23
Credit risk in dual banking systems: does competition matter? Empirical evidence Mohsin Ali Taylors Business School, Taylors University Lakeside Campus, Subang Jaya, Malaysia Mudeer Ahmed Khattak Department of Business Administration, Iqra University Islamabad Campus, Islamabad, Pakistan, and Nafis Alam School of Accounting and Finance, Asia Pacific University of Technology and Innovation (APU), Kuala Lumpur, Malaysia Abstract Purpose The study of credit risk has been of the utmost importance when it comes to measuring the soundness and stability of the banking system. Due to the growing importance of Islamic banking system, a fierce competition between Islamic and conventional banks have started to emerge which in turn is impacting credit riskiness of both banking system. Design/methodology/approach Using the system GMM technique on 283 conventional banks and 60 Islamic banks for the period of 20062017, this paper explores the important impact of size and competition on the credit risk in 15 dual banking economies. Findings The authors found that as bank competition increases credit risk seems to be reduced. On the size effect, the authors found that big Islamic banks are less risky than big conventional banks whereas small Islamic banks are riskier than small conventional banks. The results are robust for different panel data estimation models and sub-samples of different size groups. The findings of this paper provide important insights into the competition-credit risk nexus in the dual banking system. Originality/value The paper is specifically focused on credit risk in dual banking environment and tries to fill the gap in the literature by studying (1) do the Islamic and conventional banks exhibit a different level of credit risk; (2) does competition in the banking system impact the credit risk of Islamic and conventional banks and finally (3) do the big and small banks exhibit similar levels of credit risk. Keywords Credit risk, Competition, Size, Dual banking system Paper type Research paper 1. Introduction Banks have always been the central point of the economic health of the country. Since the onset of the 200708 global financial crisis, a renewed interest in the banking risk and its impact on the economy has been at the forefront in banking research (Anginer and Demirguc- Kunt, 2014). Credit risk which is the potential loss of principal due to borrowersfailure is one of the most important risks banks encounter and has been accepted as the main reason for many banking crises occurred in developing economies ( Demirg uç-Kunt and Detragiache, 2005). The crucial discussion related to the stability of the bank system is to obtain a negative trade-off between competition and bank financial fragility is that banks fully control the risk level of their asset portfolio. According to Saurina et al. (2007, p. 9), excessive competition among banks could threaten the solvency of particular institutions and, at an aggregate level, hamper the stability of the entire banking system, meaning bank failures always come Credit risk in dual banking systems The current issue and full text archive of this journal is available on Emerald Insight at: https://www.emerald.com/insight/1746-8809.htm Received 11 January 2020 Revised 20 June 2020 17 November 2020 16 February 2021 Accepted 2 May 2021 International Journal of Emerging Markets © Emerald Publishing Limited 1746-8809 DOI 10.1108/IJOEM-01-2020-0035

Credit risk in dual banking systems: does competition

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Credit risk in dual banking systems: does competition

Credit risk in dual bankingsystems: does competition matter?

Empirical evidenceMohsin Ali

Taylor’s Business School, Taylor’s University – Lakeside Campus,Subang Jaya, Malaysia

Mudeer Ahmed KhattakDepartment of Business Administration, Iqra University – Islamabad Campus,

Islamabad, Pakistan, and

Nafis AlamSchool of Accounting and Finance,

Asia Pacific University of Technology and Innovation (APU),Kuala Lumpur, Malaysia

Abstract

Purpose – The study of credit risk has been of the utmost importance when it comes to measuring thesoundness and stability of the banking system. Due to the growing importance of Islamic banking system, afierce competition between Islamic and conventional banks have started to emerge which in turn is impactingcredit riskiness of both banking system.Design/methodology/approach – Using the system GMM technique on 283 conventional banks and 60Islamic banks for the period of 2006–2017, this paper explores the important impact of size and competition onthe credit risk in 15 dual banking economies.Findings – The authors found that as bank competition increases credit risk seems to be reduced. On the sizeeffect, the authors found that big Islamic banks are less risky than big conventional banks whereas smallIslamic banks are riskier than small conventional banks. The results are robust for different panel dataestimation models and sub-samples of different size groups. The findings of this paper provide importantinsights into the competition-credit risk nexus in the dual banking system.Originality/value – The paper is specifically focused on credit risk in dual banking environment and tries tofill the gap in the literature by studying (1) do the Islamic and conventional banks exhibit a different level ofcredit risk; (2) does competition in the banking system impact the credit risk of Islamic and conventional banksand finally (3) do the big and small banks exhibit similar levels of credit risk.

Keywords Credit risk, Competition, Size, Dual banking system

Paper type Research paper

1. IntroductionBanks have always been the central point of the economic health of the country. Since theonset of the 2007–08 global financial crisis, a renewed interest in the banking risk and itsimpact on the economy has been at the forefront in banking research (Anginer and Demirguc-Kunt, 2014). Credit risk which is the potential loss of principal due to borrowers’ failure is oneof the most important risks banks encounter and has been accepted as the main reason formany banking crises occurred in developing economies (Demirg€uç-Kunt andDetragiache, 2005).

The crucial discussion related to the stability of the bank system is to obtain a negativetrade-off between competition and bank financial fragility is that banks fully control the risklevel of their asset portfolio. According to Saurina et al. (2007, p. 9), “excessive competitionamong banks could threaten the solvency of particular institutions and, at an aggregate level,hamper the stability of the entire banking system”, meaning bank failures always come

Credit risk indual banking

systems

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1746-8809.htm

Received 11 January 2020Revised 20 June 2020

17 November 202016 February 2021

Accepted 2 May 2021

International Journal of EmergingMarkets

© Emerald Publishing Limited1746-8809

DOI 10.1108/IJOEM-01-2020-0035

Page 2: Credit risk in dual banking systems: does competition

together with increased competition in the banking system. Theoretically, there are twomainassumptions regarding the relationship between competition and banking stability; thecompetition-fragility view assumes that banks in excessive competition banking systems areless concentrated and have less market power which will erode the charter values of thebanks. Banks have less market power to extract the monopoly benefits from the chartervalues which will eventually lead them to take on more risks with riskier policies such aslowering the level of capital or acquire more credit risk in the loan portfolio.

With the emergence of Islamic banking in dual banking economies which employs differentunderlying contracts for financing (Alam et al., 2017), the nature of credit risk has evolvedwithinthe banking domain. Islamic banks’ exposure to credit risk differs from conventional banks inthe sense that different Islamic banking contracts have a different level of credit risk. Forexample, in a Mudarabah (profit sharing) contract, Islamic banks do not participate in thedecision-making process giving rise to information asymmetry between the Islamic bank andentrepreneur thus exposing Islamic banks to a high level of credit risk. Credit riskmay also arisein the event borrower default, whether intentionally or unintentionally. Inmurabaha (sale-basedmarkup financing) credit risk arises when a customer fails to honor the payment obligations(Alam et al., 2017). Musharaka (profit and loss sharing) contract is exposed to the least of creditrisk among Islamic banking contract while Boumediene (2011) noted a presence of credit risk inbinding Ijarah (leasing) contract when the lessee may cancel the lease before the stipulated time.With Islamic banking becoming a significant component ofmany emerging economies (IbrahimandAlam, 2017;Akhtar et al., 2017; Azmi et al., 2019) it is important to explore the extent of creditrisk and the environment surrounding it in these dual banking economies.

Credit risk in the conventional banking sector has been exhaustively researched andhas explored various determinants like macroeconomic and bank-specific factors (Bergerand DeYoung, 1997; Louzis et al., 2012; Chaibi and Ftiti, 2015). Comparative studiesbetween conventional and Islamic banking studies have looked into the determinants ofcredit risk for both types of banks (�Cih�ak and Hesse; Boumediene, 2011; Beck et al.,2013a, b; Abedifar et al., 2013; Lassoued, 2018). Kabir et al. (2015) analysis showed mixedresults showing that the measure chosen to proxy credit risk can have a significantbearing on the results.

Most of the above literature concludes that conventional banks face increased credit risk andare riskier while some observe that there is no significant difference in credit risk between the twobanking systems. The literature also found significant cross country variations in comparativerisk analysis. This gives us a moment to ponder upon if we are comparing two equivalent entitiesand why there will be a difference in the credit riskiness of both banking system. One way toinvestigate is to look at the competitive element between the two banking system. Islamic bankinghas been growing rapidly and already occupies a major share in many countries banking system.Islamic bankingwhichhas overUSD2.2 trillion in asset covers 42%of thebanking system inGCC,24.4% in Asia and 29.1% of the MENA (ex. GCC) banking system (IFSB, 2018, see Table A1).

The statistics point to one key aspect that Islamic banking is giving conventional banks arun for their money in the above-listed jurisdictions which will be impacting the competitionbetween the two banking system (in addition to segment competition) and in turn the risk-taking behavior. The traditional view from the previous economic and finance literatureholds that there is a positive relationship between the risks that banks take and thecompetition that the banks face (Repullo, 2004). Another key consideration is the comparativesize of Islamic banks with respect to their conventional counterparts. It is noted that Islamicbanks are typically smaller than their conventional rivals, with the industry’s high fixed costsputting Shariah-compliant lenders at a disadvantage due to their relative lack of scale(IFN, 2018). In terms of banking stability, the proposition of “too big to fail” might prompt alarger bank to assume more risk and hence the relationship between size and stability isexpected to be negative.

