An-Najah National University Faculty of Graduate Studies DETERMINANTS OF BANK NET INTEREST MARGIN: EVIDENCE FROM MENA COUNTRIES By Ajyad Mojeed Ahmad Bahlaq Supervisor Dr. Islam Abdeljawad This Thesis is Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Finance, Faculty of Graduate Studies, An-Najah National University, Nablus, Palestine. 2023 ii DETERMINANTS OF BANK NET INTEREST MARGIN: EVIDENCE FROM MENA COUNTRIES By Ajyad Mojeed Ahmad Bahlaq This Thesis was Successfully Defended on 26/03/2023 and approved by iii Acknowledgment I express my gratitude to My mom who is my greatest supporter all the times, My father, brother, and amazing niece and nephew, My supervisor, Dr. Islam Abdeljawad for his guidance and support during the thesis, and whoever encouraged me through the master journey iv Declaration I, the undersigned, declare that I submitted the thesis entitled: DETERMINANTS OF BANK NET INTEREST MARGIN: EVIDENCE FROM MENA COUNTRIES I declare that the work provided in this thesis, unless otherwise referenced, is the researcher’s own work, and has not been submitted elsewhere for any other degree or qualification. v Table of Contents Acknowledgment ............................................................................................................. iii Declaration ....................................................................................................................... iv Table of Contents .............................................................................................................. v List of Tables .................................................................................................................. vii List of Figures ................................................................................................................ viii List of Appendices ........................................................................................................... ix Abstract ............................................................................................................................. x Chapter One: Introduction ............................................................................................ 1 1.1 Research Problem ....................................................................................................... 4 1.2 Research Objective ..................................................................................................... 5 1.3 Research Questions ..................................................................................................... 5 1.4 Significance of the Study ............................................................................................ 5 1.5 Research Structure ...................................................................................................... 6 Chapter Two: Theoretical Background and Hypotheses Development .................... 7 2.1 Background of the Banking Sector in the MENA Region .......................................... 7 2.2 Theoretical Framework ............................................................................................... 7 2.3 Hypothesis Development .......................................................................................... 11 2.3.1 Net Interest Margin ................................................................................................ 11 2.3.2 Size of Operation ................................................................................................... 11 2.3.3 Risk Aversion ........................................................................................................ 12 2.3.4 Credit Risk ............................................................................................................. 13 2.3.5 Liquidity Risk ........................................................................................................ 15 2.3.6 Specialization in Lending ...................................................................................... 16 2.3.7 Inflation .................................................................................................................. 16 2.3.8 Gross Domestic Product (GDP) Growth ............................................................... 18 2.3.9 Interest Rate Risk ................................................................................................... 19 2.3.10 The interaction between interest rate risk and credit risk .................................... 20 2.3.11 Islamic Banks ....................................................................................................... 21 2.3.12 Control of Corruption .......................................................................................... 21 vi 2.3.13 Rule of Law ......................................................................................................... 22 2.3.14 Regulatory Quality ............................................................................................... 22 2.3.15 Islamic Versus Conventional ............................................................................... 22 2.4 The Conceptual Framework. ..................................................................................... 27 Chapter Three: Methodology ...................................................................................... 28 3.1 Data Description ....................................................................................................... 28 3.2 Models ...................................................................................................................... 28 3.3 Estimation Methods .................................................................................................. 29 3.4 Variable Measurement .............................................................................................. 30 Chapter Four: Result and Discussions ........................................................................ 34 4.1 Descriptive Statistics ................................................................................................. 34 4.2 Correlation Analysis ................................................................................................. 35 4.3 Regression Estimation .............................................................................................. 37 4.3.1 Pooled Models ....................................................................................................... 37 4.3.2 Fixed Effect Models ............................................................................................... 39 4.3.3 Pooled Models After Adding the Moderating Variable ......................................... 47 4.3.4 Fixed Effect Models After Adding the Moderating Variable ..................... 48 4.3.5 The Generalized Method of Moments (GMM) Models ........................................ 50 4.3.6 The Generalized Method of Moments (GMM) Models Adding the Moderating Variable ........................................................................................................................... 53 Chapter Five: Conclusions ........................................................................................... 56 List of Abbreviations ...................................................................................................... 58 References ...................................................................................................................... 59 Appendices ..................................................................................................................... 65 ب ............................................................................................................................... ا?<=>; vii List of Tables Table 1: The explanatory variables, the expected sign, and its justification .................. 26 Table 2: The measure of the variables and the empirical references .............................. 33 Table 3: Descriptive Statistics ........................................................................................ 34 Table 4: Correlation matrix ............................................................................................. 36 Table 5: Pooled models ................................................................................................... 38 Table 6: Fixed effect models .......................................................................................... 40 Table 7: The generalized method of moments models ................................................... 51 Table 8: The generalized method of moments models after adding moderating variable ............................................................................................................................ 54 viii List of Figures Figure 1: Conceptual model ............................................................................................ 27 ix List of Appendices Appendix A: Bank type in the sample ............................................................................ 65 Appendix B: Variance Inflation Factor .......................................................................... 66 Appendix C: Random effect models ............................................................................... 68 Appendix D: Summary results of regressions ................................................................ 70 Appendix E: Pooled models after adding moderating variable ...................................... 71 Appendix F: Fixed effect models after adding moderating variable .............................. 73 Appendix G: Random effect models after adding moderating variable ......................... 75 x DETERMINANTS OF BANK NET INTEREST MARGIN: EVIDENCE FROM MENA COUNTRIES By Ajyad Mojeed Ahmad Bahlaq Supervisors Dr. Islam Abdeljawad Abstract The net interest margin in the dual banking system is not well documented. While previous studies tested the determinants of each bank type (Islamic and conventional banks) separately or using a dummy representing the Islamic bank. The study is concerned with the differences between Islamic and conventional banks. The study examines the factors driving net interest margins in the Middle East and North Africa using Islamic banksas a moderating variable. Also, uncover the effect of interest rate risk, its interaction with credit risk, and institutional variables that are first to investigate in the region. This study uses panel of 511 banks in twenty countries for the period (2006 - 2018). Net interest margin is the dependent variable while the size of operation, risk aversion, credit risk, liquidity risk, specialization in lending, inflation, gross domestic product, interest rate risk, the interaction between interest risk and credit risk, control of corruption, rule of law, and regulatory quality are the independent variables while Islamic banks is the moderating variable. For analyzing the data pooled models, fixed effect models, random effect models, and generalized method of moments were used. This study conclude the negative effect of the size of operation, credit risk, the interaction between interest risk and credit risk, liquidity risk, regulatory quality, and rule of law on the net interest margin while a positive effect of inflation, gross domestic product, interest rate risk, specialization in lending, and Islamic bank. In addition, it is concluded that liquidity risk affect the net interest margin of Islamic bank more than conventional banks. It is recommended banks that focus on diversifying their operations and encourage Islamic banks to adopt new risk management instruments, in addition to enhancement of the governance frameworks such as contract enforcement. Keywords: Dual banking; Islamic bank; MENA region; Net interest margin. 