s2.0 s0275531914000440 main

Research in International Business and Finance 33 (2015) 75–98

Contents lists available at ScienceDirect

Research in International Businessand Finance

journal homepage: www.elsevier.com/locate/r ibaf

Islamic versus conventional banks in the GCCcountries: A comparative study usingclassification techniques

Karim Ben Khediri a,b,c,1, Lanouar Charfeddined,e,f,∗,Slah Ben Youssefg

a CEROS, Université Paris Ouest Nanterre La Défense, Franceb College of Administrative Sciences, Najran University, KSAc FSEGN, Université de Carthage, Tunisiad Department of Finance and Economics, College of Business and Economics, P.O. Box 2713, Doha, Qatare Institut de préparationà l’administration et à la gestion, 184 Boulevard Saint-Germain, 75006 Paris,Francef Institut Supérieur de Gestion de Gabès (ISGG), University of Gabès, Tunisiag Faculté des Sciences Economiques et de Gestion de Sfax, route Aéroport km 4, BP N◦1088, 3018 Sfax,Tunisie

a r t i c l e i n f o

Article history:Received 10 December 2013Received in revised form 13 May 2014Accepted 23 July 2014Available online 1 August 2014

JEL classification:C44C45C25G21G28

Keywords:Islamic financeGCC bankingClassification techniques

a b s t r a c t

This paper contributes to the empirical literature on Islamicfinance by investigating the feature of Islamic and conventionalbanks in Gulf Cooperation Council (GCC) countries over the period2003–2010. We use parametric and non-parametric classificationmodels (Linear discriminant analysis, Logistic regression, Tree ofclassification and Neural network) to examine whether financialratios can be used to distinguish between Islamic and conven-tional banks. Univariate results show that Islamic banks are, onaverage, more profitable, more liquid, better capitalized, and havelower credit risk than conventional banks. We also find that Islamicbanks are, on average, less involved in off-balance sheet activitiesand have more operating leverage than their conventional peers.Results from classification models show that the two types of banksmay be differentiated in terms of credit and insolvency risk, oper-ating leverage and off-balance sheet activities, but not in terms

∗ Corresponding author. Tel.: +216 22998570.E-mail addresses: [email protected] (K.B. Khediri), lanouar [email protected] (L. Charfeddine),

benyoussef [email protected] (S.B. Youssef).1 Tel.: +966 538384133.

http://dx.doi.org/10.1016/j.ribaf.2014.07.0020275-5319/© 2014 Elsevier B.V. All rights reserved.

dx.doi.org/10.1016/j.ribaf.2014.07.002

http://www.sciencedirect.com/science/journal/02755319

http://www.elsevier.com/locate/ribaf

http://crossmark.crossref.org/dialog/?doi=10.1016/j.ribaf.2014.07.002&domain=pdf

mailto:[email protected]



dx.doi.org/10.1016/j.ribaf.2014.07.002

76 K.B. Khediri et al. / Research in International Business and Finance 33 (2015) 75–98

of profitability and liquidity. More interestingly, we find that therecent global financial crisis has a negative impact on the profit-ability for both Islamic and conventional banks, but time shifted.Finally, results show that Logit regression obtained slightly higherclassification accuracies than other models.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

In recent years, many conventional banks have encountered financial difficulties and failure due tothe global financial crisis of 2007–2008. In contrast, Islamic banks have successfully withstood this cri-sis. In empirical literature, most studies have attributed this success of Islamic banks to their financialregulation guided by Shariah principles which prohibits the payment or receipt of interest (riba) andencourage risk sharing (see for instance Willison, 2009; Hasan and Dridi, 2010). As a consequence, theattention of academics, policy makers and investors on Islamic banking has been largely increased inthe last few years. Actually, there are more than 300 Islamic financial institutions worldwide includingbanks, mutual funds and insurance firms. In addition, most Western international banks such as Citi-group, HSBC and others have opened Islamic windows. Several factors can explain this rapid growthof interest-free finance, including strong demand for Sharia-compliant products, improvement in thelegal and regulatory framework for Islamic finance, growing demand from conventional investors fordiversification purposes, and the capacity of the industry to innovate and develop a number of financialinstruments that meet the needs of investors (Hasan and Dridi, 2010).

In theory, there are many differences between Islamic and conventional banks. For instance,interest-bearing contracts in conventional banks are replaced in Islamic bank by return-bearing con-tracts, where the profits and losses as well as risks are shared between the creditor and the borrower.Moreover, Islamic banks collect funds through demand deposits (guaranteed and yield no return) andinvestment deposits (similar to mutual fund shares and not guaranteed a fixed return). Islamic bankshave developed free financing products based on profit and loss sharing (PLS) and markup principles.However, since all banks operate in the same competitive environment and are regulated in the sameway in most countries, it is possible that Islamic and conventional banks have similar behavior andhence adopt similar strategies. In fact, several studies show that Islamic banks are not very differentfrom conventional banks in the adoption of PLS principle. Siddiqui (2006) argues that Islamic banksare relying more on markup financing contracts rather than PLS based financing contracts. Chongand Liu (2009) and Khan (2010) find that only a small portion of Islamic banks financing is based onPLS and that Islamic deposits are not interest-free. Bourkhis and Nabi (2013) point out that in mostIslamic banks, less than 20% of total assets are dedicated to long term and risk sharing investments.They observe that Islamic banks are mimicking the commercial strategies of their conventional peersand diverging from their theoretical business model.

In parallel to the increased interest in Islamic finance, the literature on Islamic banking has beengrowing rapidly. The main sizeable body of research has explained the general Islamic principles andthe instruments used in Islamic banking (Bashir, 1983; Khan, 1985; Sundararajan and Errico, 2002;Siddiqui, 2006). Recent studies have discussed the management, regulatory and supervisory challengesrelated to Islamic banking (Murjan and Ruza, 2002; Sole, 2007; Jobst and Andreas, 2007), the efficiencyof Islamic banks using frontier analysis approaches such as Data Envelopment Analysis and StochasticFrontier Analysis (Abdull-Majid et al., 2010; Srairi, 2010; Belanes and Hassiki, 2012), the characteristicsand profitability of Islamic Banks (Karim and Ali, 1989; Srairi, 2008; Ben Khediri and Ben-Khedhiri,2009; Abedifar et al., 2013; Beck et al., 2013), and whether it is possible to distinguish between Islamicand conventional Banks (Metwally, 1997; Iqbal, 2001; Olson and Zoubi, 2008). Another strand ofliterature has studied the soundness, resilience and financial stability of Islamic banks during theglobal financial crisis (Cihak and Hesse, 2010; Hasan and Dridi, 2010; Beck et al., 2013; Caby andBoumediene, 2013; Bourkhis and Nabi, 2013).

K.B. Khediri et al. / Research in International Business and Finance 33 (2015) 75–98 77

The rapid increase of interest on Islamic banking and their resistance to the global financial crisismake it important to examine whether Islamic and Conventional Banks behave similarly or differentlybefore, during and after the crisis period. Specifically, we test whether financial ratios can be used todiscriminate between the two types of banks. This paper adds to the empirical literature on Islamicfinance in several ways. First, the current paper focuses on the GCC countries (Bahrain, Kuwait, Qatar,Saudi Arabia, and the United Arab Emirates). This region is worth studying for several reasons: the GCCregion has one of the world’s largest Islamic banking markets2; and the financial sector in the GCC isrelatively developed compared to other countries in the Middle East.3 Second, in contrast to previousstudies on Islamic finance, using only a descriptive analysis to compare Islamic and conventional banks,we employ more sophisticated and performing quantitative techniques such as the linear discriminantanalysis, Logistic regression, neural network and tree of classification models. Finally, our sampleperiod allows us to examine the consequence effects of the global financial crisis of 2007–2008 toboth Islamic and conventional banks in term of profitability, liquidity and risk ratios.

We perform a comparison between the two types of banks, using the t-test of equality of means,and some classification techniques. We find that Islamic banks are, on average, more profitable andbetter capitalized than conventional banks. Results also show that Islamic banks have higher liquidityand lower credit risk compared to their conventional peers. Interestingly, asset structure of Islamicbanks is significantly different from that of conventional banks. Islamic banks have higher operatingleverage and are less involved in off-balance sheet activities. More interestingly, we find that thefinancial crisis has a negative impact on the profitability for both Islamic and conventional banks, buttime shifted. Finally, results from classification models show that some financial ratios can be used todiscriminate between Islamic and conventional banks.

The remaining of the paper is organized as follows: Section 2 presents the main features of Islamicbanks. Section 3 presents the literature review and the proposed hypotheses. Section 4 presents dataand univariate analysis. Section 5 reports results of parametric and non-parametric classificationmodels. Finally, Section 6 concludes.

2. Islamic banking features

In the last thirty years, the interest in Islamic banking has been growing rapidly in both Muslimand non-Muslim countries. In some countries, Sudan, Iran, and Pakistan, the entire banking system iscurrently based on Islamic finance principles. In other countries Islamic banks operate side-by-sidewith conventional banks. Currently, all Islamic banks are concentrated in the Middle East and SoutheastAsia. Recently, large international banks in Europe and the United States have introduced Islamicoffices that offer separate Islamic or Sharia-compliant products within an otherwise conventionalbanking structure.

There are five principles of Islamic finance. These principles are determined by the Sharia or Islamiclaw which provides guidelines and legal framework for all aspects of life. (1) The profit and losssharing (PLS) principle is one of the most important features of Islamic finance. The provider of capitalfunds (lenders) and the entrepreneur (borrowers) share business risk in return of sharing profitsand losses. This PLS principle contradicts those of conventional finance where the rate of return offinancial assets is fixed before transaction. (2) The principle that all transactions have to be backed bya real economic transaction that involves a tangible asset. (3) The prohibition of riba (usury, which isgenerally defined as interest or excessive interest) is the most prominent features of Islamic finance.The Sharia interprets riba as a premium that must be paid by the borrower to the lender along withthe principal amount as a condition for the loan or for an extension in its maturity. (4) The prohibition

2 Islamic banks constitute an important source of financial intermediation, controlling on average 24% of the region’s bankingsystem assets (Al-Hassan et al., 2010).

