Efficiency differentials between Islamic and Conventional banks in Malaysia

Transcription

1 Faculty of Economics and Business Bachelor Thesis Efficiency differentials between Islamic and Conventional banks in Malaysia August 2013 Abstract This paper investigates efficiency differentials between Islamic and Conventional banks in Malaysia. After having elaborated on various measures of cost and profit efficiency, it was decided that the following ratios are used to measure cost and profit efficiency: Overheads, Cost-to-Income ratio, ROA, Equity ratio and ROE. Using a generalized least squared regression, it was found that Islamic banks are generally more cost efficient than Conventional banks which is mainly driven by medium sized Islamic banks. However, Islamic banks are less efficient than Conventional banks with respect to profits. Given the ROA and ROE ratio mainly large banks drive the difference between Islamic and Conventional banks whereas when looking at the equity ratio small Islamic banks are mainly driving the difference. Name: Thilo Hamann Student ID: Supervisor: Damiaan Chen MSc BSc in Economics and Business Specialization: Economics and Finance

2 Table of Contents 1. Introduction Literature Review Islamic banks VS Conventional banks Bank Efficiency Efficiency of Islamic banks VS Conventional banks Data and Methodology Methodology Sample selection and data collection Data description The Model Fixed VS Random effects Hypotheses Results st Regression nd Regression Analysis Analysis of the 1 st regression Analysis of the 2 nd regression Limitations Conclusion Bibliography Appendix Percentiles of Total Assets Correlation Matrix

3 1. Introduction Already in the early 2000s researchers found that Islamic banks are on the rise. Today, Islamic banks grow more rapidly than never. Even in the current rough financial world, the Islamic sector reached in 2010 an all time height of over 1,300 billion dollars worth of assets worldwide, which is six times as much over the past ten years (Beck et al., 2013). Moreover, Western banks such as ABN Amro or the Citibank nowadays go with the trend of Islamic banking and have Islamic subsidiaries or offer Islamic products themselves (Khan, 2010). Malaysia is one of the countries where Islamic banking is booming the most. Measured by assets, Malaysia is the most important Islamic centre with roughly 20 per cent of sharia-compliant banks whereas in other Muslim countries the average lies at around 12 per cent (Bahru, 2013). This makes Malaysia especially interesting for research of Islamic banks. The purpose of this paper is to investigate whether the growth of the Islamic banking system in Malaysia is due to a superior efficiency of Islamic banks over Conventional banks in terms of costs and profits. Over the last decades several studies about cost and profit efficiency were conducted, which mainly focused on Conventional banks and macroeconomic factors influencing the efficiency of these banks (see Beck et al., 2013; Chortareas et al., 2012; Arslan and Ergec, 2013). However, little research was done on efficiency of Islamic banks and on the difference between Islamic banks and conventional banks practices. An exception is the economic literature of Beck et al. (2013) which focuses on the efficiency and stability differentials of Islamic and conventional banks in 22 different countries. Nevertheless, this research is lacking a detailed analysis of a specific country by including several more factors affecting profit and cost efficiency of Islamic and Conventional banks than just overheads, cost income ratio, return on assets and the equity ratio. To answer the research question of this paper a panel data regression is used including 21 Conventional and 16 Islamic banks in Malaysia given a time span from 2008 to The paper is structured in the following way: The next section gives a brief theoretical background about differences between Conventional and Islamic banks, followed by efficiency measures currently used in the academic literature. Afterwards, recent academic research in the field of Islamic banking is outlined. Section 3 deals with the methodology applied in this paper. Moreover, it describes the collection of the necessary data as well as the panel data model used to measure efficiency differentials between Islamic and Conventional banks. Section 4 is devoted to presenting the results of the regression run, whereas section 5 analyses these results. The last part of this paper presents a brief summary of the findings. 3

4 2. Literature Review The rise of Islamic banks has grabbed the attention of many researchers which started investigating Islamic banks in greater detail. This section of this paper summarizes these researches briefly and gives more insights into Islamic banking practices. In the beginning, the key differences between Islamic and Conventional banking is outlined. Afterwards, several measurements of bank efficiency will be discussed which are currently used in academic literature. The focus will be put on profit and cost efficiency measurements of banks. The section finishes by a short summary of current findings in the field of efficiency differentials between Islamic and Conventional banks. 2.1 Islamic banks VS Conventional banks According to Beck et al. (2013) Islamic banking differs from Conventional banking in five main principles: On the one hand Islamic banking prohibits interest payments, called riba, engaging in speculative investments, named gharar, and lastly financing illicit sectors such as military companies, pork companies or the pornographic sector. On the other hand, Islamic banking is based on a profit and loss sharing principle between the bank and its customer as well as the principle that all transactions have to be backed by a real economic transaction that involves a tangible asset (pp ). A more detailed comparative analysis of Islamic banks and Conventional banks was carried out by Hanif (2011) who elaborates in detail on the differences and similarities between both systems. Hanif especially focuses on the different loan and deposit facilities of Islamic and Conventional banks. He concluded that the difference between both systems already lies in its foundations. Whereas Conventional banking is based on the capitalistic ideas of charging interest, Islamic banking is based on the Koran and thus sharia compliant. Furthermore, according to Hanif Islamic banks face a lot more difficulties due to the sharia compliant business such as the inability to claim interest on deposits with other banks and the central bank or the prohibition to invest in government securities. However, there are also critiques to Islamic banking which question how Islamic it actually is. Khan (2010) outlines these critiques in his economic paper. He argues that some Islamic banks just substitute different terminology for interest charges by for instance profit rates. Yousef (2004) titles this phenomenon the murabaha syndrome which means that Islamic banks closely mimic Conventional banking practices (pp ). Khan concludes after having looked at Islamic banking practices that the murabaha syndrome still persists nowadays since Islamic banks offer near identical services as Conventional banks but at a higher cost (Khan, 2010, p.818). 4

