THE COST EFFICIENCY OF ISLAMIC BANK IN INDONESIA: A STOCHASTIC FRONTIER APPROACH AND DATA ENVELOPMENT ANALYSIS

Transcription

1 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 461 THE COST EFFICIENCY OF ISLAMIC BANK IN INDONESIA: A STOCHASTIC FRONTIER APPROACH AND DATA ENVELOPMENT ANALYSIS Rafika Rahmawati 1 Abstract Entering the ASEAN Economic Community (AEC) in 2015, the Islamic banking in Indonesiais expected to have a better performance to compete sustainably with the local banks and the foreign ones. This paper measure the performance of the banks using the cost efficiency approach with a focus on the two inputs (cost of fund and cost of labor) and the two outputs (total financing and owned securities). Using Stochastic FrontierApproach (SFA) and Data Envelopment Analysis (DEA) on Islamic Banks during the period of January 2010 to December 2013, the result shows that the level of efficiency of the Islamic banks in Indonesia is not optimal. Our calculation shows different result for both method (SFA and DEA), where the highest efficiency levels using SFA method is Bank Mega Syariah, while with the DEA method is Bank Muamalat Indonesia. Furthermore, using panel estimation, this paper conform the significance of the cost of fund, the cost of labor, and the total financing on the cost efficiency of the Islamic bank; leaving some option of strategiesto improvetheircost efficiency; this includes increasing their assets, increasing deposits, and cutthe unnecessarilycosts. More strategy includes product innovation, reducing the salary of theboard of directors, and put thefundsin profitable portfolio. For the authorities, this paper has demonstrated the use of frontier approachas good alternative in assessing the performance of the banks. Keywords: cost efficiency, Islamic banks, SFA, DEA. JEL Classification: G2, C1, C3 1 Author is a student at Postgraduate Program in Management Science. Institut Pertanian Bogor; fika_annaliez@yahoo.co.id.

2 462 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 I. INTRODUCTION The ASEAN Economic Community (AEC) 2015 was a difficult period for Islamic banks in Indonesia. Coupled with the integration of financial services by ASEAN 2020, foreign banks would likely flood Indonesia. Local banks, particularly the Islamic banks have difficulty competing with foreign banks as foreign banks are able to offer credit at interest rates lower than the local banks. The Law Number 21/2008 concerning Islamic Banking, mentions three types of Islamic banks in Indonesia, i.e., the Islamic Banks (BUS), Sharia (UUS), and Bank Rakyat Syariah Financing (BPRS). The development of Islamic banking in Indonesia is still not optimal, in terms of both the number of banks, offices, and the amount of assets. The Islamic Banking Outlook 2014 recorded asset growth of Islamic Banks and Sharia Business Units per October 2013 (yoy) at 31.8% or slowed growth compared to the year 2012 to the amount of 34.1%. The deposit accounts were as many as 12.3 million (9.2% of the national level). In addition, the market share of Islamic banking to the national banking system has now reached 4.8%. The Market share of Islamic banking is still very small. Compared with Conventional Commercial Banks (BUKs), the development of Islamic Banks is still very small. In terms of numbers, there are only 11 BUS compared to 124 BUKs. In addition, the average rate of growth in the number of TPF (Third Party Funds) of BUK has been 15.94% over the last 5 years, while it s been 35.6% for BUS. In terms of growth in deposits, BUS have been experiencing greater growth; however, in terms of number of TPF, BUK is still greater. Another difference between the BUK and BUS is in the interest rate or the results intended for clients. It is known that the interest rate on average in deposits to savings at the BUK are smaller at 2.01% compared to the BUS that is equal to 5.66%. This has resulted in many people preferring to save their money in BUS. While the interest rate on average loans for working capital at BUK is 12.14%, which is less than the average margin of financing for working capital on a BUS that is equal to 14.33%. This would encourage more people to choose to borrow funds from BUK, and thus the BUS will be burdened with accumulated third party fund, which also will add tremendous obligation to BUS in paying the sharing profit. However, the biggest source of revenue is from the bank credit / financing, which is still small at BUS. In light of this, BUS must be able to manage funds efficiently in order to compete with the BUK. With an efficient BUS, the BUS will then be able to offer a percentage fee or a smaller margin for the lenders, making it a major attraction for customers who want to borrow funds from BUS. With the efficient management of the funds, BUS will be able to compete, thus the market share of Islamic banks may increase. A bank s performance is usually measured using standard accounting or financial ratios such as return on equity (ROE), return on assets (ROA), asset turnover and return on permanent capital. However, by measuring the efficiency of the accounting standards, inefficient sources