1 Chapter One Introduction Banks provide a variety of services to different sectors and customers. That helps finance household consumption needs, firms’ investment needs, and government deficits. Banks play a vital role in implementing the monetary policies. Banks constitute significant financial intermediaries that stand between lenders (savers) and borrowers (spenders) helping on transferring funds among them which leads banks to be similar or superior to financial markets in channeling funds (Garcia & Guerreiro, 2016). The intermediation function is one of the important functions provided by banks. That leads the banking system to play a fundamental role in economic growth (Maudos & De Guevara, 2004). In addition, banks need the function to support their operations (Endri, Marlina, & Hurriyaturrohman, 2021). However, performing the intermediary function comes with a cost in form of interest to the depositors and the borrowers. Since banks pay interest to depositors and at the same time charge interest to borrowers, a spread is created that is called interest margin. The interest margin is positive since banks charge interest to customers higher than pay to depositors (Tarus, Chekol, & Mutwol, 2012). The net interest margin (NIM) is significant indicator since it is known in the literature as indicator of the efficiency of performing the intermediary function (Almarzoqi & Naceur, 2015; Doliente, 2005; Fungáčová & Poghosyan, 2011; Kumari, 2014). Furthermore, the measure represents the cost of intermediation in the society (Almarzoqi & Naceur, 2015; Poghosyan, 2013). Maudos and De Guevara (2004) pointed out to the importance of performing the intermediation at a lower cost since lower margins show the social welfare of society. However, higher margin involves a trade-off, since higher margins are associated with lower efficiency and competition in the marketplace and reflect insufficient regulatory and higher information asymmetries problems that require a risk premia (Claeys & Vander Vennet, 2008). The dealership model is the most referenced model to study the net interest margindeterminants. The model presented by (Ho & Saunders) in their seminal paper in 1981. Following the model, the bank is a dealer in the credit market that provide loans 2 and ask for deposits. Since the loans and deposits arrive at a stochastic time, the bank will cover the loan demand and deposit supply at the short-term money market holding a short or long positions that presents a cost. So the bank will ask for interest spread or fees for providing it immediacy considering the uncertainty of the transaction (Ho & Saunders, 1981). The margin that results from the uncertainty of the transaction is called the pure spread. The pure spread depends on market structure, degree of risk aversion, average size of transaction, and variance of interest (the dealership model is discussed in more details in chapter 2) (Ho & Saunders, 1981). The model is augmented by later authors to consider variables that not included in the model (Allen, 1988; Angbazo, 1997; Maudos & De Guevara, 2004; McShane & Sharpe, 1985). The dealership model and its latter extension have been applied in a single country (McShane & Sharpe, 1985; Williams, 2007; Zhou & Wong, 2008), regions (Almarzoqi & Naceur, 2015; Islam & Nishiyama, 2016; Maudos & De Guevara, 2004; Poghosyan, 2010; Saunders & Schumacher, 2000; Schwaiger & Liebeg, 2008), and comparisons between developing and developed countries (Garza-Garcia, 2010). Previous studies have shown varying determinants across countries and regions (Doliente, 2005), which attributed to the contexts, economic and banking characteristics, and management practices (Kumari, 2014). For example, Garza-Garcia (2010) conducted a comparative study between developing and developed countries, Garza-Garcia (2010) confirmed higher margins in developing country samples in addition to differences among the determinants and some inconsistent results in capital adequacy, credit risk, and interaction between interest rate risk and credit risk. The Islamic finance estimated assets under management researched US$3.178 trillion at the end of December 2021 with annual growth 8.06% (ZAWYA, 2022, November 28). Islamic principles prohibit usury, materiality (financing in real assets), forbids business ethics that consider unethical or problematic, and return based on risk (World Bank, 2015). Islamic and conventional banks are different according to funding, experiences, principles, activity and regulatory structure (Zarrouk, Ben Jedidia, & Moualhi, 2016). Commercial bank's intermediation is debt-based that based in transfer the risk to the customers while Islamic banks are asset-based which are based on sharing of the risk (Hasan & Dridi, 2011). Despite that, both conventional and Islamic banks are following their local regulation and Basel rules, Islamic banks' operations must follow the Shariah 3 principle (Sun, Mohamad, & Ariff, 2017). When commercial banks price their services they used market lending and deposit rates and play with spread which differs from Islamic banks pricing which is based on the expected profit-sharing yield (Sun et al., 2017). Both banks' models are usually facing the same operational problems, which are the random time of the arrival of financing and supply of deposits, default risk, interest rate volatility, liquidity risk, and others (Hutapea & Kasri, 2010). The bank margin in Islamicbanks (Net profit margin) is identical to the net interest margin in commercial banks (Hutapea & Kasri, 2010). Islamic bank margin is the difference between what banks generate from investing and financing projects and what banks have distributed to their depositors (Bougatef & Korbi, 2018). Few studies of bank margins were performed in dual-banking countries compared to the literature performed in the conventional banking context (Bougatef & Korbi, 2018; Hutapea & Kasri, 2010; Lee & Isa, 2017; Malim & Masron, 2018; Shawtari, Ariff, & Razak, 2019; Sun et al., 2017). Lee and Isa (2017) examined the determinants of net interest margin in Malaysia, which operate under a dual banking system. They divided the sample according to the bank type (Islamic or conventional banks) and concluded the similarities with minor differences between Islamic and conventional banks determinants. They claimed that commercial banks are influenced by more variables due to the Islamic bank's small size. While Hutapea and Kasri (2010) confirmed the long-run relation between Islamic banks margins and the variables used (default risk, liquidity risk, solvency ratio, implicit cost, the opportunity cost of holding the reserve, management quality, and interest rate volatility). Hutapea and Kasri (2010) reported contrary findings between interest volatility and the margins of the two bank models (positive for commercial and negative for Islamic banks) and some changing effect of some variables when transferring the operations from conventional to Islamic models. Malim and Masron (2018)studied the net interest margin during and post-financial crisis and confirmed the higher margin in Islamic banks which indicates that Islamic banks are more conservative and dependent on the higher margins as a cushion from unfavorable market conditions. While commercial banks withstand the shocks due to the limited contagion effect in developing counties. Malim, Ibrahim, and Rasid (2017) applied the dealership model in Islamic banks for eighteen countries from the Organization of the Islamic Cooperation 4 (OIC) and emphasized that institutional variables are not effecting bank margin because they are already included in Islamic laws and religious beliefs. Ibrahim and Law (2019) claimed the existence of Islamic banks reduces the intermediation cost in dual banking systems and higher margins for Islamic banks due to the complexity of the contracts, and the unique risk that face Islamic banks compared to conventional banks such as fiduciary risk, displaced with commercial banks, rate of return. Few studies conducted on the MENA regions country as single countries (Asmar, 2018; Ben Khediri & Ben-Khedhiri, 2011; Shawtari et al., 2019)or group of countries (Al- Muharrami & Murthy, 2017; Bougatef & Korbi, 2018). The current thesis examines a period during and post financial crisis and extended the numbers of countries to twenty country more than Bougatef and Korbi (2018) who used 14 countries. 1.1 Research Problem The intermediation function is the most important function provided by banks that generate a large portion of the bank’s revenue. The higher margin considers a constraint for the deepening of the financial intermediation since lower deposit rates produce reluctant in saving flows to bank deposits and higher lending rates cause a reduction in the investment opportunities (Fungáčová & Poghosyan, 2011). The higher NIM can be seen in developing countries with different studies seeking to identify their determinants (Ahokpossi, 2013; Almarzoqi & Naceur, 2015; Kumari, 2014; Maudos & Solís, 2009) and comparative studies also confirmed the higher margins in developing countries compared to developed countries (Garza-Garcia, 2010). The higher NIM in developing countries considers a problem since banks are the main sources of funds for firms and individuals due to capital market deficiencies. Higher NIM leads to cause depression in investments and savings growth, increases the cost of financing through banks, and prohibits certain customers from using the bank's services (Ben Khediri & Ben- Khedhiri, 2011). Furthermore, higher margins reflect banks' environmental condition of inadequate regulatory and information asymmetry problems that are associated to lower of both efficiency and market competition (Claeys & Vander Vennet, 2008). As a region, most countries in the Middle East and North Africa (MENA) are dependent on the banking system to support their economic development (Bougatef & Korbi, 2018) which rise the significanceof the efficiency of the intermediation function. 5 1.2 Research Objective Since higher margins are a problem in the banking systems, the thesis investigates the determinants of net interest margin on the Middle East and North Africa using variables that mentioned on the literature to have an effect on the net interest margin. Furthermore, the thesis adds the Islamic banks to moderate the relationship between the dependent and explanatory variables. So the objective of the research: 1. Investigate the determinants of net interest margin on the Middle East and North Africa. 2. Discover the role of the Islamic bankson moderating the relationship between net interest margin and banks specific variables (size of operation, risk aversion, credit risk, liquidity risk, specialization in lending). 1.3 Research Questions On order to fulfill the research objective, we need to answer the following main questions: 1. What are the determinants of a bank's net interest margin in the MENA region? 2. What is the differences between Islamic and conventional banks net interest margin determinants? 1.4 Significance of the Study Theoretical significance: The study contributes to the literature by investigating the role of institutional variables, interest rate risk, the interaction between interest rate risk and credit risk on the net interest margins that are not being previously investigated in the Middle East and North Africa context. In addition, the research seeks to discover the moderating variable of the Islamic banksin effecting the relation between net interest margin and bank-specific variables (size of operation, risk aversion, credit risk, liquidity risk, specialization in lending). Despite that, Shawtari et al. (2019) applied the moderating variable, they study was in solo country. Practical significance: the study investigates the factors that affect the net interest margin at the bank level, macroeconomic level, and institutional level. Therefore, the study benefits bank management who care about the profitability of the bank (as NIM is an indicator of profitability). Furthermore, it benefits regulators who are concerned 6 about the efficiency and the proper conditions for the performance of the intermediationfunction. 