Bootstrap DEA Efficiencies of the GCC Islamic Banks: Sources and Comparison During 2014-2016

This paper, first, obtained three categories of efficiencies, overall bias-corrected technical efficiency (OTEBC), bias-corrected pure technical efficiency (PTEBC) and bias-corrected scale efficiency (SE) of the Islamic banks of the Gulf Cooperating Countries (GCC) during 2014-2016 using the Simar and Wilson (1998) Bootstrap data envelopment analysis (DEA). Second, decomposing the overall bias-corrected technical efficiency (OTEBC) the paper found the bias-corrected pure technical efficiency (PTEBC) and the bias-corrected scale efficiency (SE) were 91% and 59.8% respectively and thus PTEBC dominated the OTBBC (82.4%) and the SE (59.8%) of the GCC Islamic banks. Third, the paper found the sources of the inefficiency of the Islamic banks of the GCC was the DRS. Except the Islamic banks of Qatar, banks of the GCC countries were inefficient either because they operated under the IRS or DRS. DRS was the major source of inefficiency. Qatar Islamic banks demonstrated the highest level of efficiency in all three efficiency among GCC. The paper provides suggestions for future study.


INTRODUCTION
GCC, Gulf Cooperating Countries, consists of Oman, Kuwait, Qatar, Bahrain, Saudi Arabia, and United Arab Emirates (UAE). They are all oil rich countries.
Among GCC, the financial service sector, bank in particular, is a dominant sector and is an important source of income for UAE and Bahrain. The financial industry is one of the largest non-oil contributors of Bahrain's real GDP. It contributes about 17.6 of GDP in 2018. One of the main beneficiaries of the regional boom driven by demand for oil is Bahrain's banking and financial services sector, particularly Islamic banking.
The UAE domestic market banks reported Dh2.84trn ($773bn) in assets at the end of the third quarter of 2018, making the UAE's banking sector the largest in the GCC region, according to the Central Bank of the UAE (CBU). Islamic banks account for about 20.4% of overall assets as of the third quarter of 2018. Dubai Islamic Bank, founded in 1975, was seen as the first shariacompliant lender of its kind, marking the beginning of the broader Islamic finance segment.
In the GCC region, Kuwait is home to one of the oldest banking industries. The Emir of Kuwait issued a 30-year concession to a group of British investors to set up the nation's first bank in 1941. Today Kuwait has 11 domestic and 11 foreign banks and Kuwait's banking sector constitutes one of the oldest and one of the most developed financial industries in the Gulf region. side by side with conventional banks. There are currently two Islamic banks operating in the sultanate -Bank Nizwa and Alizz Islamic Bank -as well as six Islamic windows operated by conventional institutions.
Among the GCC, the Qatar Financial Center (QFC) is one of the largest and fastest-growing business and financial centers in the Middle Eastern country with a significant 33% growth and nearly 200 companies registered on QFC's platform. Because of its growing financial institutions, Islamic financé in particular, Qatar is poised to become a leading hub for the financial sector. Within the Islamic finance, the Islamic banking sector is playing an important role in making Qatar a leading interconnected Islamic finance hub. In terms of total assets, Qatar's Islamic banking sector accounts for 82% in the Islamic finance with a total value of $107bn in the first half of 2019.
The dominant sector of Saudi Arabia is oil. The banking sector's growth of Saudi Arabia is tied to the growth of the economy fueled by the oil industry. When the global financial crisis hit the oil industry, its banking sector also faced challenges. Nonperforming loans (NPLs) increased 125% in 2009 due to the rising number of credit defaults. Saudi Arabia is home of a number of dedicated Islamic Banks, as well as Islamic window operations offered through conventional banks. Al Rajhi Bank is the largest Islamic bank in Saudi Arabia, and also the largest Islamic bank internationally with assets of SAR325.2bn (USD87bn) at end-3Q15.
The growth of Islamic banking has been a phenomenon since emergence in GCC. The numbers of Islamic banks, as well as the market shares of Islamic banks, are increasing in the world. Based on Ernst and Young (2014) the total assets of the Islamic bank are around 1.7 trillion dollars. However, at the center of the growth of Islamic banking assets is the GCC. These countries provide the largest share of total assets, more than 33% of the total (Bahrini, 2017).