3 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 463 in the banking management, internal and external factors that influence inefficiencies in a bank cannot be known (Sutawijaya and Lester, 2009). In measuring the level of efficiency, there are 2 approaches. First, through a parametric approach, i.e., Stochastic Frontier Approach (SFA), Thick Frontier Approach(TFA), and the Distribution Free Approach (DFA) approach. Secondly, through a non-parametric approach, such as Data Envelopment Analysis (DEA) and the Free Disposable Hull. This paper analyzes the level of cost efficiency by using both a parametric approach,stochastic Frontier Approach (SFA), and a non-parametric approach, Data Envelopment Analysis (DEA), and analyzes the factors that influence the cost efficiency in BUS. In addition, this paper also analyzes strategies to increase cost efficiency of BUS in Indonesia using the Stochastic Frontier Approach. The second section of this paper provides a literature review; the third section presents the methodology and data used in this paper; the fourth section presents the calculations and analysis; and the fifth section describes the conclusions. II. THEORY Research on the efficiency frontier, especially in banking was started by Sherman and Gold (1985). They applied the frontier approach on the banking industry with a focus on a measurement of operational efficiency in bank savings and loans. Since then, many studies have used the frontier approach to measure the efficiency of banking (Hanim, et al., 2006). Measurement of the level of efficiency with frontier approach can be divided into two, parts, i.e. i) deterministic approach commonly called non-parametric approach, which uses Technical Mathematic Programming, or the popular Data Envelopment Analysis / DEA); and ii)the Stochastic Approach,which is classed as a parametric approach, using an Econometric Frontier. Hadad, et al (2003) measured efficiency with a parametric approach using the Stochastic Frontier Approach (SFA) and the Distribution Free Approach (DFA). Both methods can be used for banksthat are most efficient. However, the DFA efficiency score is more diverse than the SFA efficiency score using monthly data and annual data that incorporates the entire bank. Mokhtar, et al (2006) measured the level of technical efficiency and cost efficiency of Islamic banks in Malaysia with the SFA method. In measuring the technical efficiency, the input used is the total deposits and total cost, while the output used is the total asset earnings. In the measure of cost efficiency, the variable price of labor and physical capital were coupled with the price of deposits. The results showed that the average technical efficiency and cost of conventional banks were higher than Islamic banks. On the measurement of the level of efficiency of the DEA model, there are two models used to analyze the efficiency of an Economic Activity Unit (EAU). First, the model assumes

4 464 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 constant returns to scale (CRS) or the so-called CCR (Charnes-Cooper-Rhodes) model.in the model of constant return to scale each EAU are compared with the rest of EAU in the sample with the assumption that the EAU internal and external conditions are the same. According Charnes, Cooper, and Rhodes this model can show the overall technical efficiency or value of profit efficiency for each EAU. The second model is variable returns to scale (VRS) or commonly called the BCC (Bankers - Charnes - Cooper) model. In this model it is assumed that all the conditions are not the same as the EAU or it can be said that not all of EAU operate optimally. Imperfect competition, financial constraints and others may cause a company to operate below optimal scale. Research by Firdaus and Hosen (2012) concerning the measurement of the level of efficiency of the BUS with DEA method using the assumption of constant returns to scale (CRS), showed that the overall level of efficiency of BUS in Indonesia during the study period did not reachan optimal level of efficiency. In addition, research on the measurement of the efficiency by comparing parametric and non-parametric methods has been under taken by many researchers, such as Bauer et.al (1998), who examined the level of efficiency of 683 banks in the United States in 1977 to 1988 by comparing the frontier models such as DEA, SFA, TFA (thick frontier approach) and DFA (Distribution free approach). The results showed that the average level of efficiency for 12 years using DEA resulted in 21%, SFA 87.9%, TFA 80.08% and DFA 85.5%. This study showed that the level of efficiency as measured by the SFA and DEA have a weak correlation (14%). Delis et.al (2009) who studied banks in Greece in the period used DEA and SFA. The results of this study indicated the level of efficiency of DEA and SFA amounted to 63.9% and 80.5%, respectively. More research was done by Paramita (2008) with a degree of correlation equal to 32.82%. This study measured the efficiency of 1,701 BPR in Indonesia in 2007 with the SFA and DEA efficiencies of 81.2% and 8.9%,respectively. This paper analyzes the cost efficiency as calculated by two methods; first, the SFA method, and second, by the DEA method assuming constant returns to scale (CRS), followed by analysis of the input and output factors that affect the cost efficiency of BUS. III. METHODOLOGY 3.1. Data and Variables The data used is secondary data, i.e., data obtained from the information published by Bank Indonesia from January 2010 to December Data collection sources include documents from a variety of print and electronic media. The population that is the object of this research are all the banks operating under Islamic principles recorded during the period 2010 to 2013 totally 2,990 banks, including 11

