-
As
Da
Ab
c
a
ARRAA
KINFRI
1
tobrtmmS
Bnilt
(
O
h1
Journal of Financial Stability 25 (2016) 4757
Contents lists available at ScienceDirect
Journal of Financial Stability
journal homepage: www.elsevier.com/locate/jfstabil
net stable funding ratio for Islamic banks and its impact on
financialtability: An international investigation
awood Ashraf a,, Muhammad Suhail Rizwan b, Barbara LHuillier
c
Islamic Research & Training Institute (A member of Islamic
Development Bank Group), 8111 King Khalid Street, Jeddah
22332-2444, Kingdom of SaudirabiaNUST Business School, National
University of Sciences and Technology (NUST), Islamabad,
PakistanCollege of Business Administration, Prince Mohammad Bin
Fahd University, Al Khobar 31952, Kingdom of Saudi Arabia
r t i c l e i n f o
rticle history:eceived 25 November 2015eceived in revised form
21 June 2016ccepted 29 June 2016vailable online 4 July 2016
eywords:
a b s t r a c t
The Islamic Financial Services Board (IFSB) is the standard
setting body for the Islamic banking industry.The IFSB, while
endorsing the Basel III accord, modified the criteria to calculate
the Net Stable FundingRatio (NSFR) to cater for the unique aspects
of the Islamic banking industry. In this paper, we calculatedthe
modified NSFR of 136 Islamic banks from 30 jurisdictions between
2000 and 2013 and explored thepotential impact the requirements of
this ratio has on the financial stability of Islamic banks after
con-trolling for bank, country, and market-specific variables. The
empirical findings suggest that the modified
slamic bankset stable funding ratioinancial stabilityegulatory
framework
FSB
NSFR has a positive impact on the financial stability of Islamic
banks during the sample period. However,the marginal impact of the
NSFR on stability diminishes as the size of the bank increases. The
resultsremained robust after applying an alternative measure of
stability and using an alternative estimationmodel based on an
instrumental variable approach. These results validate the use of
the IFSBs modifiedNSFR for Islamic banks as a regulatory
measure.
2016 Elsevier B.V. All rights reserved.
. Introduction
The 20072009 global financial crisis highlighted weaknesses inhe
conventional banking system and drew attention to the successf the
Islamic banking model (Hasan and Dridi, 2010). The Islamicanking
sector grew at an exceptional compound annual growthate of 17%
during the period 20082013. It now accounts for morehan a quarter
of the total banking assets of 10 countries where the
ajority of the population is Muslim including five of the
oil-richembers of the Gulf Cooperative Council (GCC)1 (Islamic
Financial
ervices Board, 2015a).In response to weaknesses in the global
financial system, the
asel Committee on Banking Supervision (BCBS) introduced twoew
regulatory measures in the Basel III regulatory framework. One
s a liquidity coverage ratio (LCR) that focuses on the
short-termiquidity of banks and the other is a net stable funding
ratio (NSFR)hat aims to monitor the long-term funding stability of
banks.
Corresponding author.E-mail addresses: [email protected] (D.
Ashraf), [email protected]
M.S. Rizwan), [email protected] (B. LHuillier).1 GCC member
countries include Saudi Arabia, Kuwait, Qatar, UAE, Bahrain,
andman.
ttp://dx.doi.org/10.1016/j.jfs.2016.06.010572-3089/ 2016
Elsevier B.V. All rights reserved.
Although adoption of the Basel III accord is being phased in,it
is expected that by 2019 all requirements will be fully
imple-mented. The full impact of these new regulatory requirements
onthe banking industry is still unknown. However, there is already
agrowing literature assessing the potential impact these new
reg-ulatory measures will have on the stability of conventional
banks.This literature capitalizes on the argument that the newly
intro-duced regulatory measures (NSFR and LCR) can be calculated
usingexisting data and their potential impact on banks can be
exploredretrospectively. Yan et al. (2012), using data from a
sample of 11 UKbanks for the period 19972010, found that higher
regulatory capi-tal requirements not only reduce the probability of
a banking crisisbut also reduce the economic loss from a banking
crisis. SimilarlyJiraporn et al. (2014), using data from a sample
of 68 banks from 11East Asian countries for the period 20052009,
reported an inverserelationship between the NSFR and risk-taking
behavior of banks.King (2013), using data from a sample of banks
from 15 countries,suggested that the implementation of the NSFR has
adverse con-sequences for the economy due to the shrinking of banks
balancesheets, changes in the composition of assets or maturity
thereof.
