1

Risk and Stability in Islamic Banking

Pejman Abedifar*c, Philip Molyneux†, Amine Tarazi* * Université de Limoges, LAPE, 5 rue Félix Eboué, 87031 Limoges Cedex, France

† Bangor Business School, Bangor University, Wales, LL57 2DG, UK

Preliminary draft: do not quote without the permission of the authors

This Draft: 3 rd April 2011

Abstract

This paper investigates risk and stability features of Islamic banking using a

simultaneous modeling framework and a sample of 456 banks from 22 countries

between 2001 and 2008. We find no significant difference between Islamic and

conventional banks in terms of insolvency risk. The results on credit risk suggest that

Islamic banks write-off credits more frequently or/and have lower loan recoverability

compared to conventional banks. We also observe that Islamic banks benefit less than

conventional banks from the negative impact of asset size on both their credit and

insolvency risks. Our results are robust to different samples, estimation procedures,

risk variables and other modeling specifications.

JEL Classifications: G21; G32

Keywords: Islamic Banking, Islamic Finance, bank risk, credit risk, stability, insolvency,

Z-score

c Corresponding Author. Tel: +33 555149251

E-mail addresses: pejman.abedifar@etu.unilim.fr, p.molyneux@bangor.ac.uk, amine.tarazi@unilim.fr.

2

Introduction

The world has observed various evolutionary stages in the field of banking and

currently we see substantial growth in Islamic modes of banking and finance. According

to the Banker (2009), the assets held by Islamic banks (including the Islamic windows of

conventional banks) increased to $ 822bn from $639bn in 2008, an annual growth rate of

28.6% compared to a more modest 6.8% for conventional banks. Islamic banking has

also expanded outside the Muslim world to other continents including Europe and the

Americas1. Despite these recent developments, however, academic study of Islamic

banking remains in its infancy.

Islamic financial principles have evolved on the basis of Shariá law, which

forbids payment or receipt of Riba – the payment or receipt of interest (Obaidullah,

2005). Financing principles are governed by Islamic rules on transactions “Figh Al-

Muamelat” and follow both profit and loss sharing (PLS) and non-PLS arrangements

(such as leasing style arrangements). In addition to the prohibitions on interest, Islamic

banks also face other restrictions – such as the use of many derivatives products, because

according to Shariá all contracts should be free from excessive uncertainty “Gharar”

(Obaidullah, 2005)2.

Several papers have outlined the specific risks inherent in Islamic banking. Errico

and Farahbakhsh (1998) for instance point out that prudential supervision and regulations

governing Islamic banks should place a greater emphasis on operational risk and

information disclosure. They explain the special risks attached to PLS. For instance,

Islamic banks cannot mitigate credit risk by demanding collateral from borrowers;

moreover, they do not have enough control over the management of projects financed in

1 Imam and Kpodar (2010) investigate the determinants of the pattern of Islamic bank diffusion using

country-level data from 1992 to 2006. They find that factors including: income per capita, proportion of Muslims in the population and whether the country is an oil producer or not is related to the development of Islamic banking activity. Other factors such as economic integration with Middle Eastern countries and proximity to Islamic financial centres are also found to be important.

2 Islamic derivative products that are permissible include: spot commodity and money transactions (where exchange takes place contemporaneously or is deferred - the commodity is delivered at t+0 and the money delivered at t+1), and Salam contracts (where money is paid at t+0 and the commodity delivered at t+1). There is widespread debate as to whether futures transactions (where money and commodity payments / deliverables are deferred) are Islamic.

3

the form of Mudarabah3. Khan and Ahmad (2001) claim that sharing Islamic banks’

profit or loss with their investment account holders introduces withdrawal risk. They also

argue that different Islamic modes of finance have their own unique risk characteristics

due to the various constraints enforced by Sharia (Islamic rules). Sundararajan and Errico

(2002) suggest that the complexities of PLS modes of finance and the risks associated

with the non-PLS activities should be taken into account to establish more effective risk

management. They also point out various moral hazard issues that occur as a result of the

special relationship between Islamic banks and investment account holders. Obaidullah

(2005) argues that (deposit) withdrawal risk may persuade Islamic banks to deviate from

traditional Sharia financing principles. This occurs if banks pay competitive market

returns to investment account holders regardless of the bank’s actual performance.

A number of studies have empirically addressed these issues. Čihák and Hesse

(2010), for instance, empirically compare the stability of Islamic versus conventional

banks, using data from 20 member countries of the Organization of Islamic Conference

(OIC) between 1993 and 2004. They find that small Islamic banks are more stable than

similar-sized conventional institutions. Large Islamic banks, however, are less stable than

their conventional counterparts.4 Beck et al (2010) use a large cross-country sample of

banks (covering 141 countries over 1995 to 2007) to investigate the risk and efficiency

features of Islamic and conventional banks. They compare the two types of banks both

internationally and also in the (22) countries where they identify the banks operating side-

by-side. Overall they find that there is little difference in terms of efficiency, asset

quality, stability and business orientation of the two types of banks over the whole study

period. However, they do find that Islamic banks consistently held more capital and

liquidity during the recent banking crisis.

3 It is an Islamic mode of finance. PLS financing is mainly practiced in the form of either Mudarabah or

Musharaka. Refer to Khan (1991), Kahf and Khan (1992), Ahmad (1993) and Iqbal and Mirakhor (2007) for more details of Islamic financial instruments.

4 They employ the Z-score used by Boyd and Runkle (1993) as the stability indicator and use the Bankscope database classification to distinguish between Islamic and other types of banks. This is a limitation as Bankscope classifies banks as commercial, Islamic or other types. However an Islamic bank can be a commercial or a non-commercial bank. Such a classification is problematic: (1) In Bankscope some Islamic banks are mistakenly categorized as commercial banks. (2) Some Islamic banks are investment banks or other types that are not comparable with commercial banks. (3) The data-set also does not differentiate conventional banks with Islamic windows from Islamic or conventional banks.

4

Chong and Liu (2009), studying the Malaysian banking industry, claim that

Islamic banks, in practice, are not too different from conventional banks. Using a Granger

causality test they show that rates of return on the investment deposits of Islamic banks

are closely related to deposit rates on conventional banks’ deposits. They argue that

competition from conventional banks constrains the rates offered by Islamic banks. The

aforementioned study also notes that, on the asset side, Malaysian Islamic banks apply

PLS contracts to a miniscule 0.5% of their financings.

Baele et al (2010) examine default risk for Islamic and conventional loans using

data obtained from the Pakistani Credit Information Bureau between April 2006 and

December 2008. The sample covers all business loans outstanding over the period.

Similar to Chong and Liu (2009), Baele et al (2010) also note that PLS contracts (such as

Mudaraba and Musharakah) play a minor role in Pakistani Islamic banking amounting to

less than 2% of their (Islamic) loan sample. Using a hazard modelling approach and

controlling for a variety of factors, the main finding is that default rates on Islamic loans

are lower than for conventional loans. This, they argue, may be explained by a greater

reluctance of borrowers to default on such loans for religious reasons. Weill (2010)

investigates whether Islamic banks enjoy greater market power as their customers may be

charged higher prices due to their preference for banks offering services compatible with

their religious beliefs. Computing Lerner indices for a sample of 1,301 observations for

34 Islamic and 230 conventional banks operating in 17 OIC member countries between

2001-2007 he finds that Islamic banks have lower market power than conventional banks.

Ongena and Sendeniz-Yuncu (2010) study bank-firm relationships in Turkey using a

multinomial logit approach and a sample of 16,056 bank-firm relationships (of which 2%

relate to Islamic banks). They find that Islamic banks mainly have corporate clients that

are young, transparent, industry-focused, and have multiple-bank relationships.

This paper attempts to contribute to the aforementioned literature by investigating

credit and insolvency risk for a sample of Islamic banks, conventional banks with Islamic

windows (hereafter referred to as Islamic window banks) and traditional commercial

banks from 22 member countries of OIC over 2001 to 2008. We adopt a simultaneous

equations modeling set-up with the two-step GMM technique and a single equation

5

framework, using the fixed effect vector decomposition approach as a robustness check.

Overall we find no significant difference between Islamic banks and their conventional

counterparts in terms of insolvency risk. The results on credit risk suggest that Islamic

banks write-off loans more frequently or have lower recoverability rates compared to

traditional commercial banks. We also observe that Islamic banks benefit less than

traditional banks from risk mitigation associated with size - larger traditional commercial

banks have lower credit and insolvency risk compared to their Islamic counterparts. The

paper is organized as follows: Section I discusses the key features of Islamic finance and

risk issues and Section II outlines our methodology. Section III describes the data and

Section IV presents the results. Finally section V concludes.

I. Background on Islamic banking

This section briefly explains the key features of Islamic finance and its possible

impact on the risk and stability of banks.

I.I. Features of Islamic Finance

Islamic finance is based on Shariá principles which forbid payment or receipt of

Riba5. Riba refers to an excess to be returned on money lending. The Islamic terminology

for such a kind of lending is “Qard Al-Hasan”. It is interesting to note that Shariá

recognizes the time value of money, since according to Islamic rules the price of a good

to be sold on a deferred payment basis can be different from its current value. Interest

reflects the time value of money and the interest rate is an exchange rate across time.

While Shariá recognizes interest in business it prohibits interest on lending (Obaidullah,

2005).

Islamic finance has evolved on the basis of Islamic rules on transactions, Figh al-

Muamalat, and can mainly be categorized as: 1) Debt-based financing: the financier

purchases or has the underlying assets constructed or purchased and then this is sold to 5 There are two types of Riba: Riba in debt and Riba in exchange. For more details see Obaidullah (2005). This paper focuses only on Riba in debt.

6

the client. The sale would be on a deferred-payment basis with one or several

installments. 2) Lease-based financing: the financier purchases or has the underlying

assets constructed or purchased and then rents it to the client. At the end of the rental

period (or proportionate to the rentals) ownership would be transferred wholly or partially

to the client. 3) PLS financing: the financier is the partner of the client and the realized

profit or loss would be shared according to pre-agreed proportions (Khan and Ahmed,

2001). The first two Islamic finance methods are collectively known as Non-Profit and

Loss Sharing “Non-PLS”. Besides restrictions on Riba, Shariá has various other

prohibitions which should be taken into account. For instance, according to the Shariá all

contracts should be free from excessive uncertainty “Gharar” (Obaidullah, 2005); hence

as noted earlier, Islamic financial institutions face some restrictions on application of

financial derivatives and other types of contracts (including various forms of insurance

policies).

I.II. Are Islamic Banks Riskier than Conventional Banks?

In this section, the asset and liabilities structure of Islamic banks are analyzed

highlighting their specific risk features.

Liabilities

Islamic banks are authorized to receive deposits mainly in the following two

forms (Iqbal, et.al., 1998): current accounts6 that bear no interest but are obliged to pay

principal to holders on demand, and investment (or savings) accounts that generate a

return based on profit rates. Such rates may be adjusted according to the realized profit or

even loss which would then be shared between the Islamic bank and the investment

account holders. This PLS arrangement can (in theory at least) provide pro- cyclical

protection to banks in the event of adverse conditions – profit rates decline in bad times

and increase in good times. The extent to which investment deposits are important as a

source of funding, therefore, can have an impact on the asset portfolio of Islamic banks.

6 Deposits are received by Islamic banks in the form of “Qard Al-Hasan” or “Amanaa”.

7

Due to the obligations towards depositors as debt-holders, conventional banks aim

to allocate a considerable part of their funds to loans, and endeavor to decrease the

volatility and uncertainty of loan revenues so as to meet depositor obligations. Islamic

banks, however, have more flexibility, since they can consider investment depositors

more like equity holders. However, this flexibility may be mitigated by the fact that

Islamic banks have limited access to wholesale funding. There is a fledgling Islamic

money market (noticeably in Bahrain and Malaysia) although only the largest institutions

have access. As such Islamic banks are rather constrained from engaging in active

liability management like conventional banks.

The special relationship with investment account holders may also impact on

Islamic banks’ behavior. It may weaken their incentives for due diligence and loan

monitoring, since Islamic banks can transfer credit risk to investment account holders

who do not have the same rights as equity holders but share the same risk (Sundararajan

and Errico, 2002). Alternatively, the special relationship can discipline Islamic banks

more effectively (compared to conventional banks) since investment accounts holders

have greater incentives to monitor Islamic bank performance. In principle, Islamic

depositors are more likely to shift their deposits from poor-performing banks to those

offering higher returns or even to conventional banks7. Hence, there is greater potential

for withdrawal risk (Khan and Ahmed, 2001) and as such depositors can discipline

Islamic banks more actively.

Sharing the realized profit or loss with investment account holders may make

Islamic banks more risky. On the upside, larger payouts to investment account holders

may increase deposits and this can force bank shareholders to raise more equity capital in

order to maintain capital ratios and prevent dilution of their ownership rights. Conversely,

poor payouts may encourage deposit withdrawals leading to potential liquidity and

(ultimately) solvency problems.

7 Investment account holders may be expected to earn lower returns than conventional depositors due to

potentially greater fiduciary risk, i.e. the risk of funds mis-management by Islamic banks (AAOIFI, 1999). Fiduciary risk also includes other aspect of funds mis-management, such as deviation of Islamic banks from the Sharia principles.

8

Islamic Banking: Principles and Practice

Islamic banks, in practice, tend to deviate somewhat from the above mentioned

financing principles and operate similarly to conventional banks. Obaidullah (2005)

claims that the withdrawal risk may persuade management to vary from PLS principles

by paying competitive market returns to investment account holders regardless of realized

performance8. Chong and Liu (2009) use Malaysian data to show that investment deposit

rates of Islamic banks are closely linked to those of their conventional counterparts. They

argue that competitive pressure from conventional banks constrains the actual

implementation of PLS arrangements.

In other words, equity-holders of Islamic banks can be at risk from transferring a

part of their profits to investment account holders so as to reduce withdrawal risk. Such a

risk is known as Displaced Commercial Risk (AAOIFI, 1999). Nevertheless, in the

likelihood of crisis, management is highly likely to share realized losses with investment

account holders to avoid insolvency. This suggests that Islamic banks may have a slightly

greater capacity to bear losses compared to conventional banks. When Islamic banks are

performing well they may adjust profit rates upward but at a slower rate than realized

profitability so as to limit the level and volatility of deposit inflows.

Implicitly, investment account holders own a bond, a long position on a call

option and a short position on a put option. The strike price of the call, however, is

determined arbitrarily by Islamic banks, in the absence of supportive regulations on the

account holders’ rights. The strike price of the put is determined based on the degree of

market competitive pressures, level of incurred loss and the capital ratio of the Islamic

bank. Figure 1 illustrates how the special relationship between investment account

holders and Islamic banks work in theory and practice as compared to depositors of

conventional banks, in the absence of a deposit insurance scheme9.

8 The Governor of the Central Bank of Malaysia, Dr. Aziz, also pointed out this issue in his keynote address

at the 2nd International Conference on Islamic Banking, Kuala Lumpur, held in 2006. For more details, see http://www.bnm.gov.my/index.php?ch=9&pg=15&ac=197.

9 Some countries such as Lebanon and Indonesia have introduced a formal deposit insurance scheme and the deposits held by Islamic banks are insured in the same way as the deposits held by conventional banks. Based on the type of the scheme and the coverage ceiling, the incurred loss of depositors (investment account holders) should have different limits.

9

[FIGURE 1]

Assets

In the process of lending Islamic banks tend to apply non-PLS principles due to

the risks and complexities associated with the PLS method. For instance, under PLS

financing, Islamic banks need to determine the profit or loss sharing ratio for each project

which can be complicated due to difficulties in quantifying the characteristics of clients

and the proposed business opportunity. Revenue is not guaranteed and since they cannot

collect collateral from clients, they need to put more effort into selection and monitoring

so as to ensure that informational rents are not extracted by borrowers. Hence, for short-

term financing, it is not viable for Islamic banks to use the PLS method. Moreover, under

the Mudarabah contract, Islamic banks have limited means to control and intervene in the

management of a project10.

Aggarwal and Yousef (2000) find that Islamic banks mainly use the Non-PLS

instruments to avoid the moral hazard problem associated with PLS financing. Chong and

Liu (2009) show that in Malaysia, only 0.5% of Islamic bank finance is based on PLS

principles. Dar and Presley (2000) claim that even Mudarabah companies in Pakistan,

which are supposed to operate in the form of PLS mainly follow Non-PLS modes of

finance. (This is also emphasized by Baele et al, 2010). According to the Bank Indonesia

(2009) PLS modes of finance accounted for 35.7% in the financing of Islamic banks

operating in the country by the end of 2008. The report points-out that the use of the PLS

method in Indonesia is among the highest compared to what is practiced in other

countries. Milles and Presley (1999) also claim that PLS is only marginally practiced in

Bangladesh, Egypt, Iran, Pakistan, Philippines and Sudan. However, while Islamic banks

appear to refrain from practicing PLS modes of finance they still face possible greater

withdrawal risks than conventional banks (Khan and Ahmad, 2001; and Sundararajan and

Errico, 2002).

10 Errico and Farahbakhsh (1998), Dar and Presley (2000) and Sundarajan and Errico (2002) discuss the

complexity of the PLS method.

10

Complexity of Islamic Modes of Finance

Islamic financing agreements11, even for Non-PLS methods, are not as

straightforward as conventional loan contracts (and according to anecdotal evidence also

take longer to process). Generally, in debt-based or lease-based finance, such as

Murabaha, Islamic banks arrange for the goods/projects to be purchased and then sell or

rent them to clients. For purchase/implementation of the goods/projects, Islamic banks

normally appoint the client as their agent. Such a framework is somewhat complicated as

compared to conventional loan contracts. Sundarajan and Errico (2002) note the specific

risks attached to various Non-PLS methods, such as Salam12 and Ijara13. In the former,

Islamic banks are exposed to both credit and commodity price risks; in the latter, unlike

conventional lease contracts, Islamic banks cannot transfer ownership and therefore have

to bear all the risks until the end of the lease period14.

Another area of debate relates to the treatment of default penalties. Some

jurisdictions rule that such penalties are not authorized by Sharia15, so banks make use of

rebates instead (Khan and Ahmed, 2001). Here the mark-up on the finance arrangement

implicitly covers the return to the banks as well as a default penalty component. If the

client repays the loan in a timely manner then they will receive the rebate. While default

interest payments are typically calculated over the delayed period in conventional

banking, some Islamic banks collect the delayed penalty over the whole financing period.

In addition, Islamic banks can also face restrictions regarding the use of derivatives as

well as different types of collateral, for instance, they are not authorized to use interest-

based assets, like bonds, for security (Khan and Ahmed, 2001).

11 See Khan (1991), Khan (1992), Ahmad (1993) and Iqbal and Mirakhor (2007) for details on the features

of various Islamic financial instruments. 12 Similar in nature to futures contracts. 13 Similar in nature to conventional leasing contracts. 14 Recently, a new instrument called Ijara with Diminishing Musharaka has been developed which enables

Islamic banks to transfer the ownership of the asset in place, proportionate to the rentals paid by the lessee.