5 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 465 Islamic Banks, 23 Sharia Business Units, and 163 Bank Rakyat Syariah Financing. The sampling technique is purposive sampling with a criteria of Islamic Banks (BUS) that operate and have available monthly financial reports for the full period January 2010 to December According to these criteria, the study sample comprises 5 BUS, namely Bank Muamalat Indonesia, Bank Syariah Mandiri, Bank Mega Syariah, BRI Syariah and Bank SyariahBukopin The determination of input and output in his research was done with the asset approach as used by Muliaman D. Hadad. The input used was Personnel Expenses and Burden Sharing, while the output used was Total Financing and Securities Owned. The following operational definition of the variables are summarized in Table 1. Table 1. Variables Operational Research variable type Indicator definition of indicators Scale dependent Total Cost (TC) The sum of expenses estimated losses on commitments and contingencies Total other operating expenses + + Expenses for possible losses + nonoperating expenses). One of the bank s operating expenses, which accounted in profit and loss. millions Rupiah Independent Personnel Expense (P 1 ) millions Rupiah Independent Burden Sharing (P 2 ) Total Financing (Q 1 ) Liabilities of banks on third-party funds that have been collected by Islamic banks, this expense is recorded in the income statement. Productive assets that generate bank. Islamic bank financing, among others, in the form of financing Mudaraba and Musharaka. The bank s assets in securities which accounted for on the balance sheet on the assets side. millions Rupiah Independent millions Rupiah Independent Securities Owned (Q 2 ) millions Rupiah 3.2. Data Analysis Method The calculation of the efficiency of Islamic Banks used in this study is the cost efficiency approach, while the parametric approach calculation uses the Stochastic Frontier Approach (SFA), and the non-parametric approach uses the Data Envelopment Analysis (DEA), which calculates the deviation of the cost function estimated prior to its profit frontier. The cost efficiency ratio is defined as the minimum cost that the company can produce at a number of specific outputs, i.e., the actual cost incurred by the banking company. The smaller the actual costs used compared with the minimum cost, the level of efficiency of bank fees will be greater.

6 466 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 Stochastic Frontier Approach The Frontier 4.1 software is used to estimate the cost function using panel data on a parametric approach, Stochastic Frontier Approach (SFA). The standard function stochastic cost frontier has the general shape (log) as follows: wherec i is the total cost of the bank n; X ji is input j to the bank n; Y ki is the output k of bank n; and e i is the error, i.e., consists of two functions: where u i is the error factor that can be controlled; and v i is the random error factor that cannot be controlled. It is assumed that v is normally distributed N (0, σ 2 ) where uis half-normally v distributed, N (0, σ 2 where u = (u exp(-h (t-t)) 3 and h are parameters to be estimated. v) it i Cost-efficiency is a derivation of a function of cost, for example, the cost function has the form of the general equation (log) as follows: By using this form of the equation stochastic cost frontier, the cost equation can be written as follows: where C is the total cost or the cost efficiency; w is the number of inputs; y is the amount of output; and u and v is the error. Thus, the cost efficiency can be written as follows: (1) (2) (3) (4) (5) Data Envelopment Analysis The WDEA software is used to estimate a cost function using panel data for a non-parametric approach to Data Envelopment Analysis (DEA). Data Envelopment Analysis is a non-parametric method used in assessing the efficiency of an EAU. This study used a model assuming constant returns to scale (CRS) or a so-called CCR (Charnes-Cooper-Rhodes) model. This model was chosen based on research conducted by Suseno (2008) about the lack of correlation between the levels of efficiency of Bank Syariah (study of 10 Islamic Banks) with a scale of production. The study explained that economies of scale in the banking industry does not happen according to the company s scale due to the integrated function of a bank with other banks.

7 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 467 Panel Data Regression To identify the effect of the explanatory variables on the efficiency of a bank, this paper uses panel data estimation. Based on the variations in the assumptions formed, there are three methods for calculating the panel data regression analysis. The first method is the Common Method-Constant (the Pooled OLS Method / PLS), which is the simplest method. This estimate assumes that every object has the same intercept and slope. The second method is the Fixed Effect Method (FEM) that can differentiate the cross-section intercept with the assumption that every individual is considered to have its own characteristics. The third method is the Random Effect Method (REM). In this method, different parameters that are different between regions and eventime, are put into the error. For this reason, this method is called error component method. According to Winarno (2009), in Istianawati (2014), this method can be used if the total cross-section data is larger than the independent variables. All three of these methods will determine the most appropriate model to estimate the panel data regression parameters. Formal testing can be performed with the Chow test (F statistics) to choose the best model between PLS or FEM. Another test used is the Hausman test to select the best model among FEM or REM specifications. IV. RESULTS AND ANALYSIS 4.1. Cost Efficiency Analysisof Islamic Banks By using Stochastic Frontier Approach (SFA) and Data Envelopment Analysis (DEA), the level of efficiency of each Islamic Bankwas measured. The level of efficiency was analyzed in a cost function model with the dependent variable total cost / total cost (TC), with the input consisting of expenses of personnel (P1) and profit sharing expense (P2), and the variable output of total financing (Q1) and securities owned (Q2). The function of the resulting costs are in the form of a model frontier which is a translog model not a linear model or straight line, therefore all the variables in this study are TC, P1, P2, Q1 and Q2 modified in the form ln (Kumbhakar, 2003, in Edy Hartono). Efficiency Level Method - Stochastic Frontier Approach (SFA) Data processing was performed using the Frontier 4.1 software. The model prediction of cost efficiency in the BUS can be written as follows: ln TC= -0, ,542 lnp1 + 0,538 lnp2-0,971 lnq1 + 0,109 lnq2 In the regression equation above, the constant TC is equal to This means that if the input variables and output variables held constant, then the BUS will issue a minimum cost for a particular output level that is equal to million of total assets (e x = ).