The business model for Islamic banks is quite different fromthat
of conventional banks in terms of their asset-liability struc-ture
and product offering. The International Monetary Fund (IMF)(2011)
suggested that the business model on which banks base their
dx.doi.org/10.1016/j.jfs.2016.06.010http://www.sciencedirect.com/science/journal/15723089http://www.elsevier.com/locate/jfstabilhttp://crossmark.crossref.org/dialog/?doi=10.1016/j.jfs.2016.06.010&domain=pdfmailto:[email protected]:[email protected]:[email protected]/10.1016/j.jfs.2016.06.010
-
48 D. Ashraf et al. / Journal of Financial Stability 25 (2016)
4757
Table 1Islamic bank assets and liabilities and their
conventional counterparts. Comparative haircuts given by Basel III
for conventional banks and IFSB for Islamic banks are given inthe
last two columns respectively. These haircuts are based on the
authors understanding of quantitative guidelines for the
calculation of the NSFR published by the IFSB(for Islamic banks)
and Basel III for conventional banks.
Islamic product Conventionalcounterpart
Nature of thecontract forIslamic banks
Key features Haircut underBasel III
Haircut underIFSB
Qard-al-Hassan orwadiah
Current account Debt Resembles conventional deposits,
althoughnon-interest/return bearing. May receive a gift(wadiah)
from bank capital.
50% 50%
Qard-al-Hassan orwadiah
Saving deposits Debt Safekeeping and profit sharing of Islamic
bank(deposit) contracts.
50% 50%
Profit-sharinginvestment accounts(PSIAs)
Saving and termdeposits
Equity Structured as profit/loss sharing partnerships(mudarbah
or musharaka) or agency (wakalah)contracts.
95%
PSIA (Restricted) Saving and termdeposits
Quasi equity Funds provided by investors are invested per
accountholders instructions and not comingled with banksown assets
and so easier to trace and transfer toaccount holders. However,
assets of all such accountholders may be pooled together, so
traceability maystill be a challenge.
95% 0%
PSIA (Unrestricted) Saving and termdeposits
Hybrid Account holders give banks full discretion to invest
inany Sharah compliant assets. May be comingled withbank assets or
those of other account holders.Traceability to specific account
holders may be achallenge.
95% 9095%
Sukuk Bonds and securitizedloans
Hybrids Islamic equivalent of conventional bonds. Structuredas
certificates of participation through securitization ofspecific
assets/pool of assets.
100%
Murabahah Loans and advances Debt A sales contract whereby the
institution offeringIslamic financial services sells to a customer
a specifiedkind of asset that is already in its possession.
Sellingprice is the sum of the original price and an agreedprofit
margin.
85% 85%
Musharaka Loans and advances Equity A contract between the
institution offering Islamicfinancial services and a customer. Both
wouldcontribute capital to an enterprise. Profits generatedby that
enterprise or real estate assets are shared bythe terms of the
Musharaka agreement. Losses areshared in proportion to each
partners share of capital.
85% 50%
Ijarah Mortgages and leases Equity An agreement made by an
institution offering Islamicfinancial services to lease an asset to
a customer for anagreed period for a specified rental. An Ijarah
contractcommences with a promise to lease that is binding onthe
part of the potential lessee before entering theIjarah
contract.
50%65% 50%
Qard-al-Hassan Loans and advances Debt An interest-free loan is
given by a lender to a borrowerwith the stipulation that the latter
pays back theprinciple only.
85% 0%
Salam and istisnaa Hybrid Hybrid Salam: Agreement to purchase,
at a predeterminedprice a specified kind of commodity not
currentlyavailable to the seller, to be delivered on a
specifiedfuture date as per agreed specifications and
specifiedquality.
85%
istisnaa: A contract of sale of specified objects to becturet of
th
to the
otsamrsdBcrm
sB
Islamic banking system risk is shared between the two (Hasan
andDridi, 2010). Regulatory requirements under the BCBSs
frameworkare based upon the underlying riskiness of banks and are
designed
manufathe parobjects
perations has serious consequences for the stability of banks
andhat prior to 2008 the NSFR of investment banks declined
moreharply as compared to commercial banks. Furthermore,
Mergaertsnd Vennet (2016) while examining the impact of bank
businessodels on performance and risk of European banks found
that
etail banks perform better in terms of profitability and
stability anduggested that business model considerations should be
more fun-amentally integrated in the regulatory and supervisory
practices.eck et al. (2013) observed that Islamic banks are
generally betterapitalized compared to conventional banks. The
equity-based andisk-sharing nature of Islamic contracts helps
reduce the maturity
ismatch of assets and liabilities and enhances financial
stability.