IJOEM

Page 3: Credit risk in dual banking systems: does competition

We summarize our above discussion in three themes. First, since Islamic bankingproducts have different credit risk dynamics than their conventional counterparts, it will beinteresting to hypothesize whether Islamic and conventional banks exhibit a different level ofcredit risk. Second, competition could be an important factor for banks to define their creditrisk strategies in dual banking economies. This makes it imperative to hypothesize that if thecompetition has a bearing on the credit risk of both conventional and Islamic banks. Third,the literature has suggested that size could impact the credit risk of banks as banks withdifferent magnitude of resources at their disposal may manage their credit risks differently.This leads us to our third main hypothesis, i.e. whether big and small Islamic andconventional banks exhibit similar levels of credit risk.

Our results show that as bank competition increases, credit risk seems to reduce in thebanking system. This supports the competition-stability view which suggests that competitionlowers the lending rates in themarketswhichmakes it easier for the borrowers to repay the loansthus less risk.Thedeterminants of credit risk seem to be similar for both conventional and Islamicbanks except for the bank size. To look into the impact of bank size in more detail we split ourmain sample into big and small banks. The result of the split sample-based estimation shows thatthe big Islamic banks are less risky than big conventional bankswhereas small Islamic banks arerisky than small conventional banks. This shows the existence of inappropriate credit controlsystem in smaller Islamic banks. The results also showed that the regulatory framework is foundto be more important for Islamic banks as compared to the conventional bank as it seems toincrease Islamic banks’ credit risk. Imposing the same regulatory framework is not working wellfor Islamic banks, raising the need for a separate framework.

The remainder of the paper is organized as follows. Section 2 provides a brief review of theliterature. Section 3 discusses our data, sample selection process, variables andmethodology.Section 4 discusses the findings, while Section 5 concludes the discussion.

2. Literature reviewMost of the banking studies on credit risk in the past have been heavily concentrated onfocusing on macroeconomic determinants; bank-specific determinants or the combination oftwo (Berger andDeYoung, 1997; Louzis et al., 2012; Beck et al., 2013a, b).While the focus of thestudy was primarily on conventional banks, dual banking literature started to gainmomentum in the last 6 years given that credit risk profile for Islamic banks differs from theirconventional counterpart. The unique features of the Islamic financial contracts contribute toadditional credit risk faced by Islamic banks. In the case of murabaha transactions, theIslamic banks have a potential credit risk that it delivers the asset to the client withoutreceiving the payment in time. In the case of a non-bindingmurabaha, the client has a right toreject the delivery of the product acquired by the bank; thus, further rendering the banks tothe price and market risk.

The studies regarding the effects of competition on the bank risk-taking behavior havealways been focused on the conventional banks and lack evidence for Islamic banks (Becket al., 2013a, b; Berger et al., 2009). In another study on competition, Gonzalez et al. (2017)studied the relationship between competition and risk-taking behavior of banks in theMENAregion during the period 2005–2012. The study found that a U-shaped relationship betweencompetition and risk-taking for the banks in the MENA region. In a very recent paper,Moudud-Ul-Huq (2020) on a study on conventional banks in the BRICS economy found that incompetitive markets large banks are more efficient than small banks and there is a nonlinearrelationship between competition performance and risk. Table 1 presents the findings of thestudies on credit risk in dual banking economies.

Thus, it can be seen from the above table that almost all studies on credit risk in Islamic ordual banking system have ignored the impact of competition in determining the credit risk.

Credit risk indual banking

systems

Page 4: Credit risk in dual banking systems: does competition

The traditional view on competition risk nexus that the incentive of banks to take on morerisk is increased as competition increases. This relationship is based on the “competition-fragility view”, which is also called “charter value hypothesis” or “franchise value paradigm”(Allen and Gale, 2004). The competition-fragility view assumes that banks in excessivecompetition banking systems are less concentrated and have less market power which willerode the charter values of the banks. Contrary to this view, recent literature based on BDNmodel which is also known as “competition-stability view” by (Boyd and De Nicolo, 2005;Boyd et al., 2009) suggested that the risk-taking behavior of banks and competitionwithin thesector have a negative correlation.

Another strand of the literature highlights that banking efficiency will also have a bearingon the credit risk. In this regard, Berger and DeYoung (1997) postulated the “badmanagement”hypothesis, inwhich banks operatingwith low levels of efficiency have higher costs largely dueto inadequate credit monitoring and inefficient control of operating expenses. Declines in costand revenue efficiency will temporally precede increases in banks’ risk due to credit,operational, market and reputational problems. Berger and DeYoung (1997) also present the“cost skimping” hypothesis which suggests a trade-off between short-term cost efficiency andfuture risk-taking due to moral hazard considerations. In such cases, banks appear to be morecost-efficient as they devote fewer resources to credit risk assessment.

The literature on competition and credit risk is also inconclusive. One of the mainargument is that competition increases the risk in banking (competition-fragility view).Research on Spanish banking sector done by Jim�enez et al. (2013) concluded that low level ofbank competition results in low risk which is supported by Leroy and Lucotte (2017) andKabir andWorthington (2017). Contrary to this, there are proponents of competition-stabilityview, who suggest that competition lowers the lending rates in the markets which makes iteasier for the borrowers to repay the loans thus less risk.

Even though Kabir et al. (2015) tried to fill the gap on the comparative credit risk literaturein the dual banking system, their Merton’s distance-to-default model was not appropriate tomeasure the credit risk in Islamic Banks for the full sample as most of the Islamic banks are

AuthorsSampleperiod Data sample Key findings

Use of competition inempirical model

Abedifar et al.(2013)

1999–2009 553 banks in24 countries

Small Islamic Banks (IBs) havelower credit risk and insolvencyrisk than Conventional Banks(CBs)

No

Beck et al.(2013a,b)Khediri et al.(2015)

1995–20092003–2010

510 banks in21 countries44 CBs and 18IBs in 5 GCCnations

IBs have lower credit risk CBIBs have less credit risk than CBs

NoNo

Kabir et al.(2015)

2000–2012 156 CBs and37 IBs in 13countries

IBs have low credit risk accordingto DD measures but the Z–scoreand NPL indicate higher creditrisk than CBs

NoUse of concentrationrather thancompetition

Louhichi andBoujelbene(2016)

2005–2012 87 CBs and 30IBs in 10countries

IBs were less burdensomeregarding credit risk

No

Lassoued(2018)

2005–2015 22 CBs and 17IBs inMalaysia

IBs have a higher degree of creditrisk than CBs

NoUse of concentrationrather thancompetition

Table 1.Credit risk duties onIslamic banks/dualbanking system

IJOEM

Page 5: Credit risk in dual banking systems: does competition

not listed which significantly reduced their sample size. In the light of above discussion it canbe observed that studies of competition implications of Islamic and conventional banksremain limited and with inconclusive impact especially in the context of credit risk.

In a recent article on the theme, Alam et al. (2018) found that competition and risk arepositively related to the overall banking system and inversely related to Islamic banks. Ourpaper extends their study by including the impact of concentration on credit risk in ourbaseline model which strengthens the empirical model. Furthermore, while estimating theLerner index, we include time-trend variables, which are used as control variables to accountfor heterogeneity in the cost function across time.

3. Data and methodology3.1 DataWe examine a panel of 343 banks from 15 economies for the period 2006–2017. Out of 343total banks, 283 banks are conventional, and 60 are Islamic Banks [1]. The study uses FitchConnect database which also helps to classify the banks. The macro-level data are extractedfrom the World Development Indicator of the World Bank. To get robust results, few filtersare applied to the data: (1) Following Beck et al. (2013a, b), to avoid any double-counting, thedataset considers the available consolidated data, otherwise unconsolidated data is used, (2)the dataset is restricted to at least 4 years of banking observations. And (3) the study limitsthe data to only dual banking countries where Islamic banks and conventional banks co-existand operate together (Iran and Sudan are excluded as they have 100% Islamic banking).Considering only 15 dual banking economies, the dataset consists of following countries:Bahrain, Bangladesh, Indonesia, Jordan, Kuwait, Lebanon, Malaysia, Pakistan, Oman, Qatar,Saudi Arabia, Tunisia, Turkey, the United Arab Emirates and Yemen, and has 4,116 bank-year observations. To control for the heterogeneity in the market structure of differentcountries, we include the Herfindahl-Hirschman index (HHI) in the regression. Also, themeasure of competition is at the bank-level which controls the market power of Islamic banksaccording to the bank specialization (more on the Islamic banking share in each country isgiven in Table A1).

Additionally, we split the sample into large and small banks and re-estimate the model.Banks with total assets greater than USD 1 billion (Big) are categorized as big banks, whilebanks with total assets smaller than USD 1 billion (Small) are considered as small banks(Cihak and Hesse, 2010).