15 Islamic scholars generally consider the default penalty as the interest on debt which is prohibited by Sharia as explained in sub-section I.I.; however, it is treated differently across countries. In Iran, for instance, default penalty is a penalty for non-fulfillment of a commitment and it should not be classified as the interest on debt. In Pakistan, Islamic experts have authorized the default penalty, only if it is spent on charity (Baele et al., 2010).

11

Investment Limitations

In addition to lending, conventional banks also allocate a part of their funds to

investments. Such investments normally include purchase of bonds (as well as

instruments with shorter maturities) of different types that have risk/return features that

help manage portfolio risk. However, Islamic banks have limited options for such

investments since they are not authorized to invest in interest bearing instruments.

Alternatively they can invest in Islamic bonds, known as Sukuk16. Although (like in short-

term Islamic money markets) the asset class still remains relatively underdeveloped,

limitations on Islamic bank investment opportunities has been weakened over time due to

the expansion of alternative Islamic financing instruments.

Overall, Islamic banking is characterized by various activities that appear on the

one hand to reduce risk (e.g. PLS sharing) and on the other to increase risk (e.g. limited

tools for active balance sheet management). As such, whether Islamic banking is more or

less risky than conventional banking is an empirical question. The following section

outlines the methodology used to investigate this issue.

II. Methodology and Econometric Specifications

In order to analyze the risk features of Islamic and conventional banking we adopt

an approach similar to Altunbas et. al. (2007), Rossi et al. (2009) and Fiordelisi et. al.

(2010) that uses a modeling approach that links risk, capital and bank efficiency

relationships.

Relationships between Risk, Capital and Efficiency

Several studies investigate the relationship between risk, capital and efficiency in

banking17. Shrieves and Dahl (1992), Jacques and Nigro (1997) and Rime (2001), for

16 They are similar in nature to debt certificate, and can only be issued on the basis of the revenue which is

expected to be generated by an underlying asset. 17 See Shrieves and Dahl (1992), Jacques and Nigro (1997), Kwan and Eisenbeis (1997), Hughes and

Mester (1998) and Altunbas et al. (2007) and Fiordelisi et al. (2010).

12

instance, consider contemporaneous links between capital and risk and find there is a

positive relationship – namely, increases in risk encourage banks to grow their capital.

Alternatively, capital levels may have a positive or negative impact on the riskiness of

loans. Kahane (1977), Koehn and Santomero (1980) and Kim and Santomero (1988)

claim that banks may increase their risks in response to regulatory requirements for

higher levels of capital, since such actions by regulators limits the return-risk frontier and

therefore encourages banks to select riskier asset portfolios. Furlong and Keely (1989)

and Keely and Furlong (1990) argue that the option value of deposit insurance is

decreasing in bank’s leverage so institutions with relatively high levels of risk and

leverage have greater incentives to exploit deposit insurance subsidies incentivizing

greater risk-taking. Konishi and Yasuda (2004) use Japanese data to show that capital

adequacy requirement decrease bank risk-taking incentives.

Jensen (1986) and Harris and Raviv (1990) discuss the possible impact of capital

on inefficiency. They argue that when capital is more expensive than debt (at the margin)

management might endeavor to reduce operating costs to offset the higher financial costs

of the capital increase required by regulators. On the other hand, a fall in interest

expenses may reduce managerial attempts to control operating expenses. Kwan and

Eisenbeis (1997) find a positive link between inefficiency and capital since regulators

require inefficient banks to hold higher levels of capital. Using a simultaneous equation

approach they also show that inefficiency would increase bank risks – illustrating the

moral hazard that poorly-run banks have greater incentives for risk-taking.

Hughes and Moon (1995) and Hughes and Mester (1998, 2010) claim that risk

and capital may be determined simultaneously taking into account the level of efficiency.

The regulators may authorize an efficient bank to exhibit higher leverage. Less efficient

banks may try to increase their risk levels to reach higher levels of profit to compensate

for losses incurred due to inefficiency. Berger and DeYoung (1997) argue that a bank,

which does not efficiently monitor its loan activities, is unlikely to be very efficient in its

operations. Efficiency and risk might also move in the same direction if banks, in an

attempt to maximize short-term profit, decide to become more cost efficient by dedicating

fewer resources to loan screening and monitoring.

13

Altunbas et. al. (2007), use Zellner’s (1962) seemingly unrelated regression

(SUR) approach18 to estimate a system of equations for risk, capital and efficiency. They

show that inefficient European banks take less risk but hold more capital. They also find

a positive relationship between risk and capital, possibly because banks with a higher

level of risk are required to hold more capital. Using the Granger-causality approach,

Rossi et al. (2009) study the possible impact of diversification on risk, capital and

efficiency in Austrian banking. The results show that diversification increases cost

inefficiency, but it strengthens profit efficiency as well as banks’ capital although it

reduces risk. In a similar vein, Fiordelisi et al (2010) examine risk and efficiency in

European banking using the GMM method and a similar modeling set-up to Altunbas et

al. They estimate individual bank risk (accounting and market-based measures),

efficiency (cost and revenue) and capital models in order to test for causal relationships

and find that reductions in efficiency cause increases in bank risk, while improvements in

efficiency strengthens bank capital.

Model Specification

Following in the same vein as Altunbas et al (2007), Rossi et al. (2009) and

Fiordelisi et al (2010), we use the following system of equations to investigate risk and

stability features of Islamic versus conventional banks:

,, 1 2 , 3 , 4 , 5 , 6 7 8 , 1 9 , 1,

3 2 5 7

10, 11, 12, , , 13, 1, , , ,1 1 1 1

i ti t i t i t i t i t i t i ti t

k l m i t m t ti t k i t lk l m t

IneffRisk ISBD ISWD Size MS NII GLGETAOwnershipStructureD AgeD MacroVar YearD

, ,i t

(1)

5, , , , , , , 11 2 3 4 6 7 8,

3 2 5 7

, , 2, ,9, 10, 11, 12,, , , ,1 1 1 1

i t i t i t i t i t i t i ti t

i t m t i tk l m ti t k i t lk l m t

IneffISBD ISWD Size MS Risk ROAAETAOwnershipStructureD AgeD MacroVar YearD

(2)

5, , , ,1 2 3 4,

3 2 5 7

, , 3, ,9, 10, 11, 12,, , , ,1 1 1 1

,, , 16 7 8i t i t i t i ti t

i t m t i tk l m ti t k i t lk l m t

i ti t i tTNEARIneff Risk ETAISBD ISWD Size MS

OwnershipStructureD AgeD MacroVar YearD

(3)

Where the i subscript denotes individual banks and t denotes the time dimension.

Risk (Risk), equity capital (ETA) and inefficiency (Ineff) are modeled in equations 1 to 3,

respectively. We analyze two types of risks: Credit and insolvency risks. The former 18 This approach controls for contemporaneous correlation among the equations’ error terms.

14

primarily relates to credit quality and the latter to bank’s stability. Annex 1 illustrates our

credit and insolvency risk proxies, dependent and control variables. The effect of being

an Islamic bank is captured by a dummy variable which takes the value of one when the

bank is Islamic and zero otherwise (ISBD). Islamic window banks are also represented by

the dummy variable (ISWD). Hence, conventional banks are considered the benchmark.

Dependent Variables

We use three proxies for credit risk (Loan Risk): the ratio of problem loans to

gross loans (PLGL), the ratio of loan-loss reserves to gross loans (LLRGL) and the ratio

of loan-loss provisions to average gross loans (LLPAGL). These proxies are similar to the

variables used by Angbazo (1997), Kwan and Eisenbeis (1997), Shiers (2002), Konishi

and Yasuda (2002), Cebenoyan and Strahan (2004), Gonzalez (2005), Altunbas et al.

(2007) and Lepetit et al. (2008a). Nevertheless, these indicators of credit risk only partly

reflect the quality of the loan portfolio, since variation across banks may be due to

different internal policies regarding problem loan classification, reserve requirements and

write-off policies.

For insolvency risk analysis, we employ the Z-score measure which is widely

used in the literature as a stability indicator (see, for instance, Goyeau and Tarazi, 1992;

Boyd and Runkle, 1993; Lepetit et al., 2008a; Hesse and Čihák, 2007; Čihák et al., 2009;

Laeven and Levine, 2009; Čihák and Hesse, 2010). Using accounting information on

asset returns, its volatility and leverage, the Z-score is calculated as follows:

E ROA ETA

ZscoreSD ROA

, where E(ROA) is the expected return on assets, ETA is the equity

to asset ratio and SD(ROA) is the standard deviation of ROA. Z-score is inversely related

to the probability of a bank’s insolvency. A bank becomes insolvent when its asset value

drops below its debt. The insolvency probability can be written as P(ROA<-ETA). If we

use the standardized ROA, the probability would be equal to

ROA E ROAP Zsocre

SD ROA

.

Hence the Z-score shows the number of standard deviation that a bank’s return has to fall

below its expected value to deplete equity and make the bank insolvent. A higher Z-score

15

implies that the bank is more stable. To control for outliers and skewness of the

distribution, we consider the logarithm of the Z-score and its components.

The ratio of equity to total assets represents equity capital (ETA). Finally, using

the stochastic frontier approach, cost inefficiency (Ineff) is estimated for each bank. (The

methodology used to estimate Ineff is briefly presented in Annex 2)19.

Control Variables

A variety of control variables are also included in the estimation: bank size (Size),

market share (MS), the growth rate of gross loans (GLG), non-interest income (NII),

profitability, i.e. returns on average assets (ROAA), the share of total non-earning assets

in total assets (TNEAR), ownership structure (OwnershipStructureD), bank age or

experience level (AgeD), banking industry and macroeconomic indicators (MacroVar)

and year dummies (YearD). The rationale for their inclusion is set-out below.

The logarithm of total asset (Size) is considered as a proxy for size. Large banks

can benefit from both scale economies and diversification as claimed by Hughes et al.

(2001). At the same time, bigger banks might be more risky, since they may try and

exploit Too-Big-To-Fail safety net subsidies (Kane, 2010). Market share measured as

bank assets over total banking sector assets (MS) is used as the proxy for market power

(as in Berger, 1995).

The share of non-interest income in total operating income (NII) and the growth

rate of gross loans (GLG) are controlled for in the first (Risk) equation20. According to

previous studies, an increase in the share of non-interest income in total operating income 19 See Mohamad et al. (2008) for a cross-country study of Islamic versus conventional banks using the

stochastic frontier approach and a sample of 37 conventional and 43 Islamic banks operating in 21 OIC member countries for the 1990-2005 period. They find no significance difference in terms of efficiency between Islamic and conventional banks; however, Abdul-Majid et al. (2010) apply a distance function approach and find that Islamic banks are less technically efficient than their conventional counterparts. They set-up a sample of 558 observations covering 23 Islamic and 88 conventional banks that operate in 10 OIC member countries between 1996-2002. Beck et al (2010) use more conventional measures of bank efficiency – overhead costs and the cost-to-income ratio. Starting with a sample of 2956 banks (of which 99 are Islamic) from 141 countries between 1995 and 2007. Islamic banks appear more efficient than their conventional counterparts. However, when they examine data from the 22 countries where Islamic and conventional banks compete together they find that Islamic banks have significantly higher overhead costs but only slightly higher cost to income ratio compared to conventional banks.

20 We control for the loan growth (GLG) when we investigate the loan risk; however, for insolvency risk analysis, as we need to take into account the growth strategy of banks, GLG is replaced by the asset growth (TAG).

16

is expected to lower stability. DeYoung and Roland (2001) and Stiroh (2004, 2006,

2010), for instance, claim that the increased reliance on non-interest income has raised

the volatility of bank portfolios without increasing average profits. Lepetit et al. (2008a)

show that European banks with a higher non-interest income share in their net operating

income, exhibit a higher insolvency risk. It is also important to control for loan growth, as

a considerable increase in credit may be reflective of weaker screening standards,

relaxing collateral requirements or lowering interest rates (Dell’Ariccia and Marquez,

2006; Ogura, 2006). Clair (1992) finds a negative effect of credit expansion on non-

performing loans and the loan charge-off rate for the next year, whereas for subsequent

years a positive relationship is detected. As pointed out by Berger and Udell (2004) and

Foos et al. (2010) borrowers do not default immediately after taking loans.

Berger (1995) shows the positive impact of earnings on capital and therefore we

include returns on average assets (ROAA) as an explanatory variable in our capital

equation. The share of total non-earning assets in total assets (TNEAR) is incorporated in

the model 3 (Ineff) since higher shares may reflect inefficient asset management

(Angbazo (1997) uses a similar ratio to represent management efficiency).

Bank ownership structure should be also taken into account. La Porta et. al.

(2002) analyzes government ownership of large banks in 92 countries and shows that it

reduces efficiency. Bonin et al. (2005) investigate the impact of ownership on bank

efficiency for eleven transition countries and find that foreign-owned banks are more cost

efficient than other banks. Iannotta et al. (2007) using a sample of 181 large banks from

15 European countries claims that state-owned banks have poorer loan quality and higher

insolvency risk than other types of banks21. In our model, we classify banks into four

categories22: domestic privately-owned banks, domestic state-owned banks (StateD),

foreign-owned banks (ForeignD) and subsidiaries (SubsidiaryD). Domestic privately- 21 For a discussion of empirical estimation of ownership issues in banking see Altunbas et al (2001) and

Goddard et al (2004). More recent studies include Barry et al (2011), Taboada (2011), Forssbæck (2011) and Berger et al (2009) to name just a few.

22 We classify a bank as a state-owned bank when at least fifty percent of the equity belongs to the government. Similarly, at least fifty percent of a bank should be owned by one or more foreign entity(ies) to be classified as a foreign-owned bank. A bank which is owned by a foreign government is considered as a foreign-owned bank. We assume that although a government may decide to invest in a bank abroad based on political ties with the host country, it will not intervene in the bank’s operation as intensively as the host country’s government.

17

owned banks are used as the benchmark and hence three dummies are introduced to

represent the other banks.

State-owned banks may invest in risky projects as a result of political influence,

or/and they may also enjoy some benefits and informational rents from political bodies.

Foreign-owners can face greater risk in monitoring the bank’s activities since they may

be less familiar with the legal and judicial setting in which they operate. Alternatively,

due to such problems they may pursue relatively conservative strategies. A subsidiary

might structure a risky portfolio of loans, simply because such a portfolio can beneficially

contribute to diversification of the parent’s overall portfolio. Failure of a subsidiary may

not be viewed as undesirable in the event of a crisis if reputational risks are low.

We also consider the age of the bank by defining two dummy variables. Banks

with at least three years of operation are categorized as young banks and those which

have been operating for a period ranging from three to seven years are considered as

middle aged. Other banks, called mature banks, are considered as the benchmark. The age

of banks is expected to proxy for experience and informational advantages. Older banks

are likely to have longer term relationships and other informational advantages

(experience operating in new geographies and product markets) that are reflected in

efficiency and risk advantages. (Of course, it could be the case that younger institutions

have tougher regulatory oversight and therefore operate more cautiously).

We also introduce five country level variables to control for cross-country

variations in market structure: the Herfindahl-Hirschman Index as the market

concentration index (HHI), a dummy variable which takes the value of one for countries

with an explicit deposit insurance scheme (EDInsuranceD), the share of domestic credit

provided by the banking system in GDP (DomCredit), official exchange rate depreciation

(ERDep) and the growth rate of GDP per capita (PerCapitaGrowth). Year dummy

variables are also introduced in the model to control for time fixed effects.

Estimation method

We use the two-step GMM technique to estimate the system with robust standard

errors. All right-hand side variables, except risk, equity capital (ETA) and inefficiency

18

(Ineff) are used as instruments in all the equations. NII and GLG are only used in the risk

equation, ROAA is only used in the equity capital (ETA) equation and TNEAR is only

used in the cost inefficiency (Ineff) equation. Hence, the risk equation is just-identified,

whereas equity and inefficiency equations are over-identified. Since we are mainly

focusing on the risk equation and since the other two equations are over-identified,

simultaneous equation modeling provides us with more efficient estimation than the

single equation set-up.

III. Data and Descriptive Statistics

Bank-level data has been retrieved from the Bankscope database and the web sites

of individual banks. Country-level variables, including DomCredit, ERDep and

PerCapitaGrowth, are collected from the World Bank web-site. For EDInsuranceD, we

use the detailed data-base on the deposit insurance schemes presented in Demirguc-Kunt

et al (2005)23. The Bankscope classification for Islamic banks is incorrect in places so all

banks have been cross-checked with their websites to ensure accuracy. The sample covers

2674 observations for 456 commercial banks, across 22 country24 members of the OIC

where Islamic banking is practiced over the period 2001 to 2008. Our sample comprises

101 Islamic commercial banks, 72 commercial banks with Islamic window/branches and

283 conventional commercial banks25. For Iran and Sudan observations are only available

for Islamic banks as their banking systems are 100% Riba-free. In other countries, both

Islamic and conventional banking are authorized and practiced. The largest number of

observations is from Indonesia and the lowest from Gambia. Approximately, 19% of the

total observations are for Islamic banks, Islamic window banks represent 17% of the

sample (the remaining 64% relate to conventional banks). Islamic banks are relatively

younger than conventional banks and also the number of banks with foreign owners is

23 We assume that the deposit insurance schemes had not changed during the period of study. 24 Algeria, Bahrain, Bangladesh, Egypt, Gambia, Indonesia, Iran, Jordan, Kuwait, Lebanon, Malaysia,

Mauritania, Pakistan, Qatar, Saudi Arabia, Senegal, Syria, Sudan, Tunisia, Turkey, UAE and Yemen. 25 We remove fifty outliers and eliminate observations for countries with less than 1% of the total number

of observations (observations belonging to Iraq, Palestine and Brunei are dropped from the sample).

19

higher for Islamic compared to conventional banks (see Annex 3 for detailed cross

country and banks’ types sample specifications).

Table (1) illustrates sample descriptive statistics. It shows that relatively large

conventional banks establish Islamic windows. Islamic banks are, on average, more

capitalized and profitable than conventional banks. The lower levels of debt (as a

response to higher withdrawal risk) and implicit interest expenses of Islamic banks might

partly explain their higher profitability. Gross loans and total assets grow at higher rates

than those of conventional banks. Interestingly, the structure of the asset portfolio of

Islamic banks is significantly different from that of conventional banks. Islamic banks

have a higher ratio of net loans to total earning assets possibly because they are limited in

their investments in other earning assets (such as bonds) as discussed in section I.II. They

also allocate more funds to non-earning assets than conventional banks. The cost to

income ratios and share of non-interest income in total operating income is not

significantly different between Islamic and conventional banks.

The descriptive statistics of our risk measures show that Islamic banks have lower

levels of credit risk compared to conventional banks, when we consider the stock proxies,

i.e. LLRGL and PLGL; however the flow proxy, i.e. LLPAGL, depicts slightly higher

credit risks. In terms of insolvency risk the mean test results show that the Z-score and its

components for Islamic banks are not significantly different from those of conventional

banks, suggesting that the higher asset return volatility of Islamic banks is offset by their

higher returns and capital levels. Nevertheless, the mean test for the 3-year rolling

window Z-score shows that Islamic banks are less stable than conventional banks.