5 2.2 Bank Efficiency There are various studies existing in the current academic literature which outlined several different measures of bank efficiency. Most of them measure efficiency on a range of different financial ratios retrieved from financial statements of banks. An important issue about bank efficiency is the differentiation between cost efficiency and profit efficiency which lay the foundations of measuring efficiency of banks since these measures are based on economic maximization in reaction to market prices and competition (Berger and Mester, 1997, p.898). Cost efficiency can be defined as a measure of how close a bank s cost is to what a best practice bank s cost would be for producing the same output bundle under the same conditions (Berger and Mester, 1997, p.898). There are two common measures for cost efficiency found in economic literature: Overheads and the Cost-Income Ratio. Overheads can be defined as total operating costs divided by total assets. It measures what percentage of total assets was needed to finance the daily operation of the bank over a certain time period. Another alternative measure of cost efficiency is the Cost-to-Income Ratio which is defined as the ratio of operating costs to gross revenue, defined here as interest income of banks. It is an alternative measure of cost efficiency which shows the banks operating costs in relation to their gross revenue. Generally speaking, higher ratios of cost efficiency indicate lower levels of cost efficiency (Beck et al., 2013, p.436). Profit efficiency can be defined as a measure that determines how close a bank is to producing the maximum possible profit given a particular level of input prices and output prices (Berger and Mester, 1997, p.899). Several ratios are used to determine profit efficiency of banks. Beck et al. (2013 p. 437) use the return on assets (ROA) and the equity ratio. Whereas, Chortareas (2013, p.1227) measures profitability of banks just by the ROA ratio. A different approach is used by Brown (2003, p.47), who instead uses the return on assets and the return on equity (ROE) to determine a bank s profitability. The ROA ratio is relating net income to total assets. Conversely, the equity ratio relates shareholders equity to total assets which is another measure of how profitable a company is. Another commonly used measure for profitability which emphasizes the return to shareholders is the ROE which divides net income by total equity (Ding et al., 2010, p.83). Other authors, however, outline that there are also other financial ratios that can be used to measure profit efficiency of banks. Powers and Needles (2011) state that another useful measure of profitability is the Debt-to-Equity ratio which divides total liabilities by owner s equity. This ratio shows the proportion of a company s assets that is financed by creditors and the proportion that is financed by the owner (p.199). Ding et al. (2010) list, apart from the already mentioned ROE and ROA ratio, also the return on investment (ROI) as a measure for profitability. The ROI measures the wealth created on long term invested funds, regardless of the leverage ratio of the firm. To determine the ROI ratio Earnings before income and taxes is divided by capital employed which is average long term liabilities plus average equity (p.83). In general higher profit ratios imply that the bank is using its resources in a good manner. It is worth 5

6 noting that compared to the cost efficiency ratios profit efficiency ratios can also be negative (Berger and Mester, 1997, p.900). Concluding, there are various ratios that measure efficiency of banks. The most common measures for cost efficiency and profit efficiency were outlined in this section. These include Overheads, Cost-Income ratio, ROA, ROE, Equity ratio, Debt-to-Equity ratio and the ROI. 2.3 Efficiency of Islamic banks VS Conventional banks The raise of Islamic banks in the Arabic world led many researches to investigate Islamic banks in greater detail. One of the focuses of researches was whether the raise of Islamic banks is due to a superior efficiency of them compared to Conventional banks. Beck et al. (2013) is elaborating on that issue. In their paper they compared differences between Islamic and Conventional banks in terms of business orientation, efficiency, asset quality and stability in 22 countries using regression analysis. They find that Islamic banks do not differ much from conventional banks in their business model (p.445). However, Islamic banks have on the one hand significantly higher overhead costs compared to conventional banks but on the other hand only marginally higher cost-income ratios implying less cost efficient Islamic banks (p.436). Nevertheless, Islamic banks are significantly more profitable and better capitalized than Conventional banks (p.437). Furthermore, Islamic banks seem to perform better during crisis due to a superior asset quality (p.437). Srairi (2010) compares profit and cost efficiency of Islamic and Conventional banks in Gulf Cooperation Countries (GCC). Using a stochastic frontier approach as compared to Beck et al. he finds that Conventional banks are more cost and profit efficient than Islamic banks in GCC s which contradicts the finding of Beck et al. (p.60). He identifies three main reasons for that result. Firstly Islamic banks are usually more complex structured than Conventional banks due to higher remuneration packages to keep expertise in the Islamic bank. Secondly, due to the restriction of sharia compliant business Islamic banks are less efficient in containing costs than Conventional banks. Lastly, Islamic banks carry a lower amount of risk on their balance sheet due to more risk-averse transactions (p.60). Quershi and Shaikh (2012) carry out a non-parametric approach to look at efficiency differentials between Islamic and Conventional banks in Pakistan. Therefore, a detailed ratio regression analysis on efficiency differentials between Islamic and Conventional banks in Pakistan over a period from 2003 to 2008 is carried out. They include ratios such as ROE, ROA, Cost-to- Income and net interest margin. After running the regression the authors found that Islamic banks are more cost efficient but less profit efficient than Conventional banks in Pakistan (p.45). Given these controversial studies no clear superior banking system can be made out. Both systems seem to have its advantages and disadvantages. Moreover, a superior banking system strongly depends on country specific factors as Beck et al. (2013) points out: Cross-bank results, however, are 6