3.2 VariablesFollowing Abuzayed et al. (2018), the ratio of non-performing loans to gross loans (NPLs) isused as a proxy for credit risk. NPLs show the degree of expected losses. Higher NPLsindicates higher credit risk and shows the weakness of the bank in managing the credit risk.Besides NPLs, we also use the Z-score. This proxy has been used exhaustively in theliterature (�Cih�ak and Hesse; Demirg€uç-Kunt and Detragiache, 2011). Z-score is measured atbank-level and is considered as a powerful predictor of banks’ default (Alaeddin et al., 2019;Cihak and Hesse, 2010). It employs the accounting data to calculate the solvency risk of abank. The returns are measured using the ratio of equity to assets ratio plus return onaverage assets (ROAA) and the standard deviation of ROAA is often used to measure thechange in returns. The assumption is that a bank defaults when its capital falls to zero.Z-Score can be written as:

Z ¼ ðEQTAþ ROAAÞ=sdROAA 3 – 1

Credit risk indual banking

systems

Page 6: Credit risk in dual banking systems: does competition

The study considers three different groups of variables to address the factors that determinethe credit risk in dual banking. The first group includes the bank-specific variables which areadded to investigate the possible determinants among the banks in the sample. The secondgroup of variables is to address the banking market structure which consists of an industry-specific variable, and lastly, to address the heterogeneity among the countries in the sample,this study controls for country-specific characteristics which are gross domestic productgrowth rate, inflation rate and regulations.

3.2.1 Bank-specific controls. To control for bank and country-specific characteristics, thisstudy employs a set of bank-specific variables in the regressions. Cost efficiency is one of theimportant determinants when it comes to risks in banking. Banks start extra managerialoperations once that a loan is past due which results in increased cost incurred to monitor theloan issues. This leads to higher cost inefficiency. The empirical literature is inconclusive onthe impact of efficiency on credit risk. A study done by Chaibi and Ftiti (2015) found anegative relationship between efficiency and credit risk. The cost-efficiency ratio (cost (in)efficiency) is measured by the ratio of operating expenses (i.e. non-interest expenses) to totalassets. Return on assets (ROA) is included to control for bank profitability. Bankswith higherROAare associatedwith higher growth and resilient to adverse shocks, therefore, expected tobe negatively related to credit risk (Gulati et al., 2019). Equity ratio (EQTA) is also controlled;it is argued that banks with higher equity ratio are more likely to engage in riskier activitiesbecause they have a higher capacity for risk-taking. Following �Cih�ak and Hesse (2010) whoprovide evidence that size affects the banks’ risk, this research also controls for banks’ size(LnTA) using the log of total assets. To control the extent of bank lendingwe control for grossloans to total assets (GLTA). Loan ratio is used to control the credit exposure of the bank andwhich might have significant effects on the banks (Rashid et al., 2017). Furthermore,Birchwood et al. (2017) argue that banks that have higher loan ratios are likely to have poorperformance and be risky and unstable. Following Gulati et al. (2019), income diversification(NONIT) is proxied as non-interest income divided by total income. It is argued thatdiversified banks have less credit risk (Chaibi and Ftiti, 2015).

This research uses the Lerner index which is an inverse measure of market competition toexplore if competition determines the credit risk. Lerner index has been used by severalresearchers to measure bank competition (Amidu and Wolfe, 2013; Kabir and Worthington,2017; Weill, 2011). The index is defined as how much a bank can charge above the marginalcost. Lerner Index varies from 0 to 1. When closer to 0, the market is said to be in purecompetition, where the price intersects the marginal cost of the bank, giving no pricing powerto the bank. While closer to 1, the market is said to be monopolistic competition, where thebanks enjoy mark up in prices above marginal costs. This indicates an increase in the marketpower of the bank (see Amidu and Wolfe, 2013, for further explanation on the estimation ofLerner index). The Lerner Index is given as:

Lerner indexit ¼ Priceit − Mit

Priceit3 – 2

Following the competition-fragility hypothesis, we expect a positive relationship betweenLerner and the risk measure, which would indicate that when banks in a higher competitiveenvironment are more fragile as compared to the banks in a less competitive environment [2].

To investigate any possible differences in the business model, that is conventional andIslamic, an Islamic bank dummy (Islamic) is introduced, which takes the value of 1 for Islamicbanks and 0 for conventional banks. Besides this, to explore the effect of the recent financialcrisis 2007–2009, a crisis dummy variable is included in the model specifications. Followingthe studies of Azmi et al. (2019), dummy variable crisis takes a value of 1 for the year 2008–09and 0 otherwise.

IJOEM

Page 7: Credit risk in dual banking systems: does competition

3.2.2 Industry-specific controls. The research accounts for market concentration toexamine if the market structure determines the bank’ credit risk. Boyd and De Nicolo (2005)argue that concentration increases the lending rates. Kasman and Kasman (2015) found thatan increase in lending rates results in increased credit risk. The literature seems uncertainabout the relationship between concentration and risk (Berger et al., 2009). Therefore, thispaper uses the market concentration of total assets using the HHI. HHI is estimated bysquaring the market share of each bank and then summing the squares, which equates asfollows:

HHI ¼Xn

j¼1

ðMSnÞ2 3 – 3

Following the concentration-stability hypothesis, we expect a negative relationship betweenHHI and the riskmeasure, whichwould indicate that banks in concentratedmarkets are morestable as compared to those in a less concentrated environment.

3.2.3 Country-level controls. Following Ibrahim and Rizvi (2017), GDP growth rate andinflation rate are used to control for macro characteristics that are expected to affect the creditrisk of banks. The study employs the GDP growth rate (GDPG) where a higher GDPG isexpected to decrease the level of credit risk. Besides GDPG, the inflation rate is also controlled.It is said that inflation can result in economic ambiguity (Kelsey and Roux, 2016) where itdecreases the real earnings and results in increased NPLs. Therefore, inflation is expected toincrease credit risk. Moreover, the difference in the regulatory environment (Reg) among thecountries is also controlled. Following Klomp and De Haan (2012), we use an equal-weightedaverage of four regulatory measures (power of supervisory agencies, capital requirementsindex, private monitoring and level of restrictions on bank’s activities) to control for theregulatory environment of the countries under study. Table 2 provides the definition andsources of the variables used in this study.

3.3 Descriptive statisticsTable 3 presents descriptive statistics for our sample. Except for EQTA and impaired loans,most variables are quite stable. Conventional banks have higher ROA and are bettercapitalized looking at our statistics. Conventional banks have higher Z-score as compare toIslamic banks which shows their relative stability. Both banking systems have a similar levelof impaired loan, asset size and income diversification.

3.4 MethodologyThis paper adopts the generalized method of moments (GMM) estimator to examine thedrivers of the Credit risk. One of the reasons for preferring GMMover other techniques is thatthe number of cross-sections is significantly higher than the time series. Also, it works wellwhen there are endogeneity issues. Furthermore, the system GMM estimator is preferredbecause it corrects the biases in the difference GMM (Arellano and Bover, 1995; Blundell andBond, 1998). This methodology is appropriate when the dependent variable has a dynamicnature andwhere the control variables in themodelmight correlate with the error terms of themodel. Another reason for using system GMM is that it has small variances which improveaccuracy in the estimators (Blundell and Bond, 1998). SystemGMM is also preferredwhen thedependent variable has a dynamic nature and where the control variables in the model mightcorrelate with the error terms of the model. And since the data is bank-level, the existence ofheteroscedasticity is highly likely. The two-step system GMM also improves the quality ofthe results while looking after the issues like endogeneity, serial correlation andheteroscedasticity.

Credit risk indual banking

systems

Page 8: Credit risk in dual banking systems: does competition

This research considers all the bank-level explanatory variables as endogenous variables andall macro-level variables are considered strictly as an exogenous variable in the model. Theabsence of serial correlation is confirmed by the insignificant values of AR (1) and AR (2) toconfirm no serial autocorrelation. Besides this, since we have controlled for the bank-specificand country-specific variables, we can safely assume homoscedasticity of error terms. Wehave also confirmed homoscedasticity by plotting the error terms of our primary model. Dueto this, we do not cluster the error terms, which allows us to rely on the Sargan test to confirmthe validity of instruments (Roodman, 2009).

The following model is estimated to address the objective of this research:

NPLit ¼ γ0 þ γ1NPLijt�1 þ γ2Yit þ γ3Dummyþ γ4Macjt þ εit 3 – 4

In the abovemodel, i indicates the bank, t indicates the year, and j indicates the country. NPLjtdenotes non-performing loans, the proxy for credit risk, NPLijt�1 indicates the one-periodlag for NPLjt, Yit represents bank-specific determinants. Dummy represents the dummyvariables. Lastly, Mjt is a vector of country-level variables and εit indicates the residuals.

