENA countries Jordan, Egypt, Saudi Arabia and Bahrain, over the period 1992-2000 using a stochastic frontier approach. Islamic banks are found to be the most cost and profit efficient and investment banks the least efficient, Bahraini banks are found to be the most cost and profit efficient.
Al-Shammari (2003), investigates the cost efficiency of 72 banks operating in the GCC countries over the period 1995-1999 using a stochastic frontier approach. He finds Islamic banks
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image003.gif”>
17
.0/msohtmlclip1/01/clip_image001.gif”>
to be the most cost efficient and investment banks the least efficient, with 91% and 84% average efficiency scores, respectively. Overall, Saudi banks are found to be the most cost efficient, followed by UAE banks, with Qatari banks being least efficient (92%, 90%, and 83% respectively).
Studies outside of the GCC/MENA region lead to similar conclusions. A study by Abdul Majid et al. (2003) of Malaysian banks over the period 1993-2000 reveal Islamic banks to be more efficient, albeit marginally, than conventional banks. In a more recent study, Sufian (2007) finds domestic Islamic Malaysian banks to be more efficient than foreign Islamic banks operating in Malaysia. No comparison, however, was made with their conventional counterpart. El-Gamal and Inanoglu (2005) find Islamic banks to be among the most efficient banks operating in Turkey.
The contribution of the current study to the literature is that: 1. it, for the first time, isolates the banking system in the GCC countries in one study, and 2. it investigates the pattern of efficiency change on individual country, as well as on cross-country bases. The latter helps the banks, and the policy makers, to be aware of the effects of their decision on the prospect of future survival in an ever-growing competitive environment.
METHODOLOGY
Examination of the hypotheses set out above requires relative efficiency of banks performance over time. The (relative) efficiency is generally represented by production functions, which can be estimated and evaluated by one of the two widely used parametric and nonparametric approaches of Stochastic Frontier Analysis (SFA) and Data Envelopment Analysis (DEA), respectively. The two approaches use different techniques to envelop data more or less tightly. To that end, they make different accommodations for random noise and for flexibility in the structure of production technology. It is these two different accommodations that generate the strengths and weaknesses of the two approaches.
While DEA involves mathematical programming, SFA is based on econometric methods. This difference in methodology brings about advantages and disadvantages for each of these approaches. Perhaps the most important advantage of the parametric approach is that it has the ability of differentiating between stochastic noise and inefficiency of the production process. This advantage is, however, offset by the SFAs requirement for having a pre-determined production function, common between all the DMUs (decision making units). In addition, the SFA approach can handle only one variable as input. In cases where there are n inputs (n>1), normalization of n-1 inputs is required to make SFA analysis possible. Further, the SFA technique does not allow for direct breakdown of overall efficiency in its various components of technical, scale, and pure efficiency, and comparing each components changes over time. This breakdown is made possible through Malmquist productivity index (MPI), developed by Malmquist (1953), which is based on DEA non-parametric technique. As far as this study is concerned, a pre-determined common production function for all 6 GCC banking industries seems problematic, which would jeopardize the validity of results. Further the three inputs used in this study require a technique that can handle
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image002.gif”>
18
.0/msohtmlclip1/01/clip_image001.gif”>
multiple input problems. And finally, the use of Malmquist productivity index approach along with the DEA make the analysis harmonized across the study. For this reason we use the DEA in this study. The following is a brief outline of the DEA and Malmquist techniques.
Data Envelopment Analysis
The mathematical approach to construct production frontiers and measure efficiency relative to these constructed frontiers is called data envelopment analysis (DEA). The non-parametric approach of DEA calculates a discrete piecewise frontier determined by the set of Pareto-efficient decision-making units, DMUs – in our case banks. This results in an understanding about each DMU’s stand in relation to the rest of the DMUs. This is in contrast to the focus on the averages which is the case with parametric, statistical approaches.
The principle of DEA lies in the definition of efficiency and productivity: if a given bank, C, is capable of using X(C) input to produce Y(C) unit of output, then other banks should also be able to do the same if they were to operate efficiently. Similarly, if bank B is capable of using X(B) units of input for Y(B) unit of outputs, then other producers should also be capable of the same production schedule. Banks B, C and others can be combined to form a composite (virtual) bank with composite (virtual) inputs and composite (virtual) outputs. The non-efficient banks will be compared with this set.
As mentioned earlier, DEA does not require any assumption about the functional form of the production function in question. The only requirement in this approach is that each DMU lies on or below the frontier. Each DMU not on the frontier is scaled against a convex combination of the DMUs on the frontier facet closest to it. For each inefficient DMU, DEA identifies the sources and level of inefficiency for each of the inputs/outputs, depending on the orientation of the study. The level of inefficiency is determined by comparison to a single reference DMU or a convex combination of other referent DMUs located on the efficient frontier.