7 important variations across countries and across Islamic banks of different sizes [ ] (p.445). I believe that to evaluate efficiency differentials between Islamic and Conventional banks these country specific factors need to be taken into consideration since generalizations of efficiency differentials between Islamic and Conventional banks is not possible. Hence, this paper looks just at the case of Malaysia and determines whether the boom of Islamic banks in Malaysia is due to a superior efficiency of Islamic banks over Conventional banks. 3. Data and Methodology Having introduced the differences between Islamic and Conventional banks and summarized briefly the previous research done on the topic of efficiency differentials between Islamic and Conventional banks, a generalized least squared (GLS) regression can be set up that evaluates bank efficiency in Malaysia. This is the purpose of this section. Firstly, the methodology used for this economic research is elaborated on followed by the sample description used for this analysis as well as the method of data collection for the empirical analysis. Secondly, the data used is being described and a clear distinction between random and fixed effects is made. Afterwards, two GLS regressions for this econometric research are being set up. This section finished with two hypotheses regarding the regression outcome. 3.1 Methodology Qureshi and Shaikh (2012) state that: Traditionally financial ratios have been used to assess the financial performance of any firm including banks (p.42). That is also the case for the paper of Beck et al. whose methodology this paper will follow up on. However, there will be several variations from the ratio regression approach Beck et al. are using. Firstly, compared to Beck et al. this paper will just focus on one specific country, Malaysia, and not on 22 different countries. Secondly, this paper will look only at efficiency differentials between Islamic and Conventional banks and not, as Beck et al. did, also at the differences of the Business model, asset quality and stability. The reason for applying a slightly different methodology than Beck et al. is mainly the scope of this paper. Also, Beck et al. measure efficiency only by cost efficiency ratios, namely overheads and Cost-to-Income ratio, and use the profit efficiency ratios, ROA and the equity ratio, in a different context which they label stability. This issue is being resolved in this paper by measuring efficiency by both cost and profit efficiency ratios which are being used by Beck et al.. Moreover, I believe firstly that one more measure needs to be used to evaluate bank efficiency accurately. Hence, the ratio ROE is added to the ratio regression analysis used by Beck et al.. The main reason to include the ROE in the analysis on efficiency of banks is that various other authors, which investigated efficiency of banks, used the ROE as one of the main measures of efficiency (see Brown, 2003, Qureshi and Shaikh, 2012, Hidayat and 7

8 Abduh, 2012). A more intuitive reason to include the ROE ratio is that it measures the net income that is attributed due to the capital initially invested in the bank. Thus, it shows how well the bank can generate profit out of the capital invested which shows best how profitable the bank is. Secondly, two more variables are added to the regression set up of Beck et al. when controlling for bank size. Further details about the regression set up are discussed in section Sample selection and data collection This paper focuses on the banking system of Malaysia, thus, all licensed Conventional and Islamic banks of Malaysia are included in this analysis. A list of these licensed Conventional and Islamic banks is provided by the Central bank of Malaysia, which list 27 Conventional banks and 16 Islamic banks on their website as of now (Bank Negara Malaysia, 2013). The period for the analysis is a time span from 2008 to 2011 since the necessary data for this analysis is readily available on the web pages of all banks under consideration. For company specific reasons not all banks publish financial statements publicly that trace back a longer period than 2008 or that are closer to the current time. Moreover, banks that started business after 2011 are excluded from this analysis. This includes the following banks: BNP Paribas Malaysia Berhad, Bank of America Malaysia Berhad, India International Bank (Malaysia) Berhad, Mizuho Corporate Bank (Malaysia) Berhad, National Bank of Abu Dhabi Malaysia Berhad and Sumitomo Mitsui Banking Corporation Malaysia Berhad. This reduces the sample to 21 Conventional and 16 Islamic banks. Moreover, there are two banks, Industrial and Commercial Bank of China (Malaysia) Berhad and CIMB Islamic Bank Berhad, which have two years of missing values. On the one hand, the Industrial and Commercial Bank of China (Malaysia) Berhad just started business in Malaysia in 2010 and thus there are no ratios available for the years 2008 and On the other hand, the CIMB Islamic Bank Berhad states Annual reports on their webpage until 2009 and consequently no annual reports were found for the years 2010 and The necessary financial ratios are retrieved from annual reports. Annual reports are used in this study for one main reason: They represent the financial position of a company better than quarterly published financial statements since they are audited as compared to quarterly reports. In consequence financial fraud, if any, carried out by the bank is eliminated which results in an unbiased analysis of efficiency. Financial fraud could include for instance the overstatement of assets during the year which will result in large downward 4 th period adjustments of assets on financial statements (Healy et al., 2010). Concluding, the sample comprises 37 banks in total which lead to 144 observations given the time span of 2008 to The necessary financial ratios are collected via annual reports from the banks under consideration which can be found on their web pages. 8