[TABLE 1]

A correlation matrix is presented in annex 4 which does not suggest any major

collinearity problems among our independent variables, except for the logarithm of total

assets and market share variables. As a result, we orthogonalize the logarithm of market

share on the logarithm of total assets.

20

IV. Empirical Results

IV.I. Credit Risk

Table 2 presents the results of the simultaneous estimation of the risk, capital and

inefficiency equations using the two-step GMM method on an unbalanced panel for the

2001-2008 period. We use LLRGL as the loan risk proxy in system estimations (1) to (4).

The first estimation is obtained from the baseline model. In estimation (2), we add an

interaction term by multiplying the Islamic bank dummy variable and bank size to

capture possible differences in the relationship between risk and size among Islamic and

conventional banks. System estimation (3) is based on Hughes and Moon (1995) and

Hughes and Mester (1998). We estimate a system of two equations of risk and capital

wherein the level of inefficiency is controlled for. In this set-up, the lagged value of

inefficiency is used as the pre-determined variable. In the fourth estimation we add the

interaction term of the Islamic bank dummy and size to the system (3). In the second four

sets of estimations, i.e. (5) to (8), we estimate our models using PLGL in lieu of LLRGL.

In all eight system estimations, we perform Hansen’s (1982) J test to determine

the validity of the over-identifying restrictions; the test is often interpreted as a test of the

validity of the instruments. The estimated Chi-squared in all eight tests does not suggest

model mis-specification.

The results show that Islamic banks, on average, exhibit lower risk and higher

capital than conventional banks. However, when PLGL is used as the proxy, the

coefficients of the Islamic bank dummy variable in the credit risk and capital equations

turn out to be insignificant. Islamic banks’ level of inefficiency does not significantly

differ from that of conventional banks. Islamic window banks also exhibit lower credit

risk than their conventional counterparts. We find a negative relationship between size

and the level of credit risk, which is consistent with possible diversification and scale

economies benefits; however, the interaction term of size and Islamic bank dummy is

21

positive and significant26, suggesting that the positive impact of size on loan quality is

weaker for Islamic compared to conventional banks27.

We also find that greater risk increases the level of equity (and higher equity leads

to lower risk). This confirms the findings of Konishi and Yasuda (2004) that capital

adequacy requirements decrease banks’ risk-taking incentives. Higher risk also appears to

increase inefficiency levels whereas greater capital has the opposite influence on bank

inefficiency. However, in the two-equation set-up, we find that an improvement in cost

efficiency strengthen bank’s capital, a result similar to Fiordelisi et al (2010). Higher

shares of non-interest income increases credit risk consistent with the results obtained by

Lepetit et al (2008b) for European banks. Loan growth is associated with lower credit risk

in the following year, as also identified by Clair (1992).

In terms of ownership structure, the credit risk of state-owned banks is higher than

that of domestic privately owned banks at the ten percent significance level (a result

confirmed by Iannotta et. al. , 2007), but their capital levels and inefficiency are not

significantly different from those of domestic privately-owned banks. Foreign-owned

banks, on average, are more exposed to credit risk, but have less capital than domestic

privately-owned banks. Subsidiaries, on average, hold higher capital and are more cost

efficient than domestic privately-owned banks, but their credit risk is not significantly

different from their domestic counterparts. Young banks exhibit higher credit risk.

The results also show that an increase in banking market concentration enhances

bank risk-taking incentives. We find that banks in countries with explicit deposit

insurance schemes hold less risky loans. Banks operating in economies with a higher

share of domestic credit in GDP also exhibit a higher exposure to credit risk. Finally,

home currency depreciation is associated with lower bank credit risk.

[TABLE 2]

26 Since the interaction term is highly correlated with size, we also present the joint significant test. 27 When we employ LLRGL as the proxy, the F-statistics rejects the null hypothesis that the summation of

size and the interaction term equals to zero. However, when PLGL is used in our model, the F-statistics does not reject the null, suggesting that for Islamic banks, size has no significant impact on the loan risk.

22

Overall, despite the restrictions and limitations imposed by Islamic finance, there

is at least some evidence that Islamic banks have lower credit risks than conventional

banks – although these results cannot be confirmed when PLGL is used as the credit risk

proxy. One possible explanation for this finding could be higher withdrawal risk that

Islamic banks face compared to conventional banks (as discussed in section I). However,

the likelihood of withdrawal risk depends on depositors views that they will not be

remunerated (or will incur deposit losses) if the Islamic bank also makes a loss28.

Another possible explanation for this finding could be that Islamic banks, to some

extent, attract clients with religious concerns that are less likely to default (as partially

suggested in Baele et al , 2010). At a higher levels of asset size, Islamic banks move

toward larger clients which are less sensitive to religious concerns. This could also

explain why Islamic banks benefit less than conventional banks from the negative effect

of size on risk.

IV.II. Insolvency Risk

Since our insolvency risk proxy accounts for the degree of leverage, we are left

with two estimable equations for risk and inefficiency. Table 3 reports the simultaneous

estimation of insolvency risk and inefficiency equations, using the two-step GMM

approach. We consider the logarithm of the Z-score computed on a 3-year rolling window

as our proxy for insolvency risk. The estimated Chi-squared of Hansen’s (1982) J test

does not indicate that our insolvency risk models are mis-specified. 28 In order to evaluate the likelihood of such events, using our sample of 456 banks across 22 countries

over the 2001-2008 period, we compare the payoffs to depositors, measured by the implicit interest expense rate (IIER) to returns to equity-holder (ROAE) for Islamic and conventional banks. We find that Islamic banks with negative ROAE pay positive returns to their investment account holders. 18 Islamic banks operating in 9 countries behave in such a way. Nevertheless, at higher levels of ROAE, Islamic banks seem to pay more to investment account holders, showing that they partly share the realized profits. Hence, as figure 1 illustrates the degree of deviation from PLS principles depends on the level of earnings. We also compare the correlation between ROAE and IIER across different banks of our sample. At the 25th percentile, the correlation shows no relationship between ROAE and IIER for Islamic banks; however, when moving toward higher levels of payoffs, we find stronger and positive relationships between the payoffs to depositors and equity-holders of Islamic banks, as the correlation increases from -0.05 at the 25th percentile to 0.375 for observations above the 75th percentile. For conventional banks and Islamic window banks, at the 25th percentile, there is a negative correlation between ROAE and IIER, but for payoffs higher than this percentile, the results show no significant relationship between payoffs of depositors and equity-holders. As expected, the results show that the potential discipline imposed by depositors is replaced by stronger management monitoring of equity-holders, similar to conventional banking.

23

In the first system estimation, we find no significant difference between Islamic

and conventional banks in terms of insolvency risk which supports the finding of Beck et

al. (2010). In the second and third system estimations, we replace the stability proxy by

the logarithm of its first and second components and find no significant difference

between Islamic and conventional banks. Size depicts no significant impact on

insolvency risk. Nevertheless, when the interaction term for the Islamic bank dummy and

size is added in the fourth to the sixth system estimations, we find a negative and

significant impact of size on the stability of Islamic banks. A possible explanation could

be that for larger banks, risk management skills play a more important role than for

smaller institutions. As noted by Čihák and Hesse (2010), the risk management

limitations that Islamic banks face may explain the different size effects on stability.

Higher insolvency risk also appears to increase bank inefficiency. As expected,

asset growth is also associated with lower stability. The insolvency risk and inefficiency

level of state-owned banks are not significantly different from those of domestic

privately-owned banks. Foreign-owned banks, on average, exhibit higher risk and

inefficiency than domestic private-owned banks. Subsidiaries, on average, present no

significant difference from domestic privately-owned banks in terms of stability, but they

are more cost efficient than domestic private-owned banks. Young banks are more

inefficient and less stable than mature banks at the ten percent significance level.

The results also indicate that increases in banking market concentration have no

significant impact on bank’s stability. We find that banks in countries with explicit

deposit insurance schemes are less stable. Banks, operating in economies with higher

shares of domestic credit in GDP exhibit on average lower insolvency risk. Finally, home

currency depreciation is associated with lower bank’s stability.

[TABLE 3]

IV.III. Robustness Checks and further issues

As a robustness check and to deal with possible endogeneity issues we estimate

the risk model, using the lagged values of all accounting and macro level variables in the

right-hand-side of the risk equation. Panel data enable us to control for unobservable

24

heterogeneity across individual banks. The Hausman test rejects random effects

estimation; on the other hand, we have several dummy variables that are time invariant

such as SubsidiaryD and EDinsuranceD as well as some that rarely change over time, i.e.

ISBD, ISWD, StateD and ForeignD29. These variables have limited within variations,

since some conventional banks have established Islamic windows or have fully converted

their operation to Islamic banking. Some banks have also experienced changes in their

ownership structures, for instance they were privatized or nationalized during the period

of study. The fixed effect (FE) technique cannot estimate the time invariant variables and

it is inefficient in estimating variables with very limited within variance30. Hence, we

follow the fixed effects vector decomposition (FEVD) approach proposed by Plumper and

Troeger (2007) to capture unobservable individual-specific effects31.

Table (4) presents the results of single equation estimation of credit risk, using the

FEVD approach. In columns (1) to (5) LLRGL is used as the credit risk proxy. Column

(1), similar to our previous results, shows that Islamic banks are less exposed to credit

risk than conventional banks and size has a positive effect on credit quality. In column

(2), the interaction of size and the Islamic bank dummy variable is added in the model.

The results show that Islamic banks benefit less than their conventional counterparts from

the positive impact of size on credit quality which supports our previous finding.

Columns (3) and (4) illustrate the estimations when small and large bank sub-samples are

used32. In both specifications, Islamic banks exhibit lower credit risk than conventional

29 The average within standard deviation of ISBD, ISWD, StateD and ForeignD are 0.0011, 0.0021, 0.0021

and 0.0022 respectively. 30 When the unobservable individual effects are correlated with explanatory variables, we cannot use

random effect or pooled OLS techniques, since they generate biased and inconsistent estimates. In our model, the Hausman test suggests application of FE.

31 The FEVD technique consists of three steps: first we estimate the model using the centered variables, which is in FE estimate; then we regress the unit effects obtained from the first step on time invariant and rarely changing overtime variables. This stage splits the individual effects into two components: an explained and an unexplained part. Finally we estimate the model, using pooled OLS where the residual of the second step, representing the unexplained part of the FE vector, are incorporated.

32 Banks with total assets less than one billion US$ are classified as small. De Young, et. al. (2004) claim that small and large banks operate differently - small banks generally deal with small companies, which are relatively opaque. Large banks, however, can benefit from economies of scale, standardized products and are more transaction (as opposed to relationship) based. They mostly analyze hard information obtained from transparent firms. Hence, empirical investigation of the sub-samples might show the possible impact of different customer relationships on the credit risk of Islamic versus conventional banks.

25

banks. In column (5) the interaction of the Islamic bank dummy and capital ratio is

included in the model to investigate whether the leverage ratio has a different impact on

credit risk for Islamic banks compared to conventional banks. The results show that the

coefficient is positive but significant only at the ten percent level.

In columns (6) to (10), PLGL is used as the proxy. Column (6) shows that Islamic

banks hold less risky loans than conventional banks. We also find that size has a negative

effect on credit quality which is in contradiction with our previous findings. Nevertheless,

in column (7) we find that the negative effect of size on credit quality is stronger for

Islamic banks compared to conventional banks. Column (10) also illustrates no

significant difference between Islamic and conventional banks in terms of the relationship

between capital and risk.

Columns (11) to (15) illustrate the results when we use LLPAGL as the risk

proxy33. Despite our previous results, here we find that Islamic banks exhibit higher

credit risk than conventional banks and size has a positive influence on such risks.

Finally, in column (15) we do find that the leverage of Islamic banks has a significantly

different impact on credit quality for Islamic and conventional banks34.

[TABLE 4]

The conflicting results we find comparing stock (LLRGL & PLGL) and flow

(LLPAGL) credit risk proxies suggest that Islamic banks either write-off loans more

frequently than conventional banks or that they have lower loan recoverability skills.

Several factors may explain these two possible explanations. First, Islamic banks are

more profitable than conventional banks, and hence they may write-off loans more

frequently. Second, since Islamic banks must (in theory at least) share realized profits

with investment account holders they may be more conservative in classifying problem

loans. Third, for loans granted under the PLS framework, Islamic banks cannot require

collateral and for loans granted under Non-PLS methods, some Islamic banks face

33 We do not use LLPAGL in the main regression, since the Hansen’s J test suggests that our model with

LLPAGL as the risk proxy is mis-specified. 34 We also replace cost inefficiency estimated using the stochastic frontier approach with cost to income

ratio (ctir) and obtain similar results. We do not report the results in the paper; however these are available from the authors on request.

26

limitations in charging default penalties; as a result, they may not have as strong ability to

recover loans like their conventional counterparts. As pointed out by Baele et al. (2010),

this may persuade borrowers with a higher probability of default to prefer Islamic over

conventional loans, leading to lower loan quality (although taking into account their

limitations in being able to charge default penalties Islamic banks may put more effort

into identifying lower risk borrowers). Overall, the results on credit risk features of

Islamic banks compared to their conventional counterparts appear somewhat ambiguous.

Table (5) presents the single equation estimation of the insolvency risk model,

using the FEVD approach. Column (1) shows that the Islamic bank dummy variable is

negative and significant only at the five percent level, suggesting higher insolvency risk

of Islamic banks compared to conventional banks. We find that size has a positive effect

on bank’s stability. Column (2) shows that the risk-adjusted return of Islamic banks is not

significantly different from that of conventional banks, suggesting that higher returns of

Islamic banks is offset by their higher return volatility; however, in column (3), we find

that Islamic banks are not capitalized enough to compensate them for their higher return

volatility. In columns (4) to (6), the interaction term combining size and the Islamic bank

dummy variable is included. The results suggest that, for Islamic banks, size has a

negative effect on stability, while for conventional banks there is a positive relationship

between size and stability. Columns (7) to (9) show that the stability of small Islamic

banks is not significantly different from conventional banks of the same size; however, in

columns (10) to (12), large Islamic banks exhibit greater insolvency risk compared to

large conventional banks (at the five percent significance level). This finding is in-line

with the general result pointing to a negative link between asset size and stability for

Islamic banks.

To investigate whether the composition of total earning assets can have a

significant effect on a bank’s stability, we include in our estimations the share of net

loans in total earning assets (NLTEAR) in columns (13) to (15). Our results show that the

coefficient of NLTEAR is positive and significant at the ten percent level. In columns

(16) to (18), the squared term of NLTEAR is also added to the model. The F-statistics of

27

the joint significance test does not reject the null hypothesis that coefficients of both

NLTEAR and its squared term are equal to zero35.

[TABLE 5]

As a further robustness check, we estimate the model, using the logarithm of the

Z-score where return volatility is calculated over the whole period, for banks with at least

five consecutive observations. On the right hand side of the equation, we use the mean

value of the explanatory variables over the sample period. This approach provides us with

the between group estimation and decreases noise although we have to use cross-

sectional estimation rather than a panel data technique. Table 6 illustrates the estimations.

In column (1), we find no significant difference between Islamic and conventional banks

in terms of insolvency risk. The relationship between size and stability is also

insignificant. Columns (2) and (3) illustrate that the first and second components of

Islamic banks’ Z-scores are not significantly different from those of conventional banks.

In columns (4) to (6), the interaction term combining size and the Islamic bank dummy is

added. We find no significant difference between Islamic and conventional banks in

terms of the size effect on stability. Columns (7) to (12) show that stability as well as the

logarithms of the first and second components of the Z-score for small and large Islamic

banks are not significantly different from conventional banks of the same size. Finally, as

in our previous findings, columns (13) to (18) show that the ratio of net loans to total

earning assets and its squared term have no significant impact on bank’s stability36.

[TABLE 6]

V. Summary and Conclusion

This paper analyzes the risk and performance features of Islamic banks. The

obligations of Islamic banks towards depositors (investment account holders) are

35 We replace cost inefficiency estimated using stochastic frontier approach with cost to income ratio (ctir)

and obtain similar results. We do not report the results in the paper; however it is available from the authors on request.

36 We replace cost inefficiency estimated using stochastic frontier approach with cost to income ratio (ctir) and find that Islamic banks are more stable than conventional banks at the five percent significance level. We do not report the results in the paper; however the results are available on request.

28

different from those of conventional banks and hence they face different risks.

Conventional banks have to fulfill their obligations towards depositors irrespective of

their profits or losses whereas Islamic banks are supposed to share the realized profit or

loss with investment account holders. In practice, to avoid withdrawal risk, Islamic banks

tend to partly deviate from the PLS principles of Islamic finance. They pay a relatively

competitive rate of return to investment account holders, regardless of their realized

performance. On the asset side, it appears that Islamic banks mainly apply non-PLS

modes of Islamic finance which are in nature closer to conventional finance.

Nevertheless, Islamic banks still may face extra risks because of the complexity of

Islamic modes of finance and limitations in their funding and investment activities.

This paper investigates the credit risk and stability features of Islamic commercial

banks using a sample of 456 conventional and Islamic banks from 22 countries between

2001 and 2008. We follow the two-step GMM technique to estimate a system of

equations that model the risk, capital and efficiency features of Islamic and conventional

banks. We also use a fixed effect vector decomposition approach for our single equation

set-up as a robustness check. By controlling for various factors we find no significant

difference between Islamic banks and their conventional counterparts in terms of

stability. Our findings on the credit risk features of Islamic compared to conventional

banks appear somewhat mixed – depending on the credit risk proxy used. When we use

stock measures of credit risk (loan-loss reserves/gross loans and problem loans/gross

loans) we find that Islamic banks have lower credit risk than their conventional

counterparts but when we use a flow proxy (loan-loss provisions/gross loans) we find the

opposite relationship. It appears that Islamic banks write-off loans more frequently than

conventional banks or/and they have less latitude to recover charged-off loans as

compared to their conventional counterparts. We also observe that Islamic banks benefit

less than conventional banks from the negative impact of asset size on both their credit

and insolvency risks. Overall, in terms of macro-prudential regulation post-crisis our

results suggest no major differences between the stability of conventional and Islamic

banks so this suggests that policies aimed at reducing systemic risk in countries where

both forms of banking take place should be subject to the same macro-prudential rules. In

29

contrast, there may be a case for differential micro-prudential legislation governing credit

risk management.

30

References

Abdul-Majid, M., Saal, D.S. and Battisti, G., 2010, “Efficiency in Islamic and Conventional Banking: An International Comparison”, Journal of Productivity Analysis, Vol. 34, pp. 25-43.

Accounting and Auditing Organization for the Islamic Financial Institutions (AAOIFI), 1999, Accounting, Auditing and Governance Standards for Islamic Financial Institutions, Bahrain.

Aggarwal, R.K. and Yousef, T., 2000, “Islamic banks and investment financing”, Journal of Money, Credit and Banking, Vol. 32, No. 1 (Feb., 2000), pp. 93-120.

Ahmad, A., 1993, “Contemporary practices of Islamic financing techniques” Research Paper No. 20, IRTI, Islamic Development Bank, Jeddah.

Altunbas, Y., Evans, L. and Molyneux, P., 2001, “Ownership and Efficiency in Banking”, Journal of Money, Credit and Banking, Vol. 33 No. 4, pp. 926-954.