3 The last 30 years have been characterized by a significant structural change in the GCC financial system by implementingseveral policies that stimulate financial liberalization and financial restructuring in order to make the banking sector morecompetitive (Maghyereh and Awartani, 2012). Furthermore, banks in the GCC region have been subject to extensive reforms,such as the removal of interest rate controls, the strengthening of banking regulations and supervision, and the compliancewith Basel Accords for capital adequacy (Maghyereh and Awartani, 2014).


of gharar (excessive uncertainty) and maysar (excessive risk or gambling).4 (5) The prohibition onfinancing for illicit sectors. Islamic finance prohibits some business activities that are not in compliancewith Islamic law. These prohibited business activities can relate to food (production and sales ofalcoholic beverages, pork products, tobacco), gambling (casinos, on-line gambling, lottery schemes),entertainment (video, magazines, on-line material, strip clubs), immoral and illicit trades (prostitution,drugs).

There are a large number of different products in Islamic financing. Some of them are based on PLSsuch as mudharaba (trustee finance or profit sharing) and musharaka (equity participation or jointventure) and some other financial instruments are based on mark-up such as murabaha (cost-plusfinancing), ijara (leasing), and istisnaa (commissioned manufacture).

In order to ensure that products and business operations comply with Sharia principles in allaspects, Islamic bank is required to establish a Sharia Committee. The members of this Committeeshall have the necessary qualification and knowledge on Islamic jurisprudence (Usul al-Fiqh) and/orIslamic transaction law (Fiqh al-Mu’amalat). Good reputed and accepted personalities are also requiredto be appointed to this position.

3. Literature review and hypothesis

3.1. Literature review

In banking literature, researchers have focused on the determinants of bank performance in terms ofprofitability and efficiency. The majority of studies examine the internal and external factors that affectthe performance of Islamic and conventional banks using financial ratios and frontier approaches. Sofar, few studies have compared between Islamic and conventional banks. In this paper, we focus, inparticular, on studies which cover the banking sector in a panel of countries, including some countriesfrom the GCC region. Metwally (1997) examines the differences between the financial characteristicsof 15 interest-free banks and 15 conventional banks over the period 1992–1994. He finds that thetwo groups of banks may be differentiated in terms of liquidity, leverage and credit risk, but not interms of profitability and efficiency. Using data from 24 banks including 12 Islamic banks for theperiod 1990–1998, Iqbal (2001) find that Islamic banks are better capitalized and more profitablethan conventional banks. Olson and Zoubi (2008) examine whether financial ratios can be used todistinguish between conventional and Islamic banks in the GCC region over the period 2000–2005.They find that non-linear techniques are able to correctly distinguish between the two categoriesof banks at about a 92% success rate. Moreover, their results indicate that Islamic banks are moreprofitable but less efficient than conventional banks. Recently, Beck et al. (2013) examine the differencebetween conventional and Islamic banks on a sample of 510 banks across 22 countries over the period1995–2009. They find few significant differences in business models. However, they find that Islamicbanks are less efficient, but have higher intermediation ratios, have higher asset quality, and are bettercapitalized than conventional banks. They also find that Islamic banks perform better during crises interms of capitalization and asset quality and are less likely to disintermediate than conventional banks.In another recent study, Abedifar et al. (2013) investigates risk and stability features of Islamic bankingusing a sample of 553 banks from 24 countries between 1999 and 2009. They find that Islamic banksare, on average, more capitalized and profitable than conventional banks. They also find that smallIslamic banks that are leveraged or based in countries with predominantly Muslim populations have

4 Islamic law forbidden also gharar and maysar. Many Hadiths of the prophet (pbuh) explicitly prohibited gharar sales. Forexample, the sale of fish in the sea, birds in the sky, an unborn calf in its mother’s womb, un-ripened fruits on the tree, etc., areforbidden in Islam. An important example of gharar is insurance contract. The interdiction of that’s types of contract followsfrom the non respectability of the PLS because only the insurance company acquire all the profits when claim is not made. Theprincipal reason behind this interdiction is that gharar is a synonym of “risk” or “uncertainly”. Moreover, this ban on ghararis considered as the complete disclosure of information. Its interdiction is also viewed as an elimination of any asymmetricalinformation in a contract. In the other hand, the prohibition of maysar is explicitly prohibited in Quaran Sura 2:119, and 5:90.The term often used in Islamic law is qimar. The maysar is defined as all activities involving betting for money or property orundue speculation.


lower credit risk than conventional banks. In terms of insolvency risk, small Islamic banks also appearmore stable. Another strand of literature compares between conventional and Islamic banks usingthe frontier analysis approach. Srairi (2010) uses a stochastic frontier approach to investigate the costand profit efficiency levels of 71 commercial banks in GCC countries over the period 1999–2007. Hefinds that in terms of both cost and profit efficiency levels, the conventional banks are more efficientthan Islamic banks. Abdull-Majid et al. (2010) investigate the efficiency of a sample of Islamic andconventional banks in 10 countries for the period 1996–2002, using an output distance function, andfind that Islamic banks have moderately higher return to scale than conventional banks. Belanes andHassiki (2012) examine the efficiency of 32 Islamic and conventional banks from the MENA countriesover the period 2006–2009 and find no significant difference in the efficiency scores between thesetwo types of banks using Data Envelopment Analysis (DEA) method. All these studies cited aboveand their results are presented in Table 1. It should be noted that the use of different methodologies,measures, sample and period in these studies may explain some of their contradictory results.

3.2. Hypothesis

Our study investigates the differences between Islamic and conventional banks in terms of financialcharacteristics. Our first hypothesis is about bank profitability. Hassoune (2002) shows that Islamicbanks are more profitable than their conventional peers. He explains the outperformance of Islamicbanks by the fact that they rely, for their funding, on high amounts of non-profit bearing deposits,or non-remunerated current accounts. He considers these lower costs of funding as a market imper-fection, which constitutes a “free lunch” for Islamic banks. Hence, given the larger equity base andcurrent account deposits,5 Islamic banks interest bearing liabilities to liabilities are generally lowerthan conventional banks. Thus, Islamic banks are able to operate with strong net interest margin ina high interest rate environment because of their funding advantage, and so are more likely to bemore profitable. Abedifar et al. (2013) point out that the PLS arrangement provides pro-cyclical pro-tection to banks in the event of adverse conditions. They also argue that the religious depositors maybe more loyal and prepared to take lower returns, refusing from withdrawing deposits even if theperformance of the bank deteriorates. Therefore, the Islamic bank’s profitability is less volatile thanthat of the conventional one. Moreover, Abedifar et al. (2013) argue that the religiosity can also influ-ence the bank’s performance by encouraging borrowers to fulfill their obligations under Islamic loancontracts. Indeed, clients with religious beliefs are more likely to prefer Islamic to conventional bank-ing and to be prepared to pay rents for receiving financial services compatible with their religiousbeliefs (Abedifar et al., 2013). That is, Islamic banks may exploit the religiosity of their clients andcharge higher rates to borrowers and give lower rates to depositors. The extra rent is considered asthe price of offering Sharia-compliant products and services. Another strand of literature shows thatbanks with higher equity to assets ratio will normally have lower needs of external funding. Further,banks with higher capital ratio have reduced cost of funding since they have lower insolvency risk.Indeed, an increase in equity can lower moral hazard problems and increase the monitoring incentivesof banks (Diamond, 1984), and also can increase the banks’ risk-taking capacity (Abedifar et al., 2013).A positive relationship between profitability and bank capital has been reported in previous empir-ical studies (for instance, Hassan and Bashir, 2003; Lin et al., 2005; Pasiouras and Kosmidou, 2007;Ben Khediri and Ben-Khedhiri, 2009). Hence, given that Islamic banks employ more capital than theirconventional peers in funding their assets, we expect higher profitability for the former. Therefore,bank profitability is more likely to be higher for Islamic banks. We use the return on assets (ROA) andthe return on equity (ROE) as proxies for bank profitability. These two proxies are widely used in theempirical banking literature (for instance, Iqbal, 2001; Olson and Zoubi, 2008; Abedifar et al., 2013;Beck et al., 2013; Bourkhis and Nabi, 2013). The main empirical results of previous studies showedthat Islamic banks are more profitable than conventional banks (for instance, Olson and Zoubi, 2008;

5 Islamic banks receive deposits mainly in the following two forms current accounts that bear no interest but are obliged topay principal to holders on demand, and investment (or savings) accounts that generate a return based on profit rates. Depositsare received by Islamic banks in the form of “Qard Al-Hasan” or “Amanaa”.

80

K.B.

Khediri

et al.

/ R

esearch in

International Business

and Finance

33 (2015)

75–98Table 1Existing literature.