The Islamic Financial Services Board (IFSB) is the standard-
etting body for the Islamic banking industry. The IFSB endorsed
theasel III regulatory framework after making some adjustments
for
d or constructed, with an obligation one manufacturer or builder
to deliver the
customer upon completion.
the difference in the nature of assets and liabilities of
Islamic banks.The IFSB issued Guidance Note No. 12 which provides
guidelines forthe calculation of the NSFR for Islamic banks.2
Why is there a need for a modified NSFR for Islamic
banks?Response to this very critical question centers on the
treatment ofrisk under both banking systems. Under the conventional
bankingsystem risk transfers from lenders to borrowers while under
the
2 Guidance Note 12 on quantitative measures for liquidity risk
management ininstitutions offering Islamic financial services
[excluding Islamic insurance (Takaful)institutions and Islamic
collective investment schemes] issued by the IslamicFinancial
Services Board in 2014. Online: www.ifsb.org.
http://www.ifsb.orghttp://www.ifsb.orghttp://www.ifsb.org
-
nancial Stability 25 (2016) 4757 49
twn
nfipcImPM
attfiuaaisfcidmpc
ulwto
c3uuadg
dfietuaa
ibeitbsdh
p
D. Ashraf et al. / Journal of Fi
o adequately buffer levels of risk. However, this regulatory
frame-ork cannot remain efficient and effective if its application
does
ot take into account the risk-sharing nature of Islamic
banks.The IFSBs modified NSFR takes into account the
risk-sharing
ature of the underlying contracts of Islamic banks and modi-es
regulatory requirements for the NSFR accordingly. Table 1rovides a
detailed comparison of the BCBSs proposed NSFR foronventional banks
and the IFSBs proposed modified NSFR forslamic banks. Major
differences can be seen in regulatory require-
ents related to Profit-Sharing Investment Accounts (PSIA)
androfit-Sharing Investment Accounts-Restricted (PSIA-R),
Sukuk,usharaka, Salam, Istisnaa, and Qard-al-Hassan.3
Despite the apparent differences in the business models, therere
no studies that have investigated the impact of the NSFR onhe
stability of Islamic banks. From the sparse literature related tohe
stability of Islamic banks Cihk and Hesse (2010) compared
thenancial stability of Islamic banks with that of conventional
bankssing a data set of 19 countries and found that small Islamic
banksre financially more stable than large Islamic banks. However,
theylso noted the limitations of their research findings regarding
lim-ted availability of product-specific data on Islamic banks
financialtatements. The challenge when using standard harmonized
dataor empirical analysis is that it not only assumes that Islamic
andonventional banks share the same traits but also that their
stabil-ty is affected by similar factors. However, there are
clearly someifferences in the financial indicators of Islamic
banks. Examplesay include the nature of contracts underlying
Islamic financial
roducts, recognition of income from intermediation activities,
andomputation of capital.
This paper explores the potential impact of the IFSBs new
reg-latory measures by calculating the NSFR using existing data
and
inking it to the financial stability of Islamic banks. It
extends theork of Cihk and Hesse (2010) on stability and explores
whether
he IFSBs proposed NSFR has any potential impact on the
stabilityf Islamic banks.
For empirical estimations, this study utilizes a unique
datasetollected from the financial statements of 136 Islamic banks
from0 jurisdictions for the period 20002013. This dataset enabless
to capture the financial position of Islamic banks based on
thenderlying contracts as per the IFSBs guidelines. To the best of
theuthors knowledge there are no published studies that have used
aataset based on bank-specific variables, or have utilized the
IFSBsuidelines, to compute the NSFR for Islamic banks.
The empirical findings suggest that the NSFR measure intro-uced
by the IFSB for Islamic banks has a positive impact on thenancial
stability of Islamic banks during the sample period. How-ver, the
marginal impact of the NSFR on stability diminishes ashe size of
the bank increases. These findings remained robust aftersing an
alternative measure of financial stability and using anlternative
estimation model based on an instrumental variablepproach.
The findings of this study have significant policy
implicationsncluding the validation of the new regulatory
framework. If Islamicanks adopt the IFSBs recommended NSFR their
stability will benhanced. However, our findings also indicate that
banks operat-ng under a fully Islamic banking system are less
stable comparedo banks in a mixed banking system. We ask that these
findingse considered cautiously as countries with fully Islamic
bankingystems (Iran and Sudan) were subject to political and
economic
ifficulties during the time period under examination that
mightave biased the estimation results.
3 A detailed discussion on the difference between the BCBS and
the IFSBs pro-osed NSFR measures is provided in Section 2 of this
paper.