1.5 Research Structure The thesis is structured as followed: chapter 2 demonstrates the theoretical background and hypothesis development. Chapter 3 presents the methodology. Chapter 4 shows the results and discussions of the findings, and chapter five, the conclusion. 7 Chapter Two Theoretical Background and Hypotheses Development This chapter presents a review of the banking system in the MENA region. Then the chapter demonstrates the dealership model and summaries the main extensions added. Then it follows by the development of the hypotheses in light of the empirical evidence. The chapter concludes by presenting the conceptual model. 2.1 Background of the Banking Sector in the MENA Region The countries in the MENA region are not homogenous that vary from wealthy oil- producing countries to politically unstable countries that import oil. Countries share similarities and differences. For example, North African countries are dominated by state-owned banks with the banking sector as senior source of financing. The banking sector differs across the region in the MENA, with more developed and less risk in the Gulf Cooperation Council countries (GCC), the Middle East banking system is more efficient and profitable, While North Africa is low profitably and higher non-performing loans ratios (Abdelaziz, Rim, & Helmi, 2022). 2.2 Theoretical Framework Dealership Model The model developed by Ho and Saunders (1981) extends the hedging hypotheses and expected utility approach and the model relies on the literature of bid-ask price for security market dealer. In this model, the bank is risk averse dealer in the credit market, supplying and demanding homogenous loans and deposits (essentially banks ask for one type of deposit and provide one type of loan). The risk aversion assumption is essential for two reasons. First, uncertainty of the liquidity needs. Borrowers and lenders prefer indirect financing/ lending to direct financing for transaction cost. If the bank is not risk-averse, the bank will bear transaction costs to eliminate the margin. Second, the size of the bank and the existence of riskless instruments in the money market. The size of bank will stop the banks from engage in ad infintum to eliminate the margin (Angbazo, 1997). In addition, Angbazo (1997) mentioned traditional reasons for risk aversion: inability of management to diversity human capital, inadequate ownership diversity, governmental regulations (deposit insurance failure of resolution, etc.), that 8 create incentive problems as moral hazed and adverse selection, and the bankruptcy cost that result from default risk.Following the model, the bank determines the price of loan and deposit at the beginning of the decision period, which remain constant during the decision period. The bank seeks to maximize the expected utility of the terminal wealth (Ho & Saunders, 1981). The wealth of bank have three components: the base wealth that are allocated on devised portfolio, the net credit inventory (which is the difference of the market value of loans and deposits) that have the same maturity and mature after the decision period, and short term money market position (the differences between loan and borrowing) (Ho & Saunders, 1981). Banks set the price of loans and deposit passively while the quantity is exogenously determined, the price as followed �� = � + � ………………….(1) �� = � − …………………(2) Where: p is the bank’s opinion of the true price of loan and deposit, a and b are fees for providing loans and accepting deposits. Therefore, p, Pl and Pd are prices that are inversely related to deposit and loan rates. According to Ho and Saunders (1981), the probability of loan demand and deposit supply depends on the size of the fees. The bank can affect the loan demand and deposit supply by manipulating the fees of a and b. if bank rises b, the price is falls, and the loan rate rises leading to discouraging the loan demand. On the other hand, if a bank has an excess deposit so the bank reduces the deposit rate by raising the price of the deposit and leading to discouraging the deposit supply. The dealership model assumes that loans and deposit have long maturity that extend to after the decision period. This exposes the bank for two types of risk if the bank has unmatched loans and deposit portfolio at the end of decision period and interest rate changes. The risks are reinvestment risk and refinancing risk. First: reinvestment risk: the risk appears if the bank receives deposit at long-term rate and there is no instant loan demand. In this case, the bank will invest the deposit temporary in the short-term money market at risk-free rate. At the end of the decision period, the bank is expose to 9 reinvestment risk, if short-term interest falls. Second: refinancing risk, the risk appears when the bank receives loan demand without contemporaneous deposit. The bank will restore their shortage from the short-term money market. In the end of decision period, the bank is facing the refinancing risk if the interest rate rises (Ho & Saunders, 1981). Therefore, the bank seeks to determine the optimal spread that maximize the expected utility with considering the transaction uncertainty and interest rate risk. The spread equation: = + � = � + � � ����� ………………….(3) The � represents the symmetric of loan and deposit arrival function. Which represent the bank being risk neutral if the risk averse (R) was zero. If α is greater than β will result in a higher � ratio and spread. The higher the ratio means the bank has monopoly power and can order greater spread if thebank operates in market characterized by inelastic demand and supply function. Second term is first order risk-adjustment term with (R) the risk aversion, �� variance of loan and deposit, and � the size of bank operation. According to the model, the second term is positively related to the margin that is higher risk aversion, variance of interest, average size of operation the higher the margin. To test the validity of the theoretical model, Ho and Saunders (1981) applied a two- stage approach. In the first stage, the pure margin is estimated by regressing the net interest margin against bank imperfections that are not included in the theoretical model (default risk, opportunity cost of holding the reserve, and implicit interest). The first regression equation as follows: �� = �0 + �1 �� + �2 �� + �3 �� + �� ………………….(4) Where, Mi is the net interest margin for the bank i, δ0 is the estimated pure spread, IR is implicit interest, OC opportunity cost of required reserve, DP default premium. Ui is the error term. In the second stage, the intercept of the previous regression is regressed against the volatility of the interest rate. The constant of the second regression shows the market 10 structure and the coefficient shows the effect of interest volatility on the pure spread. The second regression equation: �� = 0 + ϒ1���+∈ � ………………….(5) Testing the model, Ho and Saunders found a significant effect of the pure margin and implicit interest in the first stage. The second stage found pure spread is positively effect by the variances of one-year bond rate. The model is extended by subsequence authors, McShane and Sharpe (1985) modified the model by attaching interest rate uncertainty instantaneous short-term money market interest rather than the rate of loan and deposit, which makes more sense since variable loan and deposit rates are predominant in the Australian context. Allen (1988) extended the model by considering the heterogeneity of loans. Taking into account the cross elasticity of demands between banks’ products can reduce the pure spread. The model was modified to take default risk and its interaction with interest rate risk by Angbazo (1997). While Maudos and De Guevara (2004) explicitly add the operating cost. The authors claimed that even in the absence of banks' risk and market power, higher margins are justified by the high cost the bank incurred. Another contribution by the authors is the Lerner index used as a direct measure of market competition. Valverde and Fernández (2007)extended the model to take into account the non- traditional activities. The authors extended Allen (1988)’s model to assume the bank has two alternatives to set the price of loans relative to the deposit rate and non-traditional relative to the deposit rate. In their paper, Entrop, Memmel, Ruprecht, and Wilkens (2015) extended the model to account for interest rate risk and expected return from maturity transformation. Banks price interest risk on loan and deposit intermediation fees separately for their individual exposure but increase those charges for deposits or reduce it for loans if there is a positive excess holding period return on long-term exposure. Islam and Nishiyama (2016) extended the Ho and Saunders (1981) model by adding the relative size variable. Furthermore, Cruz-García and Fernández de Guevara (2020) extended the model to take into consideration the deposit insurance premium and capital requirement explicitly in the model. 11 2.3 Hypothesis Development The following section is reviewing the variable definition, the findings of empirical evidence, and the hypotheses to be investigated. 2.3.1 Net Interest Margin The net interest margin measures the gap between the interest the bank pays for the providers of the funds and the interest the bank received from the users of funds (Naceur & Omran, 2011). The net interest margin is a significant indicator that shows the mix and volume of the bank’s assets and liabilities that are set to cover the intermediation cost and it’s an important element of a bank’s profitability (Angbazo, 1997). Net interest margin represents the profit from the core business since it shows the interest generated by the bank (Garcia & Guerreiro, 2016). The net interest margin is a summary of the return on the interest rate (Angbazo, 1997). Despite that lower margin indicates the existence of market competition, lower intermediation costs, and regulatory taxes. Higher margins create a degree of stability in the banking system with growing profitability and capital, which protect banks during bad economic conditions (Saunders & Schumacher, 2000). The empirical evidence references bank-specific factors, regulation and institutional environments, market structure, and macroeconomic variables to affect net interest margins (Almarzoqi & Naceur, 2015; Malim & Masron, 2018). Especially, the size of the operation, risk aversion, credit risk, liquidity risk, specialization in lending, inflation, gross domestic product (GDP growth), interest rate risk, the interaction between interest rate risk and credit risk, control of corruption, rule of law, and regulatory quality. 2.3.2 Size of Operation According to the theoretical model, banks with greater operation is exposed to more probability of losses at the same level of credit and market risk, so banks operate at higher margins (Maudos & De Guevara, 2004). Poghosyan (2010)found that higher margins are compensation for the possibility of loss per operation due to the large stake. Other studies confirmed the postive relation between size of operation and net interest margins: Maudos and Solís (2009) in Mexico in some models, Almarzoqi and Naceur 12 (2015) in the Caucasus and Central Asia countries, and Ibrahim and Law (2019) in Malaysia. Other empirical evidence claimed that banks with large operations benefit from the economics of scale (Fungáčová & Poghosyan, 2011; Maudos & De Guevara, 2004). In addition, large banks provide a variety of loans than small banks, which reduces the risk (Khan & Jalil, 2020). Moreover, large banks have better resources and sophisticated technology that led to efficiency and reduce the cost of unit of operation, which leads to lower net interest margins (Lee & Isa, 2017). In China, Zhou and Wong (2008) claimed that operaiton size is linked to net interest margins negatively since some large banks aggressively increase their credit portfolio with lower margins to reduce impaired loan ratios or the reward are based on expanding the credit sales, not on risk-adjusting performance. In emerging and low income countries, Poghosyan (2013)concluded that large banks exhibit a lower margin because of the scale effect. However, other literature finds no effect (Angori, Aristei, & Gallo, 2019; Cruz-García & Fernández de Guevara, 2020; Islam & Nishiyama, 2016; Kumari, 2014; Liebeg & Schwaiger, 2006; Rahman, Rahman, Masud, & Kaium, 2023; Schwaiger & Liebeg, 2008; Williams, 2007). Based on the discussion above, the thesis argues that size of operation effects the net interest margin negatively, the hypothesis can be written as follows: H1: there is a negative association between the size of operation and net interest margin. 