Authors Sample Methodology Variables Main results

Metwally (1997) 15 Isl. banks15 conv. banks1992–1994

Logit modelProbit modelDescriminantanalysis

Liquidity: cash to depositsLeverage: deposits to assets; equity to assetsCredit risk: funds channeled to direct investments to loanable funds; loansused to finance durable to total loans; personal loans to total loansProfitability: gross income to assets; average return on depositsEfficiency ratios: operating expenses to assets

The two groups of banks may be differentiated interms of liquidity, leverage and credit risk, but notin terms of profitability and efficiency

Iqbal (2001) 12 conv. banks12 Isl. banks1990–1998

T-test forequality ofmeans

Profitability: return on asset (ROA); return on equity (ROE)Bank capital: capital to assetsLiquidity: cash and accounts with banks to total depositsDeployment ratio: total investment to total equity and total depositsEfficiency: cost to income ratio

Islamic banks are better capitalized and moreprofitable than conventional banks

Olson and Zoubi(2008)

28 conv. banks16 Isl. banks GCCregion2000–2005

T-test forequality ofmeansLogisticregressionNeural networksk-means nearestneighbors

Profitability: ROA; ROE; profit margin; return on deposits; return onshareholders’ capital; net operating marginEfficiency: interest income to expenses; operating expense to asset; operatingincome to assets; operating expenses to revenue; asset turnover; net interestmargin; net-non interest marginAsset quality: provision to earning assets; adequacy of provisions for loans;write off ratio; loan to assets; loans to depositsLiquidity: cash to assets; cash to depositsRisk: deposits to assets; equity multiplier; equity to deposits; total liabilitiesto equity; total liabilities to shareholder capital; retained earnings to assets

Accounting ratios are good discriminators betweenIslamic and conventional banks. Islamic banks aremore profitable but less efficient than conventionalbanks

Srairi (2010) 48 conv. banks23 Isl. banks GCCregion1999–2007

stochasticfrontier analysisT-test forequality ofmeans

Profitability: net profit to average total assetsCapital adequacy: equity to total assetsCredit risk: loans to total assetsOperation cost: cost to incomeSize: natural logarithm of total assets

Conventional banks are more efficient than Islamicbanks

Belanes and Hassiki(2012)

19 conv. banks13 Isl. banks MENAregion2006–2009

DataenvelopmentanalysisWilcoxonrank-sum test

Profitability: ROA; ROE; net Interest marginLiquidity: short-term assets to short-term loansRisk: total debts to assets; reserves for losses on credits to total credits

There is no significant difference in the efficiencyscores between these two types of banks

Beck et al. (2013) Sample of 510 banksacross 22 countries1995–2009

T-test forequality ofmeans,regression

Business model: Fee income to operational income; nondeposit funding tototal funding; loans to depositEfficiency: cost to income ratio; overheads to assetsAsset quality: loss reserves to gross loans; loan loss provisions to gross loans;nonperforming loans to gross loans;Stability: z-score; ROA; equity to assets; liquid assets to deposit

There are few significant differences in businessmodels. Islamic banks are less efficient, but havehigher intermediation ratios, have higher assetquality, and are better capitalized thanconventional banks

Abedifar et al. (2013) 553 banks from 24countries1999–2009

T-test forequality ofmeans, randomeffect regression

Credit risk: loan loss reserves to gross loans; impaired loans to gross loans;loan loss provision to average gross loansInsolvency risk: z-scoreBank interest rate: net interest margin; interest income rate; interest expenserate; loan rate; deposit rateFinancial ratio: equity capital to asset ratio; ROA; ROE; net loans to totalearning assets; cost to income ratio; total assets

Islamic banks are more capitalized and profitablethan conventional banks. Islamic banks have lowercredit risk than conventional banks, specificallysmall, leveraged, or those operating in countrieswith more than 90% Muslim populations. In termsof insolvency risk small Islamic banks are morestable than small conventional banks


Abedifar et al., 2013; Beck et al., 2013; Bourkhis and Nabi, 2013). Hence, we formulate our hypothesisas follows:

H1. Islamic banks are more profitable than conventional banks.

Our second hypothesis is about liquidity. In general, banks face liquidity problem due to excesswithdrawal from current and savings accounts and bank run. If withdrawals significantly exceed newdeposits over a short period, then banks get into liquidity trouble. Cash ratios measure the bank’sability to meet its short-term obligations. Thus, higher liquidity ratios are generally associated withless risk. Islamic bank does not have enough investment opportunities since it is allowed to invest onlyin Sahria approved projects. Islamic banks also have restricted access to the inter-bank market andthe central bank (as lender-of-last resort), which challenges liquidity management. Therefore, Islamicbanks are likely to maintain high capital buffers to mitigate liquidity risk. Moreover, Hasan and Dridi(2010) explain this higher liquidity buffers by the fact that managing liquidity is more challenging inIslamic banks, given the limited capacity of many Islamic banks to attract profit sharing investmentaccounts since the return on these accounts is uncertain. Previous empirical studies showed thatIslamic banks maintain higher level of liquidity ratios compared to conventional banks (for instance,Metwally, 1997; Olson and Zoubi, 2008; Bourkhis and Nabi, 2013). We measure liquidity by the cashto assets ratio, and the cash to deposits ratio. These two proxies are widely used in the empiricalbanking literature (for instance, Iqbal, 2001; Metwally, 1997; Olson and Zoubi, 2008; Beck et al., 2013;Bourkhis and Nabi, 2013). Higher ratios denote higher liquidity. Hence, we formulate our hypothesisas follows:

H2. Islamic banks hold higher liquidity than conventional banks.

Our third hypothesis is about credit and insolvency risks. Credit risk is the possibility that a borroweror counter party will fail to meet its obligations for repayment in accordance with the conditionsstipulated in the contract. A failure to repay leads to a loss for the creditor and therefore becomes arisk for the bank. Moreover a bank is considered insolvent if its total assets value is lower than itsliabilities. The assumption that Islamic banking involves lower credit risk than conventional bankingcould be attributable to contractual arrangements (Olson and Zoubi, 2008; Baele et al., 2012; Becket al., 2013). Indeed, the PLS mechanisms allow Islamic banks to maintain their net worth and avoidthe deterioration of their balance sheets under difficult economic situations. Given that neither theprincipal nor the return of the investment deposits is guaranteed, any loss which occurs on the assetside could be totally absorbed on the liability side. Thus, the PLS allows Islamic banks to transfer thecredit risk from its asset side to its liability side (the investment deposits). Consequently, if the value ofthe assets decreases, the value of the liabilities should decrease respectively. Olson and Zoubi (2008)also point out that the default risk of not paying a return to depositors is eliminated under the PLSprinciples. This suggests that Islamic banks may have a greater capacity to bear losses compared toconventional banks (Abedifar et al., 2013). Further, The PLS principle promotes equity participationwhich in turn encourages due diligence in managing investment and active monitoring. The other typesof Islamic financial modes based on mark-up (e.g. Murabaha, Ijaras, and Istisnaa) require investors toengage in the real economy and hence that a real asset underlies the financial transaction. This featureallows the Islamic banks to have a clearer view on the allocation of its funds and to reduce theirexposure to speculative behavior. Furthermore, Islamic banks may have lower credit risk comparedto conventional banks due to the religiosity of clients that enhances loyalty and mitigates defaultand/or due to their special relationship with their depositors (Abedifar et al., 2013). Siddiqui (2006)explains the reduced credit risk by the fact that these Islamic contracts which are based on equityparticipation minimize the adverse selection and moral hazard problems. Indeed, Islamic financerequires symmetry of information and transparency in transactions since Islam prohibits excessiveuncertainty (gharar). Further, gambling (maysir) is banned, meaning that excessive risk taking is notpermitted. The tangibility of assets reduces the problem of assets substitution by engaging in otheractivities with a higher risk. Respect of these principles should decrease the moral hazard problems.Therefore, the risk level should be lower for Islamic banks than for their conventional peers. Cihakand Hesse (2010) argue that the more difficult access to liquidity put pressures on Islamic banks to bemore conservative (resulting in less moral hazard and risk taking).


Previous empirical results suggest that the two groups of banks may be differentiated in terms ofrisk (Metwally, 1997). Moreover, most empirical studies suggest that Islamic banks are less risky thanconventional banks (for instance, Abedifar et al., 2013; Beck et al., 2013).

We use four indicators of credit risk. Specifically, we use the ratio of loan loss reserves to grossloans (LLR) as a proxy for credit risk. This variable represents managers’ assessment of the quality ofthe loan portfolio, including performing and non-performing loans. LLR takes into account the pastperformance and the expectation for future performance of the existing loan portfolio. We also employthe ratio of non-performing loans to gross loans (NPL). We also use the ratio of loan to deposit (LTD),and the ratio of loans to total assets (LTA) as proxy for credit risk exposure. All ratios decrease inasset quality and credit risk. For insolvency risk, we use the ratio of equity to assets (ETA), the ratioof debt to assets (DA), the ratio of deposit to assets (DTA), and the ratio of deposit to equity (DTE).ETA is a measure of capital strength and capitalization. High equity to assets ratio is assumed to beindicator of low leverage and therefore lower risk. On contrary high debt to assets ratio, deposit toassets ratio, or deposit to equity ratio is assumed to be indicator of high leverage and therefore higherrisk of insolvency. Hence, a high (low) value of ETA (DA, DTA, or DTE) implies that the bank is morecapitalized and so more solvent. All these proxies are widely used in the empirical banking literature(for instance, Olson and Zoubi, 2008; Abedifar et al., 2013; Beck et al., 2013; Bourkhis and Nabi, 2013).Hence, we formulate our hypothesis as follows:

H3. Islamic banks are less risky than conventional banks.

