Accepted Manuscript Bank lending, deposits and risk-taking in times of crisis: A panel analysis of islamic and conventional banks Mansor H. Ibrahim, Syed Aun R. Rizvi PII: DOI: Reference: To appear in: Received date: Revised date: Accepted date: 3 S1566-0141(17)30241-8 https://doi.org/10.1016Zj.ememar.2017.12.00 EMEMAR 540 23 June 2017 12 October 2017 16 December 2017 Please cite this article as: Mansor H. Ibrahim, Syed Aun R. Rizvi, Bank lending, deposits and risk-taking in times of crisis: A panel analysis of islamic and conventional banks. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Ememar(2017), https://doi.org/10.1016Zj.ememar.2017.12.003 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. ACCEPTED MANUSCRIPT BANK LENDING, DEPOSITS AND RISK-TAKING IN TIMES OF CRISIS: A PANEL ANALYSIS OF ISLAMIC AND CONVENTIONAL BANKS Mansor H. Ibrahim School of Graduate Studies International Centre for Education in Islamic Finance (INCEIF) Lorong Universiti A 59100 Kuala Lumpur MALAYSIA Email: mansorhi@hotmail.com and mansorhi@inceif.org Syed Aun R. Rizvi Suleman Dawood School of Business Lahore University of Management Sciences (LUMS) Opp Sector U, DHA IV, Lahore, Pakistan Email: aun@rizvis.net and aun.raza@lums.edu.pk Acknowledgement: We would like to thank the two anonymous referees for helpful comments. We also would like to thank participants at the 2nd Applied Modelling Conference held at Deakin University, 2-3 February 2017. The usual disclaimer applies. ACCEPTED MANUSCRIPT BANK LENDING, DEPOSITS AND RISK-TAKING IN TIMES OF CRISIS: A PANEL ANALYSIS OF ISLAMIC AND CONVENTIONAL BANKS ABSTRACT In this study, we conduct a panel analysis of Islamic and conventional banks to ascertain whether Islamic banks are able to sustain financing supply and whether its growth is higher than conventional bank lending growth in times of stress. For concreteness, we also assess whether the sustained financing supply of Islamic banks is justified by a concomitant increase in Islamic deposit growth and is not linked to excessive risk taking. Utilizing a panel sample of 25 Islamic banks and 114 conventional banks from 10 dual-banking countries, we observe sustained financing supply by Islamic banks but significant reduction in the lending growth by conventional banks during the crisis period. The results further suggest that the financing growth of Islamic banks is higher than the lending growth of conventional banks during the crisis period. However, we find no clear evidence that the deposit growth of Islamic banks behaves differently during the period. Finally, there is no indication to suggest that Islamic banks exhibit excessive risk taking in times of stress. Our results contribute to the evidence supporting the contributive role of the Islamic banking system to financial and economic stability. Keywords: Dual-Banking System, Credit Growth, Deposit Growth, Credit Risk, Crisis JEL Classification: G21, G01, C33 1. INTRODUCTION Credit provision is essential in ensuring sustained real production and in stimulating economic activities. Thus, it is not surprising that, facing crises, policymakers resort to various mechanisms so that banks are able to maintain their credit supply. These include nationalization of distressed banks, deposit guarantees, fund injections and revision of capital requirements (Chen et al., 2016). In parallel, scholarly studies have searched for factors that allow banks to perform and function reasonably well in times of stress to inform policymakers what might be appropriate policy actions to stabilize bank credit. They have brought into their focus various bank-specific characteristics, notable among which are bank ownership (Brei and Schclarek, 2013; Cull and Martinez-Peria, 2013; Chen et al., 2016), bank types or business orientation (Chiaramonte et al., 2015; Merilainen, 2016) and bank capital (Kosak et al., 2015; Altunbas et al., 2016). In search for solutions to the problem of bank-originated financial instability, some studies have turned their attention to the Islamic banking business model. Drawn by its rapid development even during the recent global financial crisis, they make comparative assessments of Islamic banks and conventional banks using a variety of bank performance metrics. As regards Islamic bank financing, few studies are available (Hasan and Dridi, 2011; Beck et al., 2013; and Ibrahim 2016). In general, they paint a bright picture of Islamic banks for their stable or even increasing supply of credit during crisis episodes. Still, despite the alleged resiliency of Islamic banks, whether the Islamic banking sector is relatively more stable than its conventional counterparts remains contentious (Mejia et al., 2014; Kabir and Worthington, 2017). Accordingly, it is still an open empirical question whether Islamic banking can be a stabilizing force in times of stress. ACCEPTED MANUSCRIPT This paper builds upon these studies by providing further empirical assessments of Islamic and conventional banks in times of crisis. In this paper, we contend that a look at lending/financing per se would not lead to a reliable conclusion about the ability of a bank to be a stabilizing force in times of crisis. Expansion of credit, especially during crisis episodes, necessarily requires funding and entails additional risks. Further, with intensified asymmetric information problem, credit expansion during the crisis period may signify excessive risk taking. Thus, we would be more confident of a bank’s ability to be a stabilizing force if (i) it has the ability to expand credit in times of crisis and (ii) its credit expansion does not lead to excessive risk in the future. The present study considers these aspects in the analysis to arrive at a more concrete answer to the potential role of Islamic banks in the economy. Our contribution to the literature is threefold. First, we add to the strand of research on bank behaviour in times of stress by bringing into the fore the potential role of the Islamic banking sector. Second, we contribute to the on-going debate on the relative stability of Islamic banks vis-a-vis conventional banks. In the literature, there are two competing views - the stability view and the skeptic view (Ibrahim, 2016). The stability view posits that the Islamic banking sector is amiable to banking stability, which is rooted in their distinct features of being Shariahcompliant (Farooq and Zaheer, 2015)1. By contrast, emphasizing no distinct differences between Islamic banks and conventional banks in practice, the “skeptic” view doubts the Islamic banks would make a difference (Chong and Liu, 2009; Khan, 2010). We bring into the picture the financing decisions by banks to ascertain whether Islamic banks are able to stabilize credit during crisis episodes. And third, from the analytical perspective, we provide a more comprehensive evaluation of bank financing by synthesizing various aspects of banking studies to bring about concrete inferences as to whether the Islamic banking system contributes favourably to stability. We take lead from the recent contribution by Chen et al. (2016) that not only looks at loan growth during the global financial crisis but also assesses how loan increases during the crisis affect bank performance and aggregate activities. We broaden our analysis to include deposit growth and credit risk. The former is to assess, if Islamic banks did have higher credit growth, whether higher credit growth was accompanied by higher deposit growth as an indicator of their ability to expand credit. Meanwhile, the latter is to verify whether credit expansion was accompanied by higher future risk and whether it can be construed as excessive risk taking by Islamic banks. The rest of this paper is structured as follows. Section 2 provides background information and develops empirical hypotheses. Section 3 details model specifications. Section 4 presents the data and discusses estimation results. Finally, section 5 contains a summary of the main findings and concluding remarks. 2. BACKGROUND AND HYPOTHESES DEVELOPMENT To place the present study in its context, we first provide background information of the Islamic banking. The emphasis is on the (theoretical) differences of the Islamic banks from their conventional counterparts and the theory-practice divide in Islamic banking and resulting arguments whether the Islamic banking system contributes to stability. Then, on the bases of According to Farooq and Zaheer (2015), the distinct features of the Islamic banking include prohibition of interest rate, speculative activities and of excessive uncertainty and Islamic banking transactions being linked fundamentally to the real sector. 1 ACCEPTED MANUSCRIPT various contentious views and of existing literature, we develop hypotheses to be tested. 2.1 Islamic Banking The Islamic banking industry has evolved to be a material part of the global financial industry. While figures differ according to sources, the growth of Islamic banking assets has outpaced that of conventional banking assets even during the global financial crisis. Indeed, as noted by Hasan and Dridi (2011), the asset growth of Islamic banks was more than double that of conventional banks during 2007-2009, i.e. the global financial crisis period. Starting in only few Muslim-majority countries in the 1960s and 1970s, the Islamic banking presently operates in more than 75 countries. Such multinational conventional banks as Chase Manhattan, Citibank, and HSBC have also participated in the provisions of the Islamic banking services by establishing Islamic banking windows or Islamic bank subsidiaries. Islamic banking differentiates itself from the existing conventional banking through its adherence to the Islamic laws (Shari’ah). A key feature of the Islamic banking operations is the prohibition of interest rate (riba). According to Islam, charging interest rate is a form of injustice since it allows financiers to gain benefits simply by transferring their money to others without being involved in real activities. In other words, it is unlawful for money to beget money without real sector contributions (Ariff, 2014). In addition, Islamic banks must not be involved in transactions characterized by extreme uncertainties (gharar) and gambling (maysir). Thus, they must not expose themselves to such toxic assets as collateralized debt obligation (CDO) and mortgage-back security (MBS) and derivative products (Mollah et al., 2017). Moreover, the parties to Islamic financial