Altunbas, Y., Carbo, S., Gardener, E.P.M. and Molyneux, P., 2007, “Examining the relationships between capital, risk and efficiency in European banking”, European Financial Management, Vol. 13, No. 1, pp. 49-70.

Angbazo, L., 1997, “Commercial bank net interest margins, defaults risk, interest-rate risk, and off-balance sheet banking” Journal of Banking & Finance, Vol. 21, pp. 55-87.

Baele, L., Farooq, M. and Ongena, S., 2010, “Of religion and redemption: evidence from default on Islamic loans”, Working Paper.

Bank Indonesia Report, 2009, “The Lack of profit-and-loss sharing financing in Indonesia’s Islamic banks: revisited”, Center of Education and Central Banking Studies, Bank Indonesia, Jakarta, Indonesia.

Barry, T.A, Lepetit, L. and Tarazi, A., 2011, “Ownership structure and risk in publicly held and privately owned banks”, Journal of Banking & Finance, Vol. 35, pp. 1327-1340.

Battese, G.E. and Coelli, T.J., 1992, “Frontier production functions, technical efficiency and panel data: With application to Paddy farmers in India”, Journal of Productivity Analysis, Vol. 3, pp. 153-169.

Beck T., Demirgüç-Kunt, A. and Merrouche, O., 2010, “Islamic vs. Conventional Banking: Business Model, Efficiency and Stability”, World Bank Policy Research Working Paper Series, No. 5446, October.

Berger, A. N., 1995, “The relationship between capital and earnings in banking”, Journal of Money, Credit, and Banking, Vol. 27 (2), pp. 432–456.

Berger, A. N. and De Young, R., 1997, “Problem loans and cost efficiency in commercial banking”, Journal of Banking and Finance, Vol. 21 (6), pp. 849–870.

31

Berger, A. and Udell, G., 2004, “The institutional memory hypothesis and the procyclicality of bank lending behaviour”, Journal of Financial Intermediation, Vol. 13, pp. 458–495.

Berger, A.N., Hasan, I. and Zhou, M., 2009, “Bank ownership and efficiency in China: What will happen in the world’s largest nation?” Journal of Banking & Finance, Vol. 33, pp. 113-130.

Bonin, J.P., Hasan, I., Wachtel, P., 2005, “Bank performance, efficiency and ownership in transition countries”, Journal of Banking and Finance, Vol. 29, 31-53.

Boyd, J. H. and Runkle, D.E., 1993, “Size and performance of banking firms”, Journal of Monetary Economics, Vol. 31, pp. 47–67.

Cebenoyan, S.A. and Strahan, P.E., 2004, “Risk management, capital structure and lending at banks”, Journal of Banking & Finance, Vol. 28, 19-43.

Chong, B.S., and Liu, M.H., 2009, “Islamic banking: Interest-free or interest-based?”, Journal of Pacific-Basin Finance, Vol. 17, pp. 125-144.

Čihák, M. and Hesse, H., 2010, “Islamic banks and financial stability: An empirical analysis”, Journal of Financial Services Research, Vol. 38, pp. 95-113.

Čihák, M., Maechler, A., Schaeck, K. and Stolz, S., 2009, “Who disciplines bank managers?”, IMF Working Paper, No. WP/09/272.

Clair, R., 1992, “Loan growth and loan quality: Some preliminary evidence from Texas banks”, Federal Reserve Bank of Dallas Economic Review, 3rd quarter, PP. 9–22.

Dar, Humayon A., Presley, J. R., 2000, “Lack of profit loss sharing in Islamic banking: management and control imbalances”, International Journal of Islamic Financial Services, Vol. 2 (2).

Dell’Ariccia, G. and Marquez, R., 2006, “Lending booms and lending standards”, Journal of Finance, Vol. 61, pp. 2511–2546.

Demirguc-Kunt, A., Karacaovali, B. and Laeven, L., 2005, “Deposit Insurance around the World: A Comprehensive Database”, World Bank Policy Working Paper, No. 3628, Washington D.C., the World Bank.

DeYoung, R., Hunter, W.C. and Udell, G.F., 2004, “The past, present, and probable future for community banks”, Journal of Financial Services Research, Vol. 25, Nos. 2-3. p. 85-133.

DeYoung R. and Roland, K. P., 2001, “Product mix and earnings volatility at commercial banks: Evidence from a degree of total leverage model”, Journal of Financial Intermediation, Vol. 10 (1), pp. 54-84.

Errico, L. and Farrahbaksh, M., 1998, “Islamic banking: Issues in prudential regulation and supervision”, IMF Working Paper 98/30.

Foos, D., Norden, L. and Weber, M., 2010, “Loan Growth and riskiness of banks”, Journal of Banking and Finance, Vol. 34, pp. 2929-2940.

Fiordelisi, F., Marques-Ibanez, D. and Molyneux, P., 2010, “Efficiency and risk in European banking”, Journal of Banking and Finance, Article in Press.

32

Forssbæck, J., 2011, “Ownership structure, market discipline, and banks’ risk-taking incentives under deposit insurance”, Journal of Banking & Finance, Forthcoming.

Furlong, F. and Keely, M., 1989, “Bank capital regulation and asset risk”, Economic Review, Federal Reserve Bank of San Francisco, Spring, pp. 20–40.

Goddard, J.A, Molyneux, P, and Wilson J.O.S., 2004, “Dynamics of Growth and Profitability in Banking”, Journal of Money, Credit & Banking, Vol. 36, No. 6, December, pp. 1069-1090.

Gonzalez, F., 2005, “Bank regulation and risk taking incentives: An international comparison of bank risk”, Journal of Banking & Finance, Vol. 29, 1153-1184.

Goyeau, D. and Tarazi, A., 1992, “Evaluation du risque de de´faillance bancaire en Europe”, Revue d’Economie Politique, Vol. 102 (2), pp. 249–280.

Harris, M. and Raviv, A., 1990, “Capital Structure and the informational role of debt”, Journal of Finance, Vol. 45, pp. 321-349.

Hansen, L. P., 1982, Large sample properties of generalized method of moments estimators, Econometrica, Vol. 50, pp. 1029–1054.

Hesse, H. and Čihák, M., 2007, “Cooperative banks and financial stability”, IMF Working Paper, No. 07/02.

Hughes, J.P. and Mester, L., 1998, “Bank capitalization and cost: evidence of scale economies in risk management and signalling”, Review of Economics and Statistics, Vol. 80 (2), pp. 314–25.

Hughes, J.P., Mester, L. and Moon, C., 2001, “Are scale economies in banking elusive or illusive: Evidence obtained by incorporating capital structure and risk-taking into models of bank production”, Journal of Banking and Finance, Vol. 25 (12), 2001, pp. 2169–2208.

Hughes, J.P., and Mester, L., 2010, “Efficiency in Banking: Theory, Practice and Evidence”, Chapter 18 in Berger, A.N, Molyneux, P. and Wilson, J.O.S. (eds.) Oxford Handbook of Banking.pp.463-485, Oxford: Oxford University Press.

Hughes, J. P. and Moon, C., 1995, “Measuring bank efficiency when managers trade return for reduced risk”, Department of Economics Rutgers University Working Paper, No. 1995-20.

Iannota, G., Giacomo, N., Sironi, A., 2007, “Ownership structure, risk and performance in the European banking industry”, Journal of Banking and Finance, Vol. 31, pp. 2127-2149.

Imam, P. and Kpodar, K., 2010, “Islamic Banking: How Has It Diffused?” IMF Working Paper, No. 10/195.

Iqbal, M., Ausaf, A. and Khan, T., 1998, “Challenges facing Islamic banking” Occasional Paper, No. 1, IRTI, Islamic Development Bank, Jeddah.

Iqbal, Z. and Mirakhor, A., 2007, “An introduction to Islamic finance: Theory and practice” Wiley, London, ISBN 0470821884.

33

Jacques, K. and Nigro, P., 1997, “Risk-based capital, portfolio risk, and bank capital: A simultaneous equations approach”, Journal of Economics and Business, Vol. 49 (6), pp. 533–47.

Jensen, M. C., 1986, “Agency Costs of free cash flow, corporate finance and takeovers”, American Economic Review, Vol. 76 (2), pp. 323–29.

Kahane, Y., 1977, “Capital adequacy and the regulation of financial intermediaries”, Journal of Banking & Finance, Vol. 1 (2), pp. 207-18.

Kahf, M. and Khan, T., 1992, “Principles of Islamic financing” Research Paper No. 16, IRTI, Islamic Development Bank, Jeddah.

Kane, E., 2010, “Redefining and containing systemic risk”, Atlantic Economic Journal, Vol. 38, pp. 251–264.

Keely, M. and Furlong, F., 1990, “A re-examination of mean–variance analysis of bank capital regulation”, Journal of Banking and Finance, Vol. 14, pp. 69–84.

Khan, M. F., 1991, “Comparative economics of some islamic financing techniques” Research Paper No. 12, IRTI, Islamic Development Bank, Jeddah.

Khan, T. and Ahmed, H., 2001, “Risk management: An analysis of issues in Islamic financial industry”, Occasional Paper No. 5, IRTI, Islamic Development Bank, Jeddah.

Kim, D. and Santomero, A. M., 1988, “Risk in banking and capital regulation”, Journal of Finance, Vol. 43 (5), pp. 1219-33.

Koehn, M. and Santomero, A., 1980, “Regulation of bank capital and portfolio risk”, Journal of Finance, Vol. 35 (5), pp. 1235-45.

Konishi, M. and Yasuda, Y., 2004, “Factors affecting bank risk taking: Evidence from Japan”, Journal of Banking and Finance, Vol. 28, pp. 215-232.

Kwan, S. and Eisenbeis, R., 1997, “Bank risk, capitalization and operating efficiency”, Journal of Financial Services Research, Vol. 12 (2 and 3), pp. 117-31.

Laeven, L. and Levine, R., 2009, “Bank governance, regulation and risk taking”, Journal of Financial Economics, Vol. 93, pp. 259-275.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 2002, “Government Ownership of Banks”, Journal of Finance, Vol. LVII (1), pp. 265-301 .

Lepetit, L., Nys, E., Rous, P. and Tarazi, A., 2008a, “Bank income structure and risk: An empirical analysis of European banks”, Journal of Banking & Finance, Vol. 32, pp. 1452-1467.

Lepetit, L., Nys, E., Rous, P. and Tarazi, A., 2008b, “The expansion of services in European banking: Implications for loan pricing and interest margins”, Journal of Banking & Finance, Vol. 32, pp. 2325-2335.

Mills, P. S. and Presley, J.R., 1999, “Islamic finance: Theory and practice”, Macmillan, London.

34

Mohamad, S., Hassan, T. and Bader, M.K.I., 2008, “Efficiency of Conventional versus Islamic Banks: International Evidence using the Stochastic Frontier Approach (SFA)”, Journal of Islamic Economics, Banking and Finance, Vol. 4, pp. 107-130.

Obaidullah, M. 2005, “Islamic financial services” Islamic Economics Research Center, King Abdulaziz University, Jeddah, Saudi Arabia.

Ogura, Y., 2006, “Learning from a rival bank and lending boom”, Journal of Financial Intermediation, Vol. 15, pp. 535–555.

Ongena, S. and Sendeniz-Yuncu, I., 2010, “Which firms engage small, foreign, or state banks? And who goes Islamic? Evidence from Turkey”, CentER, Tilburg University, Tilburg.

Plumper, T. and Troeger, V.E., 2007, “Efficient estimation of time-invariant and rarely changing variables in finite sample panel analyses with unit fixed effects”, Journal of Political Analysis, Vol. 15, pp. 124-139.

Rime, B., 2001, “Capital requirements and bank behaviour: Empirical evidence for switzerland”, Journal of Banking and Finance, Vol. 25 (4), pp. 789–805.

Rossi, S.P.S., Schwaiger, M.S. and Winkler, G., 2009, “How loan portfolio diversification affects risk, efficiency and capitalization: A managerial behavior model for Austrian banks”, Journal of Banking & Finance, Vol. 33, pp. 2218-2226.

Shiers A.F., 2002, “Branch banking, economic diversity and bank risk”, The Quarterly Review of Economics and Finance, Vol. 42, 587-598.

Shrieves, R.E. and Dahl, D., 1992, “The Relationship between Risk and Capital in Commercial Banks”, Journal of Banking and Finance, Vol. 16 (2), pp. 439–57.

Stiroh, K., 2004, “Diversification in banking: Is non-interest income the answer?”, Journal of Money, Credit and Banking, Vol. 36 (5), pp. 853–882.

Stiroh, K., 2006, “A portfolio view of banking with interest and noninterest activities”, Journal of Money, Credit and Banking, Vol. 38 (5), pp. 1351-1361.

Stiroh, K.J., 2010. Diversification in banking, Chapter 6 in Berger, A., Molyneux, P. and Wilson, J.O.S. (eds.) Oxford Handbook of Banking, pp. 146-170, Oxford: Oxford University Press.

Sundararajan, V. and Errico, L., 2002, “Islamic financial institutions and products in the global financial system: Key issues in risk management and challenges ahead” IMF Working Paper, WP/02/192.

Taboada, A.G., 2011, “The impact of changes in bank ownership structure on the allocation of capital: International evidence,” Journal of Banking & Finance, Forthcoming.

The Banker Report, the global finance magazine of the Financial Times and Group HSBC Amanah, Issue November 2009.

Weill, L., 2010, “Do Islamic banks have greater market power?”, Bank of Finland, BOFIT, Discussion Paper, No. 2.

35

Zellner, A., 1962, “An efficient method of estimating seemingly unrelated regression and tests for aggregation bias”, Journal of the American Statistical Association, Vol. 57 (298), pp. 348–368.

36

Figure 1 Depositors’ Payoff in Islamic and Conventional Banking, in the absence of a deposit insurance scheme

The horizontal axis represents bank’s earning before paying interest expense. The vertical axis shows the interest expense to be paid to depositors (depositors’ payoff). A conventional bank incurs loss for any earnings less than B, where the earnings equal to the interest expense. Depositors of conventional banks receive interest irrespective of the realized earnings, to the extent that the possible loss does not completely deplete the capital. (Hence, the ex-ante relationship between earnings and depositors’ payoff is depicted by the horizontal line. The figure shows that the depletion occurs when earnings are negative; however, in reality depletion can happen when earnings are positive.) In theory, the realized profit or loss should be shared between depositors and equity-holders. The dashed line with a slope less than 45 degrees (α) shows that depositors payoff is proportionate to realized performance; however, in practice there is substantial evidence that Islamic banks pay a competitive rate of return, irrespective of actual performance. Also Islamic banks may adjust profit rates upward but at a slower rate than realized profitability so as to limit the level and volatility of deposit payoffs. At the time of crisis, however, Islamic banks may share the realized loss with investment account holders to avoid insolvency (the bold line illustrates the relationship). This suggests that Islamic banks may have a slightly greater capacity to bear losses compared to conventional banks. Implicitly, investment account holders own a bond, a long position on a call option and a short position on a put option. The strike price of the call is determined arbitrarily by Islamic banks, and the strike price of the put is determined based on the degree of deposit market competition, level of incurred losses and capital strength. Overall, when Islamic banks are profitable investment account holders may get P over the depositor payoffs at conventional banks, at the expense of L in the case of a scenario where losses occur. Hence, in practice the difference between depositors’ payoffs of Islamic versus conventional banks can appear mostly in the tails distribution of bank’s earnings. Displaced commercial risk illustrates the situation where equity-holders have to transfer (or sacrifices) a part of their profit or incur a portion of depositors’ loss to avoid deposits withdrawal. Fiduciary risk is the risk associated with Islamic banks deviating from Sharia principles in sharing returns between investment account holders and equity-holders. It may be that depositors do not have the relevant incentives or/and expertise to observe or take action against such deviations.

Earnings α

Conventional Banking

B

Displaced Commercial Risk

Depositors’ Payoff

Depletion of Islamic Banks’

Capital

Depletion of Conventional Banks’

Capital

Theory of Islamic Banking

Islamic Banking in Practice

Fiduciary Risk

P

L

α

Conventional Banking

B

Displaced Commercial Risk

Depositors’ Payoff

Depletion of Islamic Banks’

Capital

Depletion of Conventional Banks’

Capital

Theory of Islamic Banking

Islamic Banking in Practice

Fiduciary Risk

P

L

37

Table 1 General descriptive statistics and risk measure variables for Islamic, conventional and Islamic window banks over the 2001-2008 period.

General Descriptive Statistics Risk Measure Variables

TA (mil. $)

MS (%)

ETA (%)

ROAA (%)

ROAE (%)

IIER (%)

TAG (%)

GLG (%)

NLTEAR (%)

TNEAR (%)

NII (%) Inefficiency CTIR

(%) LLRGL

(%) PLGL (%)

LLPAGL (%)

Log zscore

Log zscorep1

Log zscorep2

Log zscore3rw

Log zscorep13rw

Log zscorep23rw

ISB

Number 518 518 508 510 513 435 407 403 518 518 473 359 485 350 179 378 54 51 54 303 286 305

Mean 3,584 5.8 14 1.8 15 4.5 36 43 61 20 20 1.6 54 4.0 5.2 1.7 2.7 0.7 2.6 3.2 1.2 3.1

SD 7,414 8.4 12 2.5 16 3.3 34 62 24 16 21 0.4 25 3.9 5.6 2.4 0.9 0.7 0.9 1.1 1.1 1.2

Min 3 0.0 -15 -8.8 -93 0.0 -24 -100 0 1 -21 1.1 10 0.0 0.0 -2.9 0.3 -1.2 0.3 -1.2 -4.1 -0.8

Max 48,400 56.1 68 13.2 92 18.9 195 386 100 86 118 3.4 171 30.9 39.5 18.8 5.5 2.6 5.4 7.6 4.6 7.6

CB

Number 1,689 1,689 1,678 1,669 1,656 1,667 1,389 1,392 1,689 1,689 1,566 1,428 1,609 1,423 1,209 1,601 186 173 187 1,038 994 1,043

Mean 3,919 6.3 12 1.3 12 5.0 20 24 54 15 19 1.7 54 8.3 9.8 1.5 2.9 0.8 2.8 3.5 1.3 3.4

SD 8,786 11.1 9 1.8 18 2.8 29 40 22 12 16 0.3 25 8.4 10.9 2.4 1.0 0.9 1.1 1.2 1.2 1.2

Min 3 0.0 -15 -9.9 -143 0.0 -65 -88 1 0 -16 1.0 2 0.0 0.0 -8.8 -1.0 -2.6 -1.2 -2.2 -4.4 -2.0

Max 76,700 97.5 69 13.7 123 19.9 220 387 100 85 114 3.1 195 47.7 59.6 16.7 5.2 2.6 5.2 9.0 5.1 9.0

T-Stat.† -0.86 -1.09 4.35*** 4.39*** 3.63*** -3.10*** 8.30*** 5.82*** 6.38*** 7.17*** 0.68 -1.09 0.16 -13.89*** -8.83*** 1.42* -1.23 -0.42 -1.33* -4.07*** -1.98** -4.29***

ISWB

Number 467 467 466 464 466 463 394 395 467 467 449 416 461 423 368 439 53 52 53 314 306 314

Mean 8,254 8.0 11 1.8 18 4.3 19 24 59 13 18 1.6 46 6.1 6.8 1.1 2.9 1.0 2.8 3.5 1.6 3.4

SD 12,300 10.1 8 1.5 12 2.5 26 37 18 9 12 0.2 22 6.1 7.9 1.8 0.7 0.9 0.7 1.1 1.3 1.0

Min 30 0.0 2 -4.0 -97 0.2 -50 -93 1 1 -12 1.0 1 0.0 0.0 -7.3 1.4 -1.4 1.3 0.6 -6.4 0.8

Max 88,200 65.1 63 11.7 63 13.2 210 384 99 61 75 2.7 188 41.1 54.6 16.4 4.4 2.3 4.3 7.3 5.7 7.1

ISB= Islamic Banks, CB= Conventional Banks, ISWB: Islamic Window Banks, TA= Total Assets, MS= Market Share, ETA=Equity to Asset Ratio, ROAA=Return on Average Assets, ROAE=Return on Average Equity, IIER= Implicit Interest Expense Rate, TAG=Total Assets Growth, GLG= Gross Loans Growth, NLTEAR=Net Loans to Total Earning Assets Ratio, TNEAR= Total Non-Earning Assets Ratio, NII= The share of Non-Interest Income in Total Operating Income, Inefficiency=Cost Inefficiency, the estimation method is presented in annex 4, CTIR=Cost to Income Ratio, LLRGL=Loan Loss Reserves on Gross Loans Ratio, PLGL=Problem Loans on Gross Loans Ratio, LLPAGL: Loan Loss Provision on Average Gross Loans Ratio, Zscore=(M_ROAA+M_ETA)/SDROAA, M_ROAA = Mean of ROAA over the sample period, M_ETA=Mean of ETA over the sample period, SDROAA= standard deviation of ROAA over the sample period (banks needs to have at least five consecutive observations), Zscorep1=M_ROAA/SDROAA, Zscorep2=M_ETA/SDROAA, Zscore3rw=(ROAA+ETA)/SDROAA3rw, SDROAA3rw= standard deviation of ROAA over 3 years (current year and two previous consecutive years), Zscorep13rw=ROAA/SDROAA3RW, Zscorep23rw=ETA/SDROAA3RW, † T-Stat. of Mean Equality Test between IB & PCB , ***, ** and * indicate significance at 1%, 5% and 10% respectively for the mean equality test between Islamic and conventional banks.