4. Data and variables

4.1. The GCC banking sector

In the GCC countries, the banking sector is relatively young compared to other countries. The firstbank dates back to no earlier than the 1950s. The GCC region has one of the world’s largest Islamicbanking markets. The depth of banking sector, as measured by domestic private credit by banks toGDP ranges from 39.26% for the Saudi Arabia to 76.26% for the Kuwait in 2010. Overall, the credit tothe private sector in all GCC countries progressively increased from 2003 to 2010. Although a majorityof banks is privately owned, the foreign ownership is limited and the role of the public sector remainssubstantial. Except for Bahrain, all GCC countries have limits on foreign ownership (Al-Hassan et al.,2010). The GCC banking sector is fairly concentrated. The three largest banks control more than ahalf of the total banking sector assets in the GCC, and the market share of the big three remains highbetween 2003 and 2010. The highest concentrated market is the Kuwait with the three largest banksaccount for 89.58% of total banking sector assets in 2010. The lowest concentrated market is the SaudiArabia, where the big three hold 51.8% of assets in 2010. This reflects entry barriers and licensingrestrictions on foreign banks, including GCC banks. The GCC Banking sector is well capitalized andprofitable. The capitalization of the banks, as measured by the bank capital to risk-weighted assets,is relatively high. The solvency ratio is above the minimum regulatory capital ratio required by BasleII and national standards in all countries. The profitability of the GCC banking sector, as measured bythe return on equity (ROE) and return on assets (ROA) is also relatively high. The banks in the GCCshow good practices of credit risk. The nonperforming loans (NPL) are low and well provisioned. TheNPL ratio ranges from 1.7% for Qatar, followed by the Saudi Arabia (3%), to 8.9% for Kuwait in 2010. Forcountries with lower levels of non-performing loans, the provisioning rates are relatively high. Theratio of provisions to NPL in 2010 is 33.9% for Kuwait, 86.3% for Qatar, and 115.7% for the Saudi Arabia.Finally, the z-score,6 measuring the banking sector soundness, is relatively high compared to average

6 The probability that a country’s banking system defaults is measured by the z-score. The indicator compares the system’sbuffers (returns and capitalization) with the system’s riskiness (volatility of returns). A higher z-score implies a lower proba-bility of insolvency, indicating that the banking sector is more stable. The z-score (or distance to default) is a ratio, defined as((ROA + (equity)/assets))/sd(ROA), where ROA is average annual return on end-year assets and sd(ROA) is the standard deviationof ROA.


of the word,7 indicating that the GCC banking sector is stable. All these indicators for the bankingsector in the GCC countries for the years 2003 and 2010 are reported in Table 2.

4.2. Data and variables

We use banks from the GCC region over the period 2003–2010 covering the 2007/2008 subprimecrisis. We use a sample that comprises only countries with both conventional and Islamic banks.Bank-level data is retrieved from the Bankscope database provided by Fitch-IBCA.8 We only includebanks with at least seven consecutive yearly observations. The final sample includes 44 conventionaland 18 Islamic banks operating in five different countries9 (Bahrain, Kuwait, Qatar, Saudi Arabia,and the United Arab Emirates), consisting of 466 bank-year observations. Table 3 lists the number ofconventional and Islamic banks and observations in each country.

In this paper, fourteen financial ratios have been considered. In Table 4, we classify these ratios intofive groups: profitability ratios (ROA, and ROE), liquidity ratios (CTA, and CTD), credit risk (LLR, NPL,LTA, LTD), insolvency risk (ETA, DA, DTA, and DTE), and asset structure ratios (FAA, OBSIA). Regardingthe later ratios, we use fixed assets to assets ratio, and off-balance sheet items to assets ratio to accountfor the operating leverage, and off-balance sheet activities, respectively. These ratios are used in theprevious empirical banking literature (for instance, Pasiouras, 2008; Srairi, 2013).

4.3. Univariate analysis

Table 5 reports the mean of financial ratios for Islamic and conventional banks, and the p-value forthe t-test of differences in means between the two groups of banks. The sample period is split intothree groups. The first one covers the pre-crisis period from 2003 to 2007, the second one covers thecrisis period from 2007 to 2008 and the third one covers the post-crisis period from 2009 to 2010.10

4.3.1. Overall periodThe univariate analysis shows that Islamic banks are significantly different from conventional banks

at 5% level with respect to the most variables used in this study. The profitability, as measured by theROA, is higher for Islamic banks than for conventional peers. The ROA of 2.82% for Islamic banks versus2.16% for conventional banks is significantly larger at the 5% level. This finding is in line with the firsthypothesis (H1). However, when we use the ROE as proxy of profitability, we do not find any significantdifference between Islamic and conventional banks.

Regarding the second hypothesis, when the differences in liquidity between Islamic and conven-tional banks are significant, Islamic banks are more liquid. We find that Islamic banks hold more cashrelative to assets. CTA averages 24.61% for Islamic banks versus 20.71% for conventional banks. Thedifference is statistically significant at the 1% level. This result corroborates the second hypothesis(H2). But, in terms of cash to deposits, it does not seem that the two types of banks are different. CTDaverages 31.19% for Islamic banks versus 31.68% for conventional banks.

Regarding the credit risk exposure, the average loans to assets ratio of Islamic banks stands at54.16% versus 55.87% for conventional banks and the average loans to deposits ratio for the twobank types are 91.51 and 71.76%, respectively. The difference is statistically significant only for theLTD ratio at 10% level. These results show that Islamic banks intermediate more of the deposits theyreceive and engage more in financing economic activity via lending compared to conventional banks,

7 The mean value of z-score for the world is 15.5 as reported in the global financial development report (2014).8 We use unconsolidated data when available and consolidated if unconsolidated data are not available, in order not to double

count subsidiaries of international banks.9 Oman is excluded from our sample since it does not have Islamic banks over our sample period. In Oman the regulator

approved Islamic banking from January, 1, 2013.10 Bank for International Settlements (2010) identify the pre-crisis period from January 2003 to June 2007 and the acute-crisis

as July 2007 to March 2009. Since quarterly data are not available, we consider 2003–2006, 2007–2008, and 2009–2010 asthe pre-crisis, the crisis and the post-crisis periods, respectively. These three periods are also considered by Bourkhis and Nabi(2013) to distinguish between the first wave of the world financial crisis and its economic wave starting from 2009.

84

K.B.

Khediri

et al.

/ R

esearch in


and Finance

33 (2015)

75–98

Table 2Banking sector indicators 2003–2010.

Bahrain Kuwait Qatar Saudi UAE

2003 2010 2003 2010 2003 2010 2003 2010 2003 2010

Domestic credit to private sector (% GDP) 47.86 67.69 67.73 76.26 29.96 44.69 28.39 39.26 36.17 75.03Concentration (%) 82.58 80.68 67.42 89.58 93.13 83.09 56.36 51.80 49.24 59.58Minimum regulatory capital ratio 12 12 12 12 10 10 8 8 10 12Bank capital ratio 21 20 23 19 - 21 20 17 20 21ROA (%) 1.30 0.90 2.03 1.65 2.30 2.60 2.27 1.75 2.18 1.44ROE (%) 12.10 7.42 18.52 12.32 16.67 18.48 21.31 13.35 14.50 10.76NPLs to total loans (%) 10.3 4.5 6.1 8.9 8.1 1.7 5.4 3 14.3 5.6Provisions to NPLs (%) 67.7 65.9 77.7 33.9 85.4 86.3 128.2 115.7 88.5 68Z-score 17.43 18.39 15.59 18.87 25.31 25.46 12.17 14.24 24.41 21.27

Sources: World Bank World Development Indicators and National authorities.Notes: This table reports the domestic credit to private sector as a percentage of GDP, the concentration, measured as the percentage share of the three largest banks in terms of assets,the minimum regulatory capital ratio, the bank capital, measured as the percentage of bank capital to risk-weighted assets, the return on assets (ROA), the return on equity (ROE), thepercentage of nonperforming loans (NPLs) to total loans, the percentage of provisions to nonperforming loans, and the z-score.


Table 3Sample description.

Conventional banks Islamic banks Islamic banks (%)

No bank No obs No bank No obs

Bahrain 8 55 7 51 46.66Kuwait 4 32 2 14 33.33Qatar 5 40 2 16 28.57Saudi Arabia 11 88 2 14 15.38UAE 15 120 5 36 25.00Total 43 335 18 131 29.50

indicating higher exposure to credit risk. In fact, Islamic banks face higher restrictions on investing innon-real sector related securities (such as bonds) since they are not authorized to invest in interestbearing instruments. For the credit loans quality measures, the average loans loss reserves to loans ofIslamic banks stands at 3.08% versus 5.19% for conventional banks and the average non-performingloans to loans ratio for the two bank types are 5.78% and 5.20%, respectively. The mean test for theloans loss reserves show that Islamic banks have significantly lower levels of credit risk comparedto conventional banks. However, we do not find any significant difference between the Islamic andconventional banks in respect to the ratio of non-performing loans to loans. In terms of solvencymeasures, the average ratios of debt to assets, and equity to assets of Islamic banks are 79.52% and20.32%, respectively, while the corresponding figures for conventional banks are 89.74 and 14.26%,respectively. Furthermore, the average deposits to assets ratio of Islamic banks stands at 68.86% versus78.73% for conventional banks and the average deposits to equity ratio for the two bank types are441.73% and 628.0%, respectively. The differences are statistically significant at 1% level. The meantest results show that Islamic banks employ more capital than their conventional peers in fundingtheir assets. These results suggest that Islamic banks are significantly better capitalized comparedto their conventional peers. Overall our findings indicate that Islamic banks have lower credit andinsolvency risks than conventional peers. These results are in line with our third hypothesis suggestingthat Islamic banks are less risky than conventional banks.

In this study, we also investigate whether some asset structure ratios behave differently acrossIslamic and conventional banks. The asset structure ratios investigated are: fixed assets to assets

Table 4Definition of variables.

Ratios Definitions

ProfitabilityROA Return on assets = Net income/Total assetsROE Return on equity = Net income/Stockholders’ equity

LiquidityCTA Cash to assets = Cash/Total assetsCTD Cash to deposits = Cash/Total customer deposits

Credit riskLLR Loans loss reserves to gross loansNPL Non-performing loans to gross loansLTA Loans to assets = Loans/Total assetsLTD Loans to deposits = Loans/Total customer deposits

Insolvency riskDA Debt to assets = Total debt/Total assetsETA Equity to assets = Total equity/Total assetsDTA Deposits to assets = Deposits/Total assetsDTE Deposits to equity = Deposits/Stockholder’s equity

Asset structureFAA Fixed assets to assets = Fixed assets/Total assetsOBSIA Off-balance sheet items to assets = Off-balance sheet items/Total assets

86

K.B.