DEA consists of a variety of models, each one suitable for different settings. The two most commonly applied models are the BCC and CCR models. The CCR model (developed by Charnes, Cooper, and Rhodes, 1978) identifies the source, and estimates the amount, of inefficiencies, as well as yielding an objective evaluation of overall efficiency. The BCC model (developed by Banker, Charnes, and Cooper) distinguishes between technical and scale efficiencies by estimating pure technical efficiency at the given scale of operation, and identifying whether increasing, decreasing, or constant returns to scale possibilities are present for further exploitation. While the BCC model identifies the most efficient banks, the CCR model identifies the most productive banks. This is why more DMUs are introduced as efficient by the BCC model than the number of productive DMUs identified by the CCR model. The following figure gives a geometrical presentation of the two models.
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image003.gif”>
19
.0/msohtmlclip1/01/clip_image005.jpg”>
DEA could be applied after deciding on the orientation of the study, which identifies whether the decision-makers have more say in determining the level and/or the type of inputs, or the level of output. The results of an input-oriented DEA indicate the degree of inefficiency in the application of inputs, holding the output fixed. The output-oriented DEA indicates the relative efficiencies of the production process of each decision making unit. The orientation of the study depends on the aim of the study and/or the nature of the industry in question. In the current study, we are seeking the cost efficiency of the banks, which moves us towards an input-oriented analysis.
The input-oriented BCC model is structured as:
Minimize: Z =2 -,.ß.S+ – ,.ß.S-
(1)
0
Subject to:
Y8 – S+ = Y
0
2X – X8 – S-= 0
0
ß.8$ 1
(2)
8, S+, S-$0
Here 8 is a N x 1vector of constants, Y and X are the output and input matrices, respectively, S is the slack variables,$ is a (1× N) row of 1s, and2 is a scalar the value of which will be the efficiency score of the ith DMU. The constraint of ß.8 $ 1 allows for variable return to scale of operation. The presence of the non-Archimedeang allows the minimization over2 to preempt the optimization involving the slacks.
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image002.gif”>
20
.0/msohtmlclip1/01/clip_image001.gif”>
Efficiency over Time
We can include the time component into the production function and measure and analyze the effects of time on the improvement of performance. This is because the production technology of almost all industries is subject to modification based on state-of-the-art technology in the industry in question. Hence the time component is in fact presenting the change in response to production technologies. With this, the production function takes the more challenging form of Y = (X1, X2, , Xn, Z1, Z2, .., Zn, e, t ), where t denotes the element of time, or in fact technology. To take account of the components of the efficiency changes over time we use the Malmquist Total factor productivity index (TFP) introduced by Malmquist (1953). According to Sufian (2006) three advantages differentiates Malmquist productivity indices (MPI) over the alternative approaches (Tornqvist and Fisher indices): Firstly, it does not require specific assumption about the profit and cost function. Secondly, it does not require data on input/output prices. Thirdly, it enables the decomposition of productivity change into technical efficiency change and changes in the best practice frontier (technical/technological change) even if only panel data is available. The shortcoming of the technique is, however, that it needs to compute the distance function. This disadvantage is solved by using DEA.
The input oriented MPI of the ith DMU is formulated as:
+
+
+
Dt(yt+1, xt+1)
Dt+1 (yt+1 ,xt+1 )1/ 2
MPIit 1
( yt1
, xt 1
, yt , xt)=
i
×
i
(3)
t
t
t
)
t+1
t
t
)
Di
(y
, x
Di
(y
, x
Where
is a distance function measuring the efficiency of the process of producing yt output
using xt input in the period t. Equation (3) could also be reformulated as:
MPIit+1 ( y t+1, xt+1 , yt , xt )= Dit+1 (yt+1 , xt+1 ) Dt(yt , xt)
.0/msohtmlclip1/01/clip_image006.gif”>142i43
EfficiencyChange
Dt(yt+1,xt+1)
Dt
(yt+1 ,xt+1 )1/ 2
×
i
×
i
(4)
t
+1
( y
t+1
, x
t+1
)
t+1
(y
t
t
Di
Di
, x
)
1444442444443
TechnolgyChange
If the MPI (or any of the two components) is greater (less) than 1, it indicates progress (regress). The first expression in (4) measures the change in ith DMUs position relative to the frontier over the two periods: other things equal, it measures how the ith DMUs relative efficiency has
changed between the two periods. In Malmquist literature this is called efficiency change, or the “catching up effect”. The second expression contains a geometric average of two alternative measures of shifts in the technology frontier between period t and period t+1. Firstly, the shift is measured in relation to year t observations. Secondly, the shift is measured in relation to year t+1 observations. This shift measures the technological change over the two period.