4. Explanation of resultsWe present our estimations for different variants of our basic model in Table 4. First, fourregression estimations have NPL as dependent variable whereas Z-score is the dependent

Variable Definition Source

Dependent variablesNon-performingloans

Non-performing loans/gross loans Fitch Connect database(FCD)

Z-score Sum of the equity ratio and total assets divided by thestandard deviation of ROAA (sdROAA)

Fitch Connect databaseAuthors’ calculation

Explanatory variables

Bank-specific variablesCost (in)efficiency(INEFF)

Operating expenses/total assets Authors’ calculation

Profitability (ROA) Net income/total assets FCDBank size (lnTA) Natural log of total assets FCDEquity ratio (EQTA) Equity/total assets FCDLoan ratio (GLTA) Gross loans/total assets FCD

Authors’ calculationLerner index An estimate of banking competition FCD

Authors’ calculationHHI An estimate of market structure Authors’ calculationDiversification(NONIT)

Non-interest income/total income Authors’ calculation

Islamic Dummy variable, equal to 1 for Islamic bank and 0 forthe conventional bank

Crisis Dummy variable, equal to 1 for the year 2008–2009 and0 otherwise

Authors’ calculation

Macroeconomic variablesRegulations (Reg) Equal-weighted average of four regulatory measures World Development

Indicators (WDI)GDP growth (GDPG) Annual GDP growth rate (WDI)Inflation (inflation) Annual inflation rate (WDI)

Table 2.Definition and sourcesof the variables used inthis study

IJOEM

Page 9: Credit risk in dual banking systems: does competition

NPLs

Z-score

INEFF

ROA

lnTA

Lerner

HHI

NONIT

GLTA

Reg

GDPG

Infl

Fullsample

Mean

6.97

27.39

0.02

1.25

8.10

0.35

0.12

0.01

0.59

0.77

4.76

5.35

Sd

11.49

24.69

0.02

2.23

1.66

0.19

0.06

0.01

0.19

0.10

3.30

3.68

ConventionalBanks

Mean

7.01

28.48

0.02

1.31

8.10

0.35

0.12

0.01

0.58

0.77

4.81

5.54

Sd

11.39

25.25

0.02

2.06

1.70

0.19

0.06

0.01

0.19

0.10

3.17

3.63

IslamicBanks

Mean

6.77

21.96

0.02

0.94

8.09

0.31

0.12

0.01

0.62

0.74

4.52

4.43

Sd

12.02

20.85

0.01

2.90

1.44

0.22

0.07

0.01

0.17

0.09

3.91

3.83

Table 3.Descriptive statistics

Credit risk indual banking

systems

Page 10: Credit risk in dual banking systems: does competition

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

NPL

NPL

NPL

NPL

Z-Score

Z-Score

Z-Score

Z-Score

L.lnpl

0.7873

***(0.033)

0.8901

***(0.046)

0.6710

***(0.029)

0.7378

***(0.019)

L.Z.score

0.9299

***(0.033)

0.7593

***(0.082)

0.8377

***(0.043)

0.9529

***(0.028)

ROA

�0.0936*

**(0.035)

�0.0735*

**(0.013)

�0.2055*

**(0.038)

�0.0902*

**(0.017)

0.0673

***(0.026)

0.0787

***(0.016)

0.1692

***(0.038)

0.0748

***(0.016)

lnTA

0.0824

*(0.047)

0.0044

(0.009)

�0.2393*

**(0.054)

�0.1083*

*(0.046)

0.1169

***(0.033)

0.0008

(0.006)

0.1081

***(0.032)

0.0323

(0.025)

NONIT

23.0655*

**(4.680)

4.1146

**(1.949)

6.8245

**(2.651)

0.0165

(2.513)

5.9957

***(2.257)

�1.6691(1.445)

1.4157

(2.274)

�1.6943(1.474)

INEFF

8.9094

*(5.197)

3.9330

*(2.072)

�3.0120(2.820)

10.7214*

**(2.854)

7.8902

**(3.350)

1.7169

(1.075)

10.4461*

**(3.671)

1.3871

(1.252)

GLTA

�0.1927(0.245)

�0.0154(0.113)

�0.2031(0.330)

�0.4185(0.278)

0.1464

(0.199)

�0.1434(0.088)

�0.1821(0.211)

0.1620

(0.113)

Reg

�0.0091(0.232)

0.2921

(0.172)*

�0.3106(0.242)

�0.5251*

*(0.227)

�0.2518(0.172)

�0.1870(0.115)

�0.6566*

**(0.199)

�0.0328(0.075)

GDPG

�0.0245*

**(0.004)

�0.0243*

**(0.005)

�0.0249*

**(0.004)

�0.0275*

**(0.004)

�0.0021(0.002)

�0.0015(0.002)

�0.0037(0.003)

�0.0036*

*(0.001)

Inflation

0.0070

(0.005)

0.0006

(0.005)

0.0064

(0.005)

�0.0069(0.005)

�0.0021(0.002)

�0.0080*

**(0.002)

�0.0148*

**(0.003)

�0.0019(0.002)

Islamic

0.0064

(0.043)

�0.0231(0.027)

�0.1310*

*(0.051)

�0.1217*

*(0.057)

�0.0250(0.033)

0.0634

**(0.028)

0.0654

**(0.032)

�0.0177(0.023)

Crisis

0.0646

*(0.038)

0.0001

(0.031)

0.0716

**(0.031)

0.0095

(0.014)

0.0358

*(0.019)

0.0142

(0.013)

Lerner

0.8115

***(0.289)

�1.0953*

**(0.252)

HHI

�3.5175*

**(0.873)

0.7494

*(0.392)

Constant

�0.5664(0.479)

�0.0328(0.216)

2.8828

***(0.657)

2.3847

***(0.576)

�0.9724*

**(0.297)

0.8780

***(0.321)

0.2252

(0.294)

�0.3774(0.279)

Observations

2,486

2,486

2,484

2,486

2,738

2,738

2,733

2,738

Instruments

67.0000

20.0000

68.0000

82.0000

46.0000

20.0000

54.0000

68.0000

Overall

318.0000

318.0000

318.0000

318.0000

335.0000

335.0000

334.0000

335.0000

AR(1)

0.0000

0.0000

0.0000

0.0000

0.0058

0.0040

0.0061

0.0057

AR(2)

0.6652

0.7750

0.8874

0.9134

0.1592

0.1942

0.8542

0.2094

Sargan

pv

0.1710

0.4739

0.4634

0.3636

0.3603

0.7490

0.0536

0.5271

Note(s):Standarderrors

arein

parantheses;*p<0.1,

**p<0.05,*

**p<0.01

Table 4.Results from systemGMM (full sample)

IJOEM

Page 11: Credit risk in dual banking systems: does competition

variable in the remaining estimations. Our results do not differ due to change in proxies ofcredit risk. The estimatedmodels have stable coefficients and fit well in the panel data setting.The Sargan tests show that the instruments are not correlated with the error terms and areappropriately valid for the models. This means that all the models are free of over-identification and are correctly specified. Additionally, AR(1) and AR(2) tests show theconsistency of the estimators used in the specifications.

We will first focus on the persistence effect of credit risk. The coefficient of the laggeddependent variables (L.lnpl and L.Z-score) consistently remains positively significant andhas a value of between 0 and 1. This suggests the persistence of credit risk which is clearevidence of the accumulation of impaired loans in the dual banking economies. In otherwords, we can also infer that the recovery of NPLs in the dual banking economies is notinstant and the banks in our sample keep on amassing major portion of their impaired loansfrom one year to the other. If we compare the magnitude of the persistence effect in our studywith the literature, we do not find much difference. We may infer that the level of persistenceof bad loans in dual banking economies is quite similar to the level of persistence in singlebanking economies (Das and Ghosh, 2007).

Now we will discuss the impact of bank-specific determinants of credit risk. Asanticipated, based on literature we find ROA to be significant but indirectly related to creditrisk across both estimations. This implies that with an increase in the profitability of thebanks in dual banking economies, they become prudent in extending financing. They seem toimpose screening and monitoring criteria for the borrowers which cause a reduction in thedefault risk. This negative impact of profitability reaffirms the findings of Louzis et al. (2012).This confirms the existence of the “bad management” hypothesis in dual banking economieswhich postulates that loss-making banks tend to generate higher NPLs and have higherdefault risk as compared to profitable banks (Berger and DeYoung, 1997).

We have added Lerner and HHI to the baseline model to investigate if market competitionand concentration determines the credit risk, respectively. The coefficient of Lerner indicatesa lower level of competition increases the credit risk of the banks and vice-versa. This showsthat the competition-stability view suggested by Boyd and De Nicolo (2005) is applicable inthe case of dual banking economies. This view suggests that competition lowers the lendingrates in the markets which makes it easier for the borrowers to repay the loans thus less risk.On the other hand, if the competition is low the banks may increase the lending rates whichwould make it difficult for the borrowers to pay back, hence, increased credit risk. On theother hand, in the case of concentration (HHI), we found that its impact on the credit risk ofbanks in our sample is negative. A highly concentrated market have fewer banks controllingover a larger share of the market. Banking market supervisors find it easier to monitorbecause of the lower number of banks (Beck et al., 2013a, b). Also, bankswithin a concentratedmarket tend to have higher profits due to lower competitive pressures therefore lower risk(Hellmann et al., 2000). This leads us to find concentration-stability view prevailing in dualbanking economies. To further confirm this market behavior, we interact with competitionand concentration measures and report the results in Table 5. The coefficient of theinteraction term in case of NPL is positive suggesting that the competition-stability viewstrengthens with the increase in market concentration. We also find similar results when werun the interaction for Z-score. To see the marginal impact of competition in a concentratedmarket we plot the impact in Figures 1 and 2 for NPL and Z-sore respectively. We find mirrorimage graphs because of the inverse interpretation of Z-score. Hence, we can infer thatcompetition within a concentrated market will decrease the credit risk of the banks.