Khediri

et al.

/ R

esearch in


and Finance

33 (2015)

75–98

Table 5Univariate analysis.

Overall period 2003–2010 Pre-crisis period 2003–2006 Crisis period 2007–2008 Post-crisis period 2009–2010

CV IB P-value CV IB P-value CV IB P-value CV IB P-value

ROA 2.16 2.82 0.022** 2.57 3.59 0.012** 1.92 3.64 0.001*** 1.57 0.56 0.035**

ROE 15.40 13.10 0.181 17.89 16.99 0.518 14.38 17.74 0.091* 11.47 1.20 0.002***

CTA 20.71 24.61 0.006*** 22.63 28.11 0.021** 20.05 21.54 0.538 17.98 20.72 0.181CTD 31.68 31.19 0.850 36.24 35.37 0.855 27.48 28.82 0.717 26.81 26.05 0.790LLR 5.19 3.08 0.000*** 5.82 3.46 0.001*** 3.41 2.66 0.333 5.72 2.85 0.001***

NPL 5.20 5.78 0.673 5.11 4.81 0.842 3.41 3.50 0.931 7.14 9.12 0.600LTA 55.87 54.16 0.367 52.97 52.87 0.974 58.47 54.73 0.254 59.05 55.85 0.384LTD 71.76 91.51 0.005*** 68.00 89.81 0.021** 74.52 99.19 0.162 76.48 86.27 0.356LTE 400.3 347.6 0.000*** 401.8 333.2 0.040** 484.8 344.6 0.001*** 462.6 376.6 0.020**

DA 89.74 79.52 0.020** 93.52 77.60 0.069* 85.94 80.25 0.000*** 86.02 82.21 0.004***

ETA 14.26 20.32 0.000*** 14.98 22.08 0.000*** 13.12 19.74 0.000*** 13.97 17.78 0.004***

DTA 78.73 68.86 0.000*** 78.80 68.58 0.000*** 79.16 67.82 0.000*** 78.16 70.50 0.020**

DTE 628.00 441.73 0.000*** 614.03 447.74 0.000*** 664.23 417.79 0.000*** 619.54 461.68 0.002***

FAA 1.05 1.59 0.001*** 0.979 1.473 0.027** 1.02 1.86 0.019** 1.21 1.50 0.421OBSIA 44.25 18.2 0.000*** 43.70 22.32 0.007*** 50.78 16.46 0.000*** 38.90 13.96 0.000***

Notes: This table reports the mean of financial ratios for Islamic and conventional banks, and the p-value for the T-test of differences in means between the two groups of banks. The T-testfor equality of means is calculated assuming unequal sample variances.

* Significance at 10% level.** Significance at 5% level.

*** Significance at 1% level.


(FAA), and off-balance sheet items to assets (OBSIA). Results indicate that Islamic banks have loweroff-balance sheet items to assets ratio, and higher fixed assets to assets ratio. FFA averages 1.59%for Islamic banks versus 1.05% for conventional banks, while OBSIA averages 18.2% for Islamic banksversus 44.25% for conventional banks. The differences between the two groups of banks are statisticallysignificant at 1% level for these two ratios, suggesting that Islamic banks have higher operating leverageand are less involved in off-balance sheet activities.

4.3.2. Pre, during and post crisis periodsIn order to investigate the evolving behavior of Islamic and conventional banks and to test the

sensitivity of our results, we repeat the univariate analysis over the pre-crisis period (2003–2006), thecrisis period (2007–2008) and the post-crisis period (2009–2010). Evidence shows that our findingsover the pre-crisis period do not differ from those over the overall period (2003–2010). Therefore,we focus in our analysis on the next two periods to gauge the impact of the first wave of the worldfinancial crisis (2007–2008) and its economic wave starting from 2009. Concerning the profitabil-ity, the mean values of ROA and ROE of Islamic banks are 3.64 and 17.74%, respectively, while thecorresponding figures for conventional banks are 1.92 and 14.38%, respectively during the crisis. Thedifferences between the two groups of banks are statistically significant. In terms of evolution, thereis no variation in profitability for Islamic banks but a slight decrease for conventional banks. Howeverafter the crisis, we observe a substantial decrease in profitability for Islamic banks. Therefore, basedon the evolution of the ROA and ROE, we conclude that the Islamic banks outperform the conven-tional banks before and during the financial crisis. However, this was reversed since 2009 as the crisishit the real economy and so conventional banks perform better. This is a clear indication that thefinancial crisis has a negative impact on the profitability for both Islamic and conventional banks, buttime shifted. Hasan and Dridi (2010) point out that business model helped Islamic bank to containthe adverse impact on profitability in 2008, while weaknesses in risk-management practices in someIslamic banks led to larger declines in profitability compared to conventional banks in 2009. Second,evidence shows that the liquidity of Islamic banks, measured by either cash to assets ratio or cash todeposits ratio, is not statistically different from that of conventional banks during and after the finan-cial crisis. These results are consistent with Beck et al. (2013) and may suggest that the two groupsof banks follow the same liquidity policy during and after the crisis. In terms of evolution, there is adecrease in cash holdings for both Islamic and conventional during and after the crisis. Third, regard-ing the asset quality, as measured by loans to assets ratio, loans to deposits ratio, and non-performingloans to loans ratio, we do not find any significant differences between Islamic and conventionalbanks during and after the financial crisis. However, in respect to the loans loss reserves to loans ratio,the difference between Islamic and conventional banks remains negative and statistically significant,indicating that Islamic banks have lower credit risk during and after the crisis. Regarding the insol-vency risk, evidence shows that leverage, as measured by debt to assets ratio, deposit to assets ratio,and deposit to equity ratio, continue to be higher for conventional banks than for Islamic banks overall periods. Moreover, Islamic banks remain better capitalized that their conventional peers. Overall,all these results may suggest that Islamic banks are less risky than their conventional peers. Finally,regarding the asset structure, evidence shows that Islamic banks have, on average, lower off-balancesheet items to assets ratio and higher fixed assets to assets ratio even during and after the financialcrisis.

5. Parametric and non-parametric methods

Parametric methods (Linear discriminant analysis and Logit model) and non-parametric methods(neural network and tree of classification) are used in the current study to determine which ratios candiscriminate between Islamic and conventional banks. In all empirical investigations, the dependentvariable is a dummy variable which takes the value 1 for Islamic banks and zero for conventional banks.All these methods can be used to assign objects to two groups, and can employ one or more predictorvariables. However, that there is a difference between LDA and logistic regression on the one hand, andthe neural network and the tree of classification, on the other. Whereas the first two parametric meth-ods use all (statistically significant) predictor variables simultaneously in the model, the second two


non-parametric methods use the predictor variables in a hierarchical and recursive manner. Further-more, to get a high rate of classification, parametric methods suppose the normality of observations,no outliers, large size of sample, and homogeneity of the variance–covariance matrix (Rubin, 1990;Robert, 2002). On the other hand, when these hypotheses are not satisfied non-parametric methodsbecome more efficient than parametric methods.

5.1. Parametric methods

Linear discriminant analysis (LDA) and logistic regression are widely used multivariate statisti-cal methods for analysis of data with categorical outcome variables. Both of them are appropriatefor the development of linear classification models. The goal of logistic regression is to find the bestfitting and most parsimonious model to describe the relationship between the outcome and a setof independent variables. Linear discriminant analysis can be used to determine which variable dis-criminates between two or more classes (Islamic or conventional banks in our study), and to derive aclassification model for predicting the group membership of new observations. Nevertheless, the twomethods differ in their basic idea. In logistic regression, unlike in the case of LDA, no assumptions aremade regarding the underlying data (normal distribution and large size of sample, homogeneity ofthe variances–covariances matrix, and no outliers). LDA has been developed for normally distributedexplanatory variables with equal covariance matrices. LDA gives better results in the case when thenormality assumptions are fulfilled, but in all other situations logistic regression should be moreappropriate. Specifically, logistic regression outperforms LDA where one of the assumptions of theLDA is not satisfied. Hence, logistic regression is more flexible and robust method in case of violationsof these assumptions.

5.1.1. Linear discriminant analysisThe Linear discriminant analysis (LDA) approach is widely used for classifying samples of unknown

classes, based on training samples with known classes (Fisher, 1936). The purpose is to classify obser-vations based on a set of variables known as predictors or input variables (financial ratios in ouranalysis). The application of this method requires some assumptions: normal distribution of data,homogeneity of the variances–covariances matrix, large size of sample and no outliers. The model isbuilt based on a set of observations for which the classes are known. This set of observations is some-times referred as the training set. Based on the training set, the technique constructs a set of linearfunctions of the predictors, known as discriminant functions, such that:

Z = + ˇ1x1 + ˇ2x2 + · · · + ˇnxn + εi

where ˇi for i ∈ [0, 1, 2, ..., n] are discriminant coefficients, x are the input variables or predictors, isa constant, and εi is the error term. These discriminant functions are used to predict the class of a newobservation with an unknown class.