Fig. 1. Islamic banks share of total banking assets by
jurisdiction.Source: Islamic Financial Services Industry Stability
Report (2015b)
The rest of this paper is organized as follows. The next
sectiondescribes the background, methodology and calculation of the
NSFRfor Islamic banks. Section three describes the hypothesis,
modeland variable development utilized in this study. Section four
pro-vides a rationale and discussion on the sample, data and
univariateanalysis used in this research with section five
providing regressionresults and section six provides robustness
checks of the empiricalfindings. Section seven will provide
concluding comments.
2. Net stable funding ratio for Islamic banks
In 2014, almost 80 percent (USD 1.63 trillion) of the
globalIslamic finance industrys assets were Islamic banking
assets(Thomson Reuters, 2015). The Islamic banking industry grew
atabout 17 percent annually between 2008 and 2013. In compari-son,
the top 1000 global banks grew by only 4.9 percent in 2012and 0.6
percent in 2013. With such rapid growth, several Islamicbanks have
become systemically important banks especially inthose economies
where Islamic banks account for over 10 percentof total bank
assets.4
Fig. 1 depicts the share of Islamic banking assets as compared
tototal banking assets in countries where the Islamic banking
sectorhas a sizeable presence. Iran and Sudan have a fully Islamic
bank-ing sector and thus 100 percent of bank assets are with
Islamicbanks. Other countries with a significant Islamic banking
sectorinclude: Saudi Arabia with 51.3%, Brunei with 41%, Kuwait
with38%, Yemen with 27.4%, Qatar with 25.1%, Malaysia with 24%,
andthe UAE with 17.4% of total domestic banking assets invested
withIslamic banks. Bangladesh, Jordan, and Pakistan are also in
dou-ble figures in terms of the percentage of total domestic bank
assetsinvested with Islamic banks. These figures highlight the
importanceof the Islamic banking sector in countries where the
majority of thepopulation is faith-adhering Muslims.
To support the Islamic banking sector legal and
regulatoryframeworks specifically designed to cater the needs of
Islamicbanks have been created. Examples include the Malaysian
Islamic
Financial Services Act 2013 which provides a legal foundation
forthe Islamic banking system to shift towards a regulatory
frame-work that caters to the needs of specific types of Sharah
contracts.
4 Islamic Financial Services Industry Stability Report
(2015b).
-
5 nancia
LgcmabcoI
Iaisli
iwaa((v(alrtb
titiIiL
ttectr
ldldtmue
tih
0 D. Ashraf et al. / Journal of Fi
ikewise, the State Bank of Pakistan launched a five-year
strate-ic plan and is finalizing details on an Islamic liquidity
frameworkonsisting of an Islamic inter-bank money market (IIMM) and
audarbah-based placement facility run by the central bank. Oman
nd Qatar recently set up a separate banking system for
Islamicanks that does not allow conventional banks to offer Islamic
finan-ial products through Islamic windows. Turkey is also in the
processf developing legal and regulatory frameworks to enhance
the
slamic banking industry.One of the major differences between
conventional banks and
slamic banks is the way assets and liabilities are structured.
Thessets and liabilities of conventional banks are structured as
debtnstruments while the assets and liabilities of Islamic banks
aretructured in more equity-like instruments. Aside from
benevolentoans (Qard-al-Hassan) the assets of Islamic banks can be
dividednto three broad categories:
Equity-like assets including partnership instruments
(Musharakaand Mudarbah),Leases (Ijarah) and,Debt-like instruments
including deferred delivery-of-products(salam for basic products
and istisnaa for manufactured or con-struction projects) and sale
plus markup (Murabahah).
The major difference between conventional and Islamic banksn
terms of liabilities is in the nature of deposits. In
comparison
ith conventional banks, deposits of Islamic banks have
guar-nteed Sharah safekeeping deposit contracts (Qard-al-Hassannd
Wadiah), non-guaranteed Sharah contracts for investmentMudarbah and
Wakalah), profit sharing investment accountsrestricted and
unrestricted), and sukuk (Islamic equivalent of con-entional
bonds). The nature of profit sharing investment accountsPSIAs) also
differs from that of conventional bank deposits. PSIAsre more like
mutual fund investments where investors bear theoss when there is a
deterioration in the value of an investment foreasons other than
management negligence. This equity-like struc-ure of liabilities
provides an extra layer of protection to Islamicanks especially
during market down turns.
To cater for the specific regulatory needs of Islamic banks,he
IFSB provides a regulatory framework for the Islamic bank-ng
industry. To remain aligned with the global banking industryhe IFSB
endorsed the Basel III regulatory framework after mak-ng
adjustments for the different nature of assets and liabilities
ofslamic banks. The IFSB issued several standards including the
Cap-tal Adequacy Framework (IFSB-15), and the Guiding Principles
oniquidity Risk Management (IFSB-12) for Islamic banks.
