“Financial performance of conventional and Islamic banks in Bahrain: a comparative study”

The main purpose of this study is to identify the variables that influence the financial performance of both types of banks, Islamic and conventional, and compare their financial performance over the period of 2003–2016. Banks listed on the Bahrain Bourse as of December 31, 2016 were used in the study, with a total of seven banks, of which three are Islamic and four are conventional. To make an appropriate comparative study, financial ratio analysis is used. Multiple regression and paired sample t-test are used to analyze the data. Return on assets (ROA) and return on equity (ROE) are considered as the basis for measuring financial performance and are set as dependent variables. The analysis of the results shows that conventional banks perform better than Islamic banks in terms of profitability. The results also show that ROA is significantly related to risk, cost of intermediation and efficiency ratios, while ROE is highly influenced by risk ratios only. Moreover, it was found out that the relationship between asset size and the performance of banks is insignificant, while the relationship between the number of branches and both ROA and ROE is significant.


INTRODUCTION
Until the fourth quarter of the twentieth century, the entire financial sector throughout the world was operating on the interest basis, which is in conflict with the injunctions of Islam.This led to the development of interest free (Shari'a compliant) banking (Hanif, Tariq, Tahir, & Momeneen, 2012).The objective of Islamic banks is the same as that of conventional banks, i.e. to make business but only in the light of Shari'ah rules.Islamic banks usually do not assure any return on deposits; they only claim that deposits will be reinvested and profits or losses will be shared between the depositor and the bank.The rationale behind prohibition of interests and the importance of the distribution of profits and losses in Islamic banking creates the relationship of financial trust and partnership between borrower, a lender and an intermediary (Yudistira, 2004;Dar & Presley, 2000).
Islamic banks have income from financing that can be treated as net interest income for the sake of comparison.In addition, Islamic banks enter into different contracts with their customers, such as Murabaha, where the bank buys the commodity for its cash price and sells it to its customers allowing them paying by installments.The cash price will definitely be lower than the amount collectible from customers.The difference represents the bank's profit from such transactions.
Murabaha is not classified as a loan, nevertheless, it is a financial asset in most Islamic banks and quite similar to Accounts Receivable in an ordinary business balance sheet.
Researchers started to make comparisons between the financial performances of both types of banks using financial ratio analysis (Saleh & Zeitun, 2007).Previous research on the financial performance shows conflicting and sometimes similar results.For example, some studies show that Islamic banks outperform conventional banks in their financial performance (Samad, 2004 Aziz, Husin, & Hashmi, 2016).However, other researchers found that conventional banks perform better than Islamic banks in terms of profitability, liquidity, credit risk management and solvency (Ayub, Sohail, & Mumtaz, 2012;Fayed, 2013;Moin, 2013;Milhem & Istaiteyeh, 2015;and Srairi, 2010).
This study seeks to provide answers to the following questions: 1. What are the variables that influence the financial performance of Islamic and conventional banks? 2. Are there significant differences in the financial performance of both types of banks?
The main purpose of the study is to conduct a comparative financial performance evaluation of conventional and Islamic banks using ratio analysis.Each type of banks will be analyzed independently so that the financial performance can be compared.
The study is important to many stakeholders interested in the financial performance of banks.Shareholders need to know the performance of the bank to assess their investment potential and future investment strategies (Aziz, Husin, & Hashmi, 2016).Furthermore, financial performance indicators help regulators track the financial industry outlook and its future growth and challenges.The study also investigates the relationship between Return on Assets (ROA) and Return on Equity (ROE) and other financial performance indicators to provide resource allocation guidelines to the banks' managers based on statistical results, rather than relying only on assumptions and theories.
The next section reviews the relevant literature on commercial banks followed by the methodology section.Results and analysis are presented in section three and the last section presents conclusions, limitations and areas for future research.