9 3.3 Data description The main data needed to evaluate efficiency of banks are the financial ratios as stated in section 2.2. These financial ratios can be retrieved from annual reports of the Islamic and Conventional banks. I will use the same ratios that measure efficiency as Beck et al. used. These are: Overheads and Cost- Income Ratio to measure cost efficiency and ROA and the Equity ratio which measure profit efficiency, nevertheless, as already stated in section 3.1 the profit efficiency ratios are used in a different context in this paper compared to Beck et al.. Moreover, I will also include one more measure of profit efficiency which lack in the paper of Beck et al. namely, ROE as mentioned in section 3.1. These financial ratios are already defined in section 2.2. Moreover, several dummy variables are included in the regression model. One is labeled Islamic which takes the value one if the bank is Islamic and the value zero if the bank is a Conventional one. The other three dummy variables have the label Large, Medium and Small which indicates the bank size. A detailed analysis about the descriptive statistics can be found in table 1 below. Variable Mean Std. Deviation Minimum Maximum Overheads Cost-to-Income Ratio ROA Equity Ratio ROE Total Assets (In Millions) Table 1 As can be seen from the table the ratios of cost and profit efficiency as well as asset size vary significantly between the banks under consideration in Malaysia. The overhead ratio of Islamic and Conventional banks is on average fairly low with 1,32 percent. Also there is not a lot of variation between Malaysian banks which is indicated by a low standard deviation of 0,52 percent. However, the Cost-to-Income Ratio differs a lot more between the banks under consideration given the high standard deviation of 13,61 percent. The average of the Cost-to-Income Ratio is with 34,36 percent relatively high. Looking at these cost efficiency ratios the Cost-to-Income Ratio is particularly interesting since it varies a lot between the banks under consideration. The substantial variation of the Cost-to-Income Ratio among banks is a sign that the interest income of banks in relation to operating cost differs significantly as compared to relation of total assets and operating costs which is captured by the Overheads ratio. The reason for this difference could be found in the Islamic concept of banking which is based on no interest as compared to the concept of Conventional banking. 9

10 The profit efficiency ratios are all fairly similar apart from the ROA which has an average of 0,85 percent and a standard deviation of just 0,81 percent. The equity ratio and ROE are in the same range around 10 percent average, with the equity ratio having an average of 10,32 percent and the ROE of 10,12 percent respectively. Also the variation in both ratios is rather similar given the standard deviation of 5,38 percent for the Equity ratio and 8,31 percent of the ROE. As can be seen from the table 1 the ROA and ROE have negative minimum values which is quite surprising. The negative values are mainly due to banks that started business in 2008 which are included in this sample. These banks do not generate enough revenues to cover their current operating expenses which yields a negative net income and hence a negative ROE and ROA ratio. Moreover, net income seems to differ substantially between banks which is indicated by the moderately high standard deviations of ROA and ROE with respect to the average of these ratios. This might be a consequence of how long a bank already operates in Malaysia and thus of how large the bank is. Large banks could generate more profits than small banks and hence have higher ROA and ROE ratios. Total assets differ notably between the different banks. The average lays around 99 Million, but the high standard deviation of 354 Million shows that there is a lot of variation. All signs are that banks are differing considerably in terms of size, which can be measured by total assets. 3.4 The Model Since the analysis includes 37 different banks and consist of a time span of four years a panel data regression is required (Stock and Watson, 2011). To analyze bank efficiency in Malaysia two different econometric regressions will be used. The first regression just looks at the effect of being an Islamic bank on being more profit or cost efficient and has a similar set up as the regression used by Beck et al.. It is run for each of the different financial ratios which represent the dependent variables. Hence, the subsequent econometric model can be set up: The following specifications apply to this model: = Overheads, Cost-Income Ratio, ROA, Equity Ratio and ROE = Islamic 1 if Islamic bank and 0 if Conventional bank The subsequent assumptions apply to this regression: Firstly, and are independent and identically distributed. Furthermore, is assumed independent of and D. Secondly, these variables 10