8 468 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 At the input variables that personnel load (lnp1), the regression coefficient of shows that if the exponent of the burden of personnel increased by 1%, then the total cost will increase by 0.542%. This shows that BUS expenses incurred to employees do not contribute positively to the BUS, and therefore the greater the burden of staffing, the higher the total burden cost borne by the BUS which would lead to inefficiencies. And for the profit sharing expense (lnp2), the regression coefficient of shows that if the exponent sharing expense increased by 1%, then the total cost will increase by 0.538%. It shows that the BUS is not optimal in managing third party funds (TPF) that exist. Where the TPF should increase revenues for BUS, it becomes a burden as the customer must pay for the return to its depositors which would then lead to inefficiencies for BUS. In the variable output for total financing (lnq1), a regression coefficient of indicates that if the exponent of total financing is increased by 1%, then the total cost would be decreased by 0.971%. It shows that the financing provided by the BUS goes well and produces a good return as well. However, the BUS must continue to improve the productivity from the financing provided so that the level of efficiency experienced by the BUS may continue to rise. And securities owned (lnq2) with a regression coefficient of shows that if the exponent of securities owned increased by 1%, then the total cost will increase by 0.109%. This shows that the securities owned by the BUS generates returns that are lower than the costs of the BUS on the securities. Table 3. Shows the results of the cost efficiency SFA method for five BUS. Table 3. Level of Cost Efficiency Commercial Bank Syariah SFA Method Level of Cost Efficiency Period BMI BSM BMS BRIS BSB March June September December March June September December March June September December March June September December Average Average Level of Efficiency SFA entire BUS =

9 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 469 The level of cost efficiency SFA approach in the overall average BUS is at or 85.38%. This means that the overall BUS efficient use fee is 85.38%, while the remaining 14.62% is wasted costs. Please also note that BMS has the highest level of cost efficiency SFA compared to other BUS and BUS as a whole that is equal to or 92.38%. BMS is known not to be a too big an acquisition target of TPF. This means the BMS is not over-burdened with the cost of funds from the depositor funds.while BRIS has the lowest level of cost efficiency SFA, which amounted to or 78.35%. The average level of cost efficiency SFA on BMI, BRIS and BSB is still below the average level of efficiency throughout BUS. Thus, in order to better compete, the third BUS should be cutting costs that are not effective. From the result of the level cost efficiency SFA, it is also shown that the overall BUS has not reached an optimal level of efficiency. Therefore, increased cost efficiencies still can be achieved. The level of cost efficiency SFA also looks at variations ranging from (BSB period June 2013) to (BSB period March 2012). It is shown that each BUS has managerial implementation strategies that vary in terms of emphasis on cost-sharing, emphasis on personnel costs, investment arrangements, and so forth. Efficiency Level Method - Data Envelopment Analysis (DEA) Below are the results of five BUS efficiency levels during the period January December 2013, using the Data Envelopment Analysis (DEA) method. Data processing was performed using the WDEA software. Table 4 shows the results of the cost efficiency DEA method on a five BUS. Table 4. Level of Cost Efficiency DEA Method Islamic Banks Level of Cost Efficiency Period BMI BSM BMS BRIS BSB March June September December March June September December March June September December March June September December Average Average Level of Efficiency DEA whole BUS = 0,9325

10 470 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 The level of the cost efficiency DEA approach in the overall average BUS is at or 93.25%. This means that overall BUS can achieve maximum output level by simply spending less than 93.25% of its input. Please also note that BMI had the highest level of DEA cost efficiency compared to other BUS (and the BUS as a whole) that is equal to or 94.87%. At the same time, BRIS had the lowest level of cost efficiency DEA that amounted to or 91.95%. BSM, BMS, and BRIS had a level of cost efficiency with the DEA method lower than the average level of efficiency in the entire BUS. From the results of the level of cost efficiency DEA it was also shown that the overall BUS had not reached an optimal level of efficiency. Therefore, increased cost efficiencies still be achieved. There are various ways to optimize the level of efficiency in BUS. One of them is by looking at the table of the target values of the WDEA output and without calculations, then the target output and input variables can be changed or adjusted appropriately. Thus, each BUS can obtain optimal efficiency levels. The level of cost efficiency DEA also looks at variations ranging from up to It was shown that each BUS has managerial implementation strategies that vary in terms of emphasis on cost-sharing, emphasis on personnel costs, investment arrangements, and so forth. 4.2 Differencesin the Level of Cost Efficiency Using with SFA and DEA Methods Based on the analysis method of SFA and DEA above, the level of cost efficiency in each BUS during the period can be seen in Table 5. Table 5. Comparison of Cost Efficiency Level of BUS using SFA and DEA BUS BMI BSM BMS BRIS BSB Cost Efficiency Level SFA 83.28% 87.96% 92.38% 78.35% 84.92% Level of Cost Efficiency DEA 94.87% 92.65% 92.86% 91.95% 93.93% From the table above, it is shown that the level of cost efficiency in the five BUS using the SFA and DEA methods were relatively good, however none of the BUS were seen to be optimal in reaching 100% efficiency. However, there is inequality of the result of cost efficiency using the SFA and DEA methods. BMS has the highest level of cost efficiency using SFA, while the BMI has the highest efficiency using the DEA method. Notably, BRIS has thelowest cost efficiency level of all five BUS, using both SFA and DEA methods.