2.3.3 Risk Aversion Risk aversion refers to the bank not accepting to hold more risk and receiving the same return amount (Khanh & Tra, 2015). The risk aversion behavior of banks can be seen by holding capital more than the mandatory capital requirement (Lee & Isa, 2017). According to the theoretical model, Banks that are more risk averse are requesting higher margins since equity financing is costly compared to debt financing (Maudos & De Guevara, 2004). Thus, higher equity holding reduce the profitability, banks have to work at higher spread (Lee & Isa, 2017). The risk averse managers ask for higher net interest margins as compensation for bearing more risk (Asmar, 2018; Fungáčová & Poghosyan, 2011; Hawtrey & Liang, 2008; Rahman et al., 2023). The risk aversion behavior is more prominent during uncertain times such as the global financial crisis as found by Angori et al. (2019). Furthermore, higher margin can be seen as a way to 13 generate more returns to fulfill the return expectation of shareholders (Aboagye, Akoena, Antwi‐Asare, & Gockel, 2008)and generate an adequate return for the increasing equity (Kumari, 2014). Khan and Jalil (2020)argued that the positive relation due to the solvency regulations that pressure lending activities so banks exhibit a net interest premium. WhileIslam and Nishiyama (2016)believed that solvent banks operate with a higher margin. In the Gulf Cooperation Council Countries, Al-Muharrami and Murthy (2017) concluded that well-capitalized banks provide a lower deposit rate since demands are financed by equity funds. Furthermore, well-capitalized banks stand on their balance sheet strength to carry more credit risk to generate more revenue. Islamic banks show risk aversion behavior since it manifests in the capital ratio, however, the positive connection means that Islamic banks pricing is based on Murabaha financing not on the profit-loss-arrangements (Malim et al., 2017). Other scholars supported the positive effect (Entrop et al., 2015; Ibrahim & Law, 2019; Khanh & Tra, 2015; Liebeg & Schwaiger, 2006; Maudos & De Guevara, 2004; Poghosyan, 2010; Schwaiger & Liebeg, 2008; Trinugroho, Agusman, & Tarazi, 2014; Williams, 2007). On contrary, Poghosyan (2013)believed that risk averse banks might not accept financing a profitable project if granting credit means greater risks. Poghosyan (2013) confirmed the findings for the two groups (low-income countries and emerging countries with a greater magnitude for emerging countries). Other empirical evidence confirmed the negative relationship (Suu, Luu, Pho, & McAleer, 2020; Zhou & Wong, 2008). While other literature found no relation (Almarzoqi & Naceur, 2015; Cruz-García & Fernández de Guevara, 2020; Endri et al., 2021). The thesis argues that risk averse banks work with higher margins as compensation for holding more risk. So the hypothesis can be written as followed: H2: There is a positive association between risk aversion and net interest margin. 2.3.4 Credit Risk Credit risk refers to the risk of the debtor will not repay the financial obligation which takes the form of losing a whole or partial of the money granted (Khanh & Tra, 2015). Angbazo (1997) added default risk variable to the dealership model and confirmed his hypothesis that banks with more risky loans are holding more risk of default, so banks impose higher margins to compensate for the risk. Kasman, Tunc, Vardar, and Okan (2010)claimed that the existence of positive relation between interest margin and credit 14 risk for all subsamples (consolidation, post-consolidation, for new and candidate members of the Europe Union (EU), and old EU members). Banks demand compensation for both expected and unexpected risks, which leads to higher net interest margins. Also, Asmar (2018) believed that banks exhibit a high net interest margin because it faces more expected and unexpected credit risk. Kumari (2014)emphasized that banks raise the net interest margin as compensation for the higher credit risk and the possibility of a loss of interest income. Schwaiger and Liebeg (2008) believed that banks acquire a positive risk premium from adjusting the price of loans and deposits in respect of the credit risk. In their paper, Claeys and Vander Vennet (2008) concluded that loan ratios positively affect the net interest margin in West Europe and a sample of EU countries. Lending wide the margins since loans are higher risk and cost-intensive assets class that demand banks to emerge those risks in the loan pricing. Those considerations are not taken into account in non-accession countries. During and post- financial crisis, Malim and Masron (2018) claimed that conventional banks do not consider diversification in their loans portfolio and charge higher margins as compensation for higher credit risk. While the crisis represents a risky environment that results in more default risks, banks impose credit policies to increase the margins through rising loan rates or reducing deposit rates. Rahman et al. (2023) claimed that higher non-performing loan ratio represents a cost that a bank passes to borrowers. Poghosyan (2013) confirmed the positiverelation for both low-income and emerging countries, banks ask for compensation for the higher risk and the effect is more pronounced in low-income countries. Other empirical studies confirmed the positive effect (Agoraki & Kouretas, 2019; Ahokpossi, 2013; Angori et al., 2019; Drakos, 2002; Hawtrey & Liang, 2008; Lee & Isa, 2017; Maudos & De Guevara, 2004; Poghosyan, 2010; Suu et al., 2020; Tarus et al., 2012; Valverde & Fernández, 2007). On contrary, Fungáčová and Poghosyan (2011) found a negative effect of credit risk on NIM. The authors justified the finding with the market discipline argument. Customers cost risky banks (having higher non-performing loans) a higher premium for depositing their savings. When the deposit rate only increases and everything else holds constant will result a low net interest margin. This confirmed by the finding of Trinugroho et al. (2014) in Indonesia after the 1997-1998 financial crisis. Other studies confirmed the negative effect of credit risk on the net interest margin. Williams (2007) asserted that banks increase their market share by writing off low asset 15 quality, which leads to higher provision of doubtful debts that are not fully priced to risk. In addition, Khan and Jalil (2020) believed that banks might follow practices to increase their market share on the loans and advance without full monitoring which leads to reduced bank margins. However, Endri et al. (2021) argued that banks prefer a lower net interest margin during the financial situation of the lender is worsened. Brock and Suarez (2000) found a negative effect for all countries in the sample except Colombia (the effect was positive), while the significant effect was for Argentina and Peru. Higher non-performing loans lead to lowers income, which in turn lowers the spread, especially when there is no adequate loan loss reserve. Furthermore, banks may grow through practicing risky strategies such as increasing deposit rates and reducing loan rates that reduce their interest spread in liberalized countries, the negative effect confirmed by Doliente (2005)for Indonesia and Thailand and positive effect in the Philippines. Other literature found no effect (Aboagye et al., 2008; Almarzoqi & Naceur, 2015; Ben Khediri & Ben-Khedhiri, 2011; Cruz-García & Fernández de Guevara, 2020; Ibrahim & Law, 2019; Islam & Nishiyama, 2016; Khanh & Tra, 2015; Liebeg & Schwaiger, 2006; Zhou & Wong, 2008). The thesis argues that banks with higher credit risk operate at the higher margin to compensate for the higher risk holding: H3: there is a positive association between credit risk and net interest margin. 2.3.5 Liquidity Risk Liquidity risk refers to the risk that the bank is not having sufficient cash to meet deposit withdrawals or customer demands leaving the bank to borrow at a higher cost (Angbazo, 1997). In his paper, Angbazo (1997) claimed that having more liquid assets (lower liquidity risk) would deprive the banks of the risk premium. Trinugroho et al. (2014) and Shawtari et al. (2019) believed that an opportunity cost is created by holding higher liquid assets. Other literature supports that the higher the liquid assets held by banks the lower the margin (Drakos, 2002; Fungáčová & Poghosyan, 2011). Moreover, banks with higher liquidity risk borrow emergency funds at higher costs (Ahokpossi, 2013). Valverde and Fernández (2007) found the higher the liquid assets, the higher the deposit loan spread. According to Malim and Normalini (2018), Islamic banks ask for higher margins as compensation of liquidity risk since using instruments that correspond to Shariah make liquidity management difficult. While Lee and Isa 16 (2017)found no relation in the baseline regression and for commercial and Islamic banks. The higher the liquidity risk, the higher the risk premium asked, which leads to rising the net interest margin. So, the hypothesis: H4: there is a positive association between liquidity risk and net interest margin. 2.3.6 Specialization in Lending Specializing in a product can reduce banks margins due to economies of scale (Maudos & Solís, 2009). Maudos and Solís (2009) confirmed the negative relationship between the specialization in lending and the net interest margin. They believed that specialized banks benefit from economies of scale to reduce the intermediation margins and the bank offer lower interest rate because of the competition in loan market with the existence of deposit cross-substitute banks which reduce the intermediation cost (Maudos & Solís, 2009). Furthermore, specialized banks are having information advantage which leads to lower margins (Valverde & Fernández, 2007). In addition, specialized banks are better at evaluating the credit positions of customers leading to reduce intermediation costs (Bougatef & Korbi, 2018). However, in the GCCs, Al- Muharrami and Murthy (2017)claimed that holding a higher percentage of loans in the bank assets (higher risky loans) generates higher interest revenue which contributes to higher net interest margin considering that return from other securities (money market and governmental securities) is little in the region. Specialized banks are experienced in evaluating the creditworthiness of customers and have better access to information that incorporates in the acceptance of loan demands, the hypothesis: H5: there is a negative association between specialization in lending and net interest margin. 