Comparative studies of the performance of Islamic and conventional banks
Iqbal (2001) found that there were significant differences between Islamic and conventional banks in terms of performance.There were no significant differences in profitability ratios, but there was a significant difference in liquidity and risk and solvency ratios.Awan (2009) Aziz, Husin, and Hashmi (2016) found that Islamic banks' performance was better in terms of efficiency and return on asset quality, but Islamic banks were struggling for advances, investment, liquidity, deposits and capital as conventional banks performed better in these areas.Yudistira (2004), Rosley and Abu Baker (2003), Samad (2004), and Kamaruddin, Safa, and Mohd (2008) found that Islamic banks showed better credit performance than conventional banks.Siraj and Pillai (2012) found that conventional banks registered growth in revenue but could not achieve improved profitability because of higher provisions for credit losses.Javaid, Anwar, Zaman, and Gafoor (2011) examined the impact of assets, loans, equity and deposits on major profitability measures (ROA) and showed that these variables had a strong impact on banks' profitability.

Profitability, risk, cost of intermediation and efficiency
Risk is a more controversial area of study because dependence on leverage increases risk but potentially increases profitability because of the low cost of deposits.However, an increase in risk may be reflected by rising losses on loan provision, which will ultimately reduce profits.It is expected that competition is high in the financial market (Malhotra, Poteau, & Singh, 2011).Thus, empirical data analysis would show growing competition in the banking industry, which is measured by the perceived increase in cost of intermediation due to the accelerated number of banks entering the Bahraini market over the last two decades.

Asset size and number of branches
The size of the bank has a great impact on its financial performance.Large banks with large assets would show higher profits because the main assets of a bank are its loans and advances to customers.Since it is expected that the competition is high in the financial market, the most profitable banks are those with larger assets rather than those who impose higher interest rates.This is consistent with the results of Kadir, Jaffar, Abdullah, and Harun (2003) and Malhotra, Poteau, and Singh (2011).However, some studies found that the relationship between profitability and bank's size was insignificant (Rashid & Jabeen, 2016;Zeitun, 2012).Moreover, Akhtar, Ali, and Sadaqat (2011) found that asset size had a negative effect on both conventional and Islamic banks, although this impact was considered to be insignificant in Islamic banks but was significant in conventional banks.In addition, large banks tend to have higher interest income and have more control over non-operating expenses, yet less favorable liquidity position when compared to smaller banks (Malhotra, Poteau, & Singh, 2011).Finally, Williams (2003) and Zeitun (2012) found that the number of bank branches was positively related to ROA.
In summary, it is clear that there are contradictions in the findings of previous research, in particular in the financial performance of banks, their liquidity and risk management.This justifies the need to continue research in the area and to choose this specific topic for Bahrain.

Research hypotheses
Based on the review of previous studies, the following research hypotheses have been put forward.
H1: Islamic banks operating in Bahrain are more profitable than conventional banks.
H2: There is a significant relationship between profitability and risk, efficiency and the cost of intermediation.
H3: Competition among banks, as measured by the increase in the cost of intermediation, is increasing in Bahrain H4: There is a significant relationship between ROA and both asset size and the number of bank branches.

Population and sample
Only commercial banks listed on the Bahrain Bourse are included in the study.This is because listed banks are a matter of interest for current and potential investors.Moreover, the majority of unlisted banks are enormous banks that activate globally and there are vast differences in resources and the performance of these banks' branches outside Bahrain.Among all listed companies, there are seven commercial banks of which three are Islamic and four are conventional.

Procedure and instrument used
Data for the study are collected from the banks' consolidated financial statements over a 14-year period (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016).To measure the financial performance, profitability ratios are used and the combined means of profitability ratios for both types of banks are compared.The use of financial ratio analysis compensates for bank disparities in assets, capital, deposits and loans.
The consolidated financial statements were collected from the official websites of respective banks and of the Central Bank of Bahrain.Thomson Reuters Eikon and Zawya Islamic database offered by the University of Bahrain Central Library was also used in the data collection process.

The dependent variables of bank financial performance
Return on assets (ROA).Investing activities, in the commercial banking industry, are in fact considered the second most significant commercial bank activity.The time value of money forces banks to invest their customers' deposits because, in the absence of a creditworthy borrower, a bank cannot keep the deposits idle: if they do so they will end up paying the depositors' interest without benefiting from their funds.Thus, eliminating all investing activities from banks' income is irrational.Instead, considering the entire net profit is more reasonable.However, comparing banks' income without adjustments is inappropriate if there is a huge variance in asset size, as various researchers found bank size as positively correlated to bank profitability because the banks' assets may reduce operational cost (Kosmidou, 2008)