Dubai, Bahrain, and Abu Dhab have become the main hub of Islamic banking in the Gulf Cooperating Countries (GCC). The rise of the Islamic banking in the GCC was tied to their rapid oil-contributed economic growth. Before the financial crisis in 2008-2009, the OPEC countries, the GCC in particular, generated windfall wealth in the world financial markets because of the huge increase in oil prices.
The Islamic banking sector was in the forefront in providing the credit needs of the nation and became an essential factor for economic growth of the GCC. Although the decline of oil price affected their economic growth, the efficiency of the banking sector, Islamic banks in particular, remains unexplored.
Banks' inefficiency increases the cost of intermediation and harms the allocation of funds and the profitability of banks leading to bank failure (Samad, 2014). The increased efficiency in banks' deposit mobilization and loan advancement are key to successful entrepreneurs for enhancing the economic growth of a country (Schumpeter, 1911).
The study of the Islamic bank technical efficiencies across the GCC and the sources of inefficiencies are important for several reasons. The efficiency of the productivity of banks including Islamic banks is of great interest to public authorities supervising and regulating banks, bank managements and bank depositors and borrowers. Each of them is interested to know bank efficiency. In a competitive market environment, bank depositors and borrowers are certainly interested to know the efficiency status of individual banks before they deposit their hard-earned savings. The borrowers of bank move to the banks which are more efficient in advancing loans.
Although the study of efficiency is important, the survey of the banking literature shows not enough evidence of Islamic banking efficiencies studies across the GCC during 2014-2016, the post global crisis period. This study is an important contribution to the Islamic banking efficiency literature by providing the comparative highlights of OTE, PTE, and SE of the GCC Islamic banking industry.
This paper provides the following sequences: A brief characteristic feature of Islamic bank is outlined in section 2. Section 3 provides the survey of literature. Data and methodology are discussed in section 4. Section 5 provides empirical results and conclusions.

KEY CHARACTERISTIC FEATURES OF ISLAMIC BANK
As the basic principles of the operation of the Islamic bank are derived from the Quran and Sunnah and the Islamic banks do not charge interest, Islamic banks are a different breed of the financial institution. The corner-stone of Muslims' way of life.
First, Islamic banks only finance the business that are permitted by the Shariah law. As Islam prohibits the consumption and production of any harmful activity, such as wine, alcohol, and destructive weaponry, Islamic banks do not finance these production and consumption, irrespective of high profit prospects.
Second, the avoidance of riba (usury) in all financial transactions is a key distinguishing feature of Islamic banks. This is because, the Quran, the Divine book of Islam strongly prohibits riba in business transactions. The Quran says: "… Allah permitted trading and forbade riba" (Quran: 2: 275). However, neither the Quran nor the Prophet of Islam did define what riba is 1 . At present, riba is interpreted as interest. The present scholars of Shariah agreed that the predetermined fixed rate of return is not permitted in the business transactions of the Islamic banking and financing world.
Third, the prohibition of riba (usury) in Islam gave birth to the rise of the profit and loss sharing (PLS) mode of production. The PLS is the most important characteristic of the Islamic banks that distinguishes the Islamic banks from the interest-based conventional banks. The key features of profit and loss sharing (PLS) are that (i) both parties (bank and borrower) share the outcome of business venture (profit or loss); unlike conventional bank equity contracts where banks do not bear the risk of financing investments, Islamic banks share the risk of investment; and (ii) unlike conventional banks' equity contracts where banks enjoy the fixed rate of return from investments, even when there are losses for the project, there is no predetermined rate of returns on investments for Islamic banks. Justice requires that both partners of business must share the risk of the business. Thus, the key features of the Islamic banking and finance are the PLS, the avoiding of fixed interest, and Shariah based business conduct.