2.3.7 Inflation Inflation refers to the rate of the increase in the price level of the same general basket of goods and services over time (Khanh & Tra, 2015). Inflation comprises a risk for loan and deposit rates. If the bank has mismatched loans and deposits, the interest rate adjustment for inflationary pressure on loans and deposits will change at different times and to different extents affecting interest margins (Al-Muharrami & Murthy, 2017). Entrop et al. (2015)asserted that passing inflation to net interest margin is important 17 economically since net interest margin contains interest income and expenses that generate from assets and liabilities with more than a year of maturity, which the contract terms are negotiated in the past.Poghosyan (2010)emphasized increasing inflation leads to higher interest margins since it considers an additional risk, which is price uncertainty. Malim et al. (2017) calimed that inflation rises the default risk so islamic banks impose higher margins. Drakos (2002) claimed that banks increase their profit in inflationary environments. Kasman et al. (2010) found that the new and candidate members of EU and in the consolidation period income increases more than cost with inflation leading to higher net interest margins while having no effect on the old members since the countries converge on a single market.The positive effect is supported by Demirgüç-Kunt and Huizinga (1999). In the Central and Eastern Europe, Agoraki and Kouretas (2019) believed that the positive effect of inflation on the net interest margin shows a failure of monetary policy, poor economic performance, and financial instability. The positive effect is supported by Poghosyan (2013) in emerging countries only, Ahokpossi (2013)in sub-Saharan Africa, Khanh and Tra (2015)in Vietnam, and Tarus et al. (2012)in Ghana. In contrast Claeys and Vander Vennet (2008)found that the expectation of reduction and the reduction of inflation have a noticeable downward effect on long-term interest rates leading to reduce NIM. In post- crisis, Malim and Masron (2018) concluded that Islamic banks try to attract customers with more favorable financing rates despite the highly volatile economic environment. Despite that inflation bring higher cost and income, Abreu and Mendes (2001)claimed that cost is more affected by inflation than revenue which led to lower net interest margins. Other empirical evidence supported the negative effect of inflation on the net interest margin (Angori et al., 2019; Khan & Jalil, 2020). Mixed results are obtained on the regional studies, Brock and Suarez (2000)found higher inflation rises the spreads in the countries sample (Chile, Peru, Bolivia, and Colombia) except for Argentina the inflation has a negative effect. However, other empirical evidence found no effect(Al- Muharrami & Murthy, 2017; Almarzoqi & Naceur, 2015; Endri et al., 2021; Ibrahim & Law, 2019; Islam & Nishiyama, 2016; Kumari, 2014; Malim & Normalini, 2018; Maudos & Solís, 2009; Rahman et al., 2023; Shawtari et al., 2019). Based on the previous discussion, the thesis expects a positive effect, since inflation represents a risk of price uncertainty, which lead to compensation for the risk: H6: there is a positive association between the inflation and net interest margin. 18 2.3.8 Gross Domestic Product (GDP) Growth Gross domestic product growth refers to the capacity of the economy to produce more goods and services over a period. It indicates an improvement in business activity, living standards, and favorable market conditions (Khanh & Tra, 2015). The growth of GDP is an important variable that affects the NIM since it affects the demand and supply of loans and deposits directly (Islam & Nishiyama, 2016). Several empirical studies confirmed the negative association. According to Entrop et al. (2015), during the expansion of the economy, banks are competing in the loan market by reducing the loan rate and credit acceptance standards. Islam and Nishiyama (2016)claimed that when the economy is in the prosperity stage and economic expansion becomes visible, banks could grow by charging less interest to customers, which lowers the margins. Valverde and Fernández (2007) claimed that GDP growth affect the spread negatively. Banks reduce interest margins after market rates rise as the economy recovers. Poghosyan (2013) supported the negative relation for the emerging countries only. Other literature confirmed the negative effect of GDP growth on the net interest margin (Agoraki & Kouretas, 2019; Angori et al., 2019; Malim et al., 2017; Shawtari et al., 2019; Suu et al., 2020; Tarus et al., 2012). Kasman et al. (2010) concluded that the effect of the gross domestic product on the net interest margin depends on the stage of transition. A negative effect during the consolidation period since transition economies face more fluctuations in economic growth while no effect after the consolidation is explained by countries gathering in the same market achieving convergence. However, for new and candidate members of the EU, Kasman et al. (2010) believed that higher risk combined with the volatility of the business cycle leads to lower interest margins. However, the old EU members have no effect. In contrast, Claeys and Vander Vennet (2008) found that GDP growth effect NIM positively in Western Europe because of increasing lending and low default. However, their findings in the Center and Eastern Europe were negative or had no effect. Claeys and Vander Vennet (2008) demonstrated the results of a transition economy by the volatility of the business cycle since high growth is interrupted by the crisis. There is minimum tendency for GDP growth to affect in speared on Latin America, the higher GDP will rise the firm capitalized value and since the default risk is low, reduce the lending cost (Brock & Suarez, 2000). Other literature confirmed the 19 positive relation (Drakos, 2002; Ibrahim & Law, 2019; Liebeg & Schwaiger, 2006; Schwaiger & Liebeg, 2008). Bougatef and Korbi (2018) found that the effect of GDP growth varies according to the bank model. Islamic banks affect positively, in favorable economic conditions, banks can profit from accepted projects to increase the return to the partnership. While a negative effect of GDP growth on net interest margin for commercial banks, during good economic conditions investors are applying for banks loans, so banks may grant loans to bad borrowers and leading to reduce interest revenue in case of customer defaults albeit the result (of commercial banks) turns insignificant during robustness checks. Other studies found no effect (Ahokpossi, 2013; Cruz-García & Fernández de Guevara, 2020; Khanh & Tra, 2015; Kumari, 2014; Malim & Normalini, 2018; Maudos & Solís, 2009; Poghosyan, 2010; Rahman et al., 2023). In good economic growth, banks are competing to increase their market share by reducing their lending rate. The hypothesis: H7: there is a negative association between GDP growth and net interest margin. 2.3.9 Interest Rate Risk In the theoretical model, Ho and Saunders (1981) believed that the variance of interest rate of loans and deposits is positively related to the pure margin for one-year bond variance while Saunders and Schumacher (2000) claimed that both short and long rate volatility affect the pure spread. Maudos and De Guevara (2004) found that banks with higher market risk expectations operate at a higher NIM. Maudos and Solís (2009) believed that banks protect themselves from high market risk by imposing higher interest margins. According to Angori et al. (2019), money market volatility increases market risk, which in turn leads to higher NIM. Khan and Jalil (2020) asserted that interest volatility causes uncertainty in the money market so banks impose higher NIM in two models. Williams (2007) found interest rate risk to affect the net interest margin positively albeit the effect becomes insignificant when controlling for the time. Schwaiger and Liebeg (2008) claimed a positive link between interest rate risk and NIM. A risk premium exists from adjusting loan and deposit prices to consider the interest rate risk. According to Rahman et al. (2023), the nature of the bank inventory of transferring funds creates uncertainty in their transactions, so banks put the price of 20 interest rate that compensates for the volatility of the interest risk. Hawtrey and Liang (2008) found interest rate volatility to affects the net interest margin positively with different choices of interest rate. However, Liebeg and Schwaiger (2006) concluded a mixed sign of interest rate risk effect on the net interest margin depending on the measure. A positive sign when the standard deviation of three months interbank and slope of the term structure are used while a negative effect for the ten-year government bond yields. Also, Valverde and Fernández (2007) confirmed the positive effect using loan and deposit spread in seven European countries. In Latin America, Brock and Suarez (2000) found a positive effect of interest rate volatility in the pure spread except for Peru and Colombia. In south Asia countries, Doliente (2005) claimed that short-term interest volatility positively affects the pure spread. However, a negative effect is obtained by Angbazo (1997) interest risk exposure affects net interest margin negatively. The higher the net short-term assets the lower the risk and the risk premium leading to a lower margin. The negative effect is supported by Garza-Garcia (2010) for developing and developed countries. Others found no effect (Cruz-García & Fernández de Guevara, 2020; Ibrahim & Law, 2019; Islam & Nishiyama, 2016; Kumari, 2014; Poghosyan, 2010). As interest rate volatility represents a risk, banks ought to ask for compensation for the exposure. The hypothesis, therefore: H8: there is a positive association between interest rate risk and net interest margin. 2.3.10 The interaction between interest rate risk and credit risk Angbazo (1997)relaxed the dealership model by including the interaction between interest rate risk and credit risk variable and claimed the effect is positive. The positive effect is supported by Garza-Garcia (2010), interest rate risk increases the credit risk which in turn rises the net interest margin. while Angori et al. (2019) and Maudos and Solís (2009) found a negative effect of the interaction between interest rate risk and credit risk on the net interest margin, the higher volatility of the money market, the higher default risk exposure, which lowers the interest margins. Banks lower the spread during higher interest rate volatility in absence of adequate provision for loan losses.Moreover, Poghosyan (2010) claimed that banks are not pricing the risk in an appropriate manner since credit risk is amplified by market risk and vice versa. Other scholars found no effect (Angbazo, 1997; Cruz-García & Fernández de Guevara, 2020; 21 Hawtrey & Liang, 2008; Liebeg & Schwaiger, 2006; Rahman et al., 2023; Williams, 2007). The hypothesis: H9: there is a negative association between the interaction between interest rate risk and credit risk and net interest margin. 2.3.11 Islamic Banks An Islamic bank means a bank providing services that are corresponding to Islamic legislation (Shariah rules). Shawtari et al. (2019) believed that Islamic banks are having lower margins for many reasons in Yemen. Murabaha is dominated to other Islamic contracts, the financing is based on personal relationships, and Yemenis culture is based on the Islamic religion that favors Islamic financing over commercial financing.All that leading Islamic banks to hold more market share and exhibit economies of scale that reduce the cost of operations and the margins. Shawtari et al. (2019)confirmed the negative relation between Islamic banks and net interest margins. In bad economic condition such as inflation or when the investment does not pay off a good return, Islamic banks share part of their profit with the depositors, which lead to lower margin. However, Ibrahim and Law (2019)claimed that Islamic banks operations are facing additional risks than commercial banks such as displacement risk to commercial bank, Islamic banks have additional governance structure to monitor the complianceof operations to Islamic principles, and the contracts are more complex than commercial banks. Islamic banks include all these differences in the margins leading to higher margins. In Palestine, Asmar (2018)found no difference between commercial and Islamic banks. Asmar (2018) attributed the finding to the small percentage of Islamic banks in Palestine. In developing the hypothesis, the thesis based on the claim of Ibrahim and Law (2019)which is Islamic banks face additional risks are not included in the dealership model and Islamic bank will ask for the compensates. So the hypothesis: H10: the level of NIM for Islamic bank is higher than conventional banks. 