The independent variables
Independent variables are divided into four types of ratios, profitability, risk and liquidity, cost of intermediation, and efficiency (Malhotra, Poteau, & Singh, 2011).
Profitability ratios other than ROA are Operating Assets Turnover (OAT) and earning assets to total assets (EATA).The operating assets turnover ratio makes it clear the ability of loans to generate revenues regardless of their cost.High operating assets turnover shows that the bank generates a reasonable amount of revenues through its assets.
Hence, if high OAT corresponds with low ROA, it is suggested that the bank could not control its operational expenses or its intermediation costs.Moreover, OAT can measure if the bank generates its revenues from lending large volume of loans or by imposing high interest rates.
Earning assets to total assets (EATA) is calculated by dividing loans plus investment over total assets.The aim of calculating this ratio is to see what percentage of assets is income producing.Usually, banks will maintain high percentage of earning assets unlike other sectors such as manufacturing where higher investment will be in plant assets.Earning assets are those assets that an entity could generate through holding or possessing them.The key feature that distinguishes earning assets from non-earning assets is the fact that an earning asset produces funds greater than its book value.Investment shares are considered one of the most common examples of earning assets, as these shares usually earn dividends.In the banking in-dustry, loans and advances represent the vast majority of earning assets, as banks receive interest as a return to hold such assets.A higher EATA ratio indicates higher profitability, thus better performance.However, banks operating in the growing phase of operations tend to have lower EATA ratio.As they establish new branches, their property, plant, and equipment value, which are non-earning assets, will increase.
Ratios that assess risk and liquidity.Ratios that assess risk are equity to assets (ETA), loans to deposits (LTD) and net loans to gross loans (NLTGL).
Loans to deposits (LTD) is one of the most controversial ratios; a lower LTD indicates a better liquidity position.Yet a very low LTD suggests that the bank is underutilizing its resources as it has a lot of funds from deposits and a low amount of loans.Regulators' attention to this ratio is steadily increasing.Therefore, it is interesting to measure the relationship between LTD and the bank's performance measured through ROA.NLTGL measures the credit risk control for a bank.A ratio closer to 1 means that the bank minimizes its uncollectable loans; therefore, the bank manages to finance creditworthy borrowers.
Cost of intermediation ratios.

Regression model
To analyze the relationships among performance variables, the Ordinary Least Squares (OLS) method was used.The OLS regression model can be written as: where Y is the dependent variable, i is an index of cross-sections, that is data of different subjects taken for a specified period of time, while t refers to time.On the other side of the equation, β0 or sometimes referred to as α is a 'constant term' meaning that if all independent variables equal zero, Y equals β0.While x s represents the independent variables, β is the coefficient of each independent variable, and, finally, µ symbolizes error.Dependent variable ROA; i represents the seven banks chosen for the study, t is the period of time from 2003 till 2016, and the x s are the independent variables.Thus the model takes the following form: A dummy variable is also included to control the effect of the bank type (Islamic = 1, conventional = 0).It should be noted that this panel is unbalanced.A balanced panel is when t is identical for all i s ; meaning that banks are examined in the exact period of time.In this study, this is not quite the case, as two banks, Khaleji Commercial Bank and Al Salam Bank, both started operating after 2003, so years (t) will be different for these two banks.