transactions must honour the sanctity of contracts to their best ability such that asymmetric information and morally hazardous behaviour are contained. Needless to state, the Islamic banks must confine their activities to those allowable by Islam. Accordingly, financing related to the production and sales of alcoholic beverages, gambling and prostitution is prohibited. As a manifestation of their conformance to the Shari’ah, Islamic banks arrange their financial transactions at both ends of the balance sheet on the basis of profit-and-loss sharing and real-sector linkages. The profit-and-loss sharing arrangement either through mudarabah or musharakah contract identifies depositors on the liability side and banks on the asset side as investors. Under these contracts, they share ex post in the outcomes of business transactions on the basis of a pre-agreed ratio. Thus, Islamic banking departs completely from its interestbased conventional counterpart where, in conventional banking, financiers are lenders deriving rewards from taking purely financial risks and not real-sector risks. Islamic banks also offer financing on the basis of sales or leasing contracts, known respectively as Murabahah and Ijarah. Under these contracts, the banks specify instalment payments over a specified period of the goods they sell or lease to clients after taking into consideration the costs and mark-up profits. Thus, in Islamic banking, interest rate is completely absent and real-sector linkages are foundational. These features of Islamic banking serve as theoretical arguments for its relative stability as compared to the conventional banking system. At the bank level, Khan (1987) demonstrates theoretically the relative soundness of an Islamic bank as compared to a conventional bank. According to Khan (1987), the real sector linkages of the Islamic bank immediately align its assets and liabilities across business cycle phases. By contrast, the separation of assets and liabilities under the conventional banking system makes it harder for the conventional bank to recover its balance sheet during adverse shocks. Accordingly, the Islamic bank would be relatively more stable. The absence of speculative and toxic financial products further firms up ACCEPTED MANUSCRIPT the resiliency of the Islamic banking system during financial shocks, as manifested during the recent global financial crisis. Ariff (2014) further mentions inter-connectivity between the real sector and financial sector, with the latter provides a supporting role for real activities, to be central to economic stability and growth. In other words, the real-sector linkages of the Islamic banking system fosters economic stability. While few would argue against the relative stability of the (theoretical) Islamic banking system, some raise doubts whether the Islamic banking is in practice markedly different from the conventional banking on three grounds. These are (i) concentration on Murabahah financing; (ii) pricing of Islamic banking products; and (iii) lack of truly distinct Shari’ah- based products. While portrayed to be a distinct feature of the Islamic banking business model, Mudarabah and Musharakah financings make up only a small percentage of Islamic banking assets. Islamic financing comprises predominantly Murabahah or sale-based financing, which is estimated to be more than 80% since the early years of Islamic banking operations in the 1970s (Al-Harran, 1995; Chong and Liu, 2009; and Khan, 2010). Moreover, the pricing of especially Murabahahbased products is also an issue since Islamic banks resort to the interest rate benchmark. Finally, some view Islamic products merely as imitation of prevailing conventional products but adapted and modified to be Shari’ah compliant. As examples, as noted by Ariff (2014), murabahah financing, bai’ bithaman ajil (BBA), al-ijarah thumma al- bai’ (AITAB) and tawarruq munazzam offered by Islamic banks are mirror images of respectively conventional fixed-rate home loans, conventional floating-rate home loans, hirepurchase, and conventional personal loans. According to Ariff (2015), to be truly distinct, the Islamic banks must leap away from Shari’ah-compliant mode to Shari’ah-based mode in the development of Islamic products. The theory and practice divide in Islamic banking is at the centre of the debate whether the adoption of the Islamic banking system can help stabilizing the prevailing fragile system. Based on the Islamic banking practices, the skeptic view sees no differences in Islamic banks. However, counter-arguments to the skeptic view are also available. The criticisms of the Islamic banking practices notwithstanding, the Islamic banking financial products possess real sector linkages and hence the real - financial decoupling is unlikely. In addition, Islamic banking differentiates itself not only on the basis of the prohibition of interest rates but also on the importance of the sanctity of contracts. The presence of Shari’ah board in addition to the standard board of governance plays a crucial role in ensuring conformity to the principles of the Shari’ah and hence curtailing Islamic banks’ excessive risk taking. The debate on whether Islamic banking is distinguishable from conventional banking and whether it is relatively more stable especially during the crisis episodes has spilled into an empirical front, with available evidence ranging from no significant difference to some key differences between them. Employing a non-parametric test, Bourkhis and Nabi (2013) find no differences in bank capital, profitability, asset quality, costs and lending activities of Islamic and conventional banks. Beck et al. (2013) and Alqahtani et al. (2016) find some differences between Islamic and conventional banks, particularly in terms of intermediation ratio and cost inefficiency. The recent study by Jawadi et al. (2016) also documents a few significant differences between the two types of banks in terms of financial risk. Evaluating convergence in performance of Islamic and conventional banks, Olson and Zoubi (2017) document evidence indicating convergence of their profitability. Still, as they demonstrate, the convergence does not apply to other financial ratios, namely net income margins and risk characteristics. Thus, Islamic banks are different. In light of these findings, it seems that the issues remain open to further empirical investigation. ACCEPTED MANUSCRIPT 2.2 Hypotheses Development In the aftermath of the recent global financial crisis, many studies have emerged to identify bank characteristics that would limit the tendency of the banking system to amplify financial shocks in a pro-cyclical manner. One interesting aspect that has been highlighted in the literature is the importance of banks’ objectives and principles in constraining bank lending behaviour. Some have argued and forwarded evidence that government-owned banks and cooperative banks are more likely willing and able to maintain their lending when facing shocks. Unlike commercial banks, these banks are not strictly profit-oriented. According to Chen et al. (2016), government-owned banks have social-welfare agendas and their liabilities are implicitly guaranteed. Accordingly, they would be more willing to continuously extend loans during crises even if they may not be profitable and would be able to do so since they have liquidity strength. Being stakeholders’ banks, cooperative banks are oriented towards long term relationships with their members or clients and, consequently, are likely to behave in a less cyclical manner (Merilainen, 2016). Islamic banks are essentially commercial banks. However, while they are profit-oriented, they may behave differently from the conventional commercial banks when facing adverse shocks. The Shari’ah principles that these banks adhere to means that the social-welfare agendas would assume priority during crunch times. The capital and asset strength of the Islamic banks, as a result of the above-mentioned Shari’ah principles, may enable them to extend financing even during crisis periods. Religion and profit-loss sharing arrangements are other potential reasons that Islamic banks would be less pro-cyclical (Mollah et al., 2017). Given the alignment between bank assets and liabilities as demonstrated by Khan (1987) and religious convictions of Islamic bank depositors (Farooq and Zaheer, 2015), Islamic banks are less immune to deposit withdrawals and hence are in a better position to maintain financing supply in the wake of adverse shocks. Several studies find support for higher financing or intermediation of Islamic banks during the global financial crisis and economic downturns. Assessing the effects of the recent global financial crisis on Islamic and conventional banks using a sample of 120 conventional and Islamic banks from 8 countries, Hasan and Dridi (2011) find the credit growth of Islamic banks to be higher. Beck et al. (2013) document higher intermediation ratio of Islamic banks as compared to that of conventional banks and the difference is even more significant during local crises. Alqahtani et al. (2016) further reaffirm that Islamic banks intermediate more. Finally, according to Ibrahim (2016) in his analysis of Malaysia’s dual banking system, Islamic financing is less pro-cyclical or even counter-cyclical as compared to conventional lending. Despite these favourable findings, it remains uncertain whether Islamic banks can maintain their credit supply in times of stress given the noted arguments that the Islamic banking is not practically different from conventional banking, which is empirically supported by Bourkhis and Nabi (2013). Due to their distinct features, Islamic banks face unique risks, e.g. Shariah compliant risk, equity investment risk, rate of return risk, and displaced commercial risk, and hence may be exposed to higher risk as compared to conventional banks (Mejia et al., 2014). Kabir et al. (2015) offer empirical evidence suggesting that Islamic banks have higher nonperforming loans while Alandejani et al. (2017) indicate that Islamic banks have a higher incidence rate of failure. These risks, coupled with the underdeveloped Islamic money markets, may affect their function as financial intermediaries during the crisis period. In views of these, we test the following hypotheses related to the financing growth of Islamic banks: H01: There is no significant reduction in the financing growth of Islamic banks during the crisis period. ACCEPTED