SURVEY OF LITERATURE
This study mainly focuses on the cross-country bank efficiency studies. Important studies on a country-bank level efficiencies include the followings: El-Gamal and Inanoglu (2004) estimated the comparative cost efficiency of the Turkish banks for the period 1990-2000 using the data envelopment analysis (DEA) method. They found that the Islamic banks were more efficient due to Islamic banks' assetbased financing. Sufian and Majid (2006) investigated the comparative efficiency of the foreign and domestic banks of Malaysia during 2001-2005. They found that banks' scale inefficiency dominated pure technical efficiency during the period. They also found that the foreign banks had higher technical efficiency than the domestic banks.
Sufian (2009) made another study in estimating the various efficiencies and the determinants of these efficiencies of Malaysian. His studies found that the efficiencies were negatively related to bank expenses and economic conditions, while the efficiencies were positively related to loan intensity. Kumer and Rachita (2008) examined the technical, pure technical and scale efficiencies of the 27 public sector banks of India just for 2004. The empirical evidence of the paper shows public sector banks operated at 88.5% level of TE i.e., the inefficiency was 11.5%. Only 7 banks were technically efficient. The regression results of the paper found that the off-balance activities positively affected the Indian bank efficiency.
Samad (2009)  Cross country-bank level bank efficiency studies include the followings: Hassan (2006) applied both parametric method (SFA) and nonparametric frontier methods (DEA) in assessing the cost, profit, allocative, technical, pure technical and scale efficiency of 43 Islamic banks in 21 countries from Middle East, Asia, Africa and Europe over the period 1995-2001 and found Islamic banks were more cost inefficient than profit inefficient meaning that Islamic banks were more efficient in profit-making and technical inefficiency dominated the scale efficiency. His study confirmed the findings of Yudistira (2004) who examined the cross-country technical efficiency of 18 Islamic banks of GCC, East Asian, African and Middle Eastern countries during the period 1997-2000and found that the overall technical inefficiency score of Islamic banks was on average just over 10%. Samad (2013) investigated the efficiency of Islamic banks using the time varying Stochastic Frontier function on the Islamic banks of 16 countries and found that the mean efficiencies between the pre global financial crisis and the post global crisis were estimated at 39 and 38% respectively and the difference was not statistically significant suggesting that the efficiencies of Islamic banks did not deteriorated during the global financial crisis.
Applying the panel DEA, Sufian (2009) estimated the technical efficiencies of the MENA Islamic banks and the Asian Islamic banks and then compared their technical efficiency over the period [2001][2002][2003][2004][2005][2006]. They found that the efficiency score of the MENA Islamic banks were higher than the technical efficiency of the Asian Islamic banks. Pure technical inefficiency was less prominent than the scale inefficiency i.e. scale inefficiency was the major source of inefficiency.
Noor and Ahmad (2012) investigated the efficiency of 78 Islamic banks operating in 25 countries during the period 1992-2009 using DEA and found that the technical efficiency of many Islamic banks in the world have increased during and after the global financial crisis period. According to them the financial crisis has decreased trust in the conventional banking system in favor of the Islamic banking model. They further found that the pure technical efficiency scores of sampled Islamic banks were higher than their scale efficiency scores which contradicted the findings of Sufian (2009) and Yudistira (2004).
Using the data of 25 Islamic banks in GCC countries for the period 2003-2009 and applying DEA, Srairi and Kouki (2012) found: (i) the overall technical inefficiency of GCC Islamic banks was the result of pure technical inefficiency (29.3%) rather than that of the scale inefficiency (17%). (ii) The overall technical efficiencies of the Islamic banks increased during and after the global financial crisis.
Applying the DEA method, Rahman and Rosman (2013)

DATA
The bank inputs, used in this study, are: (i) bank capital, (ii) employee wages, and (iii) bank deposit. The bank outputs are (i) earning assets and (ii) gross loans.All inputs and output data for the Islamic banks of all GCC countries were obtained from Bank-Scope for the period of 2014-2016. Total number of banks under this study was 34. Of these 12 from Bahrain, 9 from UAE, 4 each were from Saudi and Kuwait, 3 from Qatar, and 2 from Oman.