Hypotheses test
H1: Islamic banks operating in Bahrain are more profitable than conventional banks.
Table 1 shows the statistical summary of each category of financial performance indicators during the period of study.Four ratios are used to assess the profitability of banks: ROA, ROE, Operating assets turnover (OAT), and earning assets to total assets (EATA).The descriptive statistics show that conventional banks had a higher mean in all of these ratios except operating assets turnover.It is evident that the operating assets turnover ratio overlooks the cost associated with assets, instead it measures the ability of assets to generate income regardless of expenses.Since Islamic banks had higher operating assets turnover ratio compared to conventional banks and all other profitability indicators were lower in Islamic banks, it can be assumed that Islamic banks had the ability of generating more income through their assets.
Given the standard deviation, it is observed that the trend of conventional banks' income is more stable.Some Islamic banks had incurred losses in certain years, while all conventional banks under study showed profits in the income statements for all 14 years; this justifies the narrower range of profitability ratios in conventional banks.Besides, there is a significant difference in the ratio of earning assets to total assets in the two types of banks.
Earning assets as a share of total assets in conventional banks are higher than it Islamic banks.This gap is also reflected in ROA.The results show that Islamic banks are less profitable than conventional banks.Therefore, the null hypothesis that Islamic banks are not more profitable than conventional banks cannot be rejected.A possible explanation H2: There is a significant relationship between profitability and risk, efficiency and cost of intermediation.
A positive relationship exists between profitability and both efficiency and cost of intermediation.
Table 1 shows that conventional banks report higher reliance on leverage compared to Islamic banks.On average, equity represents 20% of total assets of Islamic banks for the 14-year period, while in conventional banks, equity is only 11%.The capital structure of conventional banks is more risky, yet the pattern is similar for all observations.Conversely, although Islamic banks showed dissimilar forms of capital, some Islamic banks showed an extremely safe capital structure with 55% equity, although in some years Islamic banks showed a much higher risk with 5% equity.
Both types of banks showed a decent risk management returned in minimizing Loss on Loans' provision.The ratio of net loans to gross loans showed that Islamic banks write-off 5.5% of their loans as uncollectable, and conventional banks writeoff 4% of their loans on average.The minimum amount of NLGL in Islamic banks is 79%, which is below average.However, it was a typical figure that cannot be generalized as the mean remained reasonably high.
As to the loans to deposits ratio, it is obvious that Islamic banks had a better liquidity position than conventional banks throughout the period of study.Nevertheless, it seems that Islamic banks underutilize their assets; as liquidity is available through the large among of deposits, loan volume is relatively low.Assembling performance variables emphasizes the proposition of underutilizing assets in Islamic banks.As previously discussed, Islamic banks show low earning assets to total assets ratio; however, associating this with the high loans to deposits ratio implies that Islamic banks raise large amounts of funds from deposits but they do not invest them in loans, and as cash is not considered as an earning asset, Islamic banks had lower EATA ratios.The reason for this investment shortage is probably the restriction imposed on Islamic banks' activities to maintain adherence to Islamic Sariah.
When considering the influential determinants of bank performance, the results show a significant correlation between profitability and risk; as risk increases, profitability decreases.While the efficiency and cost of funding ratios significantly affected ROA, its impact on ROE is insignificant.However, there is no significant relationship between asset size and both ROA and ROE.Therefore, the null hypothesis cannot be rejected.This result agrees with Rashid and Jabeen ( 2016), but it contradicts the results of Kadir et al. (2013) and Malhotra, Poteau, and Singh (2011).
H3: Competition among banks, as measured by the increase in the cost of intermediation, is increasing in Bahrain.
Table 1 shows that Islamic banks have the advantage of lower financing costs when compared to conventional banks.However, when both interest expenses and interest income are considered, conventional banks generate more income than Islamic banks.Moreover, net interest margin, which is the net interest income divided by operating assets, also supports the assumption of asset underutilization in Islamic bank, as NIT is much higher in conventional banks.Net interest income (interest spread) of conventional banks is almost three times more than the interest spread of Islamic banks.Based on this result, the null hypothesis that competition among banks as measured by the increase in the cost of intermediation is decreasing in Bahrain cannot be rejected.
H4: Asset size and the number of branches have a significant relationship with ROA and ROE.
Conventional banks were found more efficient than Islamic that is, running their business more smoothly by minimizing non-interest expense.Table 1 shows that Islamic banks incur BD 0.4 as non-interest expenses against every dinar earned as revenues, while conventional banks incur only BD 0.3.This result supports the theory that banks' asset size has a positive relationship with non-interest expenses.The ratio of non-interest income to operating income shows that Islamic banks depend on interest income in generating their revenues more than conventional banks, whereas conventional banks depend more on interest income.It is important to emphasize that income from Islamic financing activities is considered as interest income, and Islamic financing assets are treated as loans for the sake of comparison.A typical form of non-interest income is dividends from investees.The reason for Islamic banks' inefficient operations is probably the lack of experience and smaller asset size compared to conventional banks.Therefore, the null hypothesis is rejected.