MANUSCRIPT H02: There is no difference in the lending/financing growth of Islamic banks and conventional banks during the crisis period. Based on the above hypotheses, a strong case for the ability of Islamic banks to stabilize credit can be made if they can sustain financing supply in times of stress, i.e. non-rejection of H01, and the financing growth of Islamic banks is higher than the lending growth of conventional banks during the period, i.e. rejection of H02. The ability of Islamic banks to maintain financing supply during the crisis period depends crucially on the stability of their deposits. However, whether Islamic bank deposits are resilient to the crisis has not been much investigated. On one hand, the presence of displaced commercial risk, which is unique to the Islamic banking sector, means that deposit withdrawals would be highly likely in the face of adverse shocks (Khan and Ahmed, 2001; Mejia et al., 2014). On the other hand, the “religious branding” and “asset-liability” alignments of Islamic banks would mean that Islamic deposits would be less sensitive to shocks (Khan, 1987; Farooq and Zaheer, 2015). In perhaps the only study on the resiliency of Islamic deposits during a financial panic. Farooq and Zaheer (2015) document evidence supporting fewer deposit withdrawals from Islamic banks as compared to conventional banks. Indeed, some Islamic banks even attracted more deposits during the period. The findings, however, relate only to the case of Pakistan and, as a result, there is a need to look at the subject from a wider perspective. The present study draws a sample from 10 major dual-banking countries to evaluate deposit growth of Islamic and conventional banks as stated in the following hypotheses. H03: There is no significant reduction in the deposit growth of Islamic banks during the crisis period. H04: There is no difference in the deposit growth of Islamic banks and conventional banks during the crisis period. A non-rejection of H03 and rejection of H04 and findings that the deposit growth of Islamic banks is higher would further strengthen the ability of Islamic banks to stabilize credit. However, if the deposit growth of Islamic banks is more severely affected by the crisis, then there would be a concern that the Islamic banks overstretch its credit expansion, a behaviour that may likely breed future risk. From the lens of finance - growth nexus, credit expansion is applauded as a driver for economic growth. Moreover, continuous expansion of credit during crisis or recessionary periods is viewed necessary for speedy economic recovery. However, the recent global financial crisis serves as a reminder that overexpansion of credit can be disastrous. The rapid growth of bank credit is likely to worsen moral hazard and adverse selection problems and, consequently, increases the probability of a banking crisis (Schularick and Taylor, 2012). According to Corsetti et al. (1999) and Goldstein (1998), credit growth in excess of economic growth characterizes crisis-hit Asian countries during years leading up the Asian financial crisis in 1997. Following the eruption of the global financial crisis, various studies have documented a significant link between a banking crisis and rapid credit growth, see Davis et al., (2016) and references therein. Schularick and Taylor (2012), for instance, estimate the marginal effect of a 1 percentage point increase in the credit to GDP ratio to be 0.3 percentage point increase in the probability of the crisis. At the individual bank level, Foos et al. (2010) indicate the importance of loan growth in driving bank riskiness through higher loan loss provisions, lower risk-adjusted ACCEPTED MANUSCRIPT interest income and lower capital ratios. While these studies link the crises to overexpansion of credit during good times prior to crisis occurrence, Chen et al. (2016) hypothesize potential link between credit growth and bank performance during the global financial crisis. They focus on whether government banks have higher loan growth rates than private banks during the global financial crisis and how the higher loan growth by government banks is associated with their performance. Using bank-level data from 56 countries, they find significantly higher loan growth rates of government banks during the crisis period. Whether their higher loan growth rates impact positively or negatively on bank performance, however, depends on the level of corruption of the countries under study. More precisely, the bright side of government is uncovered only in the countries with low corruption level in that government banks are able to stabilize credit to speed up economic recovery without sacrificing performance. In high corruption countries, it seems that lending decisions by government banks are distorted, resulting in underperformance relative to private banks. In addition, facing heightened credit risk, banks may take risk by lending in excess of the optimal level. As suggested by Jensen and Meckling (1976), a managerial rent-seeking and a conflict of interest between shareholders and creditors may encourage managers or shareholders to make risky loans and shift the risk to depositors. This behaviour is highly likely when banks are engulfed with immense credit problems (Bernanke and Gertler, 1986; Koudstaal and Wijnbergen, 2012). In addition, during the crisis episode, the adverse selection problem is likely at work, where firms that are highly risky are the ones that actively seek financing and secure financing from banks. In a recent paper, Zhang et al. (2016) demonstrate that the lending decisions by banks do exhibit moral hazard in their examination of the Chinese commercial banking system. In the context of a dual banking system, the risk implications of Islamic financing has not been closely scrutinized. This is surprising given the rapidly growing Islamic financing, which has been far exceeding the growth of conventional loans. Moreover, by expanding financing during a period when asymmetric information is surmountable, Islamic banks may expose themselves to excessive risk. This means that the higher expansion of financing during the crisis period by Islamic banks as compared to conventional banks, if it is true, must be assessed. On the basis of the literature reviewed above, we state the following hypotheses: H05: There is no effect of financing/lending growth of Islamic and conventional banks on future bank risk. H06: There is no significant difference in the effect of financing/lending growth of Islamic and conventional banks on future bank risk. H07: Financing growth of Islamic banks during the crisis period is not related to excessive risk. We test these hypotheses to offer a more concrete answer to whether the Islamic banking sector plays a stabilizing role during the crisis period. Hypothesis 5 and hypothesis 6 are to establish the link between lending/financing growth and bank risk. Meanwhile, Hypothesis 7 is to test specifically whether Islamic banks behave in a morally hazardous manner in their expansion of credit during the crisis. 3. MODEL SPECIFICATIONS ACCEPTED MANUSCRIPT 3.1 Lending and Deposit Growth To evaluate whether Islamic banks are able to sustain loan (deposit) growth and have higher loan growth (deposit growth) than conventional banks especially during the global financial crisis, i.e. Hypothesis 1 to Hypothesis 4, we specify the following model for loan growth (deposit growth): ALft(ADft) = + PiIBit + faGFCt + ^3(lBtt + sit (1) Po x GFCt~) + $Controlit + gf where AL is the growth rate of gross loans/financing (AD is the growth rate of deposit), IB is the Islamic bank dummy, GFC is the global financial crisis dummy, Control is a vector of controlled variables comprising bank-specific, macroeconomic and regulation variables, pi- is a bankspecific effect, and st is the standard error term. In line with existing literature, the bank-specific variables include bank size, capitalization, liquidity, profitability, funding ratio, cost efficiency and credit risk. These variables are lagged once to address the endogeneity issue. As for the macroeconomic variables, we include economic growth, inflation and financial development. Finally, as for the regulation, we control for activity restrictions, capital requirement, supervisory power and private monitoring2. Since GFC is key in our analysis, its definition requires deliberation. While the span of recent US recession identified by the National Bureau of Economic Research (NBER) is December 2007 to June 2009, some view the financial crisis to start much earlier (Iley and Lewis, 2013). Indeed, in the literature, the global financial crisis has been referred as the 20072009 financial crisis (Akbar et al., 2017; Flannery et al., 2013) as well as the 2008-2009 financial crisis (Coleman and Feler, 2015; Cull and Martinez-Peria, 2013; and Feldkircher, 2014). According to Berglof et al. (2009), the effects of the crisis that started mid-2007 was confined mainly to advanced economies during the first year of the crisis. In a similar vein, Mishkin (2011) relates the first phase of the crisis from August 2007 to August 2008 to the losses in one relatively small segment of the US financial market before it turned global. This means that, for the analysis of countries other than the US and advanced economies, it is more appropriate to designate 20082009 as the crisis years. Accordingly, GFC takes the value of 1 for years 2008-2009 and 0 otherwise, in line with various studies comparing Islamic and conventional banks (Abedifar et al., 2013; Louchini and Boujelbene, 2016; Olson and Zoubi, 2017)3. The key parameters of the model are P1,P2, and P3. To clarify their interpretation, Table 1 summarizes the specification of (1) conditional on whether the bank is Islamic or conventional and whether the period is a normal or a crisis period (error terms omitted). It is clear from Table 1 that fi2 captures the difference of expected lending (deposit) growth by conventional banks during the crisis period as compared to the normal period, all else equal. If it is negative and significant, we may conclude that the expected lending (deposit) growth of conventional banks We also consider the oil price as a controlled variable given that most of the countries in our sample are oil-rich economies, as suggested by a referee. Following Alqahtani et al. (2016), we include the natural logarithm of the annual average oil price in all regressions that we perform. While Alqahtani et a (2016) find strong evidence for the significant effects of oil price on bank performance, the oil price is significant in less than half of the regressions that we perform. This is not surprising given that the countries in our sample are not confined only to oil-rich countries. More importantly, the main conclusions that we make are not materially affected by the inclusion of oil price. These results are not reported but are available upon request. Defining the global financial crisis years to be 2007-2009 as in Kabir et al. (2015) leads to insignificant effect of the crisis on bank behaviour, which may be due to considering 2007 to be the beginning of the crisis year when, as noted in the text, the effect of the crisis was confined mainly to the advanced economies during 2007. 