38

Table 2 Simultaneous Estimation of Loan Risk, Equity Capital and Cost Inefficiency

(1) (2) (3) (4) (5) (6) (7) (8)

Variables LLRGL ETA Ineff LLRGL ETA Ineff LLRGL ETA LLRGL ETA PLGL ETA Ineff PLGL ETA Ineff PLGL ETA PLGL ETA

Constant (α0) 0.353*** 0.008 1.285*** 0.359*** -0.019 1.345*** 0.328*** 0.145*** 0.342*** 0.117** 0.493*** -0.027 1.122*** 0.518*** -0.025 1.149*** 0.468*** 0.070* 0.509*** 0.074*

(5.13) (0.12) (7.35) (5.19) (-0.27) (7.31) (3.88) (3.36) (3.96) (2.43) (5.83) (-0.54) (7.56) (5.74) (-0.49) (7.52) (5.01) (1.89) (4.88) (1.92)

ISBD (α1) -0.015*** 0.027*** 0.024 -0.121** 0.213*** -0.328* -0.017*** 0.027*** -0.124*** 0.178** -0.002 0.008 0.011 -0.293*** -0.034 -0.467** -0.005 0.011 -0.332*** -0.050

(-2.85) (2.96) (1.09) (-2.52) (2.63) (-1.74) (-3.07) (3.10) (-2.59) (2.30) (-0.15) (1.09) (0.49) (-2.71) (-0.35) (-2.19) (-0.47) (1.36) (-2.72) (-0.51)

ISWD (α2) -0.012*** 0.005 -0.053*** -0.010** 0.004 -0.050*** -0.009* -0.001 -0.008 -0.002 -0.030*** 0.002 -0.045*** -0.028*** 0.003 -0.041*** -0.028*** -0.002 -0.027*** -0.002

(-2.70) (0.98) (-4.04) (-2.34) (0.69) (-3.85) (-1.84) (-0.25) (-1.56) (-0.49) (-3.91) (0.37) (-3.14) (-3.70) (0.46) (-2.93) (-3.38) (-0.47) (-3.20) (-0.38)

Size (α3) -0.016*** 0.002 0.011 -0.017*** 0.004 0.007 -0.016*** 0.001 -0.017*** 0.003 -0.018*** 0.004 0.018** -0.020*** 0.004 0.016** -0.017*** 0.004* -0.019*** 0.004

(-5.73) (0.67) (1.23) (-5.75) (1.25) (0.73) (-5.07) (0.43) (-5.08) (1.00) (-5.86) (1.60) (2.51) (-5.94) (1.47) (2.15) (-5.65) (1.80) (-5.67) (1.64)

Size X ISBD (α4) 0.007** -0.013** 0.024* 0.008** -0.010** 0.020*** 0.003 0.033** 0.023*** 0.004

(2.28) (-2.48) (1.92) (2.31) (-2.11) (2.72) (0.46) (2.22) (2.69) (0.64)

MS (α5) -0.008 -0.029*** 0.018 -0.007 -0.031*** 0.020 -0.012* -0.024*** -0.010 -0.025*** -0.024** -0.041*** 0.013 -0.024** -0.041*** 0.015 -0.027** -0.038*** -0.027** -0.038***

(-1.32) (-8.53) (1.47) (-1.15) (-8.51) (1.64) (-1.75) (-8.78) (-1.58) (-9.11) (-2.48) (-10.26) (0.86) (-2.46) (-10.16) (0.97) (-2.44) (-11.01) (-2.43) (-10.93)

Loan Risk (α6) 0.917*** 1.552*** 0.967*** 1.509*** 0.896*** 0.935*** 0.535*** 1.207*** 0.539*** 1.238*** 0.604*** 0.610***

(3.61) (2.92) (3.72) (2.81) (4.04) (4.13) (3.61) (3.47) (3.63) (3.56) (4.23) (4.29)

ETA (α7) -0.532*** -1.172*** -0.521*** -1.198*** -0.528** -0.527** -1.170*** -0.655 -1.212*** -0.659 -1.173*** -1.241***

(-2.84) (-3.58) (-2.82) (-3.66) (-2.37) (-2.37) (-4.55) (-1.62) (-4.54) (-1.62) (-3.91) (-3.87)

Ineff (α8) -0.002 -0.011 0.002 -0.017 0.012 -0.091*** 0.012 -0.092*** -0.021 -0.001 -0.019 -0.000 -0.012 -0.072*** -0.018 -0.072***

(-0.11) (-0.74) (0.11) (-1.13) (0.77) (-6.96) (0.76) (-6.99) (-0.77) (-0.06) (-0.72) (-0.01) (-0.46) (-6.12) (-0.63) (-6.20)

L.NII (α9) 0.042*** 0.042*** 0.048*** 0.050*** 0.025 0.032 0.040* 0.047**

(2.61) (2.66) (2.89) (2.98) (1.00) (1.26) (1.73) (1.99)

L.GLG (α10) -0.039*** -0.038*** -0.040*** -0.038*** -0.072*** -0.073*** -0.071*** -0.073***

(-4.58) (-4.47) (-4.46) (-4.34) (-4.04) (-4.03) (-3.87) (-3.85)

L.ROAA (α11) 1.862*** 1.893*** 1.494*** 1.533*** 2.038*** 2.034*** 1.864*** 1.858***

(7.51) (7.62) (7.33) (7.44) (5.83) (5.71) (5.52) (5.41)

L.TNEAR (α12) 1.528*** 1.543*** 1.599*** 1.605***

(18.07) (18.31) (19.11) (19.14)

StateD (α13) 0.011* -0.001 0.025 0.010* -0.001 0.023 0.007 0.005 0.007 0.005 0.033*** -0.013 0.005 0.033** -0.013 0.004 0.033** -0.011 0.033** -0.011

39

(1.79) (-0.18) (1.51) (1.65) (-0.09) (1.40) (1.20) (0.86) (1.10) (0.89) (2.62) (-1.61) (0.26) (2.55) (-1.64) (0.20) (2.50) (-1.46) (2.45) (-1.49)

ForeignD (α14) 0.033*** -0.031** -0.025 0.034*** -0.035*** -0.018 0.032*** -0.025** 0.034*** -0.028** 0.027* -0.013 0.006 0.027* -0.013 0.004 0.029* -0.015 0.029* -0.016

(3.76) (-2.47) (-0.64) (3.99) (-2.67) (-0.47) (3.73) (-2.20) (3.95) (-2.39) (1.82) (-1.19) (0.17) (1.78) (-1.24) (0.10) (1.89) (-1.32) (1.85) (-1.35)

SubsidiaryD (α15) 0.001 0.027*** -0.062*** 0.002 0.026*** -0.060*** -0.002 0.025*** -0.001 0.024*** 0.008 0.023*** -0.110*** 0.010 0.023*** -0.108*** 0.008 0.019*** 0.010 0.020***

(0.19) (4.02) (-2.91) (0.29) (3.86) (-2.82) (-0.21) (3.89) (-0.10) (3.74) (0.78) (3.73) (-5.15) (0.95) (3.78) (-5.03) (0.75) (3.15) (0.92) (3.20)

YoungD (α16) 0.039*** -0.038* 0.065 0.040*** -0.041* 0.070 0.036*** -0.027 0.036*** -0.029* 0.025 0.002 0.193*** 0.023 0.002 0.193*** 0.021 0.013 0.021 0.013

(2.68) (-1.82) (1.38) (2.68) (-1.90) (1.49) (2.68) (-1.58) (2.65) (-1.68) (1.07) (0.12) (2.77) (0.99) (0.10) (2.77) (1.00) (0.69) (0.92) (0.68)

MiddleD (α17) -0.013* 0.001 -0.069*** -0.013* 0.002 -0.070*** -0.012 -0.006 -0.012* -0.005 -0.009 0.021** -0.071*** -0.007 0.021** -0.068*** -0.009 0.016* -0.007 0.017*

(-1.77) (0.10) (-2.83) (-1.83) (0.23) (-2.90) (-1.59) (-0.78) (-1.65) (-0.67) (-0.74) (2.33) (-2.86) (-0.54) (2.39) (-2.74) (-0.72) (1.77) (-0.51) (1.84)

HHI (α18) 0.080** 0.080*** -0.418*** 0.081** 0.075*** -0.413*** 0.092*** 0.041* 0.091*** 0.038 0.208*** 0.154*** -0.403*** 0.211*** 0.153*** -0.410*** 0.207*** 0.113*** 0.210*** 0.112***

(2.24) (3.20) (-4.70) (2.28) (2.91) (-4.67) (2.61) (1.67) (2.60) (1.51) (3.66) (5.68) (-3.64) (3.67) (5.64) (-3.70) (3.55) (4.22) (3.51) (4.17)

EDInsurance (α19) -0.046*** -0.007 0.074*** -0.045*** -0.007 0.075*** -0.046*** -0.001 -0.044*** -0.001 -0.063*** -0.031*** 0.094*** -0.063*** -0.031*** 0.097*** -0.061*** -0.023*** -0.061*** -0.023***

(-4.39) (-0.75) (2.65) (-4.30) (-0.77) (2.69) (-4.32) (-0.08) (-4.21) (-0.11) (-4.24) (-4.99) (3.51) (-4.21) (-4.95) (3.66) (-3.95) (-3.50) (-3.88) (-3.47)

DomCredit (α20) 0.036*** -0.034*** 0.014 0.035*** -0.034*** 0.014 0.031*** -0.016* 0.031*** -0.017* 0.062*** -0.019* -0.011 0.061*** -0.020* -0.013 0.058*** -0.008 0.059*** -0.009

(6.20) (-2.73) (0.51) (6.11) (-2.76) (0.50) (5.85) (-1.87) (5.83) (-1.95) (5.98) (-1.87) (-0.38) (5.85) (-1.92) (-0.47) (6.06) (-0.95) (5.94) (-0.98)

ExRateDep (α21) -0.152** -0.047 -0.012 -0.151** -0.042 -0.020 -0.142* -0.083** -0.138* -0.087** 0.164* -0.077 -0.052 0.179* -0.078 -0.036 0.183** -0.115* 0.198** -0.115*

(-2.05) (-1.26) (-0.09) (-2.11) (-1.12) (-0.15) (-1.89) (-2.48) (-1.86) (-2.44) (1.77) (-1.24) (-0.28) (1.89) (-1.24) (-0.20) (2.03) (-1.73) (2.12) (-1.72)

GDPPCGrowth (α22) -0.066 0.271** 0.251 -0.065 0.277** 0.252 -0.089 0.207** -0.087 0.212** -0.155 0.243** 0.034 -0.142 0.247** 0.051 -0.168 0.195** -0.154 0.198**

(-0.73) (2.47) (0.92) (-0.73) (2.51) (0.93) (-1.02) (2.13) (-1.00) (2.16) (-1.22) (2.41) (0.13) (-1.12) (2.44) (0.19) (-1.36) (2.00) (-1.22) (2.02)

H0: α3 + α4 = 0 (F-stat.) 8.98*** 7.62*** 0.00 0.22

H0: α3 = α4 = 0 (F-stat.) 33.2*** 25.79*** 35.39*** 32.25***

Number of obs 1252 1252 1246 1246 1138 1138

1133 1133

Number of Parameters 70 73 47 49 70 73 47 49

Number of Moments 72 75 48 50 72 75 48 50

Hansen’s J chi 2 (1) 0.50 0.20 0.68 0.18 3.31 4.04 0.80 0.92

P-Value 0.779 0.905 0.411 0.668 0.191 0.133 0.371 0.337

Variable definitions: LLRGL = Loan loss reserves on gross loans, ETA = Equity to asset ratio, Ineff = Cost inefficiency, PLGL = Problem loans on gross loans, ISBD=Islamic bank dummy, ISWD= The dummy variable representing Islamic window banks, Size = Logarithm of total assets, SizeXISBD = Interaction of Size and ISBD, MS = Logarithm of market share, orthogonalized on size, Loan Risk = Either LLRG or PLGL, L.NII = Lagged value of share of noninterest income in total operating income, L.GLG = Lagged value of gross loan growth, L.ROAA = Lagged value of returns on average assets, L.TNEAR = Lagged value of total non-earning assets ratio, StateD = State-owned bank dummy, ForeignD = Foreign-owned bank dummy, SubsidiaryD = Subsidiary dummy, YoungD = Young bank dummy, MiddleD = Middle-aged bank dummy, HHI = Hirschman-Herfindahl index, EDInsurance = Explicit deposit insurance scheme, DomCredit = Domestic credit provided by banking system (% of GDP), ExRateDep = Official exchange rate depreciation, GDPPCGrowth = Annual growth of GDP per capita. We use the two-step GMM estimator with robust GMM weight matrix and standard errors. L.NII and L.GLG are only used in loan risk equation, L.ROAA is only used in equity capital (ETA) equation and L.TNEAR is only used in cost inefficiency (Ineff) equation. Loan risk equation is just-identified, but capital equity and inefficiency equations are over-identified. All right-hand side variables, except risk, equity capital and inefficiency are used as the instruments in all equations; however, in system estimations numbers (3), (4), (7) and (8), where the inefficiency

40

equation is dropped out, the lagged value of cost inefficiency is included in and used as an instrument for both loan risk and equity capital equations. LLRGL and PLGL are used the loan risk proxies in system estimation numbers (1) to (4) and (5) to (8) respectively. System numbers (1) and (5) are the baseline systems. In system numbers (2), (4), (6) and (8) the interaction term of size and ISBD are included. System numbers (3)-(4) and (7)-(8) consist of two equations of loan risk and equity capital where the lagged value of inefficiency is controlled for. Year dummies are included in the model, but not reported in the table. The chi square of Hansen’s J tests of over-identifying restrictions do not suggest that our models are mis-specified. The null hypothesis of the test is that the model is correctly specified. Robust z-statistics are reported in parentheses. ***, ** and * indicate significance at 1%, 5% and 10% respectively.