Methodology
The DEA has two versions. The DEA model originally proposed by Charnes et al. (1978)  The difference between the CCR and BCC models can be illustrated by the following graph [ Figure 1]. The line through the points Q and C represents the CRS efficiency frontier and the curve (ABCD) represents the VRS efficiency frontier. Each DMU that is on the frontier is technically efficient. For this reason, the particular DMU "F" is technically inefficient. When we refer to the CRS frontier, the distance FQ measures the technical inefficiency of the DMU "F." However, when we consider the VRS frontier, the technical inefficiency of the DMU "F" is only the distance FB. The difference between the CRS and the VRS frontiers is the distance QB which is a measure of scale inefficiency. Suppose that there are no DMUs to be evaluated. Each DMUj, j =1,…, n uses m different inputs, noted (i = 1,…, m), to produce different outputs, noted (r = 1,…, s). The technical efficiency score for a particular DMU, called DMUo, is determined by solving the following linear programming problem. The technical efficiency Figure 1: Constant returns to scale and variable return to scale efficiency frontiers (Coelli et al., 2005) score θ for a particular DMU, called DMUo, is determined by solving the following linear programming problem: θ<1 Means that the evaluated DMU is technically inefficient. θ=1 Indicates a point on the frontier and hence a technically efficient DMU. In order to estimate the efficiency scores of all the DMUs in the sample, the above problem must be solved n times, once for each DMUj, j =1, n (Coelli et al., 2005).
This paper first applied the bootstrap-data envelopment analysis (bootstrap DEA) to obtain the efficiencies, θ*: overall biascorrected technical efficiencies (BC-TE) and bias corrected pure technical efficiencies (BC-PTE). The bootstrap DEA is used because the DEA method suffers from serious shortcomings, according to Wilson (1998,2000). (i) the DEA method is deterministic. That is, the efficiency score obtained by the DEA does not allow for random error such as machine failure or power out. It thus overestimates the efficiency scores of the DMU and leads to biased efficiency (Simar and Wilson, 1998). (ii) The DEA efficiency score is a calculation and is not an estimate as it does not have statistical properties such as any confidence level attached to it and confidence estimate with confidence interval values. Bootstrap is a data-based simulation method introduced by Efron (1979). The main idea or objective of bootstrap is to simulate the data generating process (DGP) with repeated sampling. That is, it replicates repeated sampling from the data. As the replicated data set approximates the original data, the sampling distributions of the sample mean and standard deviations generated from the repeated sampling are close to the original ones. The bootstrap-DEA, introduced first by Simar and Wilson (1998) provides the estimated efficiency scores of each DMU generated from numerous repeated sampling. The bootstrap-DEA, thus, provides the bias-corrected efficiency scores together with the confidence interval at α level. So, bootstrap-DEA efficiency scores are more accurate and have statistical properties which the DEA method efficiency scores lack.
Empirically, an estimate of the radial Debreu-Farrell output-based measure of technical efficiency can be calculated and obtained by solving a linear programming problem for each data point k (k=1,…, K): Z k k k ¦ d 1 X X n 1 N kn kn , , . . ., Where Y is K × M matrix of available outputs, X is K × N matrix of available inputs. CRS specifies constant returns to scale. For variables to scale (VRS) a convexity constraint Z k k k ¦ 1 1.
θ Is a scalar and represents the efficiency score of each decision making unit (DMU). The range of ≤θ≤1, with a value of 1 indicating a point on the frontier and hence a technically efficient DMU; i.e., outputthe of the DMU cannot be increased without increasing inputs. A DMU is inefficient when the value of θ<1; that is, a given output can be produced by reducing inputs of the DMU.
Bias is calculated as follows:ˆB The bias-corrected efficiency score can be expressed as:

Input-output controversy and model selection
In a production of coal mine, output is the amount of coal and inputs are labor and capital. However, in the multiproduct firm such as the bank which produces a series of services and uses a vector of inputs, deciding inputs and outputs are not easy to determine and it has become controversial. Based on the production approach (Benston, 1965), a bank is a producer of services for the bank account holders and it produces deposit accounts and loan services with labor and capital. In this sense, the number of deposit accounts or deposits can be used as output. Depositors' income which is equivalent to interest paid to depositors is an important factor for mobilizing total deposits.