11 are assumed independent of each other for all i and t (Park, 2005, p.19). Lastly, it is assumed that the panel data is unbalanced since there are 4 missing values as outlined in section 3.2. Having set up the assumptions a generalized least squared (GLS) regression can be run in the statistical program Stata since the variance is known. The point of interest in this econometric regression lies in the coefficient. If it significantly contributes to cost and profit efficiency, then there is sufficient evidence to assume that Islamic banks have superior/inferior cost or profit efficiency than Conventional banks in Malaysia. The second regression attempts to control for bank size. As Beck et al. (2013) mention, larger banks could be more efficient due to the scale economics that these banks build up as compared to smaller banks. Moreover the authors state, larger banks have easier access to the wholesale market than smaller ones (p ). Consequently, the second econometric regression is set up in the following way: The subsequent specifications apply to this model: = Overheads, Cost-Income Ratio, ROA, Equity Ratio and ROE = Islamic 1 if Islamic bank and 0 if Conventional bank = Small 1 if the bank s asset size is smaller than 6.12 Million and 0 otherwise = Medium 1 if the bank s asset size is between 50.1 Million and 6.12 Million and 0 otherwise = Large 1 if the bank s asset size exceeds 50.1 Million and 0 otherwise There is one fundamental difference to the regression set up of Beck et al.: I assume that also the size of Conventional banks matters whereas Beck et al. just looked at different Islamic sized banks. Hence I include two more variables namely and which control for the size of Conventional banks. The borders to differentiate whether the bank is large, medium or small were chosen on the basis of percentiles of total assets which is the same methodology as in the paper of Beck et al. (2013, p.441). If the bank s total asset lies in the lower 25 percent of the sample it is considered to be small. If it is between the 75 percent and 25 percent of the sample the bank is considered medium sized and if total assets lie in the percentage above 75 percent it is considered large. A detailed table about the percentiles can be found in appendix 1. The assumptions set up in the first regression concerning the variables and the regression itself are also applying to this regression. Hence also in this model a GLS regression can be run in Stata as well. The points of interest here are the coefficients, and since they determine which Islamic banks drive the difference between superior/inferior cost and profit efficiency over Conventional 11

12 banks. In a nutshell, the different sized Islamic and Conventional banks can be compared directly given this regression. It should be noticed that, given the setup of this regression, small Conventional banks are seen as the benchmark of this regression represented by the constant. 3.5 Fixed VS Random effects When applying panel data regression, a differentiation between random and fixed effects needs to be made. Fixed effects regression is a method for controlling for omitted variables in panel data when the omitted variables vary across entities but do not change over time. Moreover it assumes differences in intercepts across groups or time periods. Random effects, however, looks at differences in error variances by assuming the same intercepts and slopes (Stock and Watson, 2011, p. 396). In this analysis a random effects regression is used since fixed effects cannot be applied with the regression set up in this paper. This has one main reason that has to do with the dummy variables regression involved in this analysis. Fixed effects panel data subtracts from each variable the total average of that variable. The problem appears when a dummy variable does not change over time since then the average is the actual number and you would obtain zeros for that variable after subtracting the average which makes the fixed effect panel data regression useless. This problem occurs in the regressions set up in section 3.4. The dummy variable Islamic will not change over time since the bank is either Conventional or Islamic forever. Hence, when subtracting the average from the variable Islamic you would obtain zeros and a fixed effect panel data regression cannot be applied. This is the case for both regressions. The statistical program Stata recognizes that and does not give any output when a fixed effect panel data regression is run for the regression in section 3.4. Thus, a random effect panel data regression needs to be used with the assumptions as stated in section 3.4 to overcome the problems with the dummy variables in the fixed effect regression. Unfortunately this paper can just provide an intuitive approach towards the issue of Fixed versus Radom effects since the mathematical reasoning lies outside the scope of the writer. 3.6 Hypotheses Since this paper follows up on the methodology used by Beck et al., I expect to find similar results as these researchers. Beck et al. found that Islamic banks in Malaysia have significantly higher Cost-to- Income ratios than Conventional banks implying lower cost efficiency for Islamic banks compared to Conventional banks without controlling for size of the banks. Furthermore, Islamic banks are less profitable than Conventional banks without controlling for size which is, however, not significant (Beck et al., 2013, p.442). 12

13 When controlling for asset size Beck et al. found that especially small Islamic banks drive the difference between Islamic and Conventional bank efficiency (Beck et al., 2013, pp ). Thus, I expect that asset size matters and that the small banks in Malaysia mainly drive efficiency differentials between Islamic and Conventional banks. 4. Results Given the methodology to analyze efficiency differentials between Islamic and Conventional banks outlined in the third section, this section intends to present the results from the two regressions presented in section st Regression The following GLS regression was run via the statistical program Stata with the data provided in section 3.3. The outcomes of the regressions are presented in table 2 below. The table should be read in the following way: On top are the five dependent variables and on the left side the independent variables are stated. Hence, the table should be read from the top to the bottom. The numbers in brackets represent the p-values of the coefficient which is an indicator of how significant that coefficient is. Overheads Cost-to-Income ROA Equity Ratio ROE Constant Islamic Dummy (0.725) (0.045) (0.244) (.031) Overall Table nd Regression In the second regression I will control for bank size as described in section 3.4. The following GLS regression is thus run in the statistical program Stata. The outcome of these regressions is presented in table 3 and should be read in the same way as table 2. Also the same specifications stated in section 4.1 apply to table 3. 13