41

Table 3 Simultaneous Estimation of Insolvency Risk and Inefficiency

(1) (2) (3) (4) (5) (6)

Risk Ineff ZP1 Ineff ZP2 Ineff Risk Ineff ZP1 Ineff ZP2 Ineff

Constant (α0) 4.302*** 0.450 1.274*** 1.209*** 4.256*** 0.414 4.070*** 0.686** 1.041** 1.271*** 4.059*** 0.599**

(10.05) (1.39) (2.76) (15.34) (9.70) (1.29) (9.56) (2.55) (2.23) (18.22) (9.25) (2.12)

ISBD (α1) 0.025 -0.039 0.074 -0.020 0.020 -0.035 2.381*** -0.923*** 2.155** -0.445* 2.011** -0.877***

(0.30) (-1.45) (0.78) (-0.84) (0.23) (-1.22) (3.23) (-3.29) (2.44) (-1.78) (2.53) (-2.99)

ISWD (α2) -0.058 -0.061*** 0.097 -0.064*** -0.045 -0.064*** -0.088 -0.052*** 0.071 -0.057*** -0.071 -0.055***

(-0.80) (-2.90) (1.14) (-3.39) (-0.61) (-2.91) (-1.21) (-2.65) (0.83) (-3.36) (-0.95) (-2.62)

Size (α3) -0.030 0.020*** 0.072*** 0.012 -0.041* 0.022*** -0.006 0.010* 0.093*** 0.009 -0.021 0.013*

(-1.42) (2.88) (3.08) (1.63) (-1.90) (2.89) (-0.28) (1.73) (3.87) (1.02) (-0.94) (1.95)

SizeXISBD (α4) -0.166*** 0.062*** -0.147** 0.030* -0.141*** 0.059***

(-3.31) (3.22) (-2.42) (1.77) (-2.61) (2.95)

MS (α5) 0.026 0.038*** 0.127*** 0.045*** 0.002 0.044*** 0.009 0.046*** 0.113*** 0.049*** -0.012 0.051***

(0.70) (3.42) (3.34) (3.48) (0.06) (3.78) (0.25) (4.61) (3.00) (4.33) (-0.33) (4.69)

Risk / ZP1 / ZP2 (α6) 0.198** 0.009 0.213*** 0.167** -0.004 0.193**

(2.44) (0.09) (2.62) (2.32) (-0.04) (2.56)

Ineff (α7) -0.266 -0.633*** -0.262 -0.314* -0.664*** -0.308*

(-1.56) (-3.35) (-1.48) (-1.83) (-3.51) (-1.74)

L_NII (α8) -0.360 -0.645* -0.312 -0.430 -0.704** -0.339

(-1.36) (-1.92) (-1.25) (-1.52) (-2.10) (-1.27)

L_TAG (α9) -0.443*** 0.007 -0.478*** -0.456*** 0.036 -0.493***

(-4.35) (0.05) (-4.56) (-4.46) (0.26) (-4.72)

L_TNEAR (α10) 1.756*** 1.695*** 1.763*** 1.774*** 1.691*** 1.788***

(14.74) (12.88) (14.12) (15.35) (13.10) (14.49)

StateD (α11) 0.026 -0.003 -0.058 0.005 0.051 -0.009 0.041 -0.009 -0.045 0.002 0.063 -0.014

(0.25) (-0.13) (-0.54) (0.26) (0.49) (-0.34) (0.42) (-0.38) (-0.42) (0.09) (0.61) (-0.54)

ForeignD (α12) -0.448*** 0.130** -0.595*** 0.044 -0.429*** 0.131** -0.477*** 0.127** -0.623*** 0.041 -0.453*** 0.133**

(-3.27) (2.30) (-3.31) (0.65) (-3.02) (2.30) (-3.42) (2.39) (-3.43) (0.61) (-3.13) (2.41)

SubsidiaryD (α13) 0.088 -0.137*** 0.145* -0.135*** 0.077 -0.133*** 0.061 -0.125*** 0.125 -0.129*** 0.054 -0.124***

(1.12) (-5.15) (1.71) (-4.92) (0.95) (-4.75) (0.79) (-5.18) (1.48) (-5.18) (0.68) (-4.70)

YoungD (α14) -0.379* 0.188** -0.207 0.067 -0.366* 0.195** -0.399** 0.181** -0.232 0.070 -0.380* 0.195**

(-1.91) (2.39) (-1.10) (1.10) (-1.83) (2.43) (-2.00) (2.42) (-1.23) (1.14) (-1.89) (2.50)

MiddleageD (α15) -0.157 -0.019 -0.110 -0.076*** -0.165 -0.017 -0.154 -0.027 -0.111 -0.078*** -0.162 -0.022

(-1.52) (-0.56) (-0.91) (-3.29) (-1.57) (-0.47) (-1.49) (-0.85) (-0.89) (-3.47) (-1.54) (-0.66)

HHI (α16) -0.455 -0.425*** -0.620* -0.530*** -0.451 -0.431*** -0.435 -0.446*** -0.600* -0.539*** -0.446 -0.446***

(-1.37) (-3.98) (-1.74) (-6.27) (-1.32) (-3.90) (-1.31) (-4.57) (-1.68) (-6.68) (-1.31) (-4.30)

EDInsurance (α17) -0.166** 0.116*** -0.090 0.113*** -0.147* 0.112*** -0.193** 0.120*** -0.113 0.116*** -0.166* 0.118***

(-1.99) (4.82) (-0.94) (6.63) (-1.72) (4.50) (-2.28) (5.30) (-1.16) (6.73) (-1.92) (4.88)

DomCredit (α18) 0.586*** -0.052 0.395*** 0.050 0.633*** -0.071 0.600*** -0.037 0.406*** 0.053 0.644*** -0.062

(7.18) (-1.00) (4.21) (1.44) (7.49) (-1.26) (7.33) (-0.78) (4.32) (1.61) (7.62) (-1.16)

ExRateDep (α19) -0.822** 0.278* -0.090 0.099 -0.943** 0.323** -0.817** 0.253** -0.067 0.092 -0.934*** 0.304**

(-2.26) (1.95) (-0.17) (1.17) (-2.53) (2.10) (-2.52) (2.14) (-0.13) (1.19) (-2.76) (2.28)

GDPPCGrowth (α20) 1.545 -0.593 2.136 -0.318 1.480 -0.560 1.414 -0.514 2.037 -0.276 1.353 -0.493

(1.20) (-1.36) (1.50) (-0.81) (1.09) (-1.21) (1.12) (-1.30) (1.44) (-0.74) (1.02) (-1.15)

H0: α3 + α4 = 0 (F-stat.) 12.35*** 0.84 9.32*** H0: α3 = α4 = 0 (F-stat.) 12.42*** 16.97*** 9.53*** Number of obs 1345 1311 1350 1345 1311 1350

42

Number of Parameters 45 45 45 47 47 47 Number of Moments 46 46 46 48 48 48 Hansen’s J chi 2 (1) 1.99 11.64 0.86 3.33 11.37 1.64 P-Value 0.159 0.001 0.353 0.068 0.001 0.201

Variable definitions: Risk = Logarithm of Zscore is used as the insolvency risk proxy. Zscore is defined as (ROAA+ETA)/SD(ROAA). ROAA = Returns on average assets at time t. ETA = Equity to asset ratio at time t. SD(ROAA) = standard deviation of ROAA over 3 years (current year and two previous consecutive years). Banks need to have three consecutive observations. Acquiring banks are excluded from the sample, since the volatility on their assets returns can be due to the acquisition. Ineff = Cost inefficiency, ZP1 = logarithm of the first component of Zscore, i.e. ROAA/SD(ROAA), ZP2 = logarithm of the second component of Zscore, i.e. ETA/SD(ROAA), ISBD=Islamic bank dummy, ISWD= The dummy variable representing Islamic window banks, Size = Logarithm of total assets, SizeXISBD = Interaction of Size and ISBD, MS = Logarithm of market share, orthogonalized on size, L.NII = Lagged value of share of noninterest income in total operating income, L.TAG = Lagged value of total assets growth, L.TNEAR = Lagged value of total non-earning assets ratio, StateD = State-owned bank dummy, ForeignD = Foreign-owned bank dummy, SubsidiaryD = Subsidiary dummy, YoungD = Young bank dummy, MiddleD = Middle-aged bank dummy, HHI = Hirschman-Herfindahl index, EDInsurance = Explicit deposit insurance scheme, DomCredit = Domestic credit provided by banking system (% of GDP), ExRateDep = Official exchange rate depreciation, GDPPCGrowth = Annual growth of GDP per capita. We use the two-step GMM estimator with robust GMM weight matrix and standard errors. L.NII and L.TAG are only used in risk / ZP1 / ZP2 equations and L.TNEAR is only used in cost inefficiency (Ineff) equation. Risk / ZP1 / ZP2 equations are just-identified, but inefficiency equation is over-identified. All right-hand side variables, except insolvency risk / ZP1 / ZP2 and inefficiency are used as the instruments in all equations. In system estimation number (1) the logarithm of Zscore is used as the insolvency risk proxy. The logarithms of the first and second components of Zscore are used in system estimation numbers (2) and (3) respectively. In system estimation numbers (4) to (6), the interaction term of size and ISBD is included in the model. Year dummies are included in the model, but not reported in the table. The chi square of Hansen’s J tests of over-identifying restrictions do not suggest that our insolvency risk models are mis-specified. The null hypothesis of the test is that the model is correctly specified. Robust z-statistics are reported in parentheses. ***, ** and * indicate significance at 1%, 5% and 10% respectively.

43

Table 4 Single Equation Estimation of Loan Risk, Using FEVD Approach

(1) (2) (3) (4) (5) (6)' (7) (8) (9) (10) (11) (12) (13) (14) (15)

Variables LLRGL LLRGL LLRGL LLRGL LLRGL PLGL PLGL PLGL PLGL PLGL LLPAGL LLPAGL LLPAGL LLPAGL LLPAGL

Constant (α0) 0.221*** 0.259*** 0.224*** 0.225*** 0.235*** -0.130*** -0.114*** -0.151** -0.073** -0.122*** -0.043*** -0.043*** -0.068*** -0.034*** -0.044*** (14.33) (15.78) (4.63) (11.19) (14.76) (-6.19) (-5.44) (-2.25) (-2.23) (-5.84) (-5.23) (-5.21) (-4.07) (-2.77) (-5.31)

ISBD (α1) -0.032*** -0.165*** -0.030*** -0.028*** -0.043*** -0.032*** -0.188*** -0.033*** -0.026*** -0.041*** 0.003*** 0.002 0.002 0.004** -0.000 (-11.66) (-13.20) (-5.73) (-8.19) (-7.02) (-7.81) (-8.96) (-4.14) (-4.82) (-4.21) (2.66) (0.27) (1.17) (2.12) (-0.04)

ISWD (α2) -0.021*** -0.018*** -0.025*** -0.019*** -0.020*** -0.044*** -0.043*** -0.050*** -0.041*** -0.043*** -0.013*** -0.013*** -0.013*** -0.012*** -0.013*** (-8.82) (-7.65) (-3.51) (-9.23) (-8.54) (-12.75) (-12.43) (-5.15) (-11.38) (-12.52) (-9.67) (-9.65) (-4.95) (-6.52) (-9.67)

Size (α3) -0.009*** -0.011*** -0.009** -0.009*** -0.010*** 0.013*** 0.012*** 0.014*** 0.009*** 0.012*** 0.006*** 0.006*** 0.008*** 0.005*** 0.006*** (-12.44) (-15.40) (-2.34) (-10.23) (-13.71) (9.29) (8.54) (2.84) (4.58) (8.86) (8.57) (8.54) (5.95) (5.03) (8.61)

Size X ISBD (α4) 0.010*** 0.011*** 0.000 (11.97) (7.82) (0.28)

MS (α5) -0.010*** -0.010*** -0.013*** -0.004* -0.010*** -0.003 -0.003 -0.001 0.002 -0.003 0.001* 0.001* 0.001 0.002*** 0.001 (-5.70) (-5.53) (-4.35) (-1.90) (-5.24) (-1.25) (-1.28) (-0.35) (0.70) (-1.27) (1.76) (1.78) (0.67) (2.60) (1.56)

TA (α6) -0.139*** -0.145*** -0.164*** -0.147*** -0.158*** -0.085** -0.091** -0.055 -0.141*** -0.093** -0.012 -0.012 -0.012 -0.007 -0.016* (-3.49) (-3.62) (-2.59) (-6.37) (-3.39) (-2.10) (-2.26) (-0.84) (-4.17) (-2.19) (-1.33) (-1.32) (-1.04) (-0.44) (-1.76)

SBDXETA (α7) 0.081* 0.078 0.029 (1.87) (1.09) (1.58)

Ineff (α8) 0.019*** 0.019*** 0.022*** 0.017*** 0.020*** 0.051*** 0.052*** 0.041*** 0.051*** 0.053*** -0.010*** -0.010*** -0.011*** -0.008*** -0.010*** (3.66) (3.65) (2.94) (3.17) (3.89) (6.83) (6.96) (2.89) (7.28) (7.11) (-4.31) (-4.31) (-3.07) (-2.77) (-4.37)

II (α9) 0.003 0.002 0.003 0.007 0.002 0.017 0.016 0.037 0.008 0.014 -0.008* -0.008* 0.005 -0.016** -0.008* (0.46) (0.30) (0.22) (0.98) (0.21) (1.46) (1.36) (1.39) (0.64) (1.20) (-1.72) (-1.69) (0.96) (-2.37) (-1.65)

GLG (α10) -0.012*** -0.011*** -0.016*** -0.006 -0.012*** -0.026*** -0.026*** -0.023** -0.026*** -0.026*** -0.001 -0.001 -0.001 -0.000 -0.001 (-3.33) (-3.15) (-2.92) (-1.14) (-3.29) (-4.39) (-4.36) (-2.49) (-3.44) (-4.37) (-0.74) (-0.73) (-0.59) (-0.13) (-0.74)

StateD (α11) 0.006** 0.004 0.010*** 0.006* 0.006** 0.044*** 0.043*** 0.054*** 0.038*** 0.044*** 0.002 0.002 0.002 0.002 0.002 (2.04) (1.34) (2.62) (1.74) (2.12) (9.28) (9.16) (5.22) (7.99) (9.31) (1.21) (1.20) (1.11) (0.99) (1.25)

ForeignD (α12) 0.044*** 0.045*** 0.036*** 0.041*** 0.043*** 0.058*** 0.058*** 0.052*** 0.045*** 0.058*** 0.007*** 0.007*** 0.008** 0.003 0.007*** (11.53) (11.93) (5.21) (9.95) (11.34) (7.90) (7.91) (3.31) (4.58) (7.88) (3.45) (3.46) (2.39) (1.44) (3.52)

SubsidiaryD (α13) -0.002 -0.002 -0.004 0.001 -0.002 -0.003 -0.003 -0.016 0.006 -0.003 0.005*** 0.005*** 0.004* 0.006*** 0.005*** (-0.66) (-0.81) (-0.53) (0.33) (-0.66) (-0.81) (-0.68) (-1.62) (1.45) (-0.80) (3.77) (3.78) (1.84) (2.88) (3.79)

YoungD (α14) -0.015* -0.015* -0.013 -0.010 -0.016* -0.021** -0.020** -0.019 -0.017*** -0.022** -0.004* -0.004 -0.006 -0.001 -0.004* (-1.84) (-1.76) (-0.83) (-1.54) (-1.95) (-2.10) (-2.04) (-0.87) (-2.74) (-2.17) (-1.65) (-1.64) (-1.16) (-0.31) (-1.69)

MiddleageD (α15) -0.015*** -0.015*** -0.015*** -0.014*** -0.016*** 0.002 0.003 0.004 -0.004 0.001 0.004** 0.004** 0.005** 0.005 0.005** (-4.79) (-4.62) (-3.23) (-2.88) (-4.99) (0.32) (0.54) (0.48) (-0.41) (0.16) (2.46) (2.46) (2.37) (1.45) (2.51)

HHI (α16) 0.007 0.007 0.016 -0.004 0.003 -0.017 -0.016 -0.033 -0.022 -0.014 -0.021*** -0.021*** -0.022*** -0.021*** -0.021*** (0.54) (0.58) (0.51) (-0.30) (0.22) (-1.03) (-0.93) (-1.07) (-0.95) (-0.82) (-6.02) (-6.04) (-3.14) (-3.51) (-5.91)

EDInsurance (α17) -0.039*** -0.039*** -0.049*** -0.031*** -0.040*** -0.008* -0.009* -0.001 -0.008* -0.009** 0.001 0.001 0.003 -0.000 0.001 (-14.35) (-14.11) (-8.67) (-9.73) (-14.64) (-1.81) (-1.94) (-0.04) (-1.87) (-2.03) (0.76) (0.77) (1.38) (-0.24) (0.82)

DomCredit (α18) 0.002 0.001 0.003 0.002 0.003 0.015*** 0.015*** 0.024*** 0.010** 0.016*** -0.008*** -0.008*** -0.009*** -0.006*** -0.008***

44

(0.81) (0.44) (0.60) (0.95) (1.19) (3.91) (3.83) (3.30) (2.30) (4.19) (-6.83) (-6.83) (-5.93) (-4.47) (-6.82)

ExRateDep (α19) -0.019*** -0.016*** -0.027*** 0.004 -0.021*** -0.001 -0.000 -0.068 0.005 -0.001 0.007*** 0.007*** 0.006*** 0.015** 0.007*** (-3.11) (-2.70) (-3.33) (0.38) (-3.07) (-0.07) (-0.01) (-0.97) (0.24) (-0.08) (3.76) (3.77) (2.94) (1.99) (3.65)

GDPPCGrowth (α20) 0.008 0.003 0.266** -0.058 0.005 -0.071 -0.076 -0.109 -0.045 -0.073 0.007 0.007 0.021 0.001 0.007 (0.16) (0.07) (2.12) (-1.13) (0.10) (-1.00) (-1.08) (-0.45) (-0.68) (-1.04) (0.46) (0.45) (0.50) (0.09) (0.44)

Number of obs 1,251 1,251 485 766 1,251 1,137 1,137 442 695 1,137 1,336 1,336 541 795 1,336

R-squared 0.756 0.757 0.783 0.722 0.758 0.747 0.747 0.760 0.737 0.747 0.362 0.362 0.465 0.294 0.364

H0: α3 + α4 = 0 (F-stat.) 3.95** 121.86*** 59.77

H0: α3 = α4 = 0 (F-stat.) 141.01*** 61.18*** 36.88***

H0: α6 + α7 = 0 (F-stat.) 12.09*** 0.05 0.51

H0: α6 = α7 = 0 (F-stat.) 8.51*** 2.42* 2.10

Variable definitions: LLRGL = Loan loss reserves on gross loans, PLGL = Problem loans on gross loans, LLPAGL = Loan loss provisions on average gross loans, ISBD = Islamic bank dummy, ISWD = The dummy variable representing Islamic window banks, Size = Logarithm of total assets, Size X ISBD = Interaction of Size and ISBD, MS = Logarithm of market share, orthogonalized on size, ETA = Equity to asset ratio, ISBDXETA = Interaction of ETA and ISBD, Ineff = Cost inefficiency, NII = Share of noninterest income in total operating income, GLG = Gross loan growth, StateD = State-owned bank dummy, ForeignD = Foreign-owned bank dummy, SubsidiaryD = Subsidiary dummy, YoungD = Young bank dummy, MiddleD = Middle-aged bank dummy, HHI = Hirschman-Herfindahl index, EDInsurance = Explicit deposit insurance scheme, DomCredit = Domestic credit provided by banking system (% of GDP), ExRateDep = Official exchange rate depreciation, GDPPCGrowth = Annual growth of GDP per capita. All the accounting and macro level variables are lagged for one period. Hausman test rejects random effect estimation; however, we have several dummy variables that are time invariant, SubsidiaryD and EDInsurance or rarely changing over time, i.e. ISBD, ISWD, StateD and ForeignD. Fixed effect technique can not estimate the time invariant variables and it is inefficient in estimating variables with very limited within variance. Hence, we follow the fixed effects vector decomposition (FEVD) approach proposed by Plumper and Troeger (2007) to capture the unobservable individual-specific effects. LLRGL, PLGL and LLPAGL are used as the loan risk proxies in columns (1) to (5), (6) to (10) and (11) to (15) respectively. Columns (1), (6) and (11) are baseline equations. In columns (2), (7) and (12) we add the lagged value of interaction term of size and ISBD. In columns (3), (8) and (13) small banks sub-sample and in columns (4), (9) and (14) large banks sub-sample are used. Finally, in columns (5), (10) and (15) the lagged value of interaction of ETA and ISBD is included in the model. Year Dummies and the unit effects captured by FEVD approach are included in the model, but not reported in the table. Robust z-statistics are reported in parentheses. ***, ** and * indicate significance at 1%, 5% and 10% respectively.