Among the fourteen financial ratios considered in this study, the LDA method selects the mostsignificant and important variables that discriminate between the two types of banks. Moreover, itprovides the Wilks’ Lambda, the canonical correlation, the Chi2, and the hit rate (or accuracy rate) ofclassification. Results reported in Table 7 shows that, for the whole period (2003–2010), five ratios cansignificantly discriminate between Islamic and conventional banks: ETA, DTA, FAA, OBSIA, and LTD.The positive and significant coefficient on the equity to assets ratio (ETA) variable confirms that Islamicbanks are better capitalized than their conventional peers, suggesting that Islamic banks are less riskythan conventional banks because of their reliance on equity. The negative coefficient on deposit toassets ratio (DTA) indicates that Islamic banks rely less on deposits than conventional banks. Thecredit risk ratio retained by the estimated LDA model is the loans to deposits ratio (LTD). This ratio hasa positive and significant coefficient at 1% level, indicating that Islamic banks are characterized by ahigher level of loans relative to deposits compared to conventional banks. This positive and significantcoefficient on the LTD indicates that Islamic banks extend more loans than conventional banks. In fact,unlike conventional banks, Islamic banks cannot allocate a part of their funds to investments in interestbearing instruments (such bonds). The higher level of loans indicates a higher exposure to credit risk

K.B.

Khediri

et al.

/ R

esearch in


and Finance

33 (2015)

75–98

89

Table 6Pearson correlation coefficients.

ROA ROE CTA CTD LLR NPL LTA LTD DA ETA DTA DTE FAA OBSIA

ROA 1ROE 0.756*** 1CTA −0.016 −0.009 1CTD 0.077* −0.005 0.858*** 1LLR −0.150*** −0.168*** 0.332*** 0.324*** 1NPL −0.204*** −0.299*** 0.253*** 0.220*** 0.824*** 1LTA −0.032 0.073 −0.506*** −0.480*** −0.340*** −0.409*** 1LTD 0.094** 0.017 −0.389*** −0.176*** −0.201*** 0.287*** 0.534*** 1DA −0.103** −0.046 0.102** 0.018 0.184*** 0.074 −0.061 −0.070 1ETA 0.506*** −0.023 0.020 0.213*** 0.112** 0.172*** −0.119*** 0.189*** −0.172*** 1DTA −0.302*** 0.051 0.162*** −0.174*** 0.002 −0.043 0.049 −0.622*** 0.128*** −0.636*** 1DTE 0.342*** −0.025 0.200*** 0.179*** −0.036 −0.091* −0.133*** −0.123*** 0.459*** 0.680*** −0.174*** 1FAA 0.112** −0.010 −0.016 0.032 −0.003 −0.042 0.012 −0.030 −0.048 0.250*** −0.120** 0.229*** 1OBSIA 0.055 0.069 0.104** 0.133*** 0.016 −0.005 0.065 0.060 0.036 0.062 −0.011 0.049 −0.049 1

* Significance at 10% level.** Significance at 5% level.

*** Significance at 1% level.


Table 7The linear discriminant analysis results.

Overall period2003–2010

Pre-crisis period2003–2006

Crisis period2007–2008

Post-crisis period2009–2010

LTD 0.279*** 0.481***

(0.000) (0.000)ETA 0.477*** 0.770*** 0.844*** 0.418***

(0.000) (0.000) (0.000) (0.000)DTA −0.296*** −0.514*** −0.443***

(0.000) (0.000) (0.000)FAA 0.260*** 0.269***

(0.000) (0.000)OBSIA −0.649*** −0.591*** −0.528*** −0.754***

(0.000) (0.000) (0.000) (0.000)

Canonical correlation 0.532 0.556 0.543 0.519Wilks’ Lambda 0.717 0.691 0.705 0.730Chi2 144.037 76.731 19.211 51.373P-value 0.000 0.000 0.000 0.000Hit rate (%) 81.20 80.20 82.80 78.40No obs. 432 208 114 113

Notes: This table reports the results from linear discriminant analysis model. P-values are reported in parentheses.*** Significance at 1% level.

for Islamic banks. This result is not consistent with our third hypothesis. The positive coefficient onfixed assets to assets ratio (FAA) and the negative coefficient on off-balance sheet items to assetsratio (OBSIA) indicate that Islamic banks have higher operating leverage and less involved in off-balance sheet activities, respectively. Finally, Results indicate that the profitability and liquidity ratioscannot discriminate between the two groups of banks. Hence, the first two hypotheses, pertaining toprofitability and liquidity, are not supported by the LDA results.

We also repeat the LDA over the sub-periods. Results over the pre-crisis period (2003–2006) aresimilar to those over the overall period, except that the ratio of loans to assets (LTD) is replaced by theratio of loans to equity (LTE). During the crisis period (2007–2008), the LDA retain only two variables-equity to assets ratio (ETA) and off-balance sheet items to assets ratio (OBSIA)-with the same sign onthe coefficients. Results over the post-crisis period (2009–2010) show that equity to assets ratio (ETA),deposits to assets ratio (DTA), and off-balance sheet items to assets ratio (OBSIA) discriminate betweenthe two groups of banks. The success rate, or classification accuracy, for these four LDA models rangesfrom 78.4% for the post crisis period to 82.8% for the crisis period.

5.1.2. Logistic regression modelsThe second parametric method is the logistic regression. This method supposes that the probability

of a dichotomous outcome is related to a set of potential predictor variables in the form:

log(

p

1 − p

)= + ˇ1x1 + ˇ2x2 + · · · + ˇnxn + εi

where p is the probability of the outcome of interest, ˛is the intercept term, ˇi for i ∈ (0, 1, . . ., n)represents the coefficient associated with the corresponding explanatory variable xi for i ∈ (0, 1, . . ., n),and εi is the error term. The dependent variable is the logarithm of two probabilities of the outcome ofinterest. These variables are usually selected for inclusion by using some form of backward or forwardstepwise regression technique (Neter et al., 1996; Pampel, 2000) though these selection techniquesmay be subject to problems. In addition, the maximization of the likelihood function is usually appliedas the convergent criterion to estimate the coefficients of corresponding parameters when the logisticregression models are used.

The classification results of logistic regression are sensitive to high correlation between the explana-tory variables. Hence, we excluded some of the explanatory variables because of the problem of


Table 8Logit model result.

Overall period2003–2010

Pre-crisis period2003–2006

Crisis period2007–2008

Post-crisis period2009–2010

LLR −14.48*** −14.89*** −15.734*

(0.001) (0.003) (0.019)LTA 2.282*** 5.598* 4.905**

(0.063) (0.093) (0.050)ETA 14.37*** 13.72*** 28.97*** 13.08**

(0.000) (0.001) (0.000) (0.014)DTA −4.923*** −5.320**

(0.009) (0.033)OBSIA −8.127*** −6.154 −9.986*** −11.26***

(0.000) (0.000) (0.000) (0.000)FAA 26.47**

(0.040)

Constant 1.394 2.720 −4.781 −3.400*

(0.481) (0.239) (0.019) (0.079)LR Chi2 183.15 72.89 62.06 46.15P-value 0.000 0.000 0.000 0.000R squared 0.389 0.348 0.458 0.352Hit rate (%) 87.27 87.98 88.60 82.30No obs. 432 208 114 113

Notes: This table reports the results from stepwise logit model. P-values are reported in parentheses.* Significance at 10% level.

** Significance at 5% level.*** Significance at 1% level.

multicolinearity. The Pearson correlation coefficients are reported in Table 6. The final selected spec-ification is obtained using backward stepwise method.

Results from logistic regressions are reported in Table 8. For the whole period (2003–2010), onlysix predictor variables out of fourteen are statistically significant and so can be used to discriminatebetween Islamic and conventional banks. These variables are: loans loss reserves to loans ratio (LLR),loans to assets ratio (LTA), equity to assets ratio (ETA), deposits to assets ratio (DTA), fixed assets toassets ratio (FAA), and off-balance sheet items to assets ratio (OBSIA). The positive coefficient on theequity to assets ratio (ETA) confirms that Islamic banks are more capitalized. The deposit to assets ratio(DTA) shows a significant and negative coefficient, which indicate that Islamic banks are characterizedby a lower level of deposit to assets ratio compared to conventional banks. The coefficient on loansto assets ratio (LTA) is positive and significant at 5% level. This result suggests that Islamic banksextend more loans than conventional banks, suggesting higher exposure to credit risk for Islamicbanks. However, the negative coefficient on loans loss reserves to loans ratio indicate that Islamicbanks have lower credit risk than conventional banks. Overall, these findings suggest that Islamicbanks are less risky compared to conventional banks, supporting our second hypothesis. Finally, thenegative coefficient on off-balance sheet items to assets ratio (OBSIA) indicates that Islamic banksare less involved in off-balance sheet activities than conventional banks. The positive coefficient onfixed assets to assets ratio (FAA) indicates that Islamic banks hold more fixed assets than conventionalbanks, suggesting higher operating leverage for the former.

We also re-estimate the logistic regressions over the sub-periods. Results indicate that three orfour variables can be used to discriminate between Islamic and conventional banks. These variablesare loans loss reserves to loans ratio (LLR), loans to assets ratio (LTA), equity to assets ratio (ETA), andoff-balance sheet items to assets ratio (OBSIA). The sings of the coefficients on these ratios remainunchanged. The success rate, or classification accuracy, for these four logistic models ranges from82.3% for the post-crisis period to 88.6% for the crisis period.


Fig. 1. Neural network with one hidden layer.

5.2. Non parametric methods

5.2.1. The neural network methodThe neural network method is an algorithmic procedure for transforming inputs into desired out-

puts using highly inter-connected networks of relatively simple processing elements. The essentialfeatures of a neural network are the nodes, the network architecture describing the connectionsbetween the nodes, and the training algorithm used to find values of the weights for a particularnetwork. The nodes are connected to one another in the sense that the output from one node canbe served as the inputs to other nodes. Each node transforms an input into an output using somespecified function that is typically monotone, but otherwise arbitrary. This function depends onparameters whose values should be determined with a training set of inputs and outputs. The net-work architecture is the organization of nodes and the types of connections allowed. The nodes arearranged in a series of layers with connections between nodes in different layers, but not betweennodes in the same layer. The layer receiving the inputs is called the input layer. The final layer pro-viding the target output signal is the output layer. Any layers between the input and output layers arehidden layers. A simple representation of a neural network with one hidden layer can be shown in theFig. 1.