14 Overheads Cost-to-Income ROA Equity Ratio ROE Constant (0.008) Islamic Dummy *Small Bank (0.725) (0.064) (0.199) (0.965) Islamic Dummy *Medium Bank (0.905) (0.081) (0.003) (0.424) (.019) Islamic Dummy *Large Bank (0.636) (0.898) (0.008) (0.180) (0.035) Medium Bank (0.501) (0.447) (0.313) (0.003) Large Bank (0.633) (0.510) (0.517) (0.005) Overall Table 3 5. Analysis After having presented the results of the GLS regression, which was established in section 3.4, an analysis of the obtained results is necessary to draw any conclusions about efficiency differentials between Islamic and Conventional banks in Malaysia. Thus, this section starts off by analyzing the obtained results from the regressions of section 4.1 and 4.2 given a significance level of 5 percent. Subsequently, limitations of the two GLS regressions are outlined. 5.1 Analysis of the 1 st regression The analysis is carried out by applying a significance level of 5 percent which is common in statistical analysis. The regression in section 4.1 attempts to provide general insights into superior or inferior cost and profit efficiency of Islamic banks over Conventional banks in Malaysia without controlling for any other factors that potentially could affect bank efficiency. Table 2 provides an overview of cost and profit efficiency differentials between Islamic and Conventional banks. As can be seen from table 2 Islamic banks are significantly more cost efficient with respect to the Cost-to-Income Ratio than Conventional banks. In numbers that means that the Cost-to-Income Ratio is 7.87 percentage points below the one of Conventional banks which shows superior cost efficiency of Islamic banks. Moreover, given the P-value of 4.5 percent the result is significant with respect to the 5 percent significance level. Nevertheless, the R 2 of the regression is with 6.93 percent quite low. By looking just at the Overheads no conclusions about superior/inferior cost efficiency of Islamic banks can be drawn since the results are one the one hand not significant and on the other hand the differences between Islamic banks and Conventional banks is far too small with percentage points. This result contradicts with the findings of Beck et al. (2013) which found 14

15 that Islamic banks are less cost efficient than Conventional banks implying a positive coefficient (p.442). This analysis shows that the reverse is true given the Cost-to-Income ratio. All three profit efficiency ratios imply that Islamic banks are less profit efficient than Conventional banks. Especially the ROA and ROE yield significant results with a P-value of and 3.1 percent respectively. By looking at the ROA it can be concluded that Islamic banks have a ROA which is 0.6 percentage points below the one of Conventional banks which yields not much of a difference. However, by looking at the ROE the difference becomes a lot more visible. Islamic banks ROE is 4.64 percentage points below Conventional banks. This shows that Islamic banks are a lot less profit efficient than Conventional banks. Also the regression with the equity ratio shows that Islamic banks are less profit efficient than Conventional banks, however, this result is not significant given the high P-value of 24.4 percent. However, the R 2 of all three regressions is fairly low since it lies between 18.7 and 8.6 percent. Nevertheless, this result coincides with the result of Beck et al. who found that Islamic banks are less profit efficient than Conventional banks. 5.2 Analysis of the 2 nd regression The analysis of the second regression, as stated in section 4.2, is carried out on the same basis as the first regression by applying a significance level of 5 percent. The regression attempts, apart from providing general insights into superior or inferior cost and profit efficiency of Islamic banks over Conventional banks in Malaysia, also to control for size of the banks which, as Beck et al. stated affects bank efficiency. Moreover, it can be determined which Islamic banks in terms of size drive the superior cost efficiency of Islamic banks and the inferior profit efficiency. It should be noticed that small Conventional banks are seen as a benchmark in this regression. The results of the regression can be found in table 3. In terms of cost efficiency the Overheads again yields statistical insignificant results since the P-values are all far above the significance level of 5 percent. Furthermore, the coefficients are far too close to zero to see any differences between Islamic and Conventional banks. Conversely, the Cost-to-Income Ratio yields some interesting insights into cost efficiency differentials between Islamic and Conventional banks. Given the results all Islamic banks are more cost efficient than Conventional banks. The Cost-to-Income Ratio of small Islamic banks is 1.06 percent below the one of small Conventional banks. Also medium sized Islamic banks are 7.7 percent more cost efficient with respect to the Cost-to-Income Ratio than medium sized Conventional banks. Large Islamic banks as well are 1.01 percent more cost efficient given the Cost-to-Income Ratio than large Conventional banks. However, this result is not statistically significant since the P-value slightly exceeds the significance level of 5 percent for small and medium sized Islamic bank with 6.4 and 8.1 percent respectively. Especially the P-value of large banks is statistically insignificant due to the high number of 89.8 percent. Furthermore, the overall R 2 for the Cost-to-Income Ratio is quite low with a 15