2.3.12 Control of Corruption The index is used to show the extent of using public power for private interest. The index covers all forms of corruption from petty to superior and extends to government capture by elites and private interests (Poghosyan, 2013). Poghosyan (2013)confirmed his hypothesis of a lower margin associated with higher control of corruption for both 22 groups understudying since countries with convenient institutional environments for business activity operate with lower margins. However Malim et al. (2017) concluded no relationship between control of corruption and the net interest margin. Better institutional conditions lower the margins, so the hypothesis: H11: there is a negative association between control of corruption and net interest margin 2.3.13 Rule of Law The index captures to what extent agents have confidence and obey the rules of society. The index covers areas of contract enforcement, courts, police, etc. (Poghosyan, 2013). Poghosyan (2013)supported the negative effect since a good governance environment leads to a lower margin for low-income and emerging countries. Moreover, Rahman et al. (2023) found that sound institutional quality induces the banking system to operate in competitive and efficient environments. While Malim et al. (2017) found no effect. Therefore, the hypothesis is as follows: H12: there is a negative association between rule of law and net interest margin. 2.3.14 Regulatory Quality The index is used to measure the government's ability to construct sound public policies and regulations with the ability to enforce them. The regulation is extended to all the areas of the economy with the banking system included in order to develop the private sector (Poghosyan, 2013). Poghosyan (2013)confirmed the negative sign for the low- income country only while Malim et al. (2017)concluded that regulatory quality is negatively affects the net interest margin when using system generalized method of moments only. The hypothesis: H13: there is a negative association between regulatory quality and net interest margin. 2.3.15 Islamic Versus Conventional Islamic banks products differ from conventional counterparties since Islamic banks products must be following sharia principles. Islamic finance suggest contract that based on the equity participation and risk sharing. Islamic bank product is cost plus sale (Murabaha), credit sale (Bay ‘‘bi-thaman ajil’’), partnership contract (Musharaka and 23 Modaraba), forward (Salam and Istisna), and benevolent loan (Qard-e-Hasna) (Siddiqi, 2006). A review of the islamic financal contract as mentioned by Siddiqi (2006): 1. Murabaha (cost plus sales): an Islamic contract in which the bank buys a desired consumer durables or real estate for the client after a client's pre-contract promises to repurchase it from the banks at cost of purchase plus a predetermined profit. 2. Bay ‘‘bi-thaman ajil’’: the same as Murabaha but with deferred payment. 3. Musharaka: the debtor participates with the bank in financing a project using his own equity, the profit is a pre-determined ratio, and the losses depend on his shares of the capital. Used to finance trade, importing, issuing letter and in agriculture and industry. 4. Modaraba: the bank finance the whole project and hold the whole risk unless the loss were debtor mismanagement or neglect of the customer. 5. Salam: a sale with delivery is deferred; the bank pays for the seller now with the obligation of the seller to deliver a specific determined commodity on the due date. The bank sells the commodity to the client for a price higher than paid for the seller in cash or deferred price. Mostly used to meet the capital requirement and cost of operation for farmer or contractor etc. 6. Ijara: the bank leases vehicles, equipment, etc. to a customer. The bank signs a contract with the seller to deliver the client the desired commodity. Also, the bank signs a lease contract with the client allowing to transfer of the ownership of the commodity after paying the installment and residual charges. Mostly used for transaction of real estate, car, computers, etc. 7. Qard-e-Hasna: the bank provides facilities for needy students or rural farmers at a limited scale those loans with a negative net present value. The conventional banks have no investment-specific risk in granting loans. Only the credit risk while market risk and interest risk are included in the loan prices. For default risk the conventional banks may garnish the wages of the earner or have the first right of the assets as collateral while Islamic banks become the owner when the debtor borrows money (as collateral) until the principle and the associated profits are outstanding (Siddiqi, 2006). Islamic bank pays rate of return that is depending on the profit while both the principal and profit are at risk if the bank suffers from a loss (Siddiqi, 2006). 24 Islamic banks face credit risk in Murabaha and Musharaka. The credit risk is the risk of non-repayment of the debtor. Credit risk arises from the information asymmetry problem since the debtor has inside information that is not accessible to banks. Murabaha risk arises from the non-performing of the project or the systematic forces (Siddiqi, 2006). Islamic banks face credit and liquidity that depend on the institutional arrangement issue (the existence of Islamic money market and center bank to set capital and liquidity requirements). For equity participation in Islamic banks are less facing moral hazard and adverse selection since joint venture and equity financing disclose company’s books value and investment (Siddiqi, 2006). It is important to study the differences between Islamic and conventional banking margins since the efficiency of the intermediation function depends on both banking models. Lee and Isa (2017)studied the size of operation on net interest margin in Malaysia, a negative effect is discovered in the baseline regression. However, when the authors separated the sample to Islamic and conventional banks, it declared that only conventional banks margins are effecting by the size of operation with a negative effect. The hypothesis: H14: Islamic banks moderate the relationship between size of operation and net interest margin. For risk aversion, Malim et al. (2017)studied the risk aversion effect on the net interest margin in the Islamic bank in the OIC countries and concluded a positive effect confirming the more risk averse the Islamic banks are, the higher the margin. Bougatef and Korbi (2018)concluded that both Islamic and conventional banks are affected positively by risk aversion. Risk-averse banks ask for higher margins and well- capitalized banks pay a lower deposit rate for depositors since the exposure to insolvency risk is lower. However, Lee and Isa (2017)claimed only commercial banks are effecting by risk aversion in Malaysia. The hypothesis: H15: Islamic banks moderate the relationship between risk aversion and net interest margin. 25 For credit risk, Lee and Isa (2017) claimed that Islamic banks are more sensitive to credit risk than commercial banks in Malaysia. However, Ibrahim and Law (2019) found that credit risk is not a determinant of the interest margin in Islamic banks in the OIC countries. In Yemen, Shawtari et al. (2019) tested the moderating role of Islamic banks and found that the interaction of credit risk with the dummy variable that represents the Islamic banks was insignificant. The hypothesis to be investigated: H16: Islamic banks moderate the relationship between credit risk and net interest margin. For liquidity risk, Lee and Isa (2017)concluded that neither Islamic nor commercial banks are effecting by liquidity risk. Shawtari et al. (2019)tested the moderating role of the liquidity risk on the bank margin in Yemeni banks and found a negative but insignificant effect. The hypothesis: H17: Islamic banks moderate the relationship between liquidity risk and net interest margin Finally, the specialization in lending. Bougatef and Korbi (2018) found conflicting results of the effect of specialization in lending on the net interest margin according to the bank type with a negative effect for commercial banks and no effect for Islamic banks. Bougatef and Korbi (2018) explained their result for commercial banks by banks that specialize in lending are able to determine the real assets quality of customers and reduce intermediation costs. However, the Islamic banks margin insignificant results are due to the fewer investment opportunity since they have to follow the sharia principles. Sun et al. (2017)found no effect of the specialization in lending on the net interest margin for Islamic and conventional banks. The hypothesis: H18: Islamic banks moderate the relationship between specialization in lending and net interest margin. Table 1 present the variable used with their expected sign and the rationality based in the previous studies. 26 Table 1 The explanatory variables, the expected sign, and its justification Variable Expected sign Rational relation Size of operations - Large bank with large operation shows economic of scale which reduce the margins Risk aversion + Banks that hold more equity are more risk-averse and hold more risk, so banks work on higher margins. Credit risk + Banks ask for compensation for the higher credit risk Liquidity risk + Banks with higher liquidity risk are gaining liquidity risk premiums. Specialization in lending - Banks specializing in lending able to distinguish creditworthy customers and grant lower margin Inflation + Inflation is risk that lead to price uncertainty, so bank ask for higher margin GDP growth - In good economic growth, banks operate at a lower margin since banks increasing their market share by lowering the loan rate. Interest rate risk + The volatility of interest rate create more risky environment and banks will ask for higher margins. Interaction between interest rate risk credit risk - During higher interest rate volatility times, having insufficient provision for loan loss lowers the margin (Angori et al., 2019) Islamic banks + Islamic banks face unique risks coming from the unique business model. Control of corruption - Better governance context work with lower margin. Rule of law - Better governance context work with lower margin. Regulatory quality - Better governance context work with lower margin. Islamic banks * Size of Operations +/- The relation to be investigated Islamic banks * Risk Aversion +/- The relation to be investigated Islamic banks * Credit Risk +/- The relation to be investigated Islamic banks * Liquidity Risk +/- The relation to be investigated Islamic banks * Specialization In Lending +/- The relation to be investigated 27 Islamic Banks 2.4 The Conceptual Framework. Figure 1 shows model used in the thesis. Figure 1 Conceptual model Moderating variable Explanatory variables Bank specific variables: Size of transaction Risk aversion Credit risk Liquidity risk Specialization in lending Macroeconomic variables: Inflation Gross domestic product Interest rate risk Institutional variables: Control of corruption Rule of law Regulatory Quality Dependent variable Net interest margin 28 Chapter Three Methodology This chapter describes the methods of the study. The chapter reviews the description of the data (duration, sample, and source). Then presents the model used in the thesis, then reviews estimation methods, and finally shows the measure of the variables with presents the empirical reference. 