11 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 471 Compared with the DEA, the SFA method requires accurate information on the prices of inputs and other exogenous variables. SFA also considers some deviation as a nuisance (noise) in the regression model. The DEA approach is referred to as deterministic frontier since it considers some deviation,which is a snapshot of inefficiency.as the DEA method does not use information such as the prices of inputs or other exogenous variables, less data is needed, fewer assumptions are required and fewer samples can be used; however, a statistical inference cannot be taken. Although the SFA is more completeness in estimating the level of efficiency, it does not mean the DEA is without an advantage. In measuring efficiency, the DEA identifies the unit used as a reference that can help to find the cause of inefficiency and a way out of inefficiency (Hadad et al: 2003). In addition, the DEA does not require a complete specification of the form function showing the observed relationship between production and distribution. On further exposure it can be seen that the distribution of data in the classical assumption in this study meets the assumptions of normality and homogeneity. This will affect the regression estimation of the SFA approach so that the proposed model can be representative. The fulfillment of the assumptions of normality and homogeneity can be caused by measurements made on each BUS and not the entire BUS. This is done so that data can be more balanced. In this study the SFA is more appropriate to measure the level of costefficiency of the five BUS. The SFA method estimates the sampling error to be infinite, especially if the input and output variables are relatively more than the number of variables observed (Hadad et al. 2003). The results of the cost efficiency SFA method can be divided into several categories based on the average value of cost efficiency that is equal to the SFA method with a standard deviation The cost efficiency SFA value can be grouped into four categories by using percentile quartile ± standard deviation is as shown in Table 6. Table 6. Category Cost Efficiency Level of Cost Efficiency < to to 0.97 > 0.97 category Not efficient Less Efficient Efficient enough Efficient

12 472 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 Then, the five BUS can be grouped as shown in Table 7. Table 7. Grouping Level of Efficiency BUS Tingkat Efisiensi Kategori Bank Muamalat Indonesia 83.28% Less Efficient Bank SyariahMandiri 87.96% Less Efficient Bank Mega Syariah 92.38% Efficient enough BRI Syariah 78.35% Less Efficient Bank SyariahBukopin 84.92% Less Efficient According to the table above, it can be concluded that BMI, BSM, BRIS, and BSB are in the category of banks that are less efficient in managing their funds. While BMS is considered in the category of banks that are fairly efficient in managing their funds. The results shows that no BUS is operating efficiently. To that end, there is a need for improvements to increase the performance efficiency of BUS The Effect of Input-Output Variableson the Efficiency of Islamic Banks In contrast to other analysis tools, dual linear regression requires stringent test requirements. After the multiple linear regression equation is formed, a classic assumption test is applied. Data normality test results showed that only BRIS was not distributed normally and does not meet the test of normality. From the autocorrelation results, it can be concluded that there is no autocorrelation, except on Bank Muamalat. From the Scatterplot diagram, which is the result of the heteroscedasticity test (not shown), it can be concluded that the regression model on the five BUS are symptom free heteroscedasticities. Based the multicollinearity test of each bank, it is shown that the independent variable on the all the BUS have problems with multicollinearities 2. Panel estimates were made on 240 cross-sectional observations of the five BUS and data time series over 48 months. In this research, the REM method was not used, due to the provision that in order to process the data with the REM method, the number of cross section data must be more than the number of variables. Therefore, this study uses FEM. After knowing the appropriate model used in this study, the next step was the Chow and Hausman test. As has been described above, the REM method cannot be used in this study, hence the only Chow test was used. The Chow test results (F = 17.1; p = 0:00) showed the specifications of FEM to be better than PLS. 2 This problem is expected to be addressed in future research.

13 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 473 Table 8. PanelData Regression Results variable coefficient Std. Error t-statistic Prob. C Personnel Expenses 12: Burden Sharing 12: Total Financing Securities Owned Fixed Effects (Cross) _BMI-C _BSM-C _BMS-C _BRIS-C _BSB-C Effects Specification Cross-section fixed (dummy variables) R-squared Mean dependent var Adjusted R-squared SD dependent var SE of regression Akaike information criterion Sum squared resid Schwarz criterion Log likelihood Hannan-Quinn criter F-statistic Durbin-Watson stat Prob (F-statistic) According to Table 8, only the Securities Owned variable had no significant effect on the Total Cost. This is indicated by the value of the probability more than the confidence level (0.05). Personnel Expenses and Burden Sharing had a significant positive effect on the Total Cost of the BUS. As for Total Financing, it had asignificant negative effect on the Total Cost on BUS. Personnel Expenses generally have a positive effect on the Total Cost. With an increase in Personnel Expenses, Total Costs will increase and will reduce the efficiency costs on BUS. Because of this, to be more efficient, the BUS must reduce Personnel Expenses. Where Personnel Expenses increasesare independent of an improvement in employee performance, then Personnel Expenses should be reduced.it was found that the salaries for directors were quite high. To that end, there is a need for an in-depth assessment of the suitability of the high salaries of the directors to the company s efficiency. Sharing the burden generally has a positive effect on the Total Cost. Increased Burden Sharing will increase the Total Financing and will reduce the efficiency costs on BUS. During this time, TPF is the only burden to BUS, which is because the BUS has an obligation to pay a return to the TPF. As described previously, the return of savings in the BUS is higher than BUK. Because of this, the BUS Shares needs to trim expenses by lowering its ratio to customer savings. What the BUS needs is not alarge number of TPF, but more, how to manage TPFs well to rotate and produce a higher return for the BUS.