Regression model and multicollinearity test
Multicollinearity test was undertaken to ensure that there was no correlation between the independent variables in the regression model.Table 2 shows the correlation coefficients.
Table 2 shows that high correlations exist among three variables (above 0.7).This points to the necessity to adjust the model as some of its independent variables are significantly correlated.There is a significant positive relationship between Operating Assets Turnover (OAT) and Net Interest Margin (NIM) with a correlation coefficient of 0.94.This result suggests that banks gener-ating greater revenues from their operating assets (mainly loans) tend to have higher interest margins; suggesting that NIM is greatly influenced by interest income rather than interest expense.Both ratios focus mainly on the ability of assets to generate income, and the statistics show an extremely positive relationship between interest income and net interest income.One of these variables is redundant, thus it should be eliminated from the model to avoid collinearity.Accordingly, OAT is excluded.
Also, there is a 0.95 correlation coefficient between asset size and interest spread, which means that larger banks tend to generate higher income than smaller banks, as they have more resources.In rare cases, bank size does not influence the interest spread, as banks hold large amount of funds financed through interest-bearing deposits to maintain liquidity.Those funds increase the size of the bank, yet it reduces interest spread as banks are obligated to pay interest expense to the depositors.However, this is not the case in Bahraini banks.Therefore, interest spread (INTS) is left out of the model as the size of the bank represents similar criteria.
Finally, a negative correlation between the type of a bank and the number of branches is also observed (-0.89).Islamic banks have lover number of branches compared to conventional banks.Bank type and the number of branches are two different criteria, and they both have an adequate value to add to the research, thus eliminating one of them will be inappropriate.Therefore, the panel

Model (1): Regression analysis of financial performance indicators, excluding the effect of bank type, with ROA as a dependent variable
Table 3 clearly shows the strength of the model used, as standard error is relatively low and the number of observations (93) is quite sufficient (generally, observations above 30 are considered statistically sufficient).Moreover, the adjusted R-squared indicates that 69% of the ROA variation is explained by the variation of the independent variables.Moreover, from the regression statistics, the validity of the model can be found because it proves the absence of collinearity, since the tolerance factor (1 -R 2 ) is more than 20%; that is about 28%.In addition, the F-test for this model is also statistically significant (Table 4).The figures in Table 4 also ensure the validity of the model used, as F-value indicates that at least one independent variable has influenced the dependent variable.
Table 5 shows that only four out of ten chosen variables were insignificant in determining bank performance and six variables have significantly influenced ROA.R-squared of the 0.467 (see Table 3) variation in profit is due to the independent variables and the remaining 0.54 variation in profit is due to other variables not included in the study.
The ratio of earning assets to total assets is positively related to ROA as banks increase their earning assets and are expected to earn more on their assets.The equity to assets ratio has a similar impact on ROA.The regression analysis suggests that a capital structure, relying on equity rather

Model (2): Regression analysis of financial performance indicators, excluding the effect of bank type, with ROE as a dependent variable
Similar to model (1), the statistics in Table 3 demonstrate the validity of the model used, since the tolerance is 0.54, which provides no collinearity.Independent variables explain about 0.40 of the variation in ROE and F-value shows that at least one independent variable has a relationship with ROE (Table 4).
Table 5 shows that ROE has a positive relationship with the earning assets to total assets ratio, the net loans to gross loans ratio, and the number of branches.All these variables have influenced ROA; however, net interest margin, equity to assets, and effiency ratios had a statistically significant influence on ROA, though their influence on ROE was immaterial and insignificant.

Model (3): Regression analysis of financial performance indicators, excluding the effect of the number of branches, with ROA as a dependent variable
Similar to the previous two models, Table 3 proves the model (3) validity, which has a high tolerance level of 1-0.74.The F-value also indicates the mod-el legitimacy.R-squared shows that about 70% of the ROA variance is explained by the model.Compared to model (1), where bank type was excluded, model (3) has a slightly higher R 2 .
Model ( 3) is almost identical to model (1), the only difference is that model (3) measures the effect of the bank's type instead of the number of branches.The dummy variable shows that conventional banks are more profitable.The reason for this is that Islamic banks underutilize their assets.