3.1 Data Description The thesis is explanatory with quantitative data. The data are secondary data extracted from the balance sheet and income statement that is derived from Bankscope except for Islamic banks which is gathered through banks' websites. The macroeconomic and institutional variables are gathered from the World Development Indicators database of World Bank. Interest rate risks are gathered from International Monetary Fund with exception of the United Arab Emirates that have been acquired from their central bank website. There is a missing data for calculating the interest rate risk in Syria. The sample consists of an unbalanced dataset for than 511 banks in 20 MENA region countries for the period of 2006 - 2018. The number of years used in the thesis consistent with Maudos and Solís (2009). Appendix A provides a summary of the numbers of banks taken in the sample by country and type (Islamic or conventional) 3.2 Models The econometrics model is based on Ho and Saunders (1981) model and following extensions. Latter studies used an augmented dealership model that gathers the theoretical bank-specific variables with country-specific variables to overcome the cross countries environments differences (macroeconomic, institutional, and regulatory) that are negligence by the theoretical model (Poghosyan, 2013). The thesis sort the independent variable to bank-specific, macroeconomic and institutional variables. Based on Maudos and De Guevara (2004), Poghosyan (2013), and Valverde and Fernández (2007), the model to be estimated in the thesis: $%��&�%'%(�) '*�& = +0 + +1 (�,%-.-�%' ��-& + +2 '�(/ 0%'(�-& + +3 1'%���'�(/ + +4 ��34����5'�(/ + +5 �%1� ��, ��-&�&�%&��&* + +6 �&.� ��-& + +7 9��*'-:�ℎ + +8 �&�%'%(�' �%'�(/ + 29 +9 �ℎ% �&�%' 1��-&�%�:%%&�&�%'%(�' �%'�(/ &�1'%���'�(/ + +10 Islamic banks + +11 1-&�'-�-.1-''4���-& + +12 '4�%-.� : + +13 '%*4� �-'534 ���5 + ��)% &�1-4&�'�%(�4))�%( + H………………….. (7) The second model, a moderatingvariable, Islamic banks, is used. The following model will be used: $%��&�%'%(�) '*�& = +0 + +1 (�,%-.-�%' ��-& + +2 '�(/ 0%'(�-& + +3 1'%���'�(/ + +4 ��34����5'�(/ + +5 �%1� ��, ��-&�&�%&��&* + +6 �&.� ��-& + +7 9��*'-:�ℎ + +8 �&�%'%(�' �%'�(/ + +9 �ℎ% �&�%' 1��-&�%�:%%&�&�%'%(�' �%'�(/ &�1'%���'�(/ + +10 �(� )�1 � &/( + +11 1-&�'-�-.1-''4���-& + +12 '4�%-.� : + +13 '%*4� �-'534 ���5 + +14 (�,%-.-�%' ��-& ∗ Islamic banks + +15 '�(/ 0%'(�-& ∗ Islamic banks + +16 1'%���'�(/ ∗ Islamic banks + +17 ��34����5'�(/ ∗ Islamic banks + +18 (�%1� ��, ��-&-&�%&��&* ∗ Islamic banks + ��)% &�1-4&�'�%(�4))�%( + H…………………. (8) The thesis follows the literature of Angbazo (1997)and Maudos and De Guevara (2004)using the single-stage approach to investigate the relation between the variables. In the single stage, the explanatory variables enter the same equation. The two-stage that conducted by Ho and Saunders (1981), Saunders and Schumacher (2000)and other authors have the drawback of requiring a long time series and cross-section to generate the pure spread and guarantee the precision of the estimation Williams (2007). 3.3 Estimation Methods In order to analyze the data STATA is used. The previous studies used panel data regression to test the data. The literature usually used pooled ordinary least square, fixed effect, random effect, generalized least squares, and generalized method of moments. In Panel data, the time and cross section are gathered in one dataset allowing us to observe the unit over time (Porter & Gujarati, 2009). The data used is unbalanced wide panel. To analyze the data three panel data estimation are used that are pooled ordinary least squares, fixed effect models, and random effect models. According to Porter and Gujarati (2009): First, pooled ordinary least squares; the model neglected the 30 heterogeneity of the cross-section so the coefficients are the same for all cross-sections. Second, the fixed effect model, the model assumes coefficients are changing between units but are time-invariant and uses dummies to consider the heterogeneity between cross-sections. Third: Random effect model, assumes the sample are drawing of large universe and intercept represent the mean value and the differences of the intercepts are included in the error terms. Hausman test is used to choose between fixed effect and random effect model. To overcome the heteroscedasticity and autocorrelation problems the robust standard error is used. Moreover, the study uses a large number of observations, which ignores the normality problems. Furthermore, the panel data is short with units are greater than time (N>T) so the unit root is not affecting the estimation. According to Cruz-García and Fernández de Guevara (2020), the model consider only the new operations of loans and deposits while net interest margin include the revenue and expenses generated from all loans and deposit outstanding for the current and previous years. The proxy include inertia that may suggest a bias in estimations. To overcome this inertia, it must be added that the lagged value of the net interest margin. So to account for endogeneity, the method of generalized method of moment will be used. Following Valverde and Fernández (2007), Maudos and Solís (2009), Trinugroho et al. (2014), Cruz-García and Fernández de Guevara (2020), and other literatures. 3.4 Variable Measurement In this thesis, four types of variable are used: Dependent variable: net interest margin for commercial banks and equivalent variable for Islamic banks. The explanatory variables: they are mentioned in the literature to have impact in the net interest margin that are bank specific, macroeconomic, institutional variables: Bank Specific Variables 1. The size of the operation defines the natural logarithm of total assets. Following Angori et al. (2019). Large operation exposed to greater losses at the same level of credit and market risk so bank ask for higher margins. 31 2. Risk aversion is captured by the equity to assets ratio. Following literature (Angori et al., 2019; Asmar, 2018; Poghosyan, 2013; Suu et al., 2020). The higher the ratio the higher the risk aversion.McShane and Sharpe (1985)argue that capitalization ratio is not satisfying proxy for risk aversion and using must interpret with caution. Since its effecting by accounting conventions, the risk of capital holding changes regarding the target market, and the capital ratio is effected by minimum capital regulation. 3. Credit risk define as non-performing loans ratio as used by Fungáčová and Poghosyan (2011), Khan and Jalil (2020), and Rahman et al. (2023). A more suitable measure for credit risk, other literature used loss provision to loan, however Schwaiger and Liebeg (2008) criticize it as a misleading measure of credit risk since its influences by other factors than credit risk such as earning management. The higher the ratio, the higher the credit risk. 4. Liquidity risk: capture the risk of not having funds to meet the customers’ demands. The liquidity risk is measure by total loan to total deposits as used by Trinugroho et al. (2014). The higher the ratio, the higher the risk and the higher the risk premium asked. Usually literature used inverse ratio of liquidity that is the liquid assets to total liabilities (Angbazo, 1997; Drakos, 2002). 5. Specialization in lending: shows the specialization of banks in the traditional activity of providing loans since nowadays banks are expanding their operations for non- traditional activity as trading, fees and commissions. It measures by total loan to total assets as used by Valverde and Fernández (2007). The higher the ratio the higher the specialization in loan granting and the lower the margins. Macroeconomic Variables 1. Inflation measure by annual inflation of the GDP defaulter as used by Demirgüç- Kunt and Huizinga (1999). The higher the rate, the higher the inflation, and the higher the net interest margins. 2. Gross domestic product: capture the annual real growth of the gross domestic product. The higher the rate, the higher the GDP growth, and the lower the net interest margin. Following Kasman et al. (2010), Kumari (2014) and other literature. 3. Interest risk volatility capture the annual standard deviation of monthly interbank rate. Following Angori et al. (2019). When the interbank money market is not 32 available the treasury bill rate is replaced as done by Poghosyan (2010). The higher the ratio, the higher the volatility of interest, and the higher the margins. Institutional Variables 1. Control of corruption: The index value range from -2.5 to 2.5. The higher the index value means higher control of corruption and the lower the margins (Poghosyan, 2013). The measure used by Malim et al. (2017)and Poghosyan (2013). 2. Rule of law: the index value range from -2.5 to 2.5. Higher variable values the higher the rule of law and the lower the margin (Poghosyan, 2013). The use of the measure follows the literature (Malim et al., 2017; Poghosyan, 2013; Rahman et al., 2023) 3. Regulatory Quality: the index value are between -2.5 and 2.5. The higher the index value the higher regulatory quality and the lower the margins (Poghosyan, 2013). Following literature (Malim et al., 2017; Poghosyan, 2013) using the index as proxy for regulatory quality. The Moderator Variable: Islamic banks. Dummy variable take a value of one if the bank is Islamic and zero otherwise. Following Asmar (2018), Ibrahim and Law (2019), and Shawtari et al. (2019). Control Variable: Time and country dummies is used in order to control for changes in legislation, tax structure, and accounting standards across the countries and over time (Agoraki & Kouretas, 2019). Table 2 present the measure of the variables used in the thesis and references the empirical studies that used the same measure. 33 Table 2 The measure of the variables and the empirical references Variable Measure Empirical Reference Net interest margin (Interest income- interest expenses) to Total assets) Equivalent items will be used for Islamic banks Aboagye et al. (2008), Almarzoqi and Naceur (2015), Angori et al. (2019), Asmar (2018), Entrop et al. (2015), Fungáčová and Poghosyan (2011), Islam and Nishiyama (2016), Khan and Jalil (2020), Lee and Isa (2017), Liebeg and Schwaiger (2006), Poghosyan (2010), and Schwaiger and Liebeg (2008). Size of operation Ln(total assets) Angori et al. (2019) Risk aversion Total equity to total assets Angori et al. (2019), Asmar (2018), Malim et al. (2017), Maudos and De Guevara (2004), and (Poghosyan, 2010, 2013) Credit risk Non-performing loan to total loan Fungáčová and Poghosyan (2011), Khan and Jalil (2020), and Rahman et al. (2023). Liquidity risk Total loan to total deposit ratio Trinugroho et al. (2014) Specialization in lending Total loan to total assets Valverde and Fernández (2007) Islamic banks Dummy variable, take value of one if the banks is Islamic and zero otherwise Asmar (2018), Ibrahim and Law (2019), and Shawtari et al. (2019). Inflation The annual inflation of GDP defaulter Demirgüç-Kunt and Huizinga (1999) GDP growth Real GDP growth rate Claeys and Vander Vennet (2008), Entrop et al. (2015), Kasman et al. (2010), Kumari (2014), (Poghosyan, 2010, 2013), Schwaiger and Liebeg (2008), Suu et al. (2020), and Tarus et al. (2012) Interest rate risk Annual standard deviationof monthly money market rates Garza-Garcia (2010) and Poghosyan (2010) Control of corruption Complied by Kaufmann, Kraay, and Mastruzzi (2010) Malim et al. (2017) and Poghosyan (2013) Rule of law Complied by Kaufmann et al. (2010) Malim et al. (2017), Poghosyan (2013), and Rahman et al. (2023) Regulatory quality Complied by Kaufmann et al. (2010) Malim et al. (2017) and Poghosyan (2013) 34 Chapter Four Result and Discussions This chapter presents the analysis of the thesis. It starts with descriptive statistics, then correlation matrix between the variables, then regression estimations using pooled models, fixed effect models, and the generalized method of moments with discussions on the findings. 