As far as variable size is concerned, in most of the estimations, it shows a negativerelationship with the credit risk. This means that as the size of the bank increases the creditrisk decreases across both proxies. Our finding is in line with those of Rajan and Dhal (2003).The possible reason for this relationship is that the bigger banks are expected to have more

Credit risk indual banking

systems

Page 12: Credit risk in dual banking systems: does competition

(1) (2)M1 M2

L.lnpl 0.6911*** (0.027)INEFF �1.8889 (2.199) 5.1839*** (0.639)ROA �0.1430*** (0.032) 0.0987*** (0.005)lnTA �0.2172*** (0.044) 0.0758** (0.008)NONIT 5.1120** (2.520) 4.5702*** (0.485)GLTA �0.5778* (0.296) 0.0722* (0.039)Lerner �0.6890 (0.479) �0.1034 (0.140)HHI �6.0601*** (1.711) 1.2501*** (0.425)Lerner*HHI 12.5929*** (4.094) �2.6242** (1.133)Reg �0.5258** (0.210) �0.2008*** (0.066)gdp growth �0.0326*** (0.004) �0.0033*** (0.001)inflation 0.0085* (0.005) �0.0038*** (0.001)Crisis �0.0013 (0.031) �0.0691** (0.030)Islamic_Dumm �0.1167** (0.052) �0.0058 (0.030)L.lnzscore 0.9532*** (0.009)Constant 3.6728*** (0.546) �0.6282*** (0.109)Observations 2,484 2,733instruments 86.0000 159.0000overall 318.0000 334.0000AR (1) 0.0000 0.0070AR (2) 0.7215 0.2730Sargan test (p-Val) 0.6444 0.5678

Note(s): Standard error are in parantheses; *p < 0.1, **p < 0.05, ***p < 0.01

–5

–4

–3

–2

–1

0

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

Eff

ects

on F

itte

d V

alu

es

HHI

Table 5.Interaction resultsfrom system GMM (fullsample)

Figure 1.Average marginaleffects of Lerner indexon Zscore (95% CIs)

IJOEM

Page 13: Credit risk in dual banking systems: does competition

resources to adopt advanced risk management systems which can better screen theborrowers. This helps them to reduce the default rate by better controlling the NPLs.Secondly, bigger banks are expected to have more opportunities to diversify their loanportfolio because of the bigger size of their resources.

We find cost efficiency (INEFF) to be positively related to the credit risk of banks. This isin line with the “skimping” hypothesis, framed by Berger and DeYoung (1997). As per thehypothesis, banks seeking higher levels of efficiency may choose to decrease the operationalcost. While doing so, they may also skimp on the resources set aside for the loan screeningand monitoring processes. The consequence of this type of strategy could be adverseselection which would deteriorate the credit quality of the bank (Adusei, 2016). Hence, in thelong run, cost efficiency may increase banks impaired loan problems.

We proxied diversification of the banks by NONINT (non-interest income to total income).Its coefficient reveals that increase in diversification increases the NPLS. In other words, morediversified banks are expected to have higher credit risk. Our finding is consistent with thefindings and explanation provided by Louzis et al. (2012). He argues that when bankmanagersventure into new areas, they tend to lose their focus and competitive advantage which as aresult increases the risk of banks. Surprisingly, credit growth (GLTA) does not seem to impactthe credit risks of the banks in our sample. One possible reason could be that the percentage ofloan to total assets did not change much to impact the credit risk of the banks.

An interesting finding from this table is that the Islamic banks are found to be less riskythan the conventional banks in our sample. This might be because Islamic banks are knownto have low leverage and Islamic banks avoid the complex financial instruments, such asfinancial derivatives and debt securitization. Such differences might give the Islamic banks acompetitive advantage over conventional banks and make them less risky (see, for example,�Cih�ak and Hesse, 2010; Sorwar et al., 2016). Furthermore, since Islamic banking is a faithsensitive form of banking, it is expected that most of its clients would be highly religiously

05

10

15

20

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

Eff

ects

on F

itte

d V

alu

es

HHI

Figure 2.Average marginal

effects of Lerner indexon NPLs (95% CIs)

Credit risk indual banking

systems

Page 14: Credit risk in dual banking systems: does competition

motivated and may tend to pay back their liabilities in time (Baele et al., 2014; Abedifaret al., 2013).

Now we shift our focus toward macro variables. Regulatory control does not give usconcluding evidence in terms of its impact on the credit risk of the banks in our sample. Crisisdummy is found to have increased credit risk in dual banking economies. This finding is inline with the literature. GDP growth is negatively impacting the credit risk implying that asthe GDP of an economy grows the credit risk of its banks tends to fall. Inflation is not found tohave any impact on the credit risk of the banks.

To further the credit risk dynamics of different types of banks, we split our sample intoconventional and Islamic banks and present our results in Tables 6 and 7 respectively.Generally, our results do not change after splitting the sample. As far as conventional banksare concerned cost efficiency and diversification increases the credit risk of conventionalbanks whereas profitability reduces the risk. The impact of bank size remained inconclusive.The most surprising factor is the GLTA coefficient which seems to also reduce credit risk forconventional banks. This may be due to the experience factor of the banks, i.e. as the bankgrows its financing portfolio, it tends to get better at screening and monitoring the advancesand hence reducing the credit risk. Competition and concentration decrease the credit risk ofconventional banks.

As far as Islamic banks are concerned, cost efficiency and diversification increase creditrisk whereas credit growth and profitability tend to reduce credit risk. The regulatoryframework is found to be more important for Islamic banks as compared to the conventionalbank as it is significantly increasing credit risk. Reason for that could be no separation of theregulatory framework for different types of bank. Imposing the same regulatory frameworkis not working well for Islamic banks, raising the need for a separate regulatory framework.The impact of competition remains the same, i.e. increase in competition reduces the creditrisk of Islamic banks and vice-versa. On the other hand in our main results, concentrationdoes not impact Islamic banks. A possible reason for that could be; Islamic bankingmarket isnot as much concentrated as conventional banks market. This leads us to infer that theIslamic banking sector is possiblymore stable as compared to their conventional counterpart.

As the findings suggest that except for the bank size, the determinants of credit risk seemto be similar for both types of banks. Our discussion on the literature also suggests size to bean important factor in determining the credit risk of the banks. To look into the impact ofbank size in more detail we split our main sample into big and small banks and show ourestimations in Tables 8 and 9. Most of the determinants seem to have a similar level ofsignificance and economic impact as the main results. The result of the split sample-basedestimations shows an interesting finding that the big Islamic banks are less risky than bigconventional banks whereas small Islamic banks are riskier than small conventional banks.This shows the existence of inappropriate credit control system in smaller Islamic banks. Asper Azmi et al. (2019), the Islamic banks are riskier as they appear to be poorly diversified.Another interesting but not so strong finding is that the regulatory framework seems toincrease credit risk for big banks whereas it is expected to reduce credit risk for smallerbanks. This may be due to the limits being put on the bigger banks with regards to per partyexposures. Another reason could be that as banks get bigger they have bigger funds forfinancing. So, to expand their portfolio, they may relax screening criteria for new customers.

To add credence to our results we conduct robustness by reestimating our last twomodelswith difference GMM. The results are presented in Table 10. We use our full sample forrobustness check. Overall, except for GLTA and bank size, the results remained quiteconsistent with ourmain findings. Secondly, we also split the sample among different regionsand did not find much deviation from our main results. The region-wise results will availableupon request.

IJOEM

Page 15: Credit risk in dual banking systems: does competition

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

NPL

NPL

NPL

NPL

Z-Score

Z-Score

Z-Score

Z-Score

L.lnpl

0.8201

***(0.064)

0.7033

***(0.051)

0.6957

***(0.036)

0.7002

***(0.021)

L.Z-score

0.7315

***(0.110)

0.4763

***(0.173)

0.9803

***(0.147)

0.8970

***(0.146)

ROA

�0.0806*

**(0.014)

�0.1543*

**(0.043)

�0.1797*

**(0.042)

�0.1580*

**(0.015)

0.0835

***(0.018)

0.0987

***(0.022)

0.0635

***(0.022)

0.0554

***(0.020)

lnTA

�0.0009(0.010)

0.2155

***(0.081)

�0.1839*

**(0.055)

0.1598

***(0.045)

0.0046

(0.007)

0.0140

(0.009)

0.0067

(0.005)

0.0029

(0.006)

NONIT

4.1550

***(1.496)

19.6324*

**(6.044)

2.7288

(2.066)

9.0379

***(1.572)

�2.0032(1.548)

�4.9533*

*(2.072)

1.2320

(2.377)

�0.3831(1.865)

INEFF

4.1459

*(2.298)

23.4069*

**(6.398)

�0.9597(2.158)

11.5326*

**(2.390)

1.7126

(1.167)

2.2876

**(1.076)

�0.1946(0.944)

1.3107

(0.835)

GLTA

�0.0714(0.156)

�1.2587*

**(0.459)

�0.5008*

(0.298)

�2.6198*

**(0.210)

�0.2080*

(0.116)

�0.3407*

*(0.141)

0.0288

(0.123)

�0.0592(0.116)

Reg

0.1433

(0.233)

�0.4600(0.375)

�0.1747(0.275)

�0.2586(0.324)

�0.1302(0.123)

�0.2115(0.134)

�0.0269(0.088)

�0.0635(0.093)

GDPG

�0.0291*

**(0.006)

�0.0178*

**(0.005)

�0.0252*

**(0.004)

�0.0182*

**(0.003)

�0.0020(0.002)

0.0005

(0.002)

�0.0042*

*(0.002)