2 3 ACCEPTED MANUSCRIPT drops during the crisis period. Meanwhile, fi2 + fl3 provides a corresponding figure for Islamic banks and hence serves a basis for testing hypothesis 1 and hypothesis 3. The comparison between Islamic banks and conventional banks is given by fl1 and fl1 + ^3, respectively for normal time and crisis time. The significance of these coefficients would mean the financing (deposit) growth of Islamic banks is different from that of conventional banks during respectively normal and crisis times, the latter of which pertains to hypothesis 2 and hypothesis 4. Specification (1) considers years surrounding the global financial crisis to be normal years. One may argue that the post-crisis period is different from the pre-crisis period. The global economy has witnessed a much volatile oil market during the post global financial crisis. This might have affected bank performance and behaviour as the dual-banking countries under study are mostly oil-rich countries. In addition, the global economy is still wary of the consequences of the European sovereign crisis that followed the global financial crisis. Accordingly, for robustness, we modify (1) by allowing the lending (deposit) growth to be potentially different during the pre-crisis, crisis, and post-crisis periods as: ALft(ADft) = fa + PiIBit + faGFCt + p3(lBit x GFCt) + 04PFCt + P5(IBit x PFCt) + $Controlit + ^ + sit (2) where PFC is the post financial crisis dummy variable taking the value of 1 after 2009 and 0 otherwise. In the empirical implementation, we estimate equations (1) and (2) using a combined sample of conventional and Islamic banks as well as separately for the two types of banks. Obviously, estimating the equations separately for conventional banks and Islamic banks means that the Islamic bank dummy and Islamic bank - crisis interaction are dropped from the equations. In this case, the coefficient of the GFC captures the lending/financing (deposit) growth during the crisis period as compared to its growth during normal time or pre-crisis period for the respective types of banks. If it is negative, we can infer that the lending/financing (deposit) growth drops during the crisis period. The results from these separate regressions would serve as a confirmatory evidence on the relative credit (deposit) growth of Islamic banks vis-a-vis conventional banks. Take note that we opt for static panel specifications of the loan and deposit growth equations. While some studies have allowed dynamics in the lending/deposit growth, our static specification follows Bhaumik et al. (2011), Cull and Martinez-Peria (2013), Kosak et al. (2015) and others. Our choice is justifiable on the basis that we find no persistence in the lending or deposit growth, i.e. the lagged dependent variables incorporated in (1) and (2) carry insignificant coefficients4. Moreover, since the loan level normally exhibits a unit root property, we have no reason to believe that its first difference would be persistent. We estimate the equations using the random-effect panel estimator with robust standard errors clustered at the bank level. With the presence of the Islamic bank dummy, the fixed-effect panel estimator would not be appropriate. Still, we experiment with the fixed-effect panel estimator when we estimate the equations separately for conventional and Islamic banks. Since the results are largely similar, they are not reported to conserve space. 4 We estimate the dynamic specifications of (1) and (2) using the system GMM estimator. ACCEPTED MANUSCRIPT 3.2 Bank Risk Next, we intend to assess whether lending growth is associated with future risk for both Islamic banks and conventional banks and whether there is difference between the two, i.e. Hypothesis 5 and Hypothesis 6. Finally, observing higher financing growth of Islamic banks during the crisis, we examine whether it is related to excessive risk, i.e. Hypothesis 7. To test hypotheses 5 and 6, we follow the dynamic specification as in Foos et al. (2010) but with some adjustments to fit our research objective, written as: NPLit =Po+ PiNPLit-i + 27=1 faj^it-j + $Controlit + gf + eit m (3) m NPLit = fa + faNPLit—i + £ P2ij^Lit—j I(IBt = 0) + £ &2jALit-j I(IBt = 1) j=1 +$Controlit + ^i+ eit j=1 (4) where NPL is the ratio of non-performing loans to gross loans, I(.) is a bank type indicator and it equals 1 when the argument is true, and other variables are as defined before. In line with Foos et al. (2010), we expect credit growth to affect bank credit risk or non-performing loans with time lags since it is unlikely that borrowers will default immediately in the first year of the loans they receive. Equation (3) restricts risk - loan growth relations to be the same for conventional banks and Islamic banks while equation (4) relaxes the restriction. Based on (3) and (4), we test the null hypotheses (i) 'E'lj=1faj = 0, (ii) 2jj=i P21j = 0 and (iii) 2'lj=iP22j = 0 that there is no future risk implications of financing/lending growth respectively for all banks, conventional banks and Islamic banks. In addition, we also test whether there is significant difference in the risk - credit growth relations of conventional and Islamic banks, i.e. 2j=i faj = 27jl=1P22j. We include a larger set of controlled variables than the one used by Foos et al. (2010). These include bank-specific variables (bank size, capitalization, liquidity), macroeconomic variables (GDP growth, inflation, financial development, and crisis dummy), and regulation (activity restrictions, private monitoring, supervisory power, and capital stringency). For Hypothesis 7, i.e. financing growth of Islamic banks during the crisis period is not related to excessive risk, we focus specifically on Islamic banks. Taking insight from Zhang et al. (2016) and Vithessonthi (2016), we specify the model as: m NPLit m = Yo + YiNPLit-i + £ Y2jMit-j + £ Y3j(Mit-j X GFCt—j) =1 +®Controlit + ^ + Eit =1 (5). In line with Zhang et al. (2016), we can infer from the link between loan growth and nonperforming loans whether Islamic financing growth during the crisis constitutes excessive risk taking. As we have noted, asymmetric information abounds in times of crisis. Prudential banks would be more cautious in their lending through, for example, more stringent lending standards. This, according to Vithessonthi (2016), would weaken the link between credit growth and credit risk. If we accept this argument, we may also state that the credit growth - risk relation would be stronger if banks take excessive risk in times of crisis. Moreover, in the presence of asymmetric information and resulting adverse selection problem, lending growth during the crisis period may likely be granted to low quality borrowers, since they are the ones actively seeking loans, or to those rejected by other banks. As a result, loan growth would be translated to higher risk in the future. On the basis of these arguments, we may infer whether Islamic banks ACCEPTED MANUSCRIPT behave prudentially or undertake excessive risk from the significance and sign of Yl>jl=1 Y3j. If l5 6 7 8jl=1 Y3j is significantly positive, then financing growth of Islamic banks during the crisis is construed as excessive risk taking. The dynamic specification of risk equations (3) to (5) renders the traditional panel estimators inappropriate. Even we can assume exogeneity of all explanatory variables, the lagged dependent variable is correlated with the bank-specific effect by construction. The standard and widely-adopted approach in dealing with this endogeneity issue is to apply the GMM estimators developed by Arellano and Bond (1991), Arellano and Bover (1995) and Blundell and Bond (1988). In line with the literature, we utilize the GMM estimators for our purpose. We also take note that the GMM estimators may suffer from instrument proliferation. In addition, they have poor finite sample properties and hence are less suitable when the number of cross-sectional unit is small (Meschi and Vivarelli, 2009). Since the estimation of equation (5) involves only 25 Islamic banks, we follow Roodman(2009) by limiting the number of instruments such that it is less than the number of cross-section units. We also apply the biasedcorrected LSDV (LSDVC) as proposed by Kiviet (1995), Judson and Owen (1999), Bun and Kiviet (1995) and Bruno (2005) to deal with the small micro panel of Islamic banks. included in the analysis. This leads the exclusion of Yemen due to the unavailability of regulation variables. By applying these sample selection criteria, we arrive at a sample of 114 conventional banks and 25 Islamic banks from 10 countries. These countries are Bangladesh, Bahrain, Indonesia, Jordan, Kuwait, Malaysia, Qatar, Saudi Arabia, Turkey and the United Arab Emirates. The data are unbalanced covering the period 2000 to 2014. We admit that other studies tend to have larger sample size. However, their sample selection criteria are less stringent. For instance, in their analysis of the 6 GCC countries, Alqahtani et al. (2016) requires that, to be included in the sample, a bank must have observations on all variables for at least one year. Meanwhile, Kabir et al. (2015) requires data availability of a least 3 consecutive years. Still, it should be noted that a small sample than ours is not uncommon (Bourkhis and Nabi, 2013). Louchini and Boujelbene (2017) employ unbalanced data covering 2005-2012 from the MENA and South-East Asian countries. Despite shorter time span, they end up with only 117 banks in their sample. The bank-level data are sourced from the Bankscope while macroeconomic and regulation variables are from respectively World Development Indicators Database and Barth et al. (2006). Table 2 provides