Model (4): Regression analysis of financial performance indicators, excluding the effect of the number of branches, with ROE as a dependent variable
Based on the statistics provided in Tables 3, 4 and 5, the model is in analyzing regressions among dependent and independent variables, since the tolerance is high enough, and F-value is significant.Moreover, Table 5 shows that the only variables that affect ROE are net loans to gross loans ratio and the bank type.Net loans to gross loans being positively related is the only common result in all four implied models, which emphasizes the importance of adequate risk management to retain high profitability levels.Conversely, the dummy variable negative relationship with ROE indicates poor performance of Islamic banks.

CONCLUSION
The study shows that Islamic banks are less profitable than conventional banks.A possible explanation is the superior experience of conventional banks due to their longer existence on the market.The results also show that Islamic banks have higher liquidity than conventional banks.This is an expected result as Islamic banks face a lack of investment products and limited investment opportunities and an advance position resulted from the interest prohibition.Moreover, Islamic banks cannot rely on borrowing money from the Central Bank due to interest prohibition in Islam.This is in addition to the larger amount of restrictions and limitations imposed on Islamic banks in terms of financing activities.Finally, Islamic banks are in a growing stage and they try to benefit more from the fund commitment.
Given these results, it can be concluded that Islamic banks have liquidity excess.On the other hand, this may indicate that Islamic banks are in a growing stage and try to get more benefits from the commitment of the funds (Usman & Khan, 2012).
The results also show a significant relationship between profitability and risk.Although efficiency and cost of funding ratios significantly affected ROA, their impact on ROE is insignificant.Thus, sound risk management and solvency indicate the strength of this type of banks to pay their debts.As to the asset size, there is no significant relationship between both ROA and ROE.This result is consistent with Rashid and Jabeen (2016).However, it contradicts the findings of Kadir et al. (2013) and Malhotra, Poteau, and Singh (2011).The number of branches is proved to have a significant influence on bank performance.
The comparison also shows that profit share given to Islamic banks' depositors is less than that of conventional banks.Perhaps this is contrary to the essence of the Islamic moral economy, where the distribution of profits must be based on justice and fairness.Instead, owners make the bulk of the profits (Aziz, Husin, & Hashmi, 2016).Thus, Islamic banks need to address this issue to increase their credibility as being fair and just.
Finally, since the financial crisis falls within the period covered by the study, it is worth noting that Islamic banks have been less affected by the financial crisis than conventional banks.Despite the decline in revenues, the impact was minimal compared to changes faced by conventional banks.Thus, it can be argued that the financial crisis has affected the performance of bank groups at different levels.
Despite the results above, the effect of bank size in both types of banks must be considered when interpreting results and making decisions based on them.The study lacks any primary data due to time constraints and difficulty in obtaining such data.In addition, the study sample includes more conventional banks than Islamic banks, and this may impose limitations on generating more accurate estimates of performance comparisons.
As Islamic and conventional banks are different in nature, the factors affecting their performance can also differ.Accordingly, a further study that examines Islamic banks and conventional banks using two different models is highly recommended.In addition, the financial performance of banks was influenced by the global financial crisis.Further research can examine the effect of the crisis on the performance of commercial banks in Bahrain.
Due to the excess liquidity of Islamic banks, they should increase their ability to use surplus cash to generate additional returns.As such, the Central Bank of Bahrain should help Islamic banks invest excess liquidity.
(Samad, 2004uity (ROE) is of great concern to investors and shareholders.ROE measures the income generated by the bank through shareholders' equity finances.If ROA is high and ROE is low for a given firm, this would indicate a heavy reliance on leverage in generating profits.It is expected to have low ROE for credit institutions, yet it is important to construct a relationship between ROA and ROE to identify whether banks with higher ROA should have lower ROE.The use of ROA and ROE as dependent variables is consistent with previous studies(Samad, 2004; Kosmidou, 2008; Siddiqui, 2008; Sufian & Habibullah, 2009; Javaid et al., 2011 and Zeitun, 2012).
. To solve this dilemma, ROA is set as the dependent variable and considered as the finest indicator to evaluate bank performance.
Ratios not included in the four categories introduced by Malhotra, Poteau, and Singh (2011), but used in other studies and proved useful are the asset size and the number of bank branches.These two variables are used to measure the size of the bank.
ing interest expense.Earning assets are those assets that the bank is expected to generate income from.These assets are usually loans and financial assets held for trading.

Table 2 .
Correlation coefficients between independent variables