14 474 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 Total Financing generally has a negative influence on the Total Cost. Increasing Total Financing will reduce the Total Cost. This suggests that the financing of the BUS should be increased, because it can generate returns that are good for the BUS so that the loads can be reduced and the level of cost efficiency will also increase. The low value of the efficiency of the BUS is caused by several factors, one of which is the high level of profit sharing or margin financing on the BUS, so that people find it expensive financing with the BUS. Therefore, to improve financing to the BUS, then the BUS must reduce the level of profit sharing or margins to be more competitive and attractive to the public. Securities Owned in general have a negative effect on the Total Cost. Increased Securities may reduce the Total Cost, so that the level of efficiency can be increased. However, for all BUS, securities do not affect the Total Cost. This is likely due to the small value of Securities inthe BUS, so it does not affect the Total Cost. Therefore, the Securities should be optimized in order to generate greater returns for the BUS. For comparison, an estimate was also performed for each bank (not the panel). The estimation results of the five BUS on input and output variables influence the cost efficiency that showed a good coefficient of determination as shown in the adjusted R Square in Table 9. Table 9. Test Results Adjusted R Square (Adj R 2) BMI BSM BMS BRIS BSB Overall BUS Adj R , Source: Secondary data are processed. The adjusted R Square value of the overall BUS regression models amounted to 0.983, or 98.3%, which showed that the independent variables (expense personnel, profit sharing expense, total financing, and Securities Owned) together can explain the dependent variable (total cost) 98.3%, and that the remaining 1.7% describes other variables not included in the empirical model. The F test for each of the samples showed fairly good model specification. The following Table 10 compares the t test results on each BUS with t test results on the overall BUS. Table 10. T Test Results Variable BMI BSM BMS BRIS BSB BUS Overall Personnel Expenses + * + + * - + * + * Burden Sharing + * + * + * + * - * + * total Financing + * + - * - * + * - * Securities - * - * - - * - + Note: * = indicates significant variables at the 5% significance level

15 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis Comparison of Costs and Efficiency Level Profitability Islamic Banks The indicators commonly used to measure the level of profitability is ROA (Return on Assets). The ROA is a ratio that indicates the overall ability of existing assets to those assets used to generate a profit. Table 11 shows the average level of profitability in the five BUS for the period 2010 to Table 11. Average Rate of Profitability of Islamic Banks ROA BMI BSM BMS BRIS BSB 1.48% 2.09% 2.77% 0.83% 0.59% The profitability results above can be grouped into five categories based on the Bank Indonesia Circular Letter No. 13/24 / DPNP dated October 25, 2011 (Table 12). Table 12. Category Profitability Level of Profitability < 0% 0% 0.5% 0.5% 1.25% 1.25% - 1.5% >1.5% Category Loss Not Profitability Profitability Less Profitability Enough High Profitability Then, the five BUS can be grouped as follows (Table 13) BUS Table 13. Profitability Groupings Level of Profitability Category Bank Muamalat Indonesia 1.48% Profitability Enough Bank SyariahMandiri 2.09% High Profitability Bank Mega Syariah 2.77% High Profitability BRI Syariah 0.83% Profitability Less Bank SyariahBukopin 0.59% Profitability Less The following Table 14 shows the results of the level of cost efficiency and the level of profitability in the five BUS from

16 476 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 Table 14. Performance Period of Islamic Banks BUS Cost Efficiency category Category Profitability Bank Muamalat Indonesia Less Efficient Profitability Enough Bank SyariahMandiri Less Efficient High Profitability Bank Mega Syariah Efficient enough High Profitability BRI Syariah Less Efficient Profitability Less Bank SyariahBukopin Less Efficient Profitability Less The table above shows that there are inconsistencies between the BUS results of the level of efficiency and profitability levels among the BMI, BSM, and BMS. The BMI has shown to have a less efficient level of cost efficiency and sufficient profitability, i.e., the BMI does not manage its funds efficiently. BMI still had not cut funds that are not important, so the BMI profitability does not reach an optimum. Based on the financial statements of BMI by May 2014,the BMI profit growth was just 3%. This happened because the cost of funds had increased. In addition, at the AGM (General Meeting of Shareholders) at the end of June 2014, there was an agreement to use the company s net profit (Rp billion)for the fiscal year 2013 after tax for bank reserves, with no dividend distribution to shareholders (Reuters, February 2015). BMI was also very aggressive to attract third party funds, from promotions in the form of prizes as well as to provide a high level of profit sharing exceeding the BI Rate. This is what should be taken into account by BMI managementin order to increase the TPF and be followed by an increase in the optimization return from managing the TPF. The BSM was shown to beless efficient and have a high profitability, i.e. the BSM does not manage the funds efficiently. However, high profitability can be obtained by BSM. This is because the BSM sets a profit sharing ratio to borrowers that is considered high. For the BSM to be more competitive, the level of cost efficiency should be increased so that the BSM can provide a lower ratio and be competitive with other banks. In addition, the BSM spends must on products that are sold in the market, such as pawn loans and gold, as well as Haj Savings. BSM s pawn loans and gold product turnover reached Rp 2.05 trillion, and the monthly turnover increased by approximately 16.25% (Metrotvnews.com, September 2014). As for the Hajj Savings, BSM also continues to expand its market share, as it wants to be known as the Hajj and Umrah bank.for this, the BSM participated in the International Islamic Expo 2013, which helped promote products and services related to Hajj and Umrah (Infobank, December 2013). It can be concluded, BSM has a product that is superior to other BUS, so BSM can earn high profits. However, such profits can actually be improved through costs efficiency by minimizing expenditures. The BMS was shown to have a pretty good level of efficiency and high profitability levels. This means the BMS, as a BUS with assets that are not large, was able to manage their funds