45

Table 5 Single Equation Estimation of Insolvency Risk, Using FEVD Approach

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)

ISBD & Size Interaction Small Banks Sub-Sample Large Banks Sub-Sample Share of Net Loans in Total Earning Assets Square form of Net Loans Share

Variables Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2

Constant (α0) -1.755*** -4.477*** -0.727*** -2.547*** -5.044*** -1.441*** -2.521*** -6.288*** -2.745*** -1.273* -3.518*** -1.281* -1.714*** -4.478*** -0.701** -1.841*** -4.540*** -0.859*** (-5.92) (-12.73) (-2.63) (-8.02) (-13.55) (-4.89) (-4.39) (-6.75) (-4.76) (-1.91) (-3.97) (-1.78) (-5.66) (-12.60) (-2.45) (-5.45) (-12.37) (-2.65)

ISBD (α1) -0.118** 0.067 -0.148** 3.007*** 2.296*** 2.536*** 0.025 -0.082 0.007 -0.165** -0.024 -0.134 -0.142** 0.047 -0.154** -0.122** 0.062 -0.135** (-1.99) (1.06) (-2.41) (6.08) (4.11) (4.94) (0.26) (-0.64) (0.07) (-2.05) (-0.24) (-1.58) (-2.40) (0.74) (-2.51) (-2.00) (0.94) (-2.13)

ISWD (α2) 0.068 0.804*** 0.067 -0.002 0.755*** 0.003 0.117 0.015 0.129 0.046 0.273*** 0.061 0.067 0.803*** 0.067 0.069 0.804*** 0.067 (1.27) (12.37) (1.20) (-0.04) (11.55) (0.05) (1.11) (0.09) (1.21) (0.72) (3.36) (0.91) (1.24) (12.40) (1.20) (1.28) (12.40) (1.20)

Size (α3) 0.061*** -0.179*** 0.013 0.120*** -0.137*** 0.067*** 0.117*** 0.312*** 0.122*** 0.045 0.094** 0.029 0.050*** -0.186*** 0.010 0.054*** -0.184*** 0.014 (4.47) (-10.07) (0.96) (8.09) (-7.29) (4.53) (2.69) (4.79) (2.78) (1.50) (2.44) (0.91) (3.64) (-10.23) (0.70) (3.83) (-9.98) (1.01)

Size X ISBD (α4) -0.226*** -0.161*** -0.195*** (-6.54) (-4.09) (-5.42)

NLTEAR (α5) 0.222* 0.166 0.051 0.483 0.318 0.406 (1.88) (1.32) (0.41) (0.87) (0.56) (0.71)

NLTEAR^2 (α6) -0.342 -0.213 -0.426 (-0.70) (-0.41) (-0.84)

MS (α7) 0.214*** 0.094*** 0.241*** 0.206*** 0.090*** 0.236*** 0.172*** 0.320*** 0.138*** 0.257*** 0.199*** 0.260*** 0.220*** 0.099*** 0.243*** 0.217*** 0.096*** 0.240*** (9.47) (3.66) (10.61) (8.95) (3.39) (10.13) (5.65) (6.62) (4.57) (5.35) (3.51) (5.38) (9.77) (3.84) (10.74) (9.61) (3.72) (10.56)

Ineff (α8) 3.247*** 5.366*** 2.933*** 3.239*** 5.364*** 2.918*** 3.171*** 2.475*** 3.200*** 3.270*** 2.794*** 3.309*** 3.239*** 5.367*** 2.930*** 3.266*** 5.378*** 2.954*** (24.73) (25.02) (23.60) (24.66) (25.00) (23.52) (19.71) (9.76) (19.92) (14.41) (9.63) (13.55) (24.66) (24.95) (23.47) (24.10) (25.07) (22.81)

NII (α9) -1.065*** -0.504*** -0.971*** -1.064*** -0.504*** -0.967*** -0.827*** -1.030*** -0.719*** -1.280*** -1.092*** -1.182*** -1.072*** -0.508*** -0.976*** -1.049*** -0.491*** -0.958*** (-7.11) (-2.82) (-6.30) (-7.23) (-2.83) (-6.34) (-3.35) (-2.95) (-2.81) (-5.77) (-3.25) (-5.07) (-7.04) (-2.83) (-6.26) (-6.82) (-2.72) (-6.07)

TAG (α10) -0.345*** -0.084 -0.358*** -0.363*** -0.097 -0.375*** -0.318*** 0.028 -0.382*** -0.388*** -0.039 -0.468*** -0.332*** -0.075 -0.355*** -0.334*** -0.077 -0.357*** (-4.60) (-0.90) (-4.80) (-4.92) (-1.05) (-5.08) (-2.98) (0.13) (-3.65) (-3.59) (-0.27) (-4.09) (-4.37) (-0.79) (-4.71) (-4.38) (-0.81) (-4.70)

StateD (α11) -0.304*** -0.064 -0.327*** -0.272*** -0.040 -0.298*** -0.150 -0.097 -0.147 -0.371*** -0.450*** -0.386*** -0.310*** -0.066 -0.328*** -0.307*** -0.065 -0.325*** (-3.91) (-0.88) (-3.92) (-3.52) (-0.55) (-3.59) (-1.35) (-0.63) (-1.33) (-3.60) (-3.49) (-3.38) (-3.99) (-0.91) (-3.94) (-3.96) (-0.89) (-3.90)

ForeignD (α12) -0.648*** -1.064*** -0.623*** -0.676*** -1.082*** -0.645*** -0.716*** -0.598*** -0.723*** -0.641*** -0.968*** -0.601*** -0.630*** -1.052*** -0.619*** -0.633*** -1.053*** -0.619*** (-7.00) (-7.87) (-6.57) (-7.20) (-7.96) (-6.71) (-5.70) (-3.29) (-5.71) (-4.66) (-4.01) (-4.27) (-6.79) (-7.75) (-6.53) (-6.87) (-7.68) (-6.58)

SubsidiaryD (α13) 0.311*** 0.233*** 0.293*** 0.311*** 0.232*** 0.297*** 0.284*** 0.338*** 0.279*** 0.288*** 0.268** 0.263*** 0.303*** 0.228*** 0.291*** 0.308*** 0.231*** 0.298*** (5.85) (4.10) (5.45) (5.83) (4.06) (5.48) (3.52) (2.92) (3.43) (3.37) (2.49) (3.02) (5.67) (3.98) (5.40) (5.78) (4.01) (5.53)

YoungD (α14) -0.562*** -0.360*** -0.642*** -0.582*** -0.373*** -0.660*** -0.514*** -0.247 -0.495*** -0.651*** -0.706*** -0.644*** -0.556*** -0.356*** -0.641*** -0.565*** -0.362*** -0.647*** (-6.78) (-4.40) (-7.58) (-7.09) (-4.62) (-7.84) (-4.22) (-1.34) (-4.07) (-5.56) (-4.77) (-5.24) (-6.76) (-4.35) (-7.57) (-6.80) (-4.41) (-7.57)

MiddleageD (α15) -0.294*** -0.008 -0.354*** -0.307*** -0.015 -0.369*** -0.268*** -0.059 -0.295*** -0.391*** -0.393*** -0.405*** -0.289*** -0.006 -0.352*** -0.294*** -0.008 -0.354*** (-4.17) (-0.09) (-5.00) (-4.33) (-0.17) (-5.20) (-2.91) (-0.44) (-3.20) (-3.41) (-2.62) (-3.54) (-4.09) (-0.07) (-4.97) (-4.15) (-0.09) (-4.99) HHI (α16) -0.136 0.358* -0.527** -0.171 0.334 -0.560*** -0.030 0.036 -0.170 -0.500 -0.318 -0.631* -0.123 0.381* -0.533** -0.150 0.361 -0.550*** (-0.68) (1.66) (-2.54) (-0.86) (1.55) (-2.72) (-0.10) (0.09) (-0.52) (-1.63) (-0.85) (-1.92) (-0.61) (1.74) (-2.54) (-0.74) (1.61) (-2.58)

EDInsurance (α17) -0.192*** -0.916*** -0.096 -0.215*** -0.935*** -0.113* -0.315*** -0.054 -0.360*** -0.126* -0.211** -0.059 -0.180*** -0.908*** -0.093 -0.182*** -0.911*** -0.093 (-3.36) (-13.56) (-1.62) (-3.77) (-13.74) (-1.90) (-2.76) (-0.36) (-3.11) (-1.76) (-2.29) (-0.76) (-3.16) (-13.47) (-1.57) (-3.16) (-13.45) (-1.55)

DomCredit (α18) -0.270*** 0.198*** -0.273*** -0.245*** 0.215*** -0.252*** -0.190** -0.446*** -0.166** -0.356*** -0.408*** -0.340*** -0.275*** 0.196*** -0.277*** -0.268*** 0.202*** -0.271*** (-4.78) (3.53) (-4.68) (-4.36) (3.84) (-4.34) (-2.46) (-3.37) (-2.13) (-4.15) (-3.77) (-3.85) (-4.67) (3.31) (-4.57) (-4.55) (3.38) (-4.48)

ExRateDep (α19) -0.136 -0.258 -0.223 -0.188 -0.294 -0.270 -0.054 0.463** -0.206 -0.784** -0.634 -1.086*** -0.137 -0.259 -0.223 -0.143 -0.263 -0.228

46

(-0.79) (-1.32) (-1.27) (-1.13) (-1.54) (-1.57) (-0.29) (2.48) (-1.07) (-2.52) (-1.49) (-3.33) (-0.79) (-1.32) (-1.26) (-0.83) (-1.34) (-1.29)

GDPPCGrowth (α20) -0.424 1.393 -0.040 -0.354 1.430 0.059 0.126 0.768 0.219 -0.069 1.908 0.179 -0.430 1.378 -0.025 -0.462 1.346 -0.056 (-0.44) (1.36) (-0.04) (-0.37) (1.40) (0.06) (0.06) (0.22) (0.10) (-0.06) (1.51) (0.15) (-0.45) (1.36) (-0.02) (-0.48) (1.32) (-0.06)

Number of obs 1,337 1,301 1,343 1,337 1,301 1,343 564 536 567 773 761 776 1,337 1,301 1,343 1,337 1,301 1,343

R-Squared 0.557 0.550 0.562 0.558 0.551 0.563 0.578 0.298 0.600 0.561 0.384 0.543 0.557 0.550 0.562 0.557 0.550 0.563

H0: α3 + α4 = 0 (F-stat.) 10.85*** 64.76*** 14.26***

H0: α3 = α4 = 0 (F-stat.) 40.04*** 52.76*** 18.59***

H0: α5 = α6 = 0 (F-stat.) 0.65 0.34 0.48

Variable definitions: Risk = Logarithm of Zscore is used as the insolvency risk proxy. Zscore is defined as (ROAA+ETA)/SD(ROAA). ROAA = Returns on average assets at time t. ETA = Equity to asset ratio at time t. SD(ROAA) = standard deviation of ROAA over 3 years (current year and two previous consecutive years). Banks need to have three consecutive observations. Acquiring banks are excluded from the sample, since the volatility on their assets returns can be due to the acquisition. ZP1 = logarithm of the first component of Zscore, i.e. ROAA/SD(ROAA), ZP2 = logarithm of the second component of Zscore, i.e. ETA/SD(ROAA), ISBD=Islamic bank dummy, ISWD= The dummy variable representing Islamic window banks, Size = Logarithm of total assets, Size X ISBD = Interaction of Size and ISBD, NLTEAR = Net loans to total earning assets ratio, NLTEAR^2 = the square form of NLTEAR, MS = Logarithm of market share, orthogonalized on size, Ineff = Cost inefficiency, NII = Share of noninterest income in total operating income, TAG = Total assets growth, StateD = State-owned bank dummy, ForeignD = Foreign-owned bank dummy, SubsidiaryD = Subsidiary dummy, YoungD = Young bank dummy, MiddleD = Middle-aged bank dummy, HHI = Hirschman-Herfindahl index, EDInsurance = Explicit deposit insurance scheme, DomCredit = Domestic credit provided by banking system (% of GDP), ExRateDep = Official exchange rate depreciation, GDPPCGrowth = Annual growth of GDP per capita. All the right-hand side variables are lagged for one year, which are in fact the median value of the 3-year rolling window. Dummy variables that are rarely changing over time , i.e. ISBD, ISWD, StateD and ForeignD, are also lagged for one period to make sure that they represent the status a bank holds in at least two periods (years) out of the 3-year rolling window. Hausman test rejects random effect estimation; however, we have several dummy variables that are time invariant, SubsidiaryD and EDInsurance or rarely changing over time. Fixed effect technique can not estimate the time invariant variables and it is inefficient in estimating variables with very limited within variance. Hence, we follow the fixed effects vector decomposition (FEVD) approach proposed by Plumper and Troeger (2007) to capture the unobservable individual-specific effects. In column (1) the logarithm of Zscore is used as the insolvency risk proxy. The logarithms of the first and second components of Zscore are used in columns (2) and (3) respectively. In columns (4) to (6), the interaction term of size and ISBD is included in the model. In columns (7) to (9) small banks sub-sample and in columns (10) to (12) large banks sub-sample are used. We add NLTEAR in columns (13) to (15). Finally, NLTEAR^2 is incorporated in columns (16) to (18). Year dummies and the unit effects captured by FEVD approach are included in the model, but not reported in the table. Robust z-statistics are reported in parentheses. ***, ** and * indicate significance at 1%, 5% and 10% respectively.

47

Table 6 Single Equation Estimation of Insolvency Risk, Cross-Section Estimation

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)

Interaction of Size & ISBD Small Banks Sub-Sample Large Banks Sub-Sample Share of Net Loans in Total Earning Assets Square form of Net Loans Share

Variables Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2 Risk ZP1 ZP2

Constant (α0) 3.941*** 0.127 3.823*** 3.872*** 0.089 3.721*** 3.366 -0.995 4.706* 2.868** 0.452 2.954** 4.130*** 0.357 4.162*** 4.374*** 0.934 4.709*** (5.25) (0.18) (5.05) (5.16) (0.13) (4.93) (1.42) (-0.49) (1.72) (2.21) (0.36) (2.25) (5.08) (0.50) (5.09) (3.82) (1.05) (3.96)

ISBD (α1) 0.256 0.167 0.234 0.941 0.673 1.240 0.121 -0.169 0.085 0.255 0.329* 0.235 0.278 0.198 0.272 0.256 0.148 0.224 (1.47) (0.97) (1.34) (0.75) (0.50) (0.95) (0.43) (-0.42) (0.29) (1.43) (1.72) (1.31) (1.57) (1.12) (1.53) (1.39) (0.87) (1.22)

ISWD (α2) -0.026 0.063 -0.050 -0.035 0.057 -0.062 -0.243 -0.226 -0.311 0.011 0.143 -0.022 -0.017 0.074 -0.033 -0.016 0.074 -0.032 (-0.22) (0.48) (-0.41) (-0.28) (0.43) (-0.51) (-1.09) (-0.76) (-1.32) (0.08) (0.94) (-0.15) (-0.14) (0.56) (-0.27) (-0.13) (0.55) (-0.26)

M_Size (α3) -0.024 0.081** -0.021 -0.018 0.085** -0.012 -0.018 0.187 -0.102 0.061 0.069 0.040 -0.020 0.086** -0.015 -0.016 0.095** -0.007 (-0.64) (2.27) (-0.53) (-0.47) (2.36) (-0.30) (-0.09) (1.06) (-0.49) (0.89) (1.14) (0.58) (-0.53) (2.36) (-0.36) (-0.44) (2.56) (-0.17)

M_Size X ISBD (α4) -0.049 -0.036 -0.072 (-0.57) (-0.40) (-0.81)

M_ NLTEAR (α5) -0.312 -0.358 -0.559 -1.238 -2.652 -2.641 (-0.77) (-1.03) (-1.18) (-0.45) (-1.39) (-0.92)

M_ NLTEAR^2 (α6) 0.860 2.132 1.940 (0.36) (1.25) (0.78)

M_MS (α7) 0.067 0.178*** 0.076 0.064 0.177*** 0.072 0.005 0.243** 0.018 0.010 0.066 0.047 0.070 0.181*** 0.080 0.069 0.180*** 0.078 (0.94) (2.85) (1.05) (0.91) (2.84) (1.01) (0.06) (2.23) (0.17) (0.08) (0.78) (0.42) (0.97) (2.90) (1.09) (0.96) (2.90) (1.07)

M_Ineff (α8) -0.647*** -0.366* -0.726*** -0.643*** -0.364* -0.718*** -0.143 -0.357 -0.393 -0.913*** -0.523* -0.902*** -0.677*** -0.404** -0.772*** -0.717*** -0.496** -0.861*** (-2.92) (-1.75) (-3.15) (-2.91) (-1.74) (-3.13) (-0.57) (-1.16) (-1.02) (-3.11) (-1.83) (-3.21) (-3.02) (-2.02) (-3.26) (-2.86) (-2.33) (-3.09)

M_TAG (α9) -0.077 0.276 -0.092 -0.081 0.270 -0.098 -0.005 1.448** 0.318 -0.275 -0.287 -0.338 -0.084 0.261 -0.109 -0.102 0.194 -0.150 (-0.23) (0.48) (-0.28) (-0.24) (0.47) (-0.29) (-0.01) (2.27) (0.37) (-0.81) (-0.47) (-1.06) (-0.25) (0.45) (-0.34) (-0.30) (0.33) (-0.47)

M_NII (α10) -0.746 -0.523 -1.224 -0.788 -0.555 -1.284 0.231 0.050 -1.800 -1.486* -1.253* -1.340* -0.803 -0.573 -1.291 -0.843 -0.641 -1.370* (-1.18) (-0.87) (-1.50) (-1.25) (-0.92) (-1.56) (0.25) (0.05) (-0.91) (-1.76) (-1.77) (-1.75) (-1.27) (-0.96) (-1.60) (-1.31) (-1.07) (-1.66)

StateD (α11) 0.181 0.016 0.243 0.182 0.017 0.244 0.627** 0.140 0.513 0.195 0.056 0.243 0.187 0.017 0.252 0.188 0.009 0.254 (0.88) (0.10) (1.25) (0.88) (0.11) (1.25) (2.36) (0.42) (1.66) (0.69) (0.30) (0.94) (0.90) (0.11) (1.28) (0.91) (0.06) (1.31)

ForeignD (α12) -0.497** -0.765*** -0.403* -0.514** -0.779*** -0.428* -0.448 -0.404 -0.112 -0.509 -1.010*** -0.490 -0.516** -0.790*** -0.441* -0.512** -0.785*** -0.432* (-2.20) (-2.82) (-1.79) (-2.19) (-2.76) (-1.84) (-1.55) (-0.98) (-0.30) (-1.61) (-2.68) (-1.61) (-2.23) (-2.87) (-1.92) (-2.22) (-2.92) (-1.90)

SubsidiaryD (α13) 0.149 0.406*** 0.191 0.143 0.402*** 0.182 0.610** 0.805*** 0.619** 0.005 0.150 -0.004 0.164 0.422*** 0.215 0.160 0.418*** 0.203 (0.90) (3.06) (1.05) (0.85) (3.00) (0.99) (2.39) (2.91) (2.32) (0.02) (0.95) (-0.02) (1.01) (3.13) (1.16) (0.97) (3.12) (1.10)

MiddleageD (α14) -0.083 -0.525 0.042 -0.099 -0.527 0.019 -0.899 -1.341** -0.574 0.477 -0.021 0.469 -0.084 -0.517 0.037 -0.065 -0.471 0.081 (-0.19) (-1.20) (0.10) (-0.23) (-1.21) (0.05) (-1.66) (-2.02) (-1.15) (0.71) (-0.03) (0.72) (-0.20) (-1.18) (0.09) (-0.15) (-1.03) (0.19)

M_HHI (α15) -0.876 -1.705** -0.764 -0.857 -1.691** -0.737 -1.982 -3.005** -1.843 -0.308 -0.995 -0.449 -1.025 -1.897*** -1.039 -0.951 -1.768** -0.872 (-1.22) (-2.54) (-1.08) (-1.21) (-2.53) (-1.04) (-1.62) (-2.48) (-1.62) (-0.38) (-1.31) (-0.58) (-1.43) (-2.79) (-1.48) (-1.34) (-2.58) (-1.23)