The classification rule of the neural network is as follows: If the output unit ≥0.5, then the obser-vation belongs to the group 1 (G1). In contrast, if the output unit ≤0.5, then the observation belongsto the group 2 (G2).

The neural network classification model is used in many areas of research. In this paper, we usethis technique in order to classify Islamic and conventional banks in two groups according to theirfinancial ratios. Table 9 shows the importance and the normalized importance of all variables in theneural network classification model. The importance of an independent variable measures how muchthe network’s predicted value varies for different values of the independent variable. The normalizedimportance is the importance values divided by highest importance value and displayed as a per-centage. For the whole period (2003–2010), results show that off-balance sheet items to assets ratio(OBSIA) scored the highest importance, followed by loans to equity (LTE), equity to assets ratio (ETA),return on assets (ROA), and fixed assets to assets ratio (FAA). OBSIA scored 15.9%, which indicates thatOBSIA strongly influence the predicted value of the model. On the other hand, ROE has the lowestimportance level of 1.2% which suggests that the profitability has no influence on the predicted value.For the pre-crisis period (2003–2006), the best discriminant variables are, in order of importance,off-balance sheet items to assets ratio (OBSIA), debt to assets (DA), loans to deposits (LTD), loans toequity (LTE), fixed assets to assets ratio (FAA), etc. For the crisis period (2007–2008), equity to assets

K.B.

Khediri

et al.

/ R

esearch in


and Finance

33 (2015)

75–98

93

Table 9Neural network results.

Overall period 2003–2010 Pre-crisis period 2003–2006 Crisis period 2007–2008 Post-crisis period 2009–2010

Importance Normolizedimportance(%)




ROA 0.039 20.60 0.031 19.50 0.049 43.60 0.035 21.40ROE 0.022 11.20 0.032 20.20 0.029 25.50 0.063 38.40CTA 0.057 29.90 0.057 36.10 0.060 53.80 0.083 51.10CTD 0.089 46.60 0.064 40.30 0.013 11.90 0.046 28.40LLR 0.113 58.80 0.039 24.60 0.082 73.00 0.043 26.10NPL 0.076 39.80 0.054 34.50 0.103 92.30 0.097 59.80LTA 0.089 46.60 0.076 48.40 0.070 62.80 0.074 45.60LTD 0.075 39.30 0.092 58.60 0.077 69.00 0.147 90.30DA 0.074 38.60 0.083 52.90 0.112 100.00 0.062 38.00ETA 0.068 35.70 0.063 39.90 0.088 78.70 0.037 22.60DTA 0.044 22.80 0.089 56.20 0.074 65.60 0.030 18.30DTE 0.062 32.30 0.047 29.70 0.101 90.00 0.024 14.60FAA 0.052 27.10 0.115 73.00 0.085 76.20 0.097 59.40OBSIA 0.192 100.00 0.158 100.00 0.055 49.30 0.163 100.00

Hit rate (%) 87.4% 81.9% 81.8% 79.7%

Notes: This table reports the neural network results, specifically the importance and normalized importance for independent variables.


Fig. 2. Classification tree.

ratio (ETA) is ranked first with the highest importance value, followed by loans to deposits (LTD), fixedassets to assets ratio (FAA), deposits to equity (DTE), off-balance sheet items to assets ratio (OBSIA),etc. Finally for the post-crisis period (2009–2010), the ratio with the highest importance is off-balancesheet items to assets ratio (OBSIA), followed by equity to assets ratio (ETA), loans to deposits (LTD),loans to equity (LTE), cash to deposits (CTD), etc. The success rate, or classification accuracy, for thesefour neural network classification models ranges from 58.3% for the crisis period to 86.3% for theoverall period.

5.2.2. The Classification treeClassification trees are among the classification techniques of data mining. Once built, a tree is in

the form of an inverted tree where each terminal node (or sheet) contains a fraction of the originalsample in which individuals are almost all in a single class. Indeed, several types of decision trees areavailable, but they differ mainly in how they choose and carry out the connections, and in how theymanage the nominal predictors. The main types of decision trees are: C&RT, CHAID and QUEST.

A new observation descends the tree from the root to a single sheet. Its trajectory in the tree isdetermined by the values of its attributes (or predictors). It is then affected to the dominant class ofthe sheet. Each sheet is a “segment” of the sample which is as homogeneous as possible compared tothe dependent variable (class).

Classification trees accept any kind of predictor: numerical, ordinal or nominal. They are fast andcan handle large volumes of data, and their decisions can be transcribed in the form of logical rules.Several types of classification trees are commercially available. Their differences are mainly in howthey choose and make connections and how they deal with nominal predictors. Decision trees areobtained by binary recursive partitioning. The parent nodes are always divided into two descendantnodes (intermediary or terminal). This process is reiterated by regarding each intermediate node as aparent node (see Fig. 2).

Thus, each branch of the tree represents a different combination of the explanatory variables, andthe ways leading a node to the sheets do not necessarily have the same number of branches.

Table 10 reports the importance and the normalized importance of all variables in the classifica-tion tree model. The importance of an independent variable measures how much the classification’spredicted value varies for different values of the independent variable. The normalized importanceis the importance values divided by highest importance value and displayed as a percentage. For thewhole period (2003–2010), results show that off-balance sheet items to assets ratio (OBSIA) scoredthe highest importance, followed by debt to assets (DA), equity to assets ratio (ETA), deposits to assetsratio (DTA), and return on assets (ROA). OBSIA scored 13.4%, which indicates that OBSIA strongly influ-ence the predicted value of the model. On the other hand, ROE has the lowest importance level of 0.6%which suggests that the profitability has weak or no influence on the predicted value. For the pre-crisisperiod, the best discriminant variables are, in order of importance, equity to assets ratio (ETA), deposits

K.B.

Khediri

et al.

/ R

esearch in


and Finance

33 (2015)

75–98

95

Table 10Classification tree Results.

Overallperiod 2003–2010 Pre-crisisperiod 2003–2006 Crisisperiod 2007–2008 Post-crisisperiod 2009–2010

Importance Normalizedimportance




ROA 0.009 6.40 0.040 22.20 0.112 46.67 0.016 12.40ROE 0.006 4.80 0.026 14.90 0.002 0.83 0.057 41.19CTA 0.015 11.40 0.036 20.00 0.000 00.00 0.018 13.95CTD 0.011 8.60 0.018 9.80 0.000 00.00 0.018 13.95LLR 0.060 44.70 0.003 1.60 0.025 11.68 0.049 37.98NPL 0.051 38.30 0.009 4.90 0.006 2.80 0.000 0.07LTA 0.046 34.60 0.010 5.60 0.000 00.00 0.016 12.40LTD 0.033 24.90 0.015 8.40 0.002 0.83 0.020 15.50DA 0.069 51.70 0.086 48.31 0.198 82.50 0.066 51.16ETA 0.061 45.70 0.032 17.90 0.214 100.00 0.109 84.49DTA 0.087 64.50 0.050 28.20 0.114 47.50 0.053 41.08DTE 0.094 70.40 0.012 6.74 0.214 100.00 0.048 37.21FAA 0.014 10.50 0.001 8.00 0.047 19.58 0.018 13.95OBSIA 0.134 100.00 0.234 100 0.186 77.50 0.129 100.00

Hit rate (%) 75.3% 80.2% 72.7% 75.0%

Notes: This table reports the classification tree results, specifically the importance and normalized importance for independent variables.


Table 11Classification accuracies and error analysis.

Correctlypredict IB (%)

Predict CB butIB (type I error)(%)

Correctlypredict CB (%)

Predict IB butCB (type errorII) (%)

Overallclassificationaccuracy (%)

Overall period (2003–2010)LDA 66.98 33.02 85.80 14.20 81.20Logit 83.82 16.18 87.91 12.09 87.27Neural Network 57.40 32.60 95.80 04.20 87.40Tree classification 63.00 37.00 80.90 19.10 75.30

Pre-crisis period (2003–2006)LDA 58.70 41.30 86.14 13.86 80.20Logit 86.96 13.04 88.11 11.89 87.98Neural Network 33.30 66.70 98.60 01.40 81.90Tree classification 63.00 37.00 85.70 14.30 80.20

Crisis period (2007–2008)LDA 76.47 23.53 85.36 14.64 82.80Logit 85.19 14.81 89.66 10.34 88.60Neural Network 68.80 31.20 87.20 12.80 81.80Tree classification 87.50 12.50 68.00 32.00 72.70

Post-crisis period (2009–2010)LDA 65.95 34.05 83.06 16.94 78.40Logit 70.83 29.17 85.39 14.61 82.30Neural Network 30.40 69.60 100.00 00.00 79.70Tree classification 31.00 69.00 95.20 04.80 75.00

to equity ratio (DTE), debt to assets (DA), off-balance sheet items to assets ratio (OBSIA), deposits toassets ratio (DTA), etc. For the crisis period (2007–2008), equity to assets ratio (ETA) is ranked firstwith the highest importance value, followed by deposits to equity ratio (DTE), debt to assets (DA), off-balance sheet items to assets ratio (OBSIA), deposits to assets ratio (DTA), etc. Finally for the post-crisisperiod (2009–2010), the ratio with the highest importance is off-balance sheet items to assets ratio(OBSIA), followed by equity to assets ratio (ETA), loans to equity (LTE), debt to assets (DA), return onequity (ROE), etc. The success rate, or classification accuracy, for these four classification tree modelsranges from 72.0% for the crisis period to 80.2% for the pre-crisis period.