definition and descriptive statistics of the variables used in the present study. On average, the Islamic banking sector has higher financing and deposit growth rates and, at the same time, higher credit risk. In line with Beck et al. (2013), Islamic banks are on average 5 DATA AND ESTIMATION RESULTS 4.1 Data Descriptions We gather bank-level data for dual-banking countries that have significant presence of the Islamic banking sector or exhibit fast-growing Islamic banking in recent years. Our initial sample comprises 296 banks (225 conventional banks and 71 Islamic banks) from 13 countries. Since our objectives are to assess lending growth, deposit growth, and risk of Islamic banks during the global financial crisis not only relative to those of conventional banks but also relative to those during the normal or pre-crisis period, we require a sufficient number of observations in the pre-crisis period such that meaningful comparison can be made. Accordingly, our sample includes countries that have both conventional and Islamic banks having relevant bank-level data available starting no later than 2005. This criterion filters out all banks from Egypt and Pakistan since Islamic banking data are available only after 2005. In addition, we also require data availability of the macroeconomic and regulation variables ACCEPTED MANUSCRIPT better capitalized but less cost efficient. As compared to conventional banks, Islamic banks are less liquid and rely more on customer deposits. While Islamic banks are on average smaller, they are more profitable. As far as the macroeconomic environments of these countries are concerned, these countries record respectable growth, averaging over 5% over the sample period. Indeed, the average annual growth exceeds 4% for all countries in the sample. However, there are large variations in the inflationary experience across these countries. As reflected by the ratio of credit to the private sector to GDP, the financial sector is not well developed. The average credit-to-GDP ratio is 56.6%. Five countries, i.e. Bangladesh, Indonesia, Qatar, Saudi Arabia, and the United Arab Emirates, have average credit-to-GDP ratio below 50% and only Malaysia records the ratio above 100%. Finally, except supervisory power, the regulation indices in these countries are on average slightly above the mid-level. We also compute the correlation coefficients among the variables and find no indication of multicollinearity problem. All correlations are below 0.60, among which only two are above 0.509. 4.2 Lending Growth This section presents the estimation results for lending/financing growth to ascertain whether Islamic banks are able to sustain financing supply and whether there is significant difference between Islamic banks and conventional banks in the provision of credit during the crisis period. These results are reported in Table 3. The key coefficients are the coefficients of the Islamic bank dummy (IB), the crisis dummy (GFC) and the Islamic bank - crisis interactions (IB * GFC), whose interpretation is explained earlier (see Table 1). These are presented in Panel (a) of the Table. In the same panel, we provide the p-values for testing the (i) null hypothesis that there is no significant difference in the financing growth of Islamic banks and the lending growth of conventional banks during the crisis (i.e. P1 + P3 = 0), and (ii) null hypothesis that there is no significant difference in the financing growth of Islamic banks during the non-crisis and crisis periods (i.e. P2 + P3 = 0), which are the focus of the present study. Panel (b) of the Table presents the coefficient estimates of the controlled variables. Regressions (1) and (2) refer to the estimation of equation (1) while regressions (3) and (4) the estimation of equation (2). The insignificance of the IB dummy in all regressions suggests no significant difference in the lending/financing growth of Islamic banks and conventional banks during the normal period. However, during the crisis period, the lending growth of conventional banks drops significantly as reflected by the significant and negative coefficients of the crisis dummy. As compared to the non-crisis years, i.e. regressions (1) and (2), the lending growth of conventional banks are lower by roughly 4 to 5 percentage points. But, when we compare to the pre-crisis years, i.e. regressions (3) and (4), the drop is roughly 8 percentage point. By contrast, the results indicate no significant reduction in the financing growth of Islamic banks. That is, we fail to reject the null hypothesis P2 + P3 = 0 in all regressions. This means that Islamic banks are able to sustain their supply of financing during the crisis. Finally, as should be expected from the above results, we find financing growth of Islamic banks to be significantly higher than the lending growth of conventional banks during the crisis period. The null hypothesis P1 + P3 = 0 is rejected in all regressions. To reaffirm our findings, we re-estimate equations (1) and (2) separately for conventional banks and Islamic banks10. The results, as reported in Table 4, echo well the conclusions made The correlation coefficients are not reported to conserve space. They are available from the authors upon request. Estimating the equations separately for conventional banks and Islamic banks relaxes the assumption of common coefficients across the two types of banks. 9 10 ACCEPTED MANUSCRIPT earlier. As may be noted from the Table, the crisis dummy carries a negative and significant coefficient for conventional banks while it is not distinguishable from zero for Islamic banks. Consistent with the results in Table 3, the lending growth of conventional banks drops by roughly 5 to 8 percentage points during the crisis period. Our finding that Islamic banks maintain their financing during the crisis is in line with earlier results by Hasan and Dridi (2011), Beck et al. (2013), and Ibrahim (2016). Turning to bank-specific controlled variables, we find bank size, liquidity, profitability, funding ratio, credit risk, and cost efficiency to be significant. Larger banks tend to have lower lending growth, regardless of whether the banks are conventional or Islamic. Likewise, Islamic and conventional banks with higher credit risk have lower credit growth. While we find the effect of bank liquidity, profitability, funding ratio and cost efficiency to be positive and significant in a combined sample of conventional and Islamic banks (Table 3), they seem to have differential effects on the lending growth of conventional and Islamic banks (Table 4). More precisely, bank profitability and funding ratio significantly affect the lending growth of only conventional banks while liquidity and cost efficiency are significant in explaining the financing growth of Islamic banks. As regards macroeconomic and regulatory variables, the results generally indicate their significance in explaining credit growth of Islamic and conventional banks. Both GDP growth and inflation affect lending growth of these banks positively. By contrast, financial development tends to slow down credit growth of especially conventional banks. We also document differential effects of regulatory variables on lending growth. While activity restrictions and capital stringency tend to depress lending growth of conventional banks, private monitoring and supervisory power tend to stimulate their credit growth. As for Islamic banks, the supervisory power significantly contributes to their financing growth. Apart from these results, we also find a slowdown in credit growth of conventional banks after the financial crisis. By contrast, Islamic banks in the sample maintain their financing supply post-crisis period, as reflected by the insignificant coefficient of the post-crisis dummy (Table 4). 4.3 Deposit Growth As in the case of lending growth, we estimate equations (1) and (2) for deposit growth using a combined sample as well as separately for Islamic banks and conventional banks. The results are reported in Table 5 and Table 6. Focusing on our main theme, we find the results quite puzzling. Using the combined sample, we find no significant difference in the deposit growth of Islamic banks and conventional banks during the global financial crisis as reflected by the non-rejection of fl1 + P3 = 0. Against this finding, it is puzzling to observe significant reduction in the deposit growth of conventional banks during the crisis as suggested by the significance of the crisis dummy and yet insignificant drop in the deposit growth of Islamic banks as indicated by nonrejection of the null P2 + P3 = 0. Then, when we estimate the equations separately for conventional banks and Islamic banks, we find significant reduction in deposit growth of both types of banks during the crisis period albeit at only 10% significance level for Islamic banks. From these results, we are unable to draw conclusion that the financing growth of Islamic banks during the crisis is supported by a concomitant growth in their deposits. Perhaps, Islamic banks draw down their liquidity holdings to maintain their financing during the crisis episode. We observe that, during the pre-crisis period, their liquid asset-to- asset ratio was 25.9%. During the crisis period, it dropped to 19.8%. At the same time, observing the ACCEPTED MANUSCRIPT increase in capitalization from 13.4% to 14.5% over the periods, we contend that Islamic banks exhibit prudential behaviour by building capital amidst higher financing growth. While these might be the answer for higher financing growth of Islamic banks during the crisis, we leave them for future research. Here, we examine next whether financing growth of Islamic banks during the crisis period is linked to their future credit risk and to excessive risk taking. As for the controlled variables, we find consistent findings for negative and significant effects of bank size and credit risk on deposit growth of both Islamic banks and conventional banks. Given that larger banks tend to have better access to alternative sources of funds, they are less likely to depend on bank deposits as a source of funds or they are less aggressive in sourcing funds from deposits. This result is not in line with Farooq and Zaheer (2015), who document insignificant relation between deposit growth and bank size for Pakistan. As regards bank credit risk, our result conforms to the finding by Farooq and Zaheer (2015) that banks with higher credit rating experience higher deposit growth. For other bank-specific variables, we find bank liquidity and funding to be significant only for conventional banks, suggesting that banks that are more liquid and have higher level of funding deposits tend to have lower deposit growth. Finally, it is very interesting to observe that Islamic bank deposits are not responsive to business cycle, macroeconomic uncertainty and the level of financial development. These results may reflect the notion of “captive customers” in Islamic banking (Azzam and Rettab, 2013). By contrast, deposit growth in conventional bank is influenced by macroeconomic cycle and financial development. 