16 value of Nevertheless, if a significance level of 10 percent would be applied the coefficients of small and medium sized Islamic banks would yield statistically significant results. In terms of profit efficiency all efficiency ratios indicate that Islamic banks of any size are less profit efficient than Conventional banks which is the same conclusion as drawn from the first regression. When looking at the ROA ratio, it becomes visible that all Islamic banks are less profit efficient than Conventional banks as indicated by the negative coefficients. Small Islamic banks are having a ROA which is 0.4 percent below the one of small Conventional banks, medium Islamic banks ROA is 0.67 percent below the one of medium sized Conventional banks and the one of large Islamic banks is 1.19 percent below the one of large Conventional banks. Thus, all sized Islamic banks have an ROA which is below the one of all sized Conventional banks. The difference between the ROA s of the banks seems quite small but it should be noticed that the ROA s average with 0.85 percent is rather low and thus the percentage difference between the banks is actually quite big. Nevertheless, the coefficients of small Islamic bank is not statistical significant at a level of 5 percent given the P-value of 19.9 percent. Also, the equity ratio yields analogous results as the ROA ratio in terms of profit efficiency differentials. Small, medium and large Islamic banks are all having an equity ratio which is below the one of small, medium and large Conventional banks with values of 7.71, 1.16 and 3.7 percent respectively. Here, however, just the coefficient of small Islamic banks yields statistically significant results at a significance level of 5 percent. In terms of the ROE very similar results can be drawn as from the regression of the ROA and equity ratio. In addition, all coefficients apart from the one of small Islamic banks are statistical significant at a level of 5 percent. Speaking in numbers small Islamic banks are having a ROE which is 5.99 percent below the one of small Conventional banks, also medium and large Islamic banks are having ROE s below the one of medium and large Conventional banks with 5.53 and 9.72 percent respectively. Lastly, it should be noticed that the overall R 2 of the profit efficiency ratios are again quite low with for the ROA, for the equity ratio and for the ROE regression. Given the correlation matrix presented in appendix 8.2 it can be seen that the issue of the dummy variable trap, perfect multicollinearity, does not apply to this regression since the correlations between the different dummy variables are too low. The correlation between Islamic*Medium and Medium is with the highest but still there is no perfect multicollinearity visible. The results of cost and profit efficiency discussed above coincide partly with the findings of Beck et al. who found that mainly small banks drive the difference between Islamic and Conventional bank efficiency. On the one hand, for cost efficiency the regression showed that the difference between Islamic and Conventional banks is caused by all Islamic banks which is, nevertheless, not statistical significant at a level of 5 percent but, however, at a level of 10 percent for small and 16

17 medium sized Islamic banks. It is worth noticing that the highest percentage difference between Islamic and Conventional banks is contributed by the medium sized Islamic banks with 7.7 percent whereas the small and large Islamic banks just differ by approximately 1 percent from small and large Conventional banks. Hence, medium sized Islamic banks seem to be the driver for the superior cost efficiency of Islamic banks. On the other hand, the profit efficiency regressions illustrated that the difference between Islamic and Conventional banks is caused by all Islamic banks as well. When looking at the ROA and ROE ratio it can be found that especially large Islamic banks drive the difference since they show the biggest difference to large Conventional banks with 1.19 and 9.72 percent respectively. However, the Equity ratio shows that small Islamic banks drive the change given the high percentage difference of 7.17 percent to small Conventional banks. Concluding, not just small Islamic banks drive the difference between Islamic and Conventional banks as stated by Beck et al.. The analysis above showed that in terms of cost efficiency especially medium sized banks play an important role. In terms of profit efficiency the regression demonstrated that the difference is mainly driven by large and small Islamic banks. 5.3 Limitations There are several problems that come along with the regression results mentioned in section 4 as well as with the analysis in section 5.1 and 5.2. Firstly, there are more factors affecting bank efficiency which are not taken into consideration in the regressions presented in this paper. Beck et al. (2013) mention for example external financial shocks, such as the financial crisis, that affect efficiency of banks. Moreover, the authors also control for bank characteristics which differ between amongst banks and which affect efficiency of banks. Lastly, Beck et al. also consider different market shares of banks. The reason is the higher the market share for Islamic banks might lead to higher potential efficiency of the Islamic banks. This is due to the fact that a higher market share of Islamic banks implies a more established sharia compliant finance which leads to repercussions with respect to regulations and competitive responses of Conventional banks (p.443). Secondly, Arslan and Ergec (2013) found that interest rates affect Islamic bank and Conventional bank efficiency even though Islamic banks are supposed to carry out interest free banking. This is evidence that the murabaha syndrome is an issue in Islamic banking and hence should be taken into consideration. Interest rate changes by the Central bank thus have an effect on deposit rates as well as on loan rates of Conventional but also the instruments of Islamic banks. Concluding, both bank types are affected by interest rate changes and therefore when measuring efficiency of Islamic and Conventional banks also interest rates should be taken into consideration. Thirdly, as Chortareas et al. (2012) financial freedom of banks also affects efficiency of banks. They found that banks having more financial freedom are more cost efficiency and, moreover, are 17