Like its conventional counterpart, the NSFR requirement underhe
IFSBs guidelines is the ratio of available stable funding (ASF)o
required stable funding (RSF). However, there are clear differ-nces
in the computation of ASF and RSF between Islamic andonventional
banks. Table 1 presents the major differences in thereatment of
assets and liabilities under both approaches and theirespective
haircuts for the calculation of the NSFR.
The difference in treatment of various categories of assets
andiabilities under the IFSB-NSFR and the BCBS-NSFR is due to
theistinctive nature of assets and liabilities of Islamic banks. On
the
iabilities side of an Islamic bank, current account deposits
andeposits under Qardal-Hassan or wadiah are debt in nature
similaro their conventional counterparts and hence need similar RSF
to
eet both the IFSB and the BCBSs requirements. However,
depositsnder profit sharing arrangements have clear differences
underach banking systems.
Islamic banks Profit Sharing Investment Accounts (PSIA) havewo
categories. One is restricted under which funds are provided
bynvestors and are invested per the account holders instructions
andence are quasi equity. The second category of PSIA is
unresticted
l Stability 25 (2016) 4757
in which account holders give banks full discretion to invest in
anySharah-compliant assets and these accounts may be comingledwith
shareholders capital. The IFSB requires a haircut of 90%95%for
unrestricted PSIA. The BCBS for conventional banks however,does not
differentiate savings deposits and requires a flat haircutof 95%
when calculating the NSFR.
A major source of difference between the BCBS and the IFSBsNSFR
is in regard to the treatment of assets. There are productsoffered
by Islamic banks which are so unique that there are nocounterparts
offered by conventional banks. One such example isMusharaka
products which are based on partnership principles.These are not
the same as conventional banks loans and advances.Under the BCBS
conventional banks are required to provide 85%stable funding
against such loans and advances but due to thepartnership nature of
the Musharaka Islamic banks are requiredto provide just 50% stable
funding as it partially qualifies as equity(risk-sharing
feature).
Another major difference under loans and advances is
Qard-al-Hassan. It is rare for conventional banks to offer such
loans andadvances to customers. However, if it does then the BCBS
requires85% stable funding. Islamic banks offer Qard-al-Hassan more
fre-quently and these loans are based upon Hassanah and the
IFSBrequires no stable funding against these loans. Sukuk, Salam
andistisnaa are unique products provided by Islamic banks. The
BCBSdoes not have any rules regarding these products but the IFSB
pro-vides for an 85% haircut for these products. Ijarah contracts
are likemortgages and leases offered in the conventional banking
system.The BCBS allows for a 50%65% haircut while the IFSB requires
50%stable funding for all Ijarah contracts.
For the calculation of RSF utilizing the IFSB guidelines, assets
andliabilities are categorized into buckets depending on their
liquid-ity. Every bucket has a haircut depending on the relative
liquidityof the asset. These haircuts range from 0% for cash
(highly liquid) to100% for fixed assets (highly illiquid). On the
other hand, haircutsalso apply to funding sources to calculate ASF.
These haircuts rangefrom 100% for regulatory capital
(non-returnable equity financing)to 0% for Sharah -compliant
hedging instruments.
Although quantitative guidelines issued by the IFSB are
compre-hensive, there are clear limitations in calculating the NSFR
usingpublically available data. The International Monetary Fund, in
itsApril 2011 Global Financial Stability Report, highlights data
issuesthat are challenging when calculating the NSFR. Most of the
stud-ies analyzing the impact of NSFRs usually apply an
approximationsapproach on the application of haircuts to various
components ofthe balance sheet when calculating the NSFR under
Basel III (King,2013; Distinguin et al., 2013; Yan et al., 2012).
These assumptionsare generally in line with the broader
interpretation of various bal-ance sheet items and are based on the
liquidity and maturity ofassets and liabilities. Similar to the
conventional approach, we usea modified IFSB approach for the
computation of the NSFR thattakes into consideration several
assumptions about maturity andliquidity in assigning haircuts as
shown below:
NSFRit =ASFitRSFit
(1)
where ASFit is the sum of
100% values of total shareholders capital (tcapit) and
mudarbahinvestment accounts (mud invit), and
50% of Mudarbah savings (mud savit), current savings
(cnsacit)
and other accounts (oth depit) that are not profit and loss
sharing.
RSFit is the sum of
-
nancia
)]2)
thtd
3
3
uvnhasic
S
tsaia
3s
vrici
aB
Zdv
D. Ashraf et al. / Journal of Fi
50% values of outstanding financing contracts based onMusharaka
and diminishing Musharaka (mshit), leasing (ijarah),hire purchase,
and ijarah muntahia bittamleek (ijait),65% of investment in
companies, funds, shares (secit) and invest-ment in Islamic bonds
(invit),85% in Murabahah, deferred sales and murabahah for
purchaseorders (mrbit), Istisnaa and parallel istisnaa (istit) and
all otherfinancings (othit), and100% of the fixed assets (net of
depreciation) (fxdit) and balanceswith banks and other institutions
(ofiit).