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image003.gif”>
21
.0/msohtmlclip1/01/clip_image001.gif”>
Technological progress is defined as the progress of the ideas that specifies all activities of an organization towards gaining economic value, or profit. It is changes in the knowledge about product technologies, as well as knowledge about the process by which goods and services are produced. It also comprises the progress in the specification of how activities are organized. In macroeconomic level, this is the most important factor insuring the economic growth of a country. At a micro level, it has the same effect on the progress of an organization (Lipsey and Carlaw, 2004).
The original MPI is based on the assumption of constant returns to scale. This assumption causes overestimation (underestimation) of productivity if the production process demonstrates decreasing (increasing) returns to scale. To solve this problem Fare et al (1994) suggest generalized MPI that includes scale index to take account of the effect of economies of scale on productivity.
Another efficiency change, namely the index of variable returns to scale (VRS) efficiency change, is the ratio of the VRS technical efficiency of period t+1 to that of period t. This also could be extracted from MPI. Moreover, the index of scale efficiency change of a DMU can be calculated as the ratio of its scale efficiency of period t+1 to that of period t (For a thorough study of Malmquist Productivity Index and its decomposition see for example Lovell, 2003)
DATA AND MODEL SPECIFICATION
Data
Balance sheet and income and expenses data of GCC banks were extracted from the BankScope database for the period 2000-2005. For missing data, whenever possible, individual bank annual reports were used. Out of 135 GCC banks reported in the BanScope database, only 56 banks (41 conventional banks and 15 Islamic banks) consistently reported the input and output variables needed for the analysis between 2000 and 2005. Development banks, mortgage banks, and special finance houses were excluded from the analysis. The UAE had the largest number of banks in the sample: 14 conventional and 3 Islamic banks (see Tables 2-7) while Bahrain had the largest number of Islamic banks (5). Oman had no full-fledged Islamic banks, but offered Islamic banking services through the Islamic windows of its conventional banks.
Selection of Input and Output
One of the most important issues in analyzing the efficiency of banking sector has proven to be the choice of input and output variables. This choice depends on the view held about the banking production function. Little agreement exists as to what a bank produces or how to measure its output. However, the two approaches of intermediation and production have been dominant in most of the mentioned studies.
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image002.gif”>
22
.0/msohtmlclip1/01/clip_image001.gif”>
The Intermediation approach views banks as mediators between the supply and the demand of funds, and as such they take deposits and other purchased funds as inputs and transforms them into loans and other assets, as outputs. The Production approach, on the other hand, views banks as producers of loans and deposits using different inputs such as labor and capital. At present, there is no agreement on the explicit definition and measurement of inputs and outputs of the industry. More precisely, researchers have always found significant difficulties in the definition and measurement of the concept of bank output (Altunbas and Molyneux, 1996). However, if we differentiate between input and output in a way that, ceteris paribus, if it is desirable to increase the quantity of the variable, it is an output, and if it is undesirable to have an increase in its value, it is an input (Charnes, Cooper, and Rhodes, 1978), then the intermediation approach become problematic as no bank would wish to have less deposits. It is for this reason that we adopt the production approach.
Accordingly, for the definition of inputs and outputs, we adopt a variation of the production approach used by Ferrier and Lovell (1990) and Cebenoyan and Register (1989) and conform more to a recent study by Ozkan-Gunay and Tektas (2006) about the Turkish banking industry (see Table1). We use three inputs and three outputs in our model. The three inputs represent resources/expenses required to operate a bank. These are: personal expenses measured by salary, other operating expenses (including expenses incurred for premises and fixed assets), and interest expense (or return to depositors in the case of Islamic banks). The three outputs represent desired outcomes. These are: earning assets, total deposits, and operating income. Banks allocate resources and control expenses and strive to maximize earning assets and income. Banks that do this better than their peers fall on the efficient frontier. Banks that utilize too much input for a given level of output, or produces too little output for a given level of input relative to their peers are considered inefficient and will fall below the efficient frontier.
.0/msohtmlclip1/01/clip_image007.gif”>.0/msohtmlclip1/01/clip_image008.gif”>.0/msohtmlclip1/01/clip_image009.gif”>
Table 1: Selected Efficiency studies and choice of Input-Output Variables
.0/msohtmlclip1/01/clip_image010.gif”>
Author
Input
Output
Approach
Country
Period
–
Personnel expenses
–
Total deposits
DEA
mix
Turkey
1990-
2001
Ozkan-Gunay and
–
Administrative expenses
–
Total loans
Tektas (2006)
–
Interest expenses
–
Total securities
–
Interest income
– Non-interest income
–
Price of Fixed Assets
–
Interbank Loans
SFA
Inter-
Germany
1993-
mediation
2003
Bos et al (2005)
–
Price of Labor
–
Customer Loans
–
Price of borrowed Funds
–
Securities
–
Equity
Academy of Banking Studies Journal, Volume 9, Number 1, 2010
.0/msohtmlclip1/01/clip_image003.gif”>