17 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 477 at optimal high levels of profitability and not be too burdened with TPF so that the level of efficiency was quite good. BMS profits at the end of 2013 increased by 13.09% compared to the previous year. BMS is known to not have a too big a target for TPF. This makes the BMS not too over burdened with the costs of funds of the TPF. BMS focused on financing that is dominated by two-wheelers joint financing, where such financing continued to experience an increase of 56.52% from the previous year (Bisnis Indonesia, February 2014). The BRIS was shown to have high levels of efficiency and less profitability. This means, the BRIS must continue to improve cost efficiency by reducing costs that are not important and cannot increase profitability. The BRIS is expected to be more efficient, with a level of profitability to also increase. BRIS income declined drastically in the first period of 2014, decreasing by 97.73%. It likely occurred for several reasons. One of them is, the majority of the financing portfolio in BRIS are murabaha (sales), in which the margin of financing had been agreed upon at the beginning, so when there is an increase margin deposits, BRIS cannot raise the sale margin of financing. The BRIS also has to bear the increased cost of funds being offset by an increase in income (Cash, August 2014). On the other hand, growth in deposits in BRIS grew only 3.19%. The BSB was shown to have a level of efficiency that was less and a level of profitability that was lacking, meaning the management of funds in BSB has not been good. The BSB was not able to minimize costs that were not important. BSB was also very selective in disbursing financing. Murabaha financing, where the return obtained by the bank is fixed, is considered to suppress profit for BSB. Therefore, the BSB should be able to place their funds in investments that provide a good profit, so the profitability of the BSB can be increased and the profitability can in turn reduce the cost burden on BSB Managerial Implications The level of price competition is getting tighter with banking demands for Islamic Banks to be able to provide competitive prices for customers. By minimizing costs, BUS will be able to provide more competitive prices. Therefore, a calculation of cost efficiency is required. The cost efficiency ratio is defined as the minimum cost that a company can produce a number of specific outputs, and the actual cost incurred by the banking company. The smaller the actual costs used compared with the minimum cost, the higher the level of cost efficiency of the bank. Below is a review of the managerial implications of the recommendations based on the research results. Bank Muamalat Indonesia BMI had a low level of cost efficiency and profitability levels that were quite good. From these results, it is known that BMI was still burdened by many costs which should be minimized. Profit sharing expenses and personnel burden significantly increased the total cost of the BMI.

18 478 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 Therefore, BMI needs to optimize TPF to be distributed in the form of financing to the TPF to be able to provide returns for BMI and not be a burden to it. Personnel Expenses should also be trimmed because they were only a burden to BMI during this time, and it was not followed by an increase in employee performance. However, the total financing of the BMI also significantly increased the total cost of the BMI. This may occur because the margin of financing was so high that the financing channeled was not optimal in providing benefits for BMI. In addition, in order to compete, the BMI gave lower margins, but these margins were not been able to cover the costs tothe BMI. Bank Syariah Mandiri BSM had a fairly good level of efficiency and a high level of profitability. From the results shown, the BSM had great assets capable of generating high profit and cost efficiency, which is quite good, although not optimal. In improving the cost efficiency, BSM needs to improve the placement of funds in profitable investments like securities. From the results of the analysis, securities significantly reduce the total cost of the BSM. In addition, profit sharing expenses also significantly increased the total cost of the BSM. Therefore, BSM needs to manage TPF well in order for the TPF to provide benefits for BSM and not be a burden. Bank Mega Syariah BMS had a level of cost efficiency that was quite good and high level of profitability. The results showed the BMS was not operating at optimal cost efficiency. Profit sharing and personnel expenses was a burden to significantly increase the total cost of the BMS. Therefore, the BMS needs to optimize TPF to be distributed in the form of financing so that the TPF is able to provide returns to BMS and not just act as a burden. Personnel expenses should also be trimmed because it is only a burden to BMS without it being followed by an increase in employee performance. Another strategy is for the BMS to increase the total financing significantly while lowering the total costs. BRI Syariah BRIS had a low level of cost efficiency and profitability, and as such, needs better management. Results of the analysis showed that the profit sharing expenses significantly increased the total cost at the BRIS. Therefore, BRIS needs to manage TPF well so that the TPF can benefit the BRIS and not just be a burden to it. Total financing and Securities Owned by the BRIS also significantly lowered the total cost of the BRIS. The BRIS should further improve financing and investment that is profitable and safe.