�0.0026(0.002)

Inflation

0.0053

(0.005)

0.0096

(0.008)

0.0121

***(0.004)

0.0184

***(0.005)

�0.0081*

**(0.003)

�0.0141*

**(0.004)

�0.0066*

**(0.002)

�0.0068*

**(0.003)

Crisis

0.0333

(0.041)

0.0039

(0.030)

0.0843

***(0.029)

0.0369

**(0.015)

0.0166

(0.014)

0.0173

(0.013)

Lerner

0.9956

***(0.303)

�0.2115*

(0.123)

HHI

�2.5336*

**(0.818)

0.0730

(0.189)

Constant

0.2634

(0.326)

�0.8454(0.965)

2.3285

***(0.677)

0.9417

(0.602)

0.9304

**(0.413)

1.7901

***(0.605)

0.0326

(0.520)

0.3056

(0.496)

Observations

2,064

2,064

2,062

2,064

2,289

2,289

2,286

2,289

Instruments

20.0000

46.0000

67.0000

88.0000

18.0000

16.0000

19.0000

19.0000

Overall

262.0000

262.0000

262.0000

262.0000

279.0000

279.0000

279.0000

279.0000

AR(1)

0.0000

0.0000

0.0000

0.0000

0.0117

0.0335

0.0048

0.0035

AR(2)

0.6757

0.6674

0.6350

0.7959

0.3044

0.5697

0.2533

0.2206

Sargan

pv0.5997

0.1349

0.7178

0.3149

0.4662

0.2708

0.6398

0.5270

Note(s):Standarderrors

arein

parantheses;*p<0.1,

**p<0.05,*

**p<0.01

Table 6.Results from system

GMM(conventional banks)

Credit risk indual banking

systems

Page 16: Credit risk in dual banking systems: does competition

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

NPL

NPL

NPL

NPL

Z-Score

Z-Score

Z-Score

Z-Score

L.lnpl

0.4243

***(0.089)

0.5973

***(0.030)

0.5217

***(0.038)

0.4827

***(0.052)

L.Z-score

0.7065

***(0.070)

0.8628

***(0.018)

0.8930

***(0.055)

0.9189

***(0.053)

ROA

�0.0888*

*(0.035)

�0.2288*

**(0.058)

�0.3567*

**(0.051)

�0.5242*

**(0.105)

0.0579

***(0.019)

0.1351

***(0.027)

0.0767

***(0.018)

0.0426

*(0.023)

lnTA

0.0133

(0.041)

�0.0156(0.044)

�0.0662*

(0.040)

0.0718

(0.057)

�0.0182(0.016)

�0.0236*

*(0.012)

�0.0185*

*(0.009)

�0.0213*

*(0.009)

NONIT

�2.4322(7.485)

9.2504

(6.184)

18.0316*

**(5.193)

30.9240*

**(7.660)

�2.0597(1.987)

�4.8400*

*(2.413)

�0.7040(2.444)

2.6515

(2.127)

INEFF

36.6691***

(11.737)

�3.3779(6.010)

�12.6895

(7.962)

�10.2358

(6.796)

4.6854

(5.471)

12.0808***

(2.089)

0.4261

(1.386)

�0.1245(1.799)

GLTA

�1.2172*

**(0.269)

�0.9402*

**(0.158)

�1.2023*

**(0.151)

�1.5337*

**(0.234)

0.1152

(0.122)

0.0034

(0.075)

0.2000

**(0.083)

0.1307

(0.091)

Reg

�0.4578(0.733)

1.0375

**(0.487)

1.3075

**(0.538)

0.9723

(0.670)

�0.5566*

(0.296)

�0.7959*

**(0.186)

�0.1747(0.173)

�0.1834(0.174)

GDPG

0.0044

(0.011)

�0.0034(0.008)

�0.0025(0.006)

0.0183

**(0.009)

0.0032

(0.004)

�0.0037(0.003)

�0.0074*

**(0.003)

�0.002

(0.0039)

Inflation

�0.0009(0.012)

0.0062

(0.006)

0.0022

(0.007)

�0.0030(0.008)

�0.0083(0.005)

�0.0149*

**(0.003)

�0.0064*

(0.004)

�0.0065*

(0.004)

Crisis

0.1619

(0.099)

�0.0261(0.074)

0.1050

(0.100)

0.0166

(0.025)

�0.0729*

*(0.030)

�0.0802*

*(0.033)

Lerner

0.5171

***(0.196)

�0.1348*

*(0.052)

HHI

�0.0446(0.689)

�0.4491*

(0.250)

Constant

1.0009

*(0.522)

0.5765

(0.391)

1.1295

**(0.465)

0.5477

(0.649)

1.1873

***(0.337)

0.8917

***(0.190)

0.4590

*(0.239)

0.4610

**(0.219)

Observations

422

422

422

422

449

449

447

449

instruments

23.0000

31.0000

35.0000

27.0000

23.0000

31.0000

19.0000

19.0000

Overall

56.0000

56.0000

56.0000

56.0000

56.0000

56.0000

55.0000

56.0000

AR(1)

0.0422

0.0011

0.0014

0.0035

0.0094

0.0056

0.0084

0.0145

AR(2)

0.0721

0.0559

0.1019

0.5372

0.3203

0.4751

0.2532

0.2471

Sargan

pv0.9331

0.3065

0.1358

0.2847

0.8455

0.8301

0.9711

0.9823

Note(s):Standarderrors

arein

parantheses;*p<0.1,

**p<0.05,*

**p<0.01

Table 7.Results from systemGMM (Islamic banks)

IJOEM

Page 17: Credit risk in dual banking systems: does competition

(1)

(2)

(3)

(4)

(5)

(6)

NPL

NPL

NPL

Z-Score

Z-Score

Z-Score

L.lnpl

0.9493

***(0.048)

0.6982

***(0.029)

0.7356

***(0.024)

L.Z-score

0.5721

***(0.096)

0.4480

***(0.137)

0.6360

***(0.033)

ROA

�0.0890*

**(0.016)

�0.2255*

**(0.045)

�0.2639*

**(0.039)

0.1339

***(0.028)

0.1723

***(0.038)

0.0756

***(0.014)

lnTA

�0.0025(0.009)

�0.3124*

**(0.057)

�0.1547*

**(0.052)

0.0264

*(0.015)

0.0302

**(0.015)

0.0300

(0.020)

NONIT

1.7581

(2.706)

�4.5633(5.646)

10.1981*

*(5.059)

�7.3228*

**(2.752)

�10.1510

***(3.611)

0.1233

(1.292)

INEFF

�0.4444(1.982)

�12.5424

(7.921)

�7.1796(6.508)

5.8963

**(2.780)

0.9793

(2.174)

1.1858

(1.886)

GLTA

0.1468

(0.115)

�0.2714(0.316)

0.2041

(0.328)

�0.2430(0.154)

�0.2288(0.156)

0.0303

(0.109)

Reg

0.6595

***(0.160)

0.1801

(0.315)

0.3917

(0.313)

�0.4897(0.300)

�0.4013(0.273)

�0.1410(0.140)

GDPG

�0.0272*

**(0.006)

�0.0276*

**(0.004)

�0.0249*

**(0.005)

�0.0014(0.002)

�0.0026(0.002)

0.0003

(0.002)

Inflation

�0.0065(0.004)

0.0100

**(0.005)

�0.0022(0.006)

�0.0096*

**(0.003)

�0.0102*

**(0.003)

�0.0029(0.002)

Crisis

0.1225

***(0.040)

0.0310

(0.036)

0.0915

**(0.037)

0.0121

(0.016)

0.0188

(0.018)

�0.0108(0.012)

Islamic

�0.0525*

(0.031)

�0.2242*

**(0.071)

�0.2523*

**(0.078)

�0.0825(0.057)

�0.1263*

*(0.058)

�0.0547(0.035)

Lerner

0.6579

**(0.280)

�0.1176(0.121)

HHI

�1.6160(1.112)

1.3429

***(0.407)

Constant

�0.2351(0.212)

3.6473

***(0.654)

2.0038

***(0.562)

1.4298

***(0.438)

1.8341

***(0.549)

0.6641

***(0.255)

Observations

2,074

2,073

2,074

2,216

2,213

2,216

Instruments

20.0000

68.0000

75.0000

14.0000

14.0000

75.0000

Overall

262.0000

262.0000

262.0000

275.0000

274.0000

275.0000

AR(1)

0.0000

0.0000

0.0000

0.0082

0.0005

0.0121

AR(2)

0.8042

0.8715

0.8682

0.3983

0.7233

0.2049

Sargan

pv0.4933

0.8407

0.3431

0.5968

0.3044

0.8370

Note(s):Standarderrors

arein

parantheses;*p<0.1,

**p<0.05,*

**p<0.01

Table 8.Results from system

GMM (big banks)

Credit risk indual banking

systems

Page 18: Credit risk in dual banking systems: does competition

NPL

NPL

NPL

Z-Score

Z-Score

Z-Score

L.lnpl

0.6712

***(0.116)

0.7281

***(0.007)

0.6885

***(0.003)

L.Z-score

�0.4772(0.290)