5.3. Comparison of the performance of classification models

The classification accuracy rate, as well as Type I (Predict conventional bank but Islamic bank) andType II (Predict Islamic bank but conventional bank) errors for the four models are reported in Table 11.In general, classification accuracy rate is the most common quantitative measure used in evaluatingthe predictive accuracy of classification models. In addition, it represents the percentage of banksthat are classified correctly. For the whole period (2003–2010), the overall classification accuracy ishigher for the Logit regression and Neural network model, with a rate of 87% compared to the LDAand classification tree models (81.2% and 75.3%, respectively). For the sub-periods, Logit regressionobtained slightly higher classification accuracies than other classification models. Overall, the differentmodels achieve high rate of classification accuracies, suggesting that financial ratios can be used todistinguish between Islamic and conventional banks.

6. Summary and implications

Islamic finance has grown rapidly over the last three decades. This rapid growth of Islamicor interest-free banking system has attracted the attention of many international policy makersand academic researchers. While most previous studies on Islamic finance have investigated andexplained the general Islamic principles and the instruments used in Islamic banking, recent


researches have investigated whether profitability, efficiency, and risks differ significantly acrossIslamic and conventional banks.

Our paper is a novelty in several ways. First, using 61 Islamic and conventional banks, this paperextends this line of research to shed light on the behavior of Islamic and conventional banks in the GCCcountries between 2003 and 2010. The aim of the current paper was to compare between the featuresof Islamic banks and conventional banks in the GCC countries using selected financial ratios and someparametric methods (linear discriminant analysis and logistic regression) and non-parametric meth-ods (neural network and the classification tree). Moreover, we contribute to Islamic finance empiricalliterature by testing three hypotheses examining the profitability, liquidity, and risk of Islamic andconventional banks.

This study documents several interesting findings. First, we show that Islamic and conventionalbanks behave somewhat differently. Mean tests results show that Islamic banks are more profitable,liquid, and capitalized, and have less credit risk than their conventional peers. Islamic banks also areless involved in off-balance sheet activities and have more operating leverage. Second, results alsoindicate that it is possible to distinguish between these two types of banks based on financial ratios.Specifically, classification models results show that the two types of banks may be differentiated interms of credit and insolvency risk, operating leverage and off-balance sheet activities, but not in termsof profitability and liquidity. Finally, the current paper provides insights into the potential of usingclassification models. Interestingly, results show that logistic regression model is more accurate andinterpretive than other models.

The current research may be extended by investigating other features of banks such as businessmodel, efficiency, and stability. Further, the question of whether Islamic and conventional banks haveor not the same behavior when operating on a small or large scale should be explored in future research.As a related issue, it is also interesting to account for the differences in ownership structure.

As a matter of policy implications, we need to draw some proposals according to the results. It isobvious that Islamic banking is different from conventional banking in terms of credit and insolvencyrisks. These differences have implications for policymakers and regulators. As a result, a particular andwell-defined regulatory and supervisory framework, and risk management tools for Islamic banking,taking into account specific features of Sharia-compliant contracts and Islamic banks, is needed fortheir effective functioning.

Acknowledgments

We would like to thank an anonymous reviewer for valuable and insightful comments that helpedus improve an earlier version of the manuscript. Any remaining errors are, of course, our own.

References

Abdull-Majid, M., Saal, D., Battisti, G., 2010. Efficiency in Islamic and conventional banking: an international comparison. J. Prod.Anal. 34, 25–43.

Abedifar, P., Molyneux, P., Tarazi, A., 2013. Risk in Islamic banking. Rev. Financ. 17, 2035–2096.Al-Hassan, A., Khamis, M., Oulidi, N., 2010. The GCC Banking Sector: Topography and Analysis. IMF Working Paper, WP/10/87.Baele, L., Farooq, M., Ongena, S., 2012. Of Religion and Redemption: Evidence from Default on Islamic Loans (Replaces CentER

DP 2010-136). Discussion Paper 2012-014. Tilburg University, Center for Economic Research.Bank for International Settlements, 2010. 80th Annual Report. Bank for International Settlements, Basel.Bashir, B.A., 1983. Portfolio management of Islamic banks: a certainty approach. J. Bank. Financ. 7, 339–354.Beck, T., Demirguc -Kunt, A., Merrouche, O., 2013. Islamic vs. conventional banking: business model, efficiency and stability. J.

Bank. Financ. 37, 443–447.Belanes, A., Hassiki, S., 2012. Efficiency in Islamic and conventional banks: a comparative analysis in the MENA region. Bank.

Mark. Invest. 120, 36–49.Ben Khediri, K., Ben-Khedhiri, H., 2009. Determinants of Islamic Bank Profitability in the MENA Region. Int. J. Monet. Econ.

Financ. 2 (2/3), 409–426.Bourkhis, K., Nabi, M.S., 2013. Islamic and conventional bank’s soundness during the 2007–2008 financial crisis. Rev. Financ.

Econ. 22, 68–77.Caby, J., Boumediene, A., 2013. The financial volatility of Islamic banks during the subprime crisis. Bank. Mark. Invest. 126,

30–39.Chong, B.S., Liu, M., 2009. Islamic banking: interest-free or interest-based? Pac. Basin Financ. J. 17, 125–144.Cihak, M., Hesse, H., 2010. Islamic banks and financial stability: an empirical analysis. J. Financ. Serv. Res. 38, 95–113.Diamond, D.W., 1984. Financial intermediation and delegated monitoring. Rev. Econ. Stud. 51, 393–414.

http://refhub.elsevier.com/S0275-5319(14)00044-0/sbref0010
























































































































































































































Fisher, R.A., 1936. The use of multiple measurements in taxonomic problems. Ann. Eugen. 7 (2), 179–188.Hasan, M., Dridi, J., 2010. The Effects of the Global Crisis on Islamic and Conventional Banks: A Comparative Study. IMF Working

Paper 10/201.Hassan, M.K., Bashir, A., 2003. Determinants of Islamic banking profitability. Paper Presented at the ERF Tenth Annual Conference,

Marrakesh, Morocco, 16–18 December.Hassoune, A., 2002. Islamic banks profitability in an interest rate cycle. Int. J. Islamic Financ. Serv. 4, 1–13.Iqbal, M., 2001. Islamic and conventional banking in the nineties: a comparative study. Islamic Econ. Stud. 8, 1–27.Jobst, Andreas, A., 2007. The Economics of Islamic Finance and Securitization. IMF Working Paper No. 07/117.Karim, R., Ali, A., 1989. Determinants of the financial strategy of Islamic banks. J. Bus. Financ. Account. 16, 193–212.Khan, M.S., 1985. Islamic Interest-Free Banking: A Theoretical Analysis. IMF Staff Papers DM/85754.Khan, F., 2010. How ‘Islamic’ is Islamic banking? J. Econ. Behav. Organ. 76, 805–820.Lin, S., Penm, J., Gong, C., Chang, C., 2005. Risk-based capital adequacy in assessing on insolvency-risk and financial performances

in Taiwan’s banking industry. Res. Int. Bus. Financ. 19, 111–153.Maghyereh, A., Awartani, B., 2012. Financial integration of GCC banking markets: a non-parametric bootstrap DEA estimation

approach. Res. Int. Bus. Financ. 26, 181–195.Maghyereh, A., Awartani, B., 2014. Bank distress prediction: empirical evidence from the Gulf Cooperation Council countries.

Res. Int. Bus. Financ. 30, 126–147.Metwally, M.M., 1997. Differences between the financial characteristics of interest-free banks and conventional banks. Eur. Bus.

Rev. 97, 92–98.Murjan, W., Ruza, C., 2002. The competitive nature of the Arab Middle Eastern banking markets. Int. Adv. Econ. Res. 8, 267–275.Neter, J., Kutner, M., Wasserman, W., Nachtsheim, C., 1996. Applied Linear Statistical Models, 4th ed. Irwin, Chicago.Olson, D., Zoubi, T.A., 2008. Using accounting ratios to distinguish between Islamic and conventional banks in the GCC region.

Int. J. Account. 43, 45–65.Pampel, F.C., 2000. Logistic Regression: A Primer, vol. 132. Sage Publications, Thousand Oaks, CA.Pasiouras, F., 2008. Estimating the technical and scale efficiency of Greek commercial banks: the impact of credit risk, off-balance

sheet activities, and international operations. Res. Int. Bus. Financ. 22, 301–318.Pasiouras, F., Kosmidou, K., 2007. Factors influencing the profitability of domestic and foreign commercial banks in the European

Union. Res. Int. Bus. Financ. 21, 222–237.Robert, P., 2002. A comparative study of the effect of position of outliers on classical and non traditional approaches to the two

group classification problem. Eur. J. Oper. Res. 136, 603–615.Rubin, P.A., 1990. A comparison of linear programming and parametric approaches to the two group discriminant problem.

Decis. Sci. 21, 373–386.Siddiqui, M.N., 2006. Islamic banking and finance in theory and practice: a survey of state of the art. Islamic Econ. Stud. 13, 1–48.Sole, J., 2007. Introducing Islamic Banks into Conventional Banking Systems. IMF Working Paper 07/175.Srairi, S.A., 2008. A comparison of the profitability of Islamic and conventional banks: the case of GCC countries. Bank. Mark.

Invest. 98, 16–24.Srairi, S.A., 2010. Cost and profit efficiency of conventional and Islamic banks in GCC countries. J. Prod. Anal. 34, 45–62.Srairi, S.A., 2013. Ownership structure and risk-taking behaviour in conventional and Islamic banks: evidence for MENA

countries. Borsa Istanb. Rev. 13, 115–127.Sundararajan, V., Errico, L., 2002. Islamic Financial Institutions and Products in the Global Financial System: Key Issues in Risk

Management and Challenges Ahead. IMF Working Paper No. 02/192.Willison, B., 2009. Technology trends in Islamic investment banking. Islamic Financ. News 6 (19), 22–23.




























































































































































































































































































































































































































































































































s2.0 s0275531914000440 main

Documents