Mathematically the above is shown as:
NSFRit =[(tcapit + mud invit) + 0.5(mud savit + oth depit +
cnsavit)]
[0.5(mshit + ijait) + 0.65(secit + invit) + 0.85(mrbit + istit +
othit) + (fxdit + ofiit(
We hypothesize that a higher NSFRit contributes positively tohe
overall stability of Islamic banks since this ensures that theyave
more available funds than is required. The following sec-ion
develops the empirical model used to test this hypothesis
andevelops control variables for empirical analysis.
. Determinants of bank stability
.1. Stability measure
Most of the empirical literature on financial stability of
banksses Z-score as a tool for the assessment of individual bank
insol-ency risk and financial stability.5 Mathematically, it
measures theumber of standard deviations of a banks
return-on-assets it wouldave to fall to deplete the sum of its
equity and income. Z-score hasdvantages over other accounting-based
financial stability mea-ures due to its capability to capture both
interest and fee-basedncome streams. Following Lepetit and Strobel
(2013)6, Z-score isalculated as:
TBLit =E(ROA)it + CARit
(ROA)i(3)
E(ROA) it is the expected return on assets, CAR it is equity
capital-o-asset ratio and (ROA) i is the volatility of
return-on-assets,ubscript i and t refers to bank and time
respectively. As it is widelyrgued in the literature that Z-score
is highly skewed we usedts logarithmic transformation in all
empirical estimations (Laevennd Levine, 2009; Schaeck and Cihk
2012).
.2. Other control variables influencing the stability of banks
andtable funding adjustments, and covariate definitions
Existing empirical literature provides a number of
explanatoryariables as to why banks may adjust their portfolio risk
to meetegulatory requirements. The explanatory variables are
dividednto two broad categories of bank-specific and
industry-specific
ovariates. In the next sub sections, the rationale for each
covariatencluded in the empirical model is considered in
detail.
5 See for example Boyd and Runkle (1993), De Nicolo (2001),
Stiroh (2004), Stirohnd Rumble (2006), Laeven and Levine (2009),
Demirgc -Kunt and Huizinga (2010),arrell et al. (2010), and De Haan
and Poghosyan (2012).6 Lepetit and Strobel (2013) compared various
methods used for calculating
-score. They suggest that an alternative measure that uses mean
and standardeviation of the return-on-assets calculated over the
full sample period and currentalues of the CAR ratio is more
robust.
l Stability 25 (2016) 4757 51
3.2.1. Bank specific variablesAmong bank-specific variables, the
size of the bank (too big
to fail) significantly influences the composition of assets and
ulti-mately the risk-taking behavior of the bank (Schwerter,
2011).Furthermore, larger banks can maintain higher liquidity
levels dueto easier access to the lender of last resort and would
be the first tobenefit from this safety net (Distinguin et al.,
2013). Similarly, largerbanks enjoy better franchise value and can
use diversification as atool for risk management (Demsetz and
Strahan, 1997). We mea-sure SIZEit as the natural log of total
assets. A negative coefficientwith SIZEit indicates the too big to
fail phenomena while a positivecoefficient with SIZEit reflects the
impact of higher franchise value,better risk management systems,
and easier access to the lenderof last resort.
A banks stability is also a function of its income sources.
Incomesources for banks have changed considerably over the past
coupleof decades. Busch and Kick (2009) concluded that fee income
ismore stable for commercial banks in Germany from 1995 to
2007.However, income diversification has been identified as one of
themajor factors that may contribute to the fragility of banks
(Ashrafet al., 2016; Khler, 2015; Ashraf and Goddard, 2012; Demirgc
-Kunt and Huizinga, 2010). NONIIit is the ratio of income from
fee-based activities to total assets. A positive relationship with
STBLitimplies diversification benefits for banks.
Bank profitability is another important driver of bank
stability.Financial institutions with strong operational
profitability enjoystable income streams (King, 2013; Jiraporn et
al., 2014; Hong et al.,2014). We use the ratio of net income to
total assets (NITAit) asour measure of profitability. We anticipate
a positive coefficient ofprofitability with bank stability.
Beck et al. (2013) reported that Islamic banks are generally
lessefficient in terms of cost efficiency compared to conventional
banks.They attributed this inefficiency to the level of maturity,
sophistica-tion, and competitive behavior of Islamic banks. We use
the inverseof the cost-to-income ratio (EFFit) to control for
efficiency. A pos-itive coefficient would imply that higher
efficiency helps Islamicbanks to become more stable.