4.1 Descriptive Statistics Table 3 shows the descriptive statistics of the variables used in the thesis. The table presents the number of observation for each variable, the mean, the standard deviation, the minimum and the maximum values for the whole sample. Table 3 Descriptive Statistics Variable Obs Mean Std. Dev. Min Max NIM 3560 .022 .026 -.15 .307 Size of Operations 3725 21.325 2.212 13.422 27.339 Risk Aversion 3716 .262 .263 -.684 1 Credit Risk 2459 .111 .171 0 1.525 Liquidity Risk 2680 2.357 20.047 0 893.226 Specialization in lending 3453 .497 .313 0 4.519 Inflation 3725 .048 .11 -.302 .465 GDP growth 3725 .03 .049 -.28 .325 interest rate risk 2932 .004 .005 0 .03 Control of corruption 3725 -.149 .743 -1.681 1.559 Rule of Law 3725 -.119 .765 -2.09 1.339 Regulatory Quality 3725 -.106 .767 -2.218 1.335 Islamic banks 3725 .23 .421 0 1 Table 3, the annual net interest margin for the thesis period is 2.2%, with a minimum value of -15% with a maximum value of 30.7%, and a standard deviation of 2.6%. The mean value is lower than reported in one-country studies conducted on the MENA countries by Asmar (2018) who found that the annual net interest margin is his sample 3.3% for the period of 2006-2016 in Palestine and Shawtari et al. (2019) who found net interest margin is 3.5% in Yemen for the period of 1996-2011. For risk aversion, its average in the Middle East and North Africa during the sample is 26.2%, which is closer to the findings of Asmar (2018) that found the risk aversion was 21.5% in Palestine. Moreover, Credit risk which is measured by non-performing loans 35 is with a mean value of 11.1%. The percentage of non-performing loans is higher than reported by Shawtari et al. (2019) who found the credit risk using the non-performing loan ratio is 21% in Yemen. While annual liquidity risk for a bank during the sample period is 235.7%. This mean that there is a higher liquidity risk in the MENA countries. For macroeconomic variables, inflation in average is 4.8% and the annual GDP growth is 3%. Finally, interest rate risk or volatility of interest is 0.4% with standard deviation of 0.05% and ranges between 0 to 3%. For the institutional variables, the mean values of control of corruption, rule of law, and regulatory quality are -1.49, -.119, and -.106 respectively. The mean values are less than the mean found in Poghosyan (2013) for the low-income and emerging county for the period of 1996-2016. In addition, the mean value is less than Malim et al. (2017) for the period of 2005-2011 for 18 OIC countries. The number of observations vary across variables; net interest margin is 3560, while size of operations, inflation, GDP growth, control of corruption, rule of law, regulatory quality, and Islamic banks is 3725. While risk aversion 3716, credit risk 2459, specialization in lending 3453, liquidity risk 2680 and interest rate risk 2932. 4.2 Correlation Analysis Table 4 shows the correlation between dependent variable and independent variables used in the investigation. Table 4 suggests that the institutional variables exhibit a higher correlation with more than 80% correlation between explanatory variables. According to Porter and Gujarati (2009), a higher zero correlation between two regressors is a sufficient way to the existence of multicollinearity. The Variance Inflation Factor (VIF) test shows a value greater than 10 so there is a multicollinearity problem. To overcome this problem, the institutional variables will not present in the same regression. The separation of institutional variables in each regression is consistent with the studies of Poghosyan (2013) and Malim et al. (2017). The result of the VIF is included in appendix B. Apart from the institutional variables, the higher correlation between the explanatory variables is less than 80% with a higher 62.3% between risk aversion and the size of operations. Table 4 shows a negative correlation between risk aversion, credit risk, inflation, interest rate risk, and the net interest margin whereas the size of operations, liquidity risk, specialization in lending, GDP growth, and institutional variables are positively correlated the net interest margin. 36 Table 4 Correlation matrix Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (1) NIM 1.000 (2) Size of Operations 0.045 1.000 (3) Risk Aversion -0.013 -0.623 1.000 (4) Credit Risk -0.130 -0.400 0.325 1.000 (5) Liquidity Risk 0.057 -0.170 0.043 -0.009 1.000 (6) Specialization in Lending 0.352 0.127 -0.175 -0.345 0.164 1.000 (7) Inflation -0.046 -0.036 -0.108 0.196 -0.016 -0.098 1.000 (8) GDP growth 0.009 0.040 0.043 -0.290 -0.006 0.047 -0.230 1.000 (9) Interest rate risk -0.032 -0.128 0.018 0.021 -0.015 -0.132 0.283 0.205 1.000 (10) Control of Corruption 0.112 0.232 -0.085 -0.322 0.025 0.330 -0.229 0.218 -0.281 1.000 (11) Rule of Law 0.069 0.205 -0.074 -0.336 0.030 0.300 -0.234 0.186 -0.343 0.927 1.000 (12) Regulatory Quality 0.086 0.176 -0.019 -0.328 -0.001 0.233 -0.329 0.240 -0.399 0.869 0.905 1.000 37 4.3 Regression Estimation 4.3.1 Pooled Models Table 5 presents the result of the pooled model regressions. For a better understanding of the impact of the explanatory variables on the net interest margin, several models are tested. The first column in table 5 is conducted without controlling for institutional variables. Then in columns 2, 3 and 4 the institutional variables (control of corruption, rule of law, and regulatory quality respectively) are presented. Taking control of the time is considered by using the time dummies in columns 5, 6 and 7 repeating the models of 2, 3, and 4 with the inclusion of the time dummies. The country dummies are dropped form the regression due to multicollinearity problem. The separation of the institutional variables is done because the VIF test suggests a multicollinearity problem. The result of the VIF test is presented in the appendix B. Finally, the columns 8 testing the effect of the interaction between the interest rate risk and the credit risk on the net interest margin. 38 Table 5 Pooled models (1) (2) (3) (4) (5) (6) (7) (8) VARIABLES NIM NIM NIM NIM NIM NIM NIM NIM Sizeof Operations -0.00117*** -0.00132*** -0.00112*** -0.00122*** -0.00119*** -0.00103*** -0.00115*** -0.00117*** (0.000136) (0.000167) (0.000173) (0.000190) (0.000160) (0.000169) (0.000183) (0.000132) Risk Aversion 0.0211*** 0.0191*** 0.0216*** 0.0207*** 0.0199*** 0.0220*** 0.0211*** 0.0212*** (0.00601) (0.00549) (0.00585) (0.00566) (0.00530) (0.00574) (0.00549) (0.00619) Credit Risk 0.00139 0.00188 0.00117 0.00175 0.00239 0.00190 0.00278 0.00103 (0.00590) (0.00605) (0.00612) (0.00627) (0.00614) (0.00618) (0.00632) (0.00919) Liquidity Risk -6.78e-05*** -6.66e-05*** -6.79e-05*** -6.59e-05*** -6.77e-05*** -6.91e-05*** -6.49e-05*** -6.78e-05*** (2.07e-05) (2.00e-05) (2.07e-05) (1.93e-05) (2.01e-05) (2.07e-05) (1.92e-05) (2.06e-05) Specialization in lending 0.0189*** 0.0177*** 0.0192*** 0.0186*** 0.0178*** 0.0191*** 0.0184*** 0.0189*** (0.00389) (0.00371) (0.00377) (0.00364) (0.00371) (0.00379) (0.00362) (0.00396) Inflation 0.0110 0.0124* 0.0107 0.0119* 0.0188** 0.0164** 0.0202** 0.0110 (0.00685) (0.00644) (0.00654) (0.00642) (0.00688) (0.00674) (0.00661) (0.00683) GDP growth 0.00248 -0.00314 0.00372 0.000154 0.0158 0.0223 0.0176 0.00242 (0.0109) (0.0115) (0.0122) (0.0130) (0.0160) (0.0162) (0.0163) (0.0108) Interest rate risk 0.584*** 0.602*** 0.577*** 0.607*** 0.665*** 0.645*** 0.696*** 0.576*** (0.107) (0.109) (0.103) (0.113) (0.110) (0.107) (0.105) (0.141) Control of Corruption 0.00130* 0.00140* (0.000701) (0.000714) Islamic banks -0.000224 -0.000326 -0.000217 -0.000169 -0.000383 -0.000267 -0.000151 -0.000228 (0.00161) (0.00164) (0.00162) (0.00158) (0.00171) (0.00168) (0.00163) (0.00163) Rule of Law -0.000390 -3.44e-05 (0.000722) (0.000750) Regulatory Quality 0.000611 0.00143 (0.00102) (0.00103) Interest rate risk*credit risk 0.0896 (1.384) Constant 0.0371*** 0.0415*** 0.0359*** 0.0384*** 0.0353*** 0.0310*** 0.0333*** 0.0371*** (0.00382) (0.00415) (0.00441) (0.00423) (0.00468) (0.00514) (0.00467) (0.00387) Time dummies NO NO NO NO YES YES YES NO Observations 1,717 1,717 1,717 1,717 1,717 1,717 1,717 1,717 R-squared 0.098 0.099 0.098 0.098 0.112 0.111 0.112 0.098 Note: Robust standard errors in parentheses, * is indication to the significant level, ***significant at 1%, ** significant at 5%, and * significant at 10%. 39 4.3.2 Fixed Effect Models Table 6 represents the fixed effect models. The first column excludes the institutional variables. In columns 2, 3, and 4, the institutional variables (control of corruption, rule of law, and regulatory quality respectively) are presented in the models. In columns 5, 6, and 7, the regressions take into consideration the dummies of time. Column 8 adds the interaction between interest rate risk and credit risk. In addition, the table shows the result of the Hausmen test. Since the chi-square is significant for all models except column 6, so the fixed effect model allows for obtaining results that are more reliable (random effect results are presented in Appendix C). The table is followed by a discussion based on the fixed effect model results in comparison with pooled models. The R squared represents the fitness of the regression equation in explaining the total variation of the explanatory variables on the dependent variable (Porter & Gujarati, 2009). The R Squared varies across models, for the fixed effect model, model 7 has the highest value at 6%. Followed by model 8 with 5.5%. Comparing model 1 with model 8, which differs only by the interaction between interest rate risk with credit risk variable. It shows an increase of the R-squared by 3.2% highlighting the importance of the variable to include in the model, while the coefficient value is statisticallyimportant with a 1% increase in the interaction variable leading to reduce the NIM by 1.089%. 40 Table 6 Fixed effect models (1) (2) (3) (4) (5) (6) (7) (8) VARIABLES NIM NIM NIM NIM NIM NIM NIM NIM Sizeof Operations -0.00178 -0.00179 -0.00169 -0.00169 -0.00464*** -0.00447*** -0.00378** -0.00441** (0.00148) (0.00149) (0.00148) (0.00146) (0.00172) (0.00170) (0.00159) (0.00171) Risk Aversion 0.00969 0.00917 0.00988 0.0111 0.00277 0.00377 0.00582 0.00411 (0.0218) (0.0220) (0.0217) (0.0212) (0.0212) (0.0208) (0.0202) (0.0209) Credit Risk -0.0155* -0.0154* -0.0156* -0.0132 -0.0136* -0.0136* -0.0127* -0.00888 (0.00862) (0.00859) (0.00862) (0.00806) (0.00746) (0.00739) (0.00712) (0.00737) Liquidity Risk -2.62e-05 -2.62e-05 -2.46e-05 -2.97e-05 -3.58e-05 -3.36e-05 -3.54e-05 -3.39e-05 (3.80e-05) (3.78e-05) (3.79e-05) (4.09e-05) (4.17e-05) (4.17e-05) (4.27e-05) (4.19e-05) Specialization in lending 0.00338 0.00359 0.00334 0.00349 0.00329 0.00304 0.00319 0.00330 (0.00502) (0.00508) (0.00500) (0.00480) (0.00451) (0.00438) (0.00443) (0.00446) Inflation 0.00334 0.00338 0.00369 0.00200 0.00604* 0.00716** 0.00428 0.00560* (0.00267) (0.00264) (0.00267) (0.00276) (0.00318) (0.00299) (0.00328) (0.00300) GDP growth 0.00630 0.00579 0.00678 0.00798 0.0132* 0.0150** 0.0135* 0.0140** (0.00610) (0.00584) (0.00619) (0.00614) (0.00698) (0.00745) (0.00706) (0.00700) Interest rate risk -0.0311 -0.0364 -0.0237 -0.0385 0.00320 0.0170 -0.00139 0.113 (0.0597) (0.0578) (0.0601) (0.0603) (0.0737) (0.0723) (0.0754) (0.103) Control of Corruption 0.00162 0.00143 (0.00176) (0.00158) Rule of Law -0.00152 -0.00265 (0.00179) (0.00199) Regulatory Quality -0.00643*** -0.00508*** (0.00164) (0.00162) Interest rate risk* credit risk -1.089* (0.620) Constant 0.0644* 0.0644* 0.0626* 0.0628* 0.126*** 0.122*** 0.108*** 0.120*** (0.0345) (0.0346) (0.0344) (0.0340) (0.0388) (0.0384) (0.0360) (0.0386) Time dummies NO NO NO NO YES YES YES NO Hausmen test 27.54*** 27.60*** 27.31** 30.21*** 38.56** 24.05 33.14** 39.74*** R-squared 0.023 0.024 0.024 0.040 0.052 0.053 0.061 0.055 Note: robust standard errors in parentheses, * shows the significant level, *** at 1%, ** at 5%, and * at 10%. 41 Size of the operation is found to be negative in all estimations of the pooled model. However, in the fixed effect models was only significant when the dummies are included. The findings support the first hypothesis, there is a negative significant effect. The interpretation is that large banks are able to distribute their operations to different agents granting a lower margin (Almarzoqi & Naceur, 2015), the result was obtained by the authors in Azerbaijan and U