Under the intermediation approach, first used by Sealey and Lindley (1977), the bank is a financial intermediary which collects deposits from the savers and channels funds to borrowers. It treats earning assets as outputs and deposits as inputs. In this sense, loans, investments in securities, and advances are the outputs of a bank and labor, capital, deposits, and expenses related to them are inputs of a bank.
Using the definition of Sealey and Lindley (1977), this paper estimated the following model using the bootstrap DEA: loan i = β 0 +β 1 CAP i +β 2 salary+β 3 Deposit i (4) Where loan i = total loans + total earning assets. They are considered as output.
Inputs of the bank are: Fixed capital (CAP), employee salary (EMEXPSE), and total deposits (DEPOST).The descriptive statistics of inputs and output variables are provided in Table 1.
All variables were in US thousand dollars. Table 1 shows that the mean of all variables increased during 2014-2016.

EMPIRICAL RESULTS
The empirical results of all efficiencies (Overall-technical efficiency (OTE 2 ), overall-bias-corrected technical efficiencies (OTECB), and biases (BIAS)) are presented in Table 2.
The decomposition of the overall technical efficiencies (OTEBC) into pure technical efficiencies (PTEBC), known as managerial efficiency, and scale efficiency (SE) are presented in Table 3.
The sources of inefficiencies across the GCC countries' Islamic banks are presented in Tables 2-8.
The average overall DEA technical efficiency (OTE) of the Gulf Islamic banks, in Using a 90% confidence interval, (C.I) estimate, Table 2 shows that the average bias in estimating technical efficiency is 3.7%.
Over-all bias-corrected technical efficiency (OTEBC) is decomposed into pure bias corrected technical efficiency (PTEBC) and scale efficiency SE). The pure bias corrected technical efficiency (PTEBC) is known as managerial efficiency. It represents the efficiency of management in using resources of firms. On the other hand, scale efficiency (SE)/inefficiency represents whether firms operate below or above the optimum capacity level of firms.
The decomposition of bank bias-corrected technical efficiency into PTEBC and SE is presented in Table 3 for understanding the sources of inefficiency.    The SE efficiencies of the Islamic bank of GCC countries marginally increased suggesting that the scale inefficiency decreased during 2014-2016.
When the three efficiencies, (OTEBC), PREBC, and SE, were compared, it showed that the managerial efficiency dominated both OTEBC and SE. The average BTEBC, 91%, was higher than those of both OTEBC, 84.4% and SE, 59.8%.
The average inefficiency of the Gulf Islamic banks due to inefficiency of the optimum loan size was 42.4 (100-57.5) and it dominated the inefficiency of due managerial inefficiency. The average PTEBC was 90.6% suggesting the average managerial inefficiency (PTEBC) of GCC banks was 9.4% compared to 17.6% inefficiency in OTEBC. This result confirms the finding of Srairi and Kouki (2012).
When three efficiencies, OTEBC, PTEBC, and SE, were compared, it was found that the dominant source of inefficiency was scale inefficiency. As the average scale efficiency (SE) of the GCC banks during 2014-2016 was 59.8%, the average scale inefficiency was (1-0.59.8) = 40.2% compared to the average managerial inefficiencies of 9%.
A comparison of efficiencies, OTEBC, PTEBC, and SE, across the Islamic banks of the GCC countries during 2014-2016 is presented in Tables 4-6, respectively. Table 4 shows the rank of efficiency level of the Islamic banks among the GCC. Ranking shows Qatar first, Kuwait second, and the rest of the countries are the same. The mean efficiency of the Islamic banks of Qatar, 88%, was the highest among the GCC Islamic banks. The average overall technical efficiency, OTEBC, of the Qatar Islamic banks was 88% and was higher than the GCC average of 82% during 2014-2016.
The average overall technical efficiency of the Islamic banks of Kuwait was 84% and was the second highest among the GCC Islamic banks during the period.