19 The Cost Efficiency Of Islamic Bank In Indonesia: A Stochastic Frontier Approach and Data Envelopment Analysis 479 Bank Syariah Bukopin BSB had a low level of cost efficiency and profitability. The results of analysis showed that the profit sharing expenses and securities owned by the BSB significantly increased the total cost of the BSB. Therefore, the BSB needs to manage the TPF well in order for the TPF to provide a benefit and not become a burden. In addition, the placement of assets in investments, such as securities, must also be taken to ensure that these investments are able to produce profits for the BSB. The load of personnel was shown to significantly lower the total cost of the BSB. This means that the burden of the existing personnel at the BSB can improve employee performance, so that output can be more optimal. However, in the long term, the increase of the personnel does not necessarily increase the level of efficiency. For that, BSB needs to continue to be cautious in making decisions. Islamic Banks in Overall Overall, the level of cost efficiency in the BUS was still less efficient. Income generated also was not optimal. Some of the factors that cause thisis the lack of capital or assets of the BUS. The BUS needs to increase its assets in order to continue to expand the market. The TPF on BUS acquisition should also be optimized. The BUS needs to be competitive in providing returns for customers on the TPF. However, BUS also must be rational in providing a return, so as not to experience losses. Another important thing for the BUS, is for the conventional commercial bank, as the parent of the BUS, to seriously encourage the business development of subsidiaries. The BUS must create innovative products that are appealing to customers and according to their needs, so that rational clients do not turn to conventional banks. In addition, the BUS also has to cut costs that are not important. For example, the salary of directors of large banks need pruning and such cuts ought to be re-allocated to other things that are more useful for the banks Policy Implications for Bank Indonesia From the analysis that has been done, it can be concluded that in the face of Asean Economic Community (AEC), Islamic banking in Indonesia, especially the BUS, are not ready to compete with other banks from ASEAN countries. As prices were not competitive, it would be increasingly difficult for BUS in Indonesia to obtain TPF-financing and channel financing. Indonesia is the world s biggest market for Islamic banking. It should be a driving force for the BUS to be more efficient in managing their funds and dare to offer more competitive prices so the market share of the great nation can be enjoyed. Surely the people of Indonesia will be more rationale and not emotional in their choice. So if the BUS cannot provide an attractive price, then the public will easily turn to other banks, either other Islamic banks or other conventional banks that are more profitable.

20 480 Bulletin of Monetary, Economics and Banking, Volume 17, Number 4, April 2015 Here are some recommendations to the Bank Indonesia policy based on the research results: 1. One of the missions of the BI is to encourage a national financial system that works efficiently. Therefore, the need for BI is to calculate the level of efficiency by using an frontier approach (SFA and DEA) as an alternative or support in assessing the performance of bank that have traditionally used ROA ratios. With the frontier approach, the factors that cause inefficiencies can be known and can be addressed, thus increasing the tracking for results to be achieved for the efficiency performance of banking. 2. Criteria need to be made on the raw value of the level of efficiency. This can bring about a common understanding of a value of efficiency for banks. 3. BI should treat provide Islamic banks and conventional banks differently. In terms of the distribution of funds, Islamic banks are not only faced with the question of profits, but also the distribution of funds that should be run in accordance with Sharia. Therefore, the central bank needs to develop Islamic financial instruments that Islamic banks can further increase its investments. 4. There is a need for awards giving activity to banks that manage their funds efficiently. This activity is expected to encourage national banks to continue to be more efficient, so as to compete with banking in Indonesia and foreign banks from abroad. V. CONCLUSION The results of this analysis on the level of cost efficiency in Islamic Banks (Bank Muamalat Indonesia, Bank SyariahMandiri, Bank Mega Syariah, BRI Syariah and Bank Syariah Mega) for the period , provides the following conclusions: 1. In the analysis of the BUS level of efficiency from period, the value of the average level of cost efficiency with SFA and DEA models was shown on each BUS as follows the BMI at 83.28% and 94.87%; the BSM at 87.96% and 92.65%; the BMS at 92.38% and 92.86%; the BRIS at 78.35% and 91.95%; the BSB at 84.92% and 93.93%; and the overall BUS was at 85.38% and 93.25%. 2. From these results in mind there are different levels of cost efficiency as calculated by the SFA and DEA methods. In this study, SFA was more appropriate to use. 3. The regression showed that the variables of Personnel Expenses, Profit Sharing Expenses, Total Financing and Securities Owned had a significant effect on the efficiency of the BMI. At BSM just Profit Sharing Expenses and Securities Owned had a significant effect on the level of efficiency. While at BMS, Personnel Expenses, Profit Sharing Expenses and Total Funding had a significant effect on efficiency. As for the BRIS, Load Sharing, Total Financing and Securities Owned had a significant effect on the level of efficiency. At BSB, Personnel