4.4 Credit Risk We first estimate equation (3) by setting the lag order of the credit/financing growth to 311. The first-difference and system GMM results, reported in Table 7, provide evidence that there are lagged effects of credit growth on credit risk. More precisely, credit growth bears significant risk implications after two years, as reflected by significant coefficients of credit growth lagged two and three years. This finding conforms well to Foos et al. (2010). As regards the controlled variables, capitalization, liquidity, economic growth and inflation appear significantly and negatively related to credit risk while the remaining controlled variables are not significant. Thus, the better-capitalized and more liquid banks are less risky. Moreover, economic growth tends to improve credit risk. Finally, under an inflationary environment, banks are likely to be more cautious and accordingly have better credit risk. Having verified that credit growth anticipates future risk, we proceed to assessing whether the credit growth - risk relations are the same for Islamic banks and conventional banks as specified in (4) as a further check of the link between credit growth and future risk. The results from the GMM estimators are given in Table 8. The credit growth - risk relations of conventional banks mimic closely those reported in Table 7. As for Islamic banks, only financing growth lagged two years is statistically significant. When we test the null hypothesis 27=1^21/ = 27= 1 P22j, however, we fail to find significant difference between the credit growth - risk relations of conventional and Islamic banks. As for the controlled variables, the results are in conformity to those documented previously. The key result that we have established here is that the financing growth of Islamic banks does in general lead to future risk. The next question is: Does financing growth of Islamic banks during the crisis, which we find earlier to be higher than the credit growth of conventional banks We also experiment with the lag order of 4 as in Foos et al. (2010) but find credit/financing growth lagged 4 to be insignificant. 11 ACCEPTED MANUSCRIPT and which we observe to be indistinguishable from the level recorded prior to the crisis, amounts to excessive risk taking? To this end, we estimate equation (5) using a sample of only Islamic banks. The results using the system GMM and bias-corrected LSDV estimators are provided in Table 912. The lag order of financing growth is set to 2, in line with the finding we document in Table 8 that only financing growth lagged two years is significantly different from zero. Regressions (1) and (2) of Table 9 relook at the link between credit risk and past credit growth of Islamic banks. As in the combined sample, the coefficient of financing growth lagged two years is positive and statistically significant. The results of estimating equation (5) using the system GMM are given in regressions (3) and (4). While we find financing growth to be related to future risk, we find no evidence that the crisis makes Islamic banks to be more prudent or to be riskier. As reflected by the insignificance of the credit growth - crisis interactions, financing growth during the crisis does not affect risk differentially beyond its documented effect during the normal period. In other words, there is no evidence of excessive risk taking by Islamic banks. Using the biased-corrected LSDV, we find that the once-lagged financing growth turns significant. Similar to the system GMM estimation results, the financing growth - crisis interactions are statistically insignificant. Thus, our conclusion that Islamic financing during the crisis is not the result of excessive risk-taking stands. Among the controlled variables, only capitalization remains robust. Indeed, the negative coefficient of capital suggests that capital adequacy requirements would serve as a risk-mitigating mechanism. In addition, we also note the significance of private monitoring and supervisory power in respectively reducing and increasing bank risk. These results strengthen the ability of Islamic banks to play a stabilizing role during the crisis period. Facing the crisis, Islamic banks sustain their financing supply and have higher financing growth as compared to conventional banks. While we are not able to verify whether there is a parallel increase in Islamic deposits, we find no evidence that Islamic banks exhibit excessive risk taking in their financing expansion during the crisis. 5. CONCLUSION Whether a nascent but fast-growing Islamic banking sector has the ability to be a stabilizing force in times of crisis is a subject that has attracted substantial research attention especially since the global financial crisis. We contribute to this area of research by analysing lending growth, deposit growth and risk of Islamic and conventional banks covering 10 major dualbanking countries from 2000 to 2014. Our main aim is to empirically evaluate whether financing and deposit growth of Islamic banks is resilient to the crisis and whether the Islamic financing growth during the crisis is related to future risk and can be construed as excessive risk-taking. The evaluation is made in comparison not only to normal years but also to conventional banks. From the analysis, we observe no significant reduction in Islamic financing growth during the crisis period. In addition, there is strong evidence that the growth of Islamic financing is higher than the growth of conventional lending during the period. However, we are unable to draw conclusion that the financing growth of Islamic banks during the crisis is supported by a concomitant growth in their deposits. Finally, while we find evidence that financing/lending growth of both Islamic and conventional banks is related to future risk, we find no evidence to suggest that the sustained and higher financing growth of Islamic banks during the crisis period The first-difference GMM estimator and using longer lags yield similar results. These results are not reported to conserve space. 12 ACCEPTED MANUSCRIPT is related to excessive risk taking. That is, there is no increased financing growth - risk relations beyond their observed relation during the normal period. The implication from these findings is straightforward. Given that the Islamic banking system has the ability to contribute positively to financial and economic stability, its development should be further bolstered. Still, safeguard measures are needed to be in place even for the Islamic banking sector. Although we find no evidence that Islamic banks behave in a morally hazardous manner, they do face risk from their intermediation activities. Particularly, encouraging the build-up of bank capital should be facilitated as it would constrain Islamic bank risk. Having stated these, we believe that the ability of the Islamic banking system to inject the much needed financial stability requires further investigation. In the present study, focusing specifically on the effects of the global financial crisis on Islamic and conventional banks, we define the crisis period to be common for all countries in the sample. The documented better performance of Islamic banks may be due to the fact that they are less integrated and less exposed to developments in the global financial market as compared to the longer-established conventional banks and not due to the Islamic principles underlying their operations. Thus, extending the present analysis by assessing the implications of country-level indicators of financial crises as in Doumpos et al. (2015) on bank performance in a dual banking system would provide a firmer evidence for the stabilizing role of the Islamic banking system. In addition, our analysis on the moral hazard behaviour of Islamic banks relies exclusively on the strengthened link between bank financing and bank risk. To be more concrete, this must be further evaluated and extended to whether Islamic banks really share risk or they, as conventional banks do, shift risk. Finally, a more focus and thorough analysis on the determinants of Islamic bank risk is required since, by the very nature of banking, Islamic banks do face risk. The fact that there are risks unique only to Islamic banks such as the Shariahcompliant risk or displaced business risk, the issue of risk determinants requires deeper evaluation. ACCEPTED MANUSCRIPT REFERENCES Abedifar, P., Molyneux, P., Tarazi, A., 2013. Risk in Islamic banking. Review of Finance 17(6), 2035-2096. Akbar, S., Rehman, S., Liu, J., Shah, S. Z. A., 2017. Credit supply constraints and financial policies of listed companies during 2007-2009 financial crisis. Research in International Business and Finance 42, 559-571. Al-Harran, S., 1995. Leading Issues in Islamic Banking and Finance. Petaling Jaya: Pelanduk Publications. Alandejani, M., Kutan, A.M., Samagardi, N., 2017. Do Islamic banks fail more than conventional banks? Journal of International Financial Markets, Institutions & Money. Alqahtani, F., Mayes, D.G., Brown, K., 2016. Economic turmoil and Islamic banking: evidence from the Gulf Cooperation Council. Pacific-Basin Finance Journal, 39, 44-56. Altunbas, Y., Tommaso, C. D., Thornton, J., 2016. Do better-capitalized banks lend less? Evidence from European banks. Finance Research Letters 17, 246-250. Arellano, M., Bond S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies 58, 277-297 Arellano, M., Bover O., 1995. Another look at the instrumental variable estimation of errorcomponents models. Journal of Econometrics 68, 29-51 Ariff, M., 2014. Wither Islamic Banking? The World Economy 37(6), 733-746. Ariff, M., 2015 Islamic banks in Malaysia: industry at crossroads. MGCC Perspectives 21(4), 12-14. Azzam, A., Rettab, B., 2013. Market power versus efficiency under uncertainty: conventional versus Islamic banking