The average efficiencies (OTEBC) of the UAE, Oman, Saudi, and Bahrain Islamic banks were 81% and was the same as the average of GCC.
Although Bahrain and UAE were the front runner in introducing the Islamic banking operation and were the main hub of the Islamic banks, the average OTEBC of these countries were behind the technical efficiency of Qatar and Kuwait. One possible reason for this was that the number of banks of these countries under this study was very few compared to the number of banks from Bahrain and UAE. Table 5 shows the ranking of pure technical efficiency level of the Islamic banks among the GCC. Ranking shows Qatar was the first, Kuwait was second, and the rest of the countries were the same.
The mean pure efficiency of the Islamic banks of Qatar, 95%, was the highest among the GCC Islamic banks. The average pure technical efficiency, PTEBC, of the Qatar Islamic banks, 95%, was higher than the GCC average efficiency of 91% during 2014-2016. The management inefficiency level of the Qater Islamic banks was 5% (100-95=5%).
The average pure technical efficiency of the Islamic banks of Kuwait was 93% and was the second highest among the GCC Islamic banks during the period. The management inefficiency level of the Kuwait Islamic banks was 5% (100-93=7%).
The average pure technical efficiency of the Islamic banks of Saudi Arabia and UAE was 92% and 91% respectively and was the second highest among the GCC Islamic banks during the period suggesting management inefficiency level of the Saudi and UAE Islamic banks was 8% and 9% respectively.
The management efficiency level found to be lowest for the Islamic banks of Bahrain and Oman. The average pure technical efficiency of the Islamic banks of Bahrain and Oman was 92% for both countries.
Although Bahrain and UAE were the front runner in introducing the Islamic banking operation and were the main hub of the Islamic banks, the average OTEBC of these countries were behind the technical efficiency of Qatar and Kuwait. One possible reason for this was that the number of banks of these countries under this study was very few compared to the number of banks from Bahrain and UAE.    3 out of 4 banks, i.e. 75% of the Islamic banks of Saudi Arabia was scale inefficient and the major source of inefficiency was due to the DRS.
In sum, among two sources of inefficiency, IRS and DRS, the major source of the inefficiency of the Islamic banks of the GCC was due to the DRS. Most Islamic banks operated under the DRS during the 2014-2016.

Policy Prescriptions
The paper found that among three inefficiencies, scale inefficiency dominated the GCC Islamic banks, the paper provides policy prescriptions. Bank managements of the GCC banks should emphasize more on improving scale efficiency as it is a dominant source of inefficiency among three inefficiencies.  Third, decomposition of the sources of scale efficiency/inefficiency found that among two sources of inefficiency, IRS and DRS, the major source of the inefficiency of the Islamic banks of the GCC was due to the DRS. Most Islamic banks operated under the DRS during the 2014-2016. On average, 59.7% of the GCC banks were inefficient due to the reason that they were operating under DRS. 11.3% of the GCC Islamic banks were scale inefficient due to the IRS. On the other hand, only 27.3% of the banks were scale efficient as they operated under the CRS.
Fourth, a comparison of efficiencies of the Islamic banks among the Gulf countries found (i) the Islamic banks of Qatar were the most efficient in all three efficiencies, OTEBC, PTEBC, and SE. All three banks of Qatar were scale efficient in all 3 years, 2014-2016. The OTEBC and PTEBC of the Qatar Islamic banks 88% and 95% respectively and were highest among the GCC.
(ii) Islamic banks of Saudi Arabia were the least scale efficient country among the GCC. The scale efficiency (SE) of the Saudi Islamic banks was 45.8%.
As the number of years, 2014-2016, are short, results of the study of the Islamic bank efficiency should be interpreted and considered cautiously. Similarly, results of the Islamic banks of Qatar should be counted cautiously because there are only three Islamic banks of Qatar under this study.
As the number of years consisted of only 3 years and the number of banks of Qatar and Oman was limited, this paper suggests future study should include more Islamic banks and more extended periods.
Bank managements of the GCC banks should emphasize more on improving scale efficiency as it is a dominant source of inefficiency among three inefficiencies.