3.2.2. Country-specific control variablesThe economic outlook of
a country plays an important role in
the stability of its financial institutions. Credit demand by
corpo-rations and credit supply by the financial sector shrinks
during adown turn in the economy resulting in a poor performance in
thefinancial sector (Lowe and Rohling, 1993). Literature suggests
thatthe financial performance of a bank is influenced by business
cycles(Laker, 1999; St. Clair, 2004; Jokipii and Monnin, 2013). To
controlfor the impact of business cycles we use GDP growth (GDPjt)
as amacroeconomic control variable.
The banking industry has been transformed considerably overthe
past couple of decades due to the deregulation of banking
activ-ities, financial innovation, and technical advancements. This
hasled to higher merger and acquisition activities and hence
competi-tion within both the domestic and international banking
industries(Goddard et al., 2007). Higher competition within the
banking sec-tor may lead to higher risk-taking (Boyd et al., 2006;
Uhde andHeimeshoff, 2009). To control for the impact of competition
on bankstability, we used 5-banks asset concentration (CONCjt) as a
proxyfor competition.
Religiosity in a society could affect the stability of financial
insti-tutions. We added a new country-specific variable to
determine ifthere is any potential impact of religiosity on the
financial stability
of Islamic banks. Ahmed and Gouda (2014) developed a
consti-tutional religiosity index that measures the extent to which
theconstitution of a country is developed under Islamic
fundamen-tal principles. We used the constitutional religiosity
index (RELGjt)
-
52 D. Ashraf et al. / Journal of Financial Stability 25 (2016)
4757
Table 2Descriptive statistics of all the non-dummy variables
used in this study. Definitions are in column 2. Data is sourced
from the Islamic banking information system (IBIS). Thetime period
is from 2000 to 2013.
Variable Definition Obs Mean Std. dev. Min Max
STBLit Financial stability 1226 2.32 0.96 0.16 3.82NSFRit Net
stable funding ratio 1234 1.01 0.34 0.20 1.78SIZEit Log of total
assets 1234 14.02 1.95 2.00 18.48NITAit Net income/total assets
1234 0.01 0.02 0.11 0.05NONIIit Non-interest income/total assets
1234 0.01 0.01 0.00 0.04EFFit Bank efficiency 1182 3.94 4.48 1.02
19.14
0.05 0.04 0.15 0.2669.42 31.04 0.81 100.00
9.88 8.02 0.00 26.00
cr
wivoi
4
isItcdtt
citIiaIlduI2aI
btIdWrtpbtt
Table 3Pairwise correlation matrix of all the non-dummy
variables used in this study.Definitions are in column 2 of Table
2. Data is sourced from the Islamic bankinginformation system
(IBIS). The time period is from 2000 to 2013.
STBLit NSFRit SIZEit NITAit NONIIit EFFit GDPjt CONCjt
RELGjt
NSFRit 0.060** 1SIZEit 0.096** 0.197***1NITAit 0.116***
0.134***0.137*** 1NONIIit0.237***0.046*** 0.310***0.180***1EFFit
0.023 0.159***0.179***0.178***0.246***1GDPjt 0.008 0.068** 0.008
0.190***0.039 0.064** 1CONCjt0.082** 0.122*** 0.369***0.001
0.232***0.122***0.032 1RELGjt 0.047* 0.163***0.319***
0.184***0.0010.122***0.0110.618***1* p < 0.1.
GDPjt Gross domestic product-growth rate 1219 CONCjt 5-Banks
assets/bank industry assets 1203 RELGjt Index of constitutional
religiosity 1234
reated by Ahmed and Gouda (2014) in our model to control
foreligiosity.
To control for the impact of the 2007-2009 global financial
crisise included a dummy variable that takes the value of one
dur-
ng 2007-2009, zero otherwise. We also use another
dichotomousariable (FULLjt) to control if conventional and Islamic
banks areperating side-by-side. FULLjt is equal to one if the
banking system
s fully Islamic or zero otherwise.