0.8989

***(0.186)

0.9405

***(0.069)

ROA

�0.0949*

**(0.017)

�0.0369*

**(0.006)

�0.0824*

**(0.002)

0.1139

***(0.022)

0.0399

***(0.015)

0.0295

***(0.009)

lnTA

�0.1691*

**(0.046)

�0.0701*

**(0.011)

0.0398

***(0.003)

�0.4815*

**(0.137)

�0.0530(0.052)

�0.0458*

(0.025)

NONIT

7.1821

***(2.173)

4.0216

***(0.241)

7.1276

***(0.072)

�8.2577*

**(3.062)

0.3249

(1.748)

1.5704

*(0.950)

INEFF

�0.4723(3.158)

13.2149*

**(0.156)

16.6950*

**(0.102)

�5.6249(3.638)

�0.3161(1.303)

�0.7508(1.073)

GLTA

�0.5348(0.326)

�0.8911*

**(0.048)

�2.3017*

**(0.010)

�0.5979(0.427)

�0.0755(0.182)

�0.0085(0.087)

Reg

�0.6140(0.651)

�0.5264*

(0.280)

�0.1202(0.244)

�0.6382(1.099)

0.0118

(0.367)

�0.0638(0.170)

GDPG

�0.0034(0.014)

0.0107

***(0.001)

0.0210

***(0.001)

0.0108

(0.013)

�0.0053(0.005)

�0.0027(0.003)

Inflation

0.0385

***(0.015)

0.0324

***(0.002)

0.0298

***(0.000)

�0.0463*

**(0.012)

�0.0089(0.007)

�0.0096*

*(0.004)

Crisis

0.0009

(0.079)

�0.0518(0.037)

0.0058

(0.018)

0.1051

(0.097)

�0.0043(0.037)

�0.0024(0.026)

Islamic

0.0862

(0.112)

0.1838

***(0.042)

0.1283

***(0.036)

�0.1605(0.181)

�0.0807(0.059)

�0.0604(0.042)

Lerner

0.2756

***(0.037)

0.0626

(0.195)

HHI

�2.9185*

**(0.464)

0.0491

(0.225)

Constant

2.0874

**(0.866)

1.0250

***(0.234)

1.1263

***(0.232)

8.5875

***(2.135)

0.6772

(0.921)

0.5256

(0.410)

Observations

412

411

412

522

520

522

Instruments

21.0000

68.0000

89.0000

19.0000

20.0000

22.0000

Overall

112.0000

112.0000

112.0000

131.0000

130.0000

131.0000

AR(1)

0.0414

0.0005

0.0006

0.3058

0.0000

0.0000

AR(2)

0.7476

0.4083

0.4619

0.2020

0.9861

0.9731

Sargan

pv0.8306

0.9786

0.5768

0.6747

0.2504

0.6836

Note(s):Standarderrors

arein

parantheses;*p<0.1,

**p<0.05,*

**p<0.01

Table 9.Results from systemGMM (small banks)

IJOEM

Page 19: Credit risk in dual banking systems: does competition

5. ConclusionThis study was intended to study the credit risk of both Islamic and conventional banks indual banking economies. Along with that, we wanted to see whether bank size andcompetition has a role to play. On the full sample basis, we find cost efficiency, bank size anddiversification to increase the credit risk of the banks in dual banking economies. On the otherhand, profitability seems to decrease credit risk. Islamic banks are found to be less risky ascompared to their conventional counterparts. When we split the sample between Islamic andconventional banks, we find that except for the bank size, the determinants of credit risk seemto be similar for both conventional and Islamic banks. To look into the impact of bank size inmore detail we split our main sample into big and small banks. The result of the split sample-based estimation shows that the big Islamic banks are less risky than big conventional bankswhereas small Islamic banks are riskier than small conventional banks.

Our results have several policy implications. First, they suggest regulating the level ofcompetition in countries with a dual banking system and how smaller banks need to beprotected which are more prone to the credit risk. Our paper also has regulatory implicationsas the current regulatory framework is found to be more significant in increasing credit riskfor Islamic banks as compared to their conventional counterparts. A possible reason for thatcould be no separation of the regulatory framework for different types of banks. Imposing thesame regulatory framework is not working well for Islamic banks and hence raising the needfor a separate regulatory framework. In support of the competition-stability view, we findcompetition to decrease the credit risk of banks. The impacts of both competition andconcentration on credit risk remain similar for conventional banks but the concentration isfound to be insignificant for Islamic banks.

Auxiliary to market structure, our findings provide support to policymakers andregulators to consider the development of special regulations for Islamic banks with respectto their size. We find smaller Islamic banks to be less safe. Policymakers and regulators need

(1) (2) (3) (4)NPL NPL Z-Score Z-Score

L.lnpl 0.7049*** (0.109) 0.7058*** (0.054)L.Z-score 0.0488 (0.073) 0.4054*** (0.051)INEFF 10.3614** (5.249) 10.1429*** (3.520) 15.8661*** (4.591) 3.7888** (1.599)ROA �0.1219*** (0.022) �0.0917*** (0.016) 0.1875*** (0.046) 0.0764*** (0.014)lnTA 0.0218 (0.135) 0.0365 (0.053) 0.0673 (0.058) 0.0129 (0.032)NONIT �2.4730 (4.039) 0.8087 (2.410) 4.9559** (2.214) 2.4891 (1.531)GLTA �0.8393** (0.340) �0.6552*** (0.209) �0.6318** (0.272) �0.3846*** (0.142)Reg 0.4696 (0.456) 0.9753** (0.450) 0.2282 (0.278) �0.0548 (0.153)GDPG �0.0232*** (0.006) �0.0236*** (0.005) �0.0026 (0.003) �0.0028** (0.001)Inflation �0.0076 (0.006) 0.0028 (0.005) 0.0002 (0.004) 0.0027 (0.002)Islamic �0.0805 (0.026)Crisis 0.0650* (0.037) 0.0626* (0.036) �0.0286 (0.025) �0.0280** (0.013)Lerner 0.6668** (0.289) �1.4335*** (0.409)HHI �0.3337 (0.807) 2.4102*** (0.743)Observations 2,088 2,168 2,399 2,403instruments 19.0000 19.0000 45.0000 59.0000Overall 299.0000 299.0000 323.0000 324.0000AR (1) 0.0001 0.0000 0.3693 0.0170AR (2) 0.8574 0.8398 0.2294 0.2696Sargan Test (p-Val) 0.8744 0.4502 0.9711 0.7450

Note(s): Standard error are in parantheses; *p < 0.1, **p < 0.05, ***p < 0.01

Table 10.Robustness: results

from difference GMM(full sample)

Credit risk indual banking

systems

Page 20: Credit risk in dual banking systems: does competition

to put additional guidelines to safeguard the interest of investors and depositors. Frominvestors and depositors viewpoint big Islamic banks are found to be less risky and maybe apreferred selection.

Notes

1. The data on Islamic banking windows is not available, therefore we use consolidated financialstatements instead; however, since the Fitch Connect database does not provide the classification offull-fledged Islamic banks and Islamic banking subsidiaries, the subsidiaries are treated as full-fledged Islamic banks.

2. It is important to note that competition and concentration are two different aspects of a marketenvironment. For instance, a highly concentrated market does not necessarily mean that there is nocompetition among the banks (Leroy and Lucotte, 2017). Studies include both Lerner (competition)and HHI (concentration) to have a complete view of the banking market.

References

Abedifar, P., Molyneux, P. and Tarazi, A. (2013), “Risk in Islamic banking”, Review of Finance, Vol. 17No. 6, pp. 2035-2096.

Abuzayed, B., Al-Fayoumi, N. and Molyneux, P. (2018), “Diversification and bank stability in theGCC”, Journal of International Financial Markets, Institutions and Money, Vol. 57, pp. 17-43.

Adusei, M. (2016), “Does entrepreneurship promote economic growth in Africa?”, African DevelopmentReview, Vol. 28 No. 2, pp. 201-214.

Akhtar, B., Akhter, W. and Shahbaz, M. (2017), “Determinants of deposits in conventional and Islamicbanking: a case of an emerging economy”, International Journal of Emerging Markets, Vol. 12No. 2, pp. 296-309.

Alaeddin, O., Khattak, M.A. and Abojeib, M. (2019), “Evaluating stability in dual banking system :comparison between conventional and Islamic banks in Malaysia”, Humanities and SocialSciences Reviews, Vol. 7 No. 2, pp. 510-518.

Alam, N., Gupta, L. and Shanmugam, B. (2017), Islamic Finance: A Practical Perspective, Springer, NewYork, NY.

Alam, N., Hamid, B.A. and Tan, D.Y. (2018), “Does competition make banks riskier in dual bankcompetition?”, Borsa Istanbul Review, Vol. 19 No. S1, pp. S34-S43.

Allen, F. and Gale, D. (2004), “Competition and financial stability”, Journal of Money, Credit, andBanking, Vol. 36 No. 3, pp. 453-480.

Amidu, M. and Wolfe, S. (2013), “Does bank competition and diversification lead to greater stability?Evidence from emerging markets”, Review of Development Finance, Vol. 3 No. 3, pp. 152-166.

Anginer, D. and Demirguc-Kunt, A. (2014), “Has the global banking system become more fragile overtime?”, Journal Financial Stability, Vol. 13 No. 2014, pp. 202-213.

Arellano, M. and Bover, O. (1995), “Another look at the instrumental variable estimation of error-components models”, Journal of Econometrics, Vol. 68 No. 1, pp. 29-51.