18 having advantages in overall efficiency. Hence financial freedom should also be taken into consideration when measuring bank efficiency. Lastly, the set up of the regression itself causes some trouble when the assumptions stated do not hold for the random effect regression described in section 3.4. Especially the assumption that is assumed independent of and D is crucial for a random effect regression. However, this might not necessarily be the case since there are a lot more factors affecting efficiency as stated above. These are all included in the error term and thus it is well possible that they are correlated with the variables used in this regression. This would imply that a random effect regression cannot be used and a fixed effect regression needs to be applying which is, however, in this setup not possible as stated in section Conclusion The aim of this paper was to investigate whether Islamic banks are more efficient with respect to cost and profits than Conventional banks in Malaysia. The issue is addressed by applying a GLS regression with cost efficiency and profit efficiency ratios as dependent variables. In the first regression just a dummy variable for Islamic banks is used as independent variable and in the second regression a combination of bank size and the dummy variables is applied to control for bank size. The results are partly congruent with the current research done in the field of Islamic banking. Generally, the research found that Islamic banks are statistically more cost efficient than Conventional banks. When controlling for bank size it was found that especially medium sized Islamic banks drive the superior cost efficiency of Islamic banks even though the contribution of medium sized Islamic banks is not significant at a significance level of 5 percent but, nevertheless, at a significance level of 10 percent. When looking at profit efficiency it was found that overall Islamic banks are less profit efficient along all three financial ratio measures used in this research. In terms of profit efficiency it can be concluded that small banks seems drive the inferior profit efficiency of Islamic banks when looking at the equity ratio whereas when looking at the ROE and ROA large Islamic banks are especially less profit efficient than large Conventional banks. These results are significant at a level of 5 percent. However, a major limitation of this research is the omission of potential factors that affect bank efficiency such as interest rates, financial freedom, external financial shocks, bank characteristics or market share which need to be addressed in future research in this field. Moreover, the panel data regression applied underlies a lot of assumptions which can cause problems when they are violated. Thus, the regression should be further specified in a way that a fixed effect regression analysis can be applied. Hence the results obtained by this research should be applied with caution by the reader. Future research should address the issues mentioned above and compare the results with the ones obtained by this research to see whether controlling for several other factors leads to different conclusions of efficiency differentials between Islamic and Conventional banks in Malaysia. 18

19 7. Bibliography Abduh, M., Hidayat, S.E., Does Financial Crisis Give Impacts on Bahrain Islamic BankingPerformance? A Panel Regression Analysis. International Journal of Economics and Finance, 4(7), pp Arslan, B., Ergec, E., Impact of interest rates on Islamic and conventional banks: the case of Turkey. Applied Economics, 45(17), pp Bahru, J., Banking on the ummah. The Economist. Retrieved from: May. Bank Negara Malaysia, List of licensed banks in Malaysia. Retrieved from: June. Beck, T., Demirgüc-Kunt, A., Merrouche, O., Islamic vs. conventional banking: Business model, efficiency and stability. Journal of Banking & Finance, 37(2), pp Berger, A.N., Mester, L.J., Inside the black box: What explains differences in the efficiencies of financial institutions? Journal of Banking & Finance, 21(7), pp Chortareas, G.E., Girardone, C. and Ventouri, A., Financial freedom and bank efficiency: Evidence from the European Union. Journal of Banking & Finance, 37(4), pp Ding, Y., Lebas, M.J., Stolowy, H., Financial accounting and reporting a global perspective. Hepshire: Cengage Learning. Hanif, M., Differences and Similarities in Islamic and Conventional Banking. International Journal of Business and Social Sciences, 2(2), pp Healy, P.M., Palepu, K.G., Peek, E., Business Analysis and Valuation. Hepshire: Cengage Learning. Khan, F., How Islamic is Islamic Banking? Journal of Economic Behavior & Organization, 76(3), pp Needles, B.E., Powers, M., Financial accounting principles. Hepshire: Cengage Lerning. Park, H.M., Linear regression models for panel data using SAS, Stata, LIMDEP, and SPSS. Indiana University. Qureshi, M.A., Shaikh, M., Efficiency of Islamic and Conventional Banks in Pakistan: A Non-parametric Approach. International Journal of Business and Management, 7(7), pp Srairi, S., 2010, Cost and profit efficiency of conventional and Islamic banks in GCC countries. Journal of Productivity Analysis, 34(1), pp Stock, J.H., Watson, M.M., Introduction to Econometrics. London: Pearson Education. Yousef, T.M., The murabaha syndrome in Islamic finance: laws, institutions and politics. In: Henry, C.M., Wilson, R. (Eds.), The Politics of Islamic Finance. Edinburg University Press, Edinburgh, pp