EDInsurance (α16) -0.057 -0.078 -0.115 -0.069 -0.084 -0.132 0.225 0.515 -0.406 -0.071 -0.245 -0.052 -0.073 -0.085 -0.141 -0.083 -0.105 -0.162 (-0.28) (-0.44) (-0.60) (-0.35) (-0.47) (-0.70) (0.61) (1.08) (-0.67) (-0.34) (-1.23) (-0.26) (-0.37) (-0.48) (-0.74) (-0.40) (-0.59) (-0.82)

M_DomCredit (α17) 0.824*** 0.405*** 0.848*** 0.826*** 0.406*** 0.851*** 0.557* 0.128 0.748** 0.875*** 0.610*** 0.936*** 0.796*** 0.370** 0.799*** 0.792*** 0.357** 0.792*** (5.05) (2.75) (5.13) (5.06) (2.75) (5.14) (1.74) (0.37) (2.03) (3.93) (3.47) (4.18) (4.97) (2.37) (4.70) (5.03) (2.27) (4.70)

M_ExRateDep (α18) -2.284** -0.456 -2.762*** -2.194** -0.387 -2.629*** -5.331*** -1.260 -7.085*** -1.661* -0.306 -1.681* -2.325** -0.489 -2.821*** -2.286** -0.363 -2.728***

48

(-2.44) (-0.47) (-2.80) (-2.39) (-0.40) (-2.74) (-3.33) (-0.82) (-3.57) (-1.66) (-0.25) (-1.72) (-2.50) (-0.51) (-2.89) (-2.42) (-0.37) (-2.75)

M_GDPPCGrowth (α19) 2.826 7.146 6.748 2.613 6.775 6.426 -7.404 -9.728 8.759 9.973 12.006 10.570 4.135 8.254 8.953 3.987 8.085 8.549 (0.31) (1.08) (0.77) (0.29) (1.01) (0.73) (-0.45) (-0.46) (0.37) (0.98) (1.42) (1.07) (0.47) (1.25) (0.99) (0.45) (1.25) (0.97)

Number of Obs 231 223 232 231 223 232 81 78 82 150 145 150 231 223 232 231 223 232

R-squared 0.213 0.192 0.213 0.214 0.192 0.215 0.433 0.325 0.339 0.222 0.213 0.241 0.215 0.196 0.222 0.216 0.202 0.225

H0: α3 + α4 = 0 (F-stat.) 0.63 0.30 0.93

H0: α3 = α4 = 0 (F-stat.) 0.36 2.78 0.47

H0: α5 = α6 = 0 (F-stat.) 0.30 1.20 0.78

Variable definitions: Risk = Logarithm of Zscore is used as the insolvency risk proxy. Zscore is defined as (M_ROAA+M_ETA)/SD(ROAA). M_ROAA = Mean of returns on average assets over the sample period. M_ETA = Mean of Equity to asset ratio over the sample period. SD(ROAA) = standard deviation of ROAA over the sample period. Banks need to have at least five consecutive observations. Acquiring banks are excluded from the sample, since the volatility on their assets returns can be due to the acquisition. ZP1 = logarithm of the first component of Zscore, i.e. M_ROAA/SD(ROAA), ZP2 = logarithm of the second component of Zscore, i.e. M_ETA/SD(ROAA), ISBD=Islamic bank dummy, ISWD= The dummy variable representing Islamic window banks, M-Size = Logarithm of mean of total assets, M_Size X ISBD = Interaction of M_Size and ISBD, M_NLTEAR = Mean of net loans to total earning assets ratio, M_NLTEAR^2 = The square form of MNLTEAR, M_MS = Logarithm of mean of market share, orthogonalized on size, M_Ineff = Mean of cost inefficiency, M_NII = Mean of share of noninterest income in total operating income, M_TAG = Mean of total assets growth, StateD = State-owned bank dummy, ForeignD = Foreign-owned bank dummy, SubsidiaryD = Subsidiary dummy, YoungD = Young bank dummy, MiddleD = Middle-aged bank dummy, M_HHI = Mean of Hirschman-Herfindahl index, EDInsurance = Explicit deposit insurance scheme, M_DomCredit = Mean of Domestic credit provided by banking system (% of GDP), M_ExRateDep = Mean of official exchange rate depreciation, M_GDPPCGrowth = Mean of annual growth of GDP per capita.. In column (1) the logarithm of Zscore is used as the insolvency risk proxy. The logarithms of the first and second components of Zscore are used in columns (2) and (3) respectively. In columns (4) to (6), the interaction term of M_Size and ISBD is included in the model. In columns (7) to (9) small banks sub-sample and in columns (10) to (12) large banks sub-sample are used. We add M_NLTEAR in columns (13) to (15). Finally, NLTEAR^2 is incorporated in columns (16) to (18). Robust z-statistics are reported in parentheses. ***, ** and * indicate significance at 1%, 5% and 10% respectively.

49

Annex 1 Description of Variables Used in the Analysis

Loans’ Risk Proxy Description

Problem Loans / Gross Loans (PLGL)

Problem Loans (PL) increases when a bank classifies a specific loan or a part of loans portfolio as the bad loans. It decreases when either a bank re-assesses a problem loan or a part of the portfolio of loans as the good loans or when a bank writes off a loan or a part of portfolio of loans.

Loan Loss Reserves / Gross Loans (LLRGL)

Loan Loss Reserves (LLR) is considered for the whole loans portfolio, and not only for the Problems Loans. The managers assess the quality of the loans portfolio and determine the required reserves. Then the current level of LLR will be adjusted to reach to the required level. The adjustment will be reflected in the Loan Loss Provision stipulated in the income statement. When a bank decides to write off a loan, the loan amount would be deducted from the LLR.

Loan Loss Provisions / Average Gross Loans (LLPAGL)

Loan Loss Provision (LLP) is the incurred cost by banks as a result of adjusting the LLR or writing off a loan. Hence, despite of PL and LLR which are stock. LLP is flow and is stipulated in the income statement. It is possible to have a negative LLP in one period, when the required loan loss reserve is lower than the current reserve.

Insolvency Risk Proxy

Z-score [E(ROAA)+ETA]/ SD(ROAA), wherein E(ROAA) stands for Expected Returns on Average Assets and ETA stands for Equity to Asset ratio. SD(ROAA) represents volatility of Returns on Average Assets (ROAA).

Size Logarithm of Total Assets controls for size

Market Share Logarithm of market share of total assets.

Financial Ratios

Capital Equity to Asset Ratio is the proxy

Net Loans to Total Earning Assets Ratio The share of net loans in total earning assets, representing the degree of diversity in total earning assets of a bank

Total Non Earning Asset Ratio Share of total non-earning assets in total assts

ROAA Stands for Returns on Average Assets

Non-Interest Income Represented by Net Comm. & Trading Income to Total Operating Income Ratio

Asset Growth Calculated based on the annual growth of total assets

Growth of Gross Loans Calculated based on the annual growth of gross loans

Inefficiency Using the stochastic frontier approach, the cost inefficiency is estimated for each bank. The model and methodology is briefly presented in Annex 2.

Cost to Income Ratio Stands for cost to income ratio, an alternative proxy for inefficiency

Ownership Structure

State-Owned Bank Dummy Takes one, in case the bank is state-owned, and zero otherwise.

Foreign-Owned Bank Dummy Takes one, in case the bank is Foreign-owned, and zero otherwise.

Subsidiary Dummy Takes one, in case the bank is subsidiary, and zero otherwise.

Banks’ Age Dummy

Young Bank Dummy Takes one, in case the bank operates for at most three years, and zero otherwise.

Middle-Aged Bank Dummy Takes one, in case the bank operates between three to seven years, and zero otherwise.

Country Level Variables

HHI 2

j,t i,t,j i,t,j1 i=1

HHI = TotalAssets TotalAssetsn n

i

. Hirschman-Herfindahl index (HHI) is a proxy for

50

market concentration. It takes the value between zero and one. Higher values show that the market is more concentrated.

EDInsurance Dummy Takes one, in case the country adopts explicit deposit insurance scheme, and zero otherwise.

DomCredit The share of domestic credit extended by banking system in GDP (%)

ERDep Official Exchange Rate Depreciation

PerCapitaGrowth The annual growth of GDP per Capita

Year Dummies Seven dummy variables are supposed to control for the years effects.

51

Annex 2 Estimating Cost Inefficiency

To estimate the cost inefficiency, we adopt the translog cost function which is the same

as the model applied by Altunbas et al. (2007). Two outputs (loans and other earning

assets) and three input prices (wage. interest expense rate and other operating expenses

price) are considered in the model. The empirical specification which is used to estimate

the cost inefficiency can be expressed as follows:

2 32

, , ,1 1 , ,1 1

2 2 3 3 2 3

, ,, , , , , ,, , , , , ,1 1 1 1 1 1

1ln ( ) ln ( ) ln2

1 ln ln ln ln ln ln2

i t k i tj kj kj i tj k

i t i tk i t m i t m i tjl km jmj i t l i t j i tj l k m j m

TC C t t t

V U

Qt P

Q Q QP P P

Wherein ,ln i tTC is the natural logarithm of total cost of the bank i at the time of t;

, ,ln

j i tQ is the natural logarithm of output vector of the bank i at the time of t. Loans and

Other Earning Assets are considered as the output;

, ,ln k i tP is the natural logarithm of input prices vector of the bank i at the time of t, and

consists of wage, interest expense rate and other operating expenses price; wage is

obtained by dividing personnel expenses on fixed assets, as a proxy for the number of

employees. Interest expense rate is the interest expense divided by deposits, short term

funding and other funding. Other operating expenses price is calculated as the ratio of

other operating expenses on total assets.

,i tV is the random variable which are assumed to be normally distributed with the

expected value of zero. 2(0, )v

N .

,i tU is assumed to be equal to exp( ( ))i

t TU as proposed by Battese and Coelli (1992),

wherein iU is the non-negative random variable representing the cost inefficiency. It is

assumed to have a normal distribution of 2( , )u

N truncated at zero. is a parameter to

be estimated.

52

Annex 3 Cross Country and Banks’ Types Sample Specifications

Table A Number of Islamic, conventional and Islamic window banks across 22 countries, over the 2001-2008 period.

Country Islamic bank Islamic Window Bank Conventional Bank Total

Banks Observations Banks Observations Banks Observations Banks Observations

Algeria 1 8 2 10 10 52 13 70 Bahrain 10 60 10 52 1 8 21 120 Bangladesh 5 37 9 55 20 163 34 255 Egypt 2 16 4 31 23 134 29 181 Gambia 1 6 0 0 6 25 7 31 Indonesia 2 12 9 51 52 324 63 387 Iran 11 67 0 0 0 0 11 67 Jordan 2 15 0 0 9 66 11 81 Kuwait 3 14 1 8 5 40 9 62 Lebanon 2 7 2 16 40 217 44 240 Malaysia 14 36 10 69 17 115 41 220 Mauritania 1 8 3 16 3 23 7 47 Pakistan 7 30 11 75 14 74 32 179 Qatar 2 16 2 9 5 38 9 63 Saudi Arabia 4 17 6 55 0 0 10 72 Senegal 1 4 0 0 10 59 11 63 Sudan 17 87 0 0 0 0 17 87 Syria 2 2 0 0 8 32 10 34 Tunisia 1 8 0 0 14 82 15 90 Turkey 4 13 0 0 26 96 30 109 UAE 5 29 3 20 14 106 22 155 Yemen 4 26 0 0 6 35 10 61 Total 101 518 72 467 283 1689 456 2674

53

Annex 3 Cross Country and Banks’ Types Sample Specifications

Table B The mean value of five banking system and macroeconomic indicators across 22 countries, over the 2001-2008 period.

Countries HHI EDIsurance DomCredit (%) ERDep (%) PerCapitaGrowth (%)

Algeria 0.33 1 15 -1.8 2.5 Bahrain 0.26 1 72 0.0 4.2 Bangladesh 0.09 1 54 3.5 4.1 Egypt 0.17 0 101 6.3 2.9 Gambia 0.38 0 26 8.5 1.9 Indonesia 0.12 1 46 2.2 3.9 Iran 0.21 0 46 64.1 4.0 Jordan 0.48 1 101 0.0 4.7 Kuwait 0.18 0 74 -1.6 4.4 Lebanon 0.09 1 181 0.0 3.1 Malaysia 0.11 1 149 -1.6 3.1 Mauritania 0.21 0 24 -23.9 2.0 Pakistan 0.12 0 41 3.7 2.5 Qatar 0.30 0 50 0.0 4.0 Saudi Arabia 0.14 0 49 0.0 1.5 Senegal 0.23 0 23 -5.5 1.6 Sudan 0.26 1 14 -2.4 5.2 Syria 0.76 0 32 -12.5 1.6 Tunisia 0.16 0 72 -1.2 3.8 Turkey 0.26 1 48 13.2 3.1 UAE 0.10 0 60 0.0 2.5 Yemen 0.20 0 7 2.7 1.0

HHI: Hirschman-Herfindahl index, EDInsurance: Explicit Deposit Insurance Scheme Dummy, DomCredit: Domestic Credit Provided by Banking System as the percentage of GDP, ERDep: Official Exchange Rate Depreciation, PerCapitaGrowth: Annual Growth Rate of GDP Per Capita

Table C Ownership structure and age (experience level) of Islamic, conventional and Islamic window banks

Islamic bank Islamic Window Bank Conventional Bank Total

Banks Observations Banks Observations Banks Observations Banks Observations

State-owned Banks 10 64 8 51 36 241 54 356 Foreign-owned Banks 17 88 6 35 18 98 41 221 Subsidiaries 15 74 12 75 78 437 105 586 Private-owned Banks 58 291 46 306 152 914 256 1511 Total 100 517 72 467 284 1690 456 2674 Young Banks 19 104 4 30 11 106 34 240 Middle Aged Banks 18 77 3 36 26 136 46 248 Matured Banks 64 337 65 401 246 1447 376 2186

Total 101 518 72 467 283 1689 456 2674

State-owned banks: state ownership > 50%. Foreign-owned banks: foreign ownership > 50%. Subsidiaries: parent ownership = 100%. Private-owned banks: domestic private ownership > 50%. Young banks: operating less than 3 years. Middle aged banks: operating between 3 to 7 years. Matured banks: operating more than 7 years. The information is mainly obtained from banks web-sites.

54

Annex 4 Correlation Matrix

a b c d e f g h i j k l m n o p q R s t u v w x y z

(a) Loan Loss Reserves on Gross Loan 1

(b) Problem Loans on Gross Loan 0.74 1

(c) Loan Loss Provision on Average Gross Loans 0.30 0.23 1

(d) Islamic Bank Dummy -0.18 -0.12 0.04 1

(e) Islamic Window Bank Dummy -0.07 -0.10 -0.06 -0.23 1

(f) Log of Asset -0.19 -0.18 -0.12 -0.08 0.23 1

(g) Log of Market Share -0.06 -0.03 0.08 0.02 0.16 0.60 1

(h) Asset Growth -0.20 -0.26 -0.02 0.20 -0.06 -0.04 -0.03 1

(i) Gross Loan Growth -0.22 -0.27 0.02 0.16 -0.04 -0.03 -0.01 0.62 1

(j) Equity Asset Ratio 0.00 -0.06 -0.02 0.11 -0.07 -0.30 -0.31 0.06 0.04 1

(k) Net Loan to Total Earning Assets Ratio -0.34 -0.24 -0.09 0.11 0.07 0.10 0.06 0.00 0.08 -0.14 1

(l) Total Non-Earning Assets Ratio 0.02 0.09 0.08 0.18 -0.08 -0.21 -0.09 0.02 -0.01 0.07 -0.07 1

(m) Return on Average Assets -0.23 -0.33 -0.22 0.09 0.06 0.08 0.04 0.20 0.12 0.31 0.01 -0.16 1

(n) Non-Interest Income Ratio 0.15 -0.01 0.13 0.03 -0.04 -0.02 0.22 -0.01 -0.01 -0.02 -0.05 0.06 -0.08 1

(o) Inefficiency 0.22 0.22 0.15 -0.01 -0.09 0.11 0.03 -0.15 -0.11 -0.25 -0.13 0.56 -0.24 -0.01 1

(p) Cost to Income Ratio 0.17 0.24 -0.04 0.03 -0.12 -0.27 -0.15 -0.11 -0.04 -0.11 -0.11 0.18 -0.57 0.04 0.16 1

(q) State-Owned Bank Dummy 0.05 0.09 0.07 -0.01 -0.03 0.15 0.19 -0.03 -0.03 -0.11 0.03 -0.06 -0.02 -0.01 -0.02 0.01 1

(r) Foreign-Owned Bank Dummy 0.15 0.12 0.04 0.16 -0.01 -0.08 -0.03 0.00 0.00 0.05 -0.08 0.02 -0.06 0.07 0.09 0.05 -0.12 1

(s) Subsidiary Dummy 0.04 -0.05 0.00 -0.09 -0.07 -0.16 -0.18 0.01 -0.02 0.11 -0.02 0.08 0.01 0.14 -0.06 -0.01 -0.21 -0.16 1

(t) Young Bank Dummy -0.08 -0.10 0.04 0.19 -0.04 -0.17 -0.14 0.31 0.33 0.20 -0.05 0.10 0.00 -0.03 -0.03 0.03 -0.10 0.03 0.06 1

(u) Middle-Age Bank Dummy -0.07 -0.03 0.07 0.09 -0.03 -0.15 -0.07 0.09 0.08 0.01 0.04 0.01 0.00 -0.01 -0.10 0.00 -0.05 0.00 0.03 -0.10 1

(v) HHI 0.05 0.03 0.11 0.12 -0.10 -0.06 0.26 0.09 0.12 0.06 -0.17 0.06 0.05 0.17 -0.16 0.02 -0.01 -0.02 0.02 0.15 0.04 1

(w) Explicit Deposit Insurance Scheme -0.12 0.03 -0.08 -0.06 -0.04 -0.10 -0.34 -0.08 -0.06 -0.03 -0.03 0.24 -0.06 -0.32 0.24 0.05 -0.15 -0.03 -0.02 -0.05 0.02 -0.10 1

(x) Domestic Credit Provided by Banking System, as a percentage of GDP 0.19 0.23 -0.14 -0.18 0.01 0.18 -0.17 -0.16 -0.20 -0.08 -0.20 0.14 -0.14 -0.04 0.32 0.00 -0.17 0.01 0.03 -0.12 -0.09 -0.24 0.29 1

(y) Official Exchange Rate Depreciation 0.02 0.07 0.04 0.09 -0.02 0.00 -0.03 0.01 0.05 -0.07 0.00 -0.09 0.03 -0.01 -0.01 0.00 0.09 0.00 -0.04 0.02 0.01 -0.05 -0.05 -0.01 1

(z) GDP per Capita Growth Rate -0.14 -0.14 -0.09 0.09 -0.03 0.07 -0.04 0.10 0.07 0.05 0.05 0.00 0.07 -0.09 0.05 -0.04 -0.02 0.01 -0.03 0.00 -0.03 -0.04 0.20 -0.02 -0.01 1