in the GCC. Applied Economics 45, 2011-2022. Barth, J.R., Caprio, G., Levine, R., 2006. Rethinking Bank Regulation: Till Angels Govern. New York: Cambridge University Press. Beck, T., Demirguc-Kunt, A. Merrouche, O., 2013. Islamic vs. conventional banking: business model, efficiency and stability. Journal of Banking & Finance 37 (2), 433-447. Berglof, E., Korniyenko, Y., Plekhanov, A., Zettelmeyer, J., 2009. Understanding the crisis in emerging Europe. EBRD Working Paper No. 109. Bernanke, B., Gertler, M. 1986. Banking and Macroeconomic Equilibrium. Princeton University, Woodrow Wilson School of Public and International Affairs. Bhaumik, S.K., Dang, V., Kutan, A.M., 2011. Implications of bank ownership for the credit channel of monetary policy transmission: Evidence from India. Journal of Banking & Finance 35, 2418-2428. Blundell, R., Bond S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87, 115-143. ACCEPTED MANUSCRIPT Bourkhis, K., Nabi, M.S., 2013. Islamic and conventional banks’ soundness during the 20072008 financial crisis. Review of Financial Economics 22(2), 68-77. Brei, M., Schclarek, A. 2013. Public bank lending in times of crisis. Journal of Financial Stability 9, 820-830. Bruno, G. S. F., 2005. Estimation and inference in dynamic unbalanced panel-data models with a small number of individuals. Stata Journal 5, 473-500. Bun, M.J.G., Kiviet, J.F., 2003. On the diminishing returns of higher order terms in asymptotic expansions of bias. Economics Letters 79, 145-152. Chiaramonte, L., Poli, F., Oriani, M.E., 2015. Are cooperative banks a lever for promoting bank stability? Evidence from the recent financial crisis in OECD countries. European Financial Management 21(3), 491-523. Chen, Y.-S., Chen, Y., Lin, C.-Y., Sharma, Z., 2016. Is there a bright side to government banks? Evidence from the global financial crisis. Journal of Financial Stability 26, 128-143. Chong, B.S., Liu, M.-H., 2009. Islamic banking: interest-free or interest-based? Pacific-Basin Finance Journal 17, 125-144. Coleman, N., Feler, L., 2015. Bank ownership, lending, and local economic performance during the 2008-2009 financial crisis. Journal of Monetary Economics 71, 50-66. Corsetti, G., Pesenti, P., Roubini, N., 1999. What caused the Asian currency and financial crisis? Japan and the World Economy 11, 305-373. Cull, R., Martinez-Peria, M.S., 2013. Bank ownership and lending patterns during the 20082009 financial crisis: Evidence from Latin America and Eastern Europe. Journal of Banking & Finance 37, 4861-4878. Davis, J.C., Mack, A., Phoa, W., Vandenabeele, A., 2016. Credit booms, banking crises, and the current account. Journal of International Money and Finance 60, 360-377. Doumpos, M., Gaganis, C., Pasioras, F., 2015. Central bank independence, financial supervision structure and bank soundness: An empirical analysis around the crisis. Journal of Banking & Finance 61, S69-S83. Farooq, M., Zaheer, S., 2015. Are Islamic banks more resilient during financial panics? Pacific Economic Review 20(1), 101-124. Feldkircher, M., 2014. The determinants of vulnerability to the global financial crisis 2008 to 2009: credit growth and other sources of risk. Journal of International Money and Finance 43, 19-49. Flannery, M.J., Kwan, S.H., Nimalendran, M., 2013. The 2007-2009 financial crisis and bank opaqueness. Journal of Financial Intermediation 22(1), 55-84. Foos, D., Norden, L., Weber, M., 2010. Loan growth and riskiness of banks. Journal of Banking & Finance 34, 2929-2940. Goldstein, M., 1998. The Asian financial crisis: causes, cures, and systemic implications. Policy Analysis in International Economics No. 55. Washington D.C.: Institute for International Economics. ACCEPTED MANUSCRIPT Hasan, M., Dridi, J., 2011. The effects of the global crisis on Islamic and conventional banks: a comparative study. Journal of International Commerce, Economics and Policy 2(2), 163200. Ibrahim M., 2016. Business cycle and bank lending procyclicality in a dual banking system. Economic Modelling 55, 127-134. Iley, R.A., Lewis, M.K., 2013. Global Finance after the Crisis: The United States, China and the New World. Cheltemham and Northampton: Edward Elgar Publishing. Jawadi, F., Cheffou, A.I., Jawadi, N., 2016. Do Islamic and conventional banks really differ? A panel data statistical analysis. Open Economies Review 27, 293-302. Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: managerial behaviour, agency costs and ownership structure. Journal of Financial Economics 3(4), 305-360. Judson, R.A., Owen, A.L., 1999. Estimating dynamic panel data models: a guide for macroeconomists. Economics Letters, 65, 9-15. Kabir, M.N., Worthington, A.C., 2017. The ‘competition-stability/fragility’ nexus: A comparative analysis of Islamic and conventional banks. International Review of Financial Analysis, 50, 111-128. Kabir, M.N., Worthington, A., Gupta, R., 2015. Comparative credit risk in Islamic and conventional bank. Pacific-Basin Finance Journal 34, 327-353. Khan, F., 2010. How ‘Islamic’ is Islamic banking? Journal of Economic Behavior & Organization. 76, 805-820. Khan, M.S., 1987. Islamic interest-free banking: A theoretical analysis. In M.S. Khan and A. Mirakhor (eds), Theoretical Studies in Islamic Banking and Finance. North Haledon: Islamic Publications International, 15-35. Khan, T., Ahmed, H., 2001. Risk management: an analysis of issues in Islamic financial industry. Occasional Paper No. 5, IRTI, Islamic Development Bank, Jeddah. Kiviet, J.F., 1995. On Bias, Inconsistency and Efficiency of Various Estimators in Dynamic Panel Data Models. Journal of Econometrics 68, 53-78. Kosak, M., Li. S., Loncarski, I., Marinc, M., 2015. Quality of bank capital and bank lending behaviour during the global financial crisis. International Review of Financial Analysis 37, 168-183. Koudstaal, M., Wijnbergen, S.V. 2012. On risk, leverage and banks: do highly leveraged banks take on excessive risk? SSRN Working Paper 2170008. Louchini, A., Boujelbene, Y., 2016. Credit risk, managerial behaviour and macroeconomic equilibrium within dual banking systems: Interest-free vs. interest-based banking industries. Research in International Business and Finance 38, 104-121. Mejia, A.L., Aljabrin, S., Awad, R., Norat, M., Song, I., 2014. Regulation and supervision of Islamic banks. IMF Working Paper #WP/14/219. Merilainen, J.-M., 2016. Lending growth during the financial crisis and the sovereign debt crisis: The role of bank ownership type. Journal of International Financial Markets, Institutions and Money 41, 168-182. ACCEPTED MANUSCRIPT Meschi, E., Vivarelli, M., 2009. Trade and income inequality in developing countries. World Development 37(2), 287-302. Mishkin, F.S., 2011. Over the cliff: from the Subprime to the global financial crisis. Journal of Economic Perspectives 25(1), 49-70. Mollah, S., Hassan, M.K., Farooque, O.A., Mobarek, A., 2017. The governance, risk-taking, and performance of Islamic banks. Journal of Financial Services Research 51(2), 195-219. Olson, D., Zoubi, T., 2017. Convergence in bank performance for commercial and Islamic banks during and after the global financial crisis. The Quarterly Review of Economics and Finance 65, 7187. Roodman D., 2009. A Note on the Theme of Too Many Instrument. Oxford Bulletin of Economics and Statistics 71(1) 135-158 Schularick, M.H., Taylor, A.M., 2012. Credit booms gone bust: monetary policy, leverage cycles, and financial crises, 1870-2008. American Economic Review 102(2), 1029-1061. Vithessonthi, C., 2016. Deflation, bank credit growth, and non-performing loans: Evidence from Japan. International Review of Financial Analysis 45, 295-305. Zhang, D., Cai, J., Dickinson, D.G., Kutan, A.M., 2016. Non-performing loans, moral hazard and regulation of the Chinese commercial banking system. Journal of Banking & Finance 63, 48-60. ACCEPTED MANUSCRIPT Table 1: Conditional Models Bank/Period Conventional Islamic Normal Crisis 00 + 0Controlit (0o + 02) + $Controlit (£0 + 0i) + $Controlit (0o + 0i + 02 + 03~) + $Controlit Table 2: Variable Definition and Descriptions (a) Bank-specific Variables Variable Loan Growth Deposit Growth Credit Risk Capitalization Liquidity Size Profit Funding Definition Logarithmic difference of gross loans, AL Logarithmic difference of customer deposits, AD Ratio of non-performing loans to gross loans, NPL Equity to asset ratio, EQA Ratio of liquid assets to total assets, LIQA Natural logarithm of total assets, Ln(TA) Return on average assets, ROAA Ratio of customer deposits to total liabilities, FUNDING Cost to income ratio, CIR All Banks Mean Std D 0.156 0.201 IBs CBs Mean Std D Mean J0.182 0.210 0.151 Std D 0.199 0.142 0.246 0.183 0.435 0.34 0.186 6.792 10.53 7.242 10.24 6.918 10.60 12.61 24.87 7.815 15.70 13.03 21.85 8.434 12.53 11.11 25.48 7.686 16.40 15.17 1.612 15.01 1.419 15.20 1.650 1.709 67.32 2.072 16.07 1.797 71.16 2.890 1.691 18.07 66.56 1.867 15.54 Ratio Cost 46.16 33.14 51.34 48.20 45.13 29.13 Efficiency (b) Macroeconomic and Regulatory Variables Variable Definition Mean Std Dev Economic Growth rate of real gross domestic product, AGDP 5.400 3.674 Growth Inflation CPI inflation rate, INF 5.933 7.506 Ratio of credit to the private sector to GDP, FMD 56.573 31.642 Financial Development Activity The degree to which banks are restricted or permitted 2.567 0.671 Restrictions to activities related to securities, insurance and real estate markets and to ownership and control of nonfinancial firms, AR Private 5.778 1.648 Monitoring The degree to which regulations facilitate incentives and ability for private monitoring of banks, PM Supervisory 11.044 2.490 Power The degree to which the country’s bank supervisory agency has the power to take specific actions, SUP The degree of capital stringency regulation, CR 4.806 1.774 Capital Stringency ACCEPTED MANUSCRIPT Table 3: Bank Lending Growth - All Banks Independent Variables Regressions (1) (a) Key Variables IB GFC IB x GFC (2) (3) (4) 0.941 (0.504) -3.909*** (0.002) 4.366 (0.176) 0.410 (0.776) -5.443*** (0.000) 4.589 (0.136) -0.958 (0.661) -7.927*** (0.000) 6.339* (0.060) -1.623 (0.463) -7.805*** (0.000) 6.642** (0.044) 0.0488 0.8762 0.0511 0.7645 0.0446 0.6093 0.0536 0.7088 -1.456*** (0.000) -0.037 (0.778) 0.139*** (0.003) 1.844*** (0.000) 0.101*** (0.008) -0.466*** (0.000) 0.105*** (0.000) 0.639*** (0.000) 0.410*** (0.001) -0.069*** (0.001) -1.824*** (0.000) -0.026 (0.850) 0.091* (0.065) 1.532*** (0.000) 0.112*** (0.004) -0.457*** (0.000) 0.095*** (0.000) 0.530*** (0.000) 