Azmi, W., Ali, M., Arshad, S. and Rizvi, S.A.R. (2019), “Intricacies of competition, stability, anddiversification: evidence from dual banking economies”, Economic Modelling, Vol. 83,pp. 111-126.

Baele, L., Farooq, M. and Ongena, S. (2014), “Of religion and redemption: evidence from default onIslamic loans”, Journal of Banking and Finance, Vol. 44, pp. 141-159.

Beck, T., De Jonghe, O. and Schepens, G. (2013a), “Bank competition and stability: cross-countryheterogeneity”, Journal of Financial Intermediation, Vol. 22 No. 2, pp. 218-244.

Beck, T., Demirg€uç-Kunt, A. and Merrouche, O. (2013b), “Islamic vs. conventional banking: businessmodel, efficiency and stability”, Journal of Banking and Finance, Vol. 37 No. 2, pp. 433-447.

IJOEM

Page 21: Credit risk in dual banking systems: does competition

Berger, A.N. and DeYoung, R. (1997), “Problem loans and cost efficiency in commercial banks”,Journal of Banking and Finance, Vol. 21 No. 6, pp. 849-870.

Berger, A.N., Klapper, L.F. and Turk-Ariss, R. (2009), “Bank competition and financial stability”,Journal of Financial Services Research, Vol. 35 No. 2, pp. 99-118.

Birchwood, A., Brei, M. and Noel, D.M. (2017), “Interest margins and bank regulation in CentralAmerica and the Caribbean”, Journal of Banking and Finance, Vol. 85, pp. 56-68.

Blundell, R. and Bond, S. (1998), “Initial conditions and moment restrictions in dynamic panel datamodels”, Journal of Econometrics, Vol. 87 No. 1, pp. 115-143.

Boumediene, A. (2011), “Is credit risk really higher in Islamic banks?”, Journal of Credit Risk, Vol. 7,pp. 97-129.

Boyd, J.H. and De Nicolo, G. (2005), “The theory of bank risk taking and competition revisited”, TheJournal of Finance, Vol. LX No. 3, pp. 1329-1343.

Boyd, J.H., De Nicolo, G. and Jalal, A.M. (2009), “Bank competition, risk, and asset allocations”,International Monetary Fund Working Paper No. WP/09/143, Washington, DC.

Chaibi, H. and Ftiti, Z. (2015), “Credit risk determinants: evidence from a cross-country study”,Research in International Business and Finance, Vol. 33, pp. 1-16.

�Cih�ak, M. and Hesse, H. (2010), “Islamic banks and financial stability: an empirical analysis”, Journalof Financial Services Research, Vol. 38, pp. 95-113.

Das, A. and Ghosh, S. (2007), “Determinants of credit risk in Indian state-owned banks: an empiricalinvestigation”, Economic Issues, Vol. 12 No. 2, pp. 27-46.

Demirg€uç-Kunt, A. and Detragiache, E. (2005), “Cross-country empirical studies of systemic bankdistress: a survey”, National Institute Economic Review, Vol. 192 No.1, pp. 68-83.

Demirg€uç-Kunt, A. and Detragiache, E. (2011), “Basel core principles and bank soundness, doescompliance matter?”, Journal of Financial Stability, Vol. 7 No. 4, pp. 179-190.

Gonz�alez, L.O., Razia, A., B�ua, M.V. and Sestayo, R.L. (2017), “Competition, concentration and risktaking in banking sector of MENA countries”, Research in International Business and Finance,Vol. 42, pp. 591-604.

Gulati, R., Goswami, A. and Kumar, S. (2019), “What drives credit risk in the Indian banking industry?An empirical investigation”, Economic Systems, Vol. 43 No. 1, pp. 42-62.

Hellmann, T.F., Murdock, K.C. and Stiglitz, J.E. (2000), “Liberalization, moral hazard in banking, andprudential regulation: are capital requirements enough?”, American Economic Review, Vol. 90No. 1, pp. 147-165.

Ibrahim, M.H. and Alam, N. (2017), “Islamic economics and Islamic finance in the world economy”,World Economy, Vol. 3 No. 2017, pp. 1-6.

Ibrahim, M.H. and Rizvi, S.A.R. (2017), “Do we need bigger Islamic banks? An assessment of bankstability”, Journal of Multinational Financial Management, Vol. 40, pp. 77-91.

IFN (Islamic Financial services Board) (2018), “IFN Fintech huddle report 2018”, available at https://redmoneyevents.com/main/framework/assets/2018/reports/fintech.pdf (accessed 20 May 2019).

IFSB (Islamic Financial services Board) (2018), “Islamic financial services industry stability report2018”, available at https://www.ifsb.org/download.php?id54811&lang5English&pg5/index.php (accessed 20 May 2019).

Jim�enez, G., Lopez, J.A. and Saurina, J. (2013), “How does competition affect bank risk-taking?”, Journalof Financial Stability, Vol. 9 No. 2, pp. 185-195.

Kabir, M.N. and Worthington, A.C. (2017), “The ‘competition–stability/fragility’ nexus: a comparativeanalysis of Islamic and conventional banks”, International Review of Financial Analysis, Vol. 50,pp. 111-128.

Kabir, M.N., Worthington, A. and Gupta, R. (2015), “Comparative credit risk in Islamic andconventional bank”, Pacific-Basin Finance Journal, Vol. 34, pp. 327-353.

Credit risk indual banking

systems

Page 22: Credit risk in dual banking systems: does competition

Kasman, S. and Kasman, A. (2015), “Bank competition, concentration and financial stability in theTurkish banking industry”, Economic Systems, Vol. 39 No. 3, pp. 502-517.

Kelsey, D. and Roux, S.L.E. (2016), “Dragon slaying with ambiguity:theory and experiments”, Journalof Public Economic Theory, Vol. 19 No. 1, pp. 1-20.

Khediri, K.B., Charfeddine, L. and Youssef, S.B. (2015), “Islamic versus conventional banks in the GCCcountries: a comparative study using classification techniques”, Research in InternationalBusiness and Finance, Vol. 33, pp. 75-98.

Klomp, J. and De Haan, J. (2012), “Banking risk and regulation: does one size fit all?”, Journal ofBanking and Finance, Vol. 36 No. 12, pp. 3197-3212.

Lassoued, M. (2018), “Comparative study on credit risk in Islamic banking institutions: the case ofMalaysia”, The Quarterly Review of Economics and Finance, Vol. 70 No. 201, pp. 267-278.

Leroy, A. and Lucotte, Y. (2017), “Is there a competition-stability trade-off in European banking?”,Journal of International Financial Markets, Institutions and Money, Vol. 46, pp. 199-215.

Louhichi, A. and Boujelbene, Y. (2016), “Credit risk, managerial behaviour and macroeconomicequilibrium within dual banking systems: interest-free vs. interest-based banking industries”,Research in International Business and Finance, Vol. 38, pp. 104-121.

Louzis, D.P., Vouldis, A.T. and Metaxas, V.L. (2012), “Macroeconomic and bank-specific determinantsof non-performing loans in Greece: a comparative study of mortgage, business and consumerloan portfolios”, Journal of Banking and Finance, Vol. 36 No. 4, pp. 1012-1027.

Moudud-Ul-Huq, S. (2020), “Does bank competition matter for performance and risk-taking? Empiricalevidence from BRICS countries”, International Journal of Emerging Markets, Vol. ahead-of-printNo. ahead-of-print.

Rajan, R. and Dhal, S.C. (2003), “Non-performing loans and terms of credit of public sector banks inIndia: an empirical assessment”, Reserve Bank of India Occasional Papers, Vol. 24 No. 3,pp. 81-121.

Rashid, A., Yousaf, S. and Khaleequzzaman, M. (2017), “Does Islamic banking really strengthenfinancial stability? Empirical evidence from Pakistan”, International Journal of Islamic andMiddle Eastern Finance and Management, Vol. 10 No. 2, pp. 130-148.

Repullo, R. (2004), “Capital requirements, market power, and risk-taking in banking”, Journal ofFinancial Intermediation, Vol. 13 No. 2, pp. 156-182.

Roodman, D. (2009), “A note on the theme of too many instruments”, Oxford Bulletin of Economics andStatistics, Vol. 71 No. 1, pp. 135-158.

Sorwar, G., Pappas, V., Pereira, J. and Nurullah, M. (2016), “To debt or not to debt: are Islamic banksless risky than conventional banks?”, Journal of Economic Behavior and Organization, Vol. 132,pp. 113-126.

Weill, L. (2011), “Do Islamic banks have greater market power?”, Comparative Economic Studies,Vol. 53 No. 2, pp. 291-306.

IJOEM

Page 23: Credit risk in dual banking systems: does competition

Appendix

Corresponding authorMohsin Ali can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:www.emeraldgrouppublishing.com/licensing/reprints.htmOr contact us for further details: [email protected]

Country name Islamic banking share in the country(%)

Bahrain 14Bangladesh 20Indonesia 5Jordan 15Kuwait 40Lebanon 2Malaysia 25Oman 12Pakistan 11Qatar 25Saudi Arabia 53Tunisia 5Turkey 6United Arab Emirates 20

Source(s): IFSB, 2018

Table A1.Islamic banking share

for year 2017

Credit risk indual banking

systems