0.304** (0.013) -0.071*** (0.001) -2.672*** (0.004) 0.698 (0.108) 0.458** (0.040) -0.693** (0.016) -0.931*** (0.010) 0.114 (0.412) 0.082* (0.089) 1.465*** (0.000) 0.114*** (0.002) -0.513*** (0.000) 0.102*** (0.000) 0.568*** (0.000) 0.399*** (0.001) -0.057*** (0.004) -1.405*** (0.000) 0.083 (0.561) 0.062 (0.223) 1.358*** (0.001) 0.121*** (0.001) -0.490*** (0.000) 0.095*** (0.000) 0.526*** (0.000) 0.327*** (0.008) -0.064*** (0.002) -2.486*** (0.007) 0.279 (0.511) 0.349 (0.107) -0.264 (0.374) -6.002*** (0.000) 4.434 (0.116) 27.589*** (0.001) 1766 139 0.215 Tests (p-values) Pl+P3 = 0 p2+p3 = o (b) Controlled Variables Ln(TA)t-1 EQAt-1 LIQAt-1 ROAAt-1 FUNDINGt-1 NPLt-1 CIRt-1 AGDP INF FMD AR PM SUP CR PFC IB x PFC Constant Observations Number of Banks R2 p-values in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01 21.722*** (0.003) 1766 139 0.187 31.424*** (0.000) 1766 139 0.207 -7.654*** (0.000) 4.109 (0.143) 16.942** (0.021) 1766 139 0.207 ACCEPTED MANUSCRIPT Table 4: Bank Lending Growth - Separate Samples Independent __ Conventional Banks Variables (1) (2) (a) Key Variables -5.495*** -7.874*** GFC (0.000) (0.000) -1.864*** (0.000) -0.109 (0.390) 0.082 (0.108) 1.634** (0.012) 0.116*** (0.004) -0.421*** (0.000) 0.061 (0.142) 0.540*** (0.000) 0.331** (0.026) -0.083*** (0.001) -3.263*** (0.002) 0.770* (0.078) 0.370* (0.084) -0.799** (0.017) -1.392*** (0.000) 0.017 (0.899) 0.045 (0.401) 1.395** (0.033) 0.122*** (0.002) -0.455*** (0.000) 0.059 (0.146) 0.532*** (0.000) 0.366** (0.013) -0.074*** (0.001) -3.039*** (0.003) 0.317 (0.465) 0.229 (0.285) -0.272 (0.431) -6.098*** (0.000) 32.807*** (0.000) 1466 114 0.218 Islamic Banks (3) (4) 0.551 (0.804) 0.506 (0.863) -2.992** (0.022) 0.467 (0.228) 0.344*** (0.003) 0.551 (0.344) 0.041 (0.684) -0.799*** (0.000) 0.112*** (0.002) 0.660* (0.067) 0.345* (0.085) 0.032 (0.501) -0.662 (0.669) -1.384 (0.290) 2.293*** (0.001) 0.489 (0.530) -2.978** (0.014) 0.468 (0.234) 0.344*** (0.005) 0.546 (0.353) 0.041 (0.687) -0.800*** (0.000) 0.112*** (0.002) 0.661* (0.067) 0.346* (0.080) 0.032 (0.504) -0.652 (0.680) -1.392 (0.295) 2.292*** (0.001) 0.496 (0.527) -0.111 (0.973) 19.468 (0.437) 300 25 0.262 (b) Controlled Variables Ln(TA)t-1 EQAt-1 LIQAt-1 ROAAt-1 FUNDINGt-1 NPLt-1 CIRt-1 AGDP INF FMD AR PM SUP CR PFC Constant Observations Number of Banks R2 37.198*** (0.000) 1466 114 0.208 p-values in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01 19.615 (0.448) 300 25 0.262 ACCEPTED MANUSCRIPT Table 5: Bank Deposit Growth - All Banks Independent Variables (a) Key Variables IB Regression (1) (2) (3) (4) 4.188*** (0.001) -2.689** (0.042) -0.595 (0.868) 3.844*** (0.006) -5.153*** (0.001) 0.023 (0.995) 0.347 (0.927) -6.954*** (0.000) 3.248 (0.357) 0.072 (0.984) -7.892*** (0.000) 3.758 (0.282) 0.2735 0.3658 0.2486 0.1995 0.2913 0.2421 0.2676 0.1947 -2.115*** (0.000) 0.084 (0.776) -0.032 (0.450) 0.821* (0.083) -0.157* (0.076) -0.562*** (0.002) 0.081* (0.095) 0.312* (0.079) 0.178 (0.163) -0.048** (0.031) -2.461*** (0.000) 0.052 (0.858) -0.087** (0.039) 0.755* (0.095) -0.165* (0.068) -0.573*** (0.002) 0.086* (0.077) 0.227 (0.304) 0.189 (0.117) -0.050** (0.039) -0.115 (0.889) 2.193*** (0.000) -0.390 (0.384) -0.276 (0.485) -1.646*** (0.000) 0.238 (0.432) -0.091** (0.031) 0.482 (0.316) -0.144 (0.106) -0.609*** (0.001) 0.081 (0.116) 0.243 (0.168) 0.164 (0.205) -0.036* (0.091) -2.048*** (0.000) 0.170 (0.555) -0.120*** (0.005) 0.596 (0.181) -0.155* (0.087) -0.607*** (0.001) 0.088* (0.086) 0.222 (0.305) 0.211* (0.092) -0.043* (0.070) 0.028 (0.973) 1.746*** (0.003) -0.505 (0.240) 0.178 (0.674) -6.897*** (0.000) 7.973 (0.228) 53.972*** (0.000) 1762 139 0.114 GFC IB x GFC Tests (p-values) Pl+P3 = 0 p2+p3 = o (b) Controlled Variables Ln(TA)t-1 EQAt-1 LIQAt-1 ROAAt-1 FUNDINGt-1 NPLt-1 CIRt-1 AGDP INF FMD AR PM SUP CR PFC IB x PFC Constant Observations Number of Banks R2 55.299*** (0.000) 1762 139 0.092 p-values in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01 57.211*** (0.000) 1762 139 0.106 -8.234*** (0.000) 8.022 (0.242) 51.552*** (0.000) 1762 139 0.107 ACCEPTED MANUSCRIPT Table 6: Bank Deposit Growth - Separate Samples Independent Variables (a) Key Variables GFC (b) Controlled Variables Ln(TA)t-1 EQAt-1 LIQAt-1 ROAAt-1 FUNDINGt-1 NPLt-1 CIRt-1 AGDP INF FMD AR PM SUP CR Conventional Banks (1) (2) __ Observations Number of Banks R2 (4) -4.065*** (0.002) -6.480*** (0.000) -7.691* (0.079) -8.279* (0.090) -2.887*** (0.000) -0.203** (0.034) -0.093** (0.035) 0.719 (0.195) -0.154* (0.070) -0.397*** (0.000) -0.016 (0.711) 0.413*** (0.004) 0.266** (0.045) -0.054** (0.033) -1.395* (0.059) 1.598*** (0.000) -0.062 (0.809) -0.232 (0.369) -2.332*** (0.000) -0.067 (0.536) -0.131*** (0.003) 0.479 (0.390) -0.142* (0.093) -0.430*** (0.000) -0.019 (0.652) 0.404*** (0.004) 0.305** (0.019) -0.043* (0.076) -1.182 (0.118) 1.173*** (0.003) -0.216 (0.394) 0.278 (0.276) -6.068*** (0.000) 65.022*** (0.000) 1463 114 0.149 -3.855** (0.018) 1.084 (0.239) 0.364 (0.209) -1.740 (0.326) -0.546 (0.148) -1.905*** (0.008) 0.143** (0.030) -0.553 (0.400) 0.443 (0.140) 0.031 (0.700) 2.215 (0.343) 2.773* (0.081) 0.676 (0.587) 0.685 (0.580) -3.664** (0.015) 1.102 (0.227) 0.355 (0.211) -1.798 (0.312) -0.549 (0.152) -1.919*** (0.009) 0.143** (0.032) -0.545 (0.412) 0.453 (0.114) 0.033 (0.679) 2.335 (0.336) 2.659 (0.152) 0.660 (0.590) 0.777 (0.558) -1.461 (0.772) 68.341** (0.040) 299 25 0.195 PFC Constant Islamic Banks (3) 71.237*** (0.000) 1463 114 0.136 p-values in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01 70.263* (0.054) 299 25 0.195 ACCEPTED MANUSCRIPT Table 7: Risk and Bank Lending Growth First-Difference GMM Independent Variables (1) (2) NPL-1 0.641*** 0.638*** (0.000) (0.000) ALt-i 0.001 -0.003 (0.929) (0.776) ALt-2 0.014*** 0.011** (0.008) (0.038) ALt-3 0.013*** 0.010** (0.000) (0.010) Ln(TA)t-1 -0.239 -0.016 (0.673) (0.976) EQAt-1 -0.237* -0.236* (0.081) (0.080) LIQAt-1 -0.021 -0.028 (0.344) (0.244) AGDP -0.053** -0.053** (0.016) (0.012) INF -0.057** -0.046* (0.044) (0.054) FMD 0.007 0.005 (0.699) (0.784) GFC 0.225 0.190 (0.222) (0.376) AR -0.022 (0.944) PM 0.010 (0.963) SUP 0.093 (0.534) CR 0.009 (0.914) Constant Observations Number of Banks P(Hansen) P(AR1) P(AR2) AA1267 139 0.109 0.005 0.976 p-values in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01 1267 139 0.176 0.005 0.886 System GMM (3) 0.742*** (0.000) 0.003 (0.772) 0.025*** (0.000) 0.024*** (0.000) 0.525 (0.362) -0.172 (0.141) -0.035* (0.072) -0.084** (0.012) -0.042 (0.153) -0.025 (0.155) 0.258 (0.228) -2.919 (0.779) 1406 139 0.074 0.005 0.925 (4) 0.738*** (0.000) 0.001 (0.957) 0.023*** (0.000) 0.021*** (0.003) 0.673 (0.282) -0.165 (0.159) -0.040** (0.035) -0.081*** (0.009) -0.035 (0.173) -0.022 (0.220) 0.269 (0.271) 0.174 (0.784) -0.086 (0.818) 0.142 (0.577) 0.008 (0.932) -6.852 (0.584) 1406 139 0.152 0.005 0.983 ACCEPTED MANUSCRIPT Table 8: Risk and Conventional and Islamic Lending Growth Independent ___ First-Difference GMM Variables (1) (2) NPL-1 0.646*** 0.644*** (0.000) (0.000) ALt-i I(IB = 0) -0.002 -0.005 (0.827) (0.511) ALt-2 I(IB = 0) 0.013** 0.010** (0.013) (0.033) ALt-3 I(IB = 0) 0.015*** 0.013*** (0.001) (0.008) ALt-1 I(IB = 1) 0.023 0.018 (0.457) (0.513) ALt-2 I(IB = 1) 0.024* 0.018 (0.098) (0.203) ALt-3 I(IB = 1) 0.007 0.003 (0.429) (0.717) Ln(TA)t-1 -0.230 -0.082 (0.681) (0.876) EQAt-1 -0.216 -0.214 (0.154) (0.142) LIQAt-1 -0.019 -0.026 (0.420) (0.289) AGDP -0.057** -0.059*** (0.011) (0.006) INF -0.057** -0.043* (0.043) (0.073) FMD 0.008 0.007 (0.650) (0.684) GFC 0.233 0.221 (0.229) (0.319) AR 0.002 (0.994) PM -0.013 (0.952) SUP 0.097 (0.494) CR 0.013 (0.884) Constant System GMM (3) 0.747*** (0.000) -0.001 (0.912) 0.022*** (0.000) 0.023*** (0.001) 0.037 (0.234) 0.044*** (0.006) 0.024 (0.103) 0.519 (0.335) -0.145 (0.275) -0.033* (0.077) -0.087*** (0.009) -0.043 (0.101) -0.024 (0.171) 0.268 (0.181) -3.282 (0.747) (4) 0.745*** (0.000) -0.003 (0.713) 0.020*** (0.003) 0.022*** (0.004) 0.033 (0.267) 0.041*** (0.010) 0.022 (0.136) 0.639 (0.292) -0.139 (0.321) -0.038** (0.046) -0.088*** (0.006) -0.034 (0.195) -0.020 (0.269) 0.290 (0.235) 0.233 (0.731) -0.106 (0.794) 0.139 (0.594) 0.005 (0.961) -6.874 (0.583) P-value: 27=1^2/ = 27=1^3/ 0.5340 0.6190 0.1801 0.1913 Observations Number of Banks P(Hansen) P(AR1) P(AR2) 1267 139.000 0.096 0.007 0.866 1267 139.000 0.152 0.006 0.782 1406 139.000 0.073 0.006 0.960 1406 139.000 0.143 0.006 0.879 p-values in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01 ACCEPTED MANUSCRIPT Table 9: Risk and Financing Growth of Islamic Banks Independent System GMM Variables (1) (2) (3) NPL-1 1.034*** 0.946*** 1.038*** (0.000) (0.000) (0.000) ALt-i 0.034 0.026 0.035 (0.208) (0.259) (0.371) ALt-2 0.037*** 0.023** 0.045** (0.000) (0.016) (0.010) ALt-i x GFCt-i -0.007 (0.791) ALt-2 x GFCt—2 -0.024 (0.657) Ln(TA)t-1 0.492 -0.080 0.407 (0.669) (0.972) (0.729) EQAt-1 -0.229*** -0.164 -0.244* (0.006) (0.131) (0.051) LIQAt-1 0.033 0.025 0.031 (0.685) (0.603) (0.741) AGDP -0.116 -0.039 -0.108 (0.318) (0.638) (0.332) INF -0.018 -0.006 -0.020 (0.350) (0.830) (0.387) FMD 0.003 0.031 0.011 (0.952) (0.508) (0.766) GFC -0.582 0.368 -0.843 (0.546) (0.633) (0.597) AR 1.823 (0.170) PM -1.881** (0.027) SUP 1.218** (0.015) CR 0.057 (0.755) Constant -6.609 -7.521 -5.585 (0.659) (0.825) (0.750) Observations r* 251 251 251 Number of 25 25 25 Banks P(Hansen) 0.977 0.999 0.963 P(AR1) 0.077 0.118 0.078 P(AR2) 0.843 0.862 0.978 (4) 0.990*** (0.000) 0.012 (0.688) 0.031*** (0.003) -0.006 (0.772) 0.004 (0.915) 0.183 (0.907) -0.205* (0.058) 0.001 (0.975) -0.003 (0.976) -0.002 (0.968) 0.025 (0.484) 0.064 (0.969) -0.693 (0.761) -1.360** (0.048) 1.163** (0.024) 0.106 (0.549) -7.021 (0.787) 251 25 LSDVC (5) (6) 1.094*** 1.075*** (0.000) (0.000) 0.036*** 0.027** (0.008) (0.045) *** 0.036 0.027** (0.008) (0.049) 0.016 0.003 (0.507) (0.898) -0.026 -0.030 (0.252) (0.192) 0.474 0.462 (0.393) (0.480) -0.130** -0.163*** (0.014) (0.002) 0.018 0.004 (0.567) (0.910) -0.039 -0.038 (0.626) (0.633) -0.037 -0.045 (0.437) (0.343) 0.058* 0.082** (0.092) (0.022) -0.517 -0.084 (0.443) (0.907) 1.294** (0.014) -1.039*** (0.001) 0.783*** (0.000) -0.229 (0.111) 251 25 251 25 1.000 0.053 0.935 p-values in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01 HIGHLIGHTS • Investigate the lending growth, deposit growth and credit risk of Islamic and conventional banks during the crisis • • • ACCEPTED MANUSCRIPT Lending growth of Islamic banks is resilient to the crisis No different in deposit growth of Islamic and conventional banks during the crisis No evidence of excessive risk taking by Islamic banks in their expansion of lending