. Sample, data and univariate analysis
This section describes the sources of data used for our
empiricalnvestigation and its univariate analysis. We obtained our
financialtatement data from the Islamic Banks and Financial
Institutionsnformation usually referred as Islamic Banking
Information Sys-em(IBIS) database for all Islamic banks with at
least three yearsonsecutive data.7 IBIS collects and provides
financial statementata on Islamic banks based on underlying Islamic
financial con-racts for their assets and liabilities. Existing
studies that analyzedhe stability of Islamic banks have used the
Bankscope database.8
Although the Bankscope database is considered to be the
mostomprehensive database for banking it has some limitations whent
comes to Islamic banking (Cihk and Hesse, 2010). Limitationso
Bankscope data include differences in variable definitions byslamic
and conventional banks. For example, what is or is notncluded in
capital for Islamic banks or how to measure (the equiv-lent of)
interest income? Second, for the calculation of the NSFR ofslamic
banks, we need data on Islamic products based on under-ying Islamic
financial contracts and Bankscope does not have suchata on these
accounts in a long time series. For these reasons wesed the dataset
from IBIS as it is a database specifically designed for
slamic banks and have data on Islamic financial statements
since000. The use of the IBIS data set makes this study more
valuables it utilizes Islamic bank data measured and reported
through theslamic banks reporting framework and is thus more
reliable.
The initial data set for this study consisted of 173 Islamicanks
for which the financial data was available from IBIS duringhe
period 20002013. We then dropped all observations whereslamic banks
showed no deposits or financing activities. We alsoropped all banks
with less than three years of continuous data.e lost some
observations due to missing data or obviously incor-
ect data. In addition, we dropped data for banks from Iraq dueo
war and political instability during this time. After this
filteringrocess, we were left with an unbalanced panel data of 136
Islamic
anks from 30 countries with a total of 1226 bank-year
observa-ions. The data for macroeconomic variables was downloaded
fromhe World Bank website. Finally, due to the presence of large
out-
7 The data is available online at http://www.ibisonline.net.8
See for example Cihk and Hesse (2010).
** p < 0.05.*** p < 0.01.
liers, we winsorized all continuous variables at the 1 st and
99thpercentile.
Table 2 reports the descriptive statistics of the sample. The
meanof STBLit is 2.32 suggesting that an additional 43% of the
volatility ofreturn-on-assets is required to deplete the equity of
Islamic banks.Among explanatory variables Islamic banks have, on
average, NSFRof 1.01 suggesting that Islamic banks are sufficiently
funded overthe sample period. On average, the return-on-asset
(NITAit) forIslamic banks and income from fee-based activities
(NONIit) is 1%.The low ROA and income from fee-based activities can
be attributedto the fact that most Islamic banks started their
operations dur-ing the sample period and were investing heavily to
compete withconventional banks.
Generally, country-specific variables are within the normalrange
but do exhibit some differences. The growth in Gross Domes-tic
Product (GDPjt) shows a mean value of 5% however, countriesdo
reflect some degree of dispersion with a minimum negative of15% to
a maximum of 26%. The concentration of the banking sectoris
measured by 5-banks asset concentration (CONSjt). This showsthat,
on average, the five biggest banks hold 78% of the bankingsectors
total assets in those countries where Islamic banks oper-ate. On
average constitutional religiosity stands at 9.88% suggestingthat
countries that offer Islamic finance are more likely to follow
asecular constitution.
The pairwise correlation matrix for the main variables is
pre-sented in Table 3. Generally, the correlation is in line with
ourexpectations. Factors that can adversely affect the stability of
bankson a stand-alone basis include the NSFR, income from
fee-basedactivities, and the concentration of banking assets in the
five biggestbanks. Among the covariates that enhance the resilience
of banksare the size of banks and profitability.
4.1. Empirical methodology
The theoretical literature investigating the impact of
regula-tory requirements on the stability of banks use dynamic
models
http://www.ibisonline.nethttp://www.ibisonline.nethttp://www.ibisonline.nethttp://www.ibisonline.net
-
nancial Stability 25 (2016) 4757 53
tteTbccbTm
bpetfij
S
wacc
w(ip
bivfappmdmh
5
pdcit
a
b
c
tw
Table 4Estimation results based on a dynamic panel
autoregressive model as specified in Eq.(4). Definitions are in
column 2 of Table 2. Data is sourced from the Islamic
bankinginformation system (IBIS). The time period is from 2000 to
2013. Standard errors inparentheses.
(1) (2) (3)Variables STBLit STBLit STBLit
NSFRit 0.1304*** 0.1468*** 1.0869***
(0.0439) (0.0442) (0.2526)SIZEit 0.0448*** 0.0546*** 0.0053
(0.0098) (0.0102) (0.0191)NITAit 0.6666*** 0.6143** 0.4567*
(0.2544) (0.2533) (0.2558)NONIIit 3.3897** 4.2146***
4.0676***
(1.4690) (1.4801) (1.4797)EFFit 0.7226*** 0.7260***
0.6494***
(0.0810) (0.0826) (0.1897)Global financial crisis (dummy) 0.0019
0.0016
(0.0229) (0.0229)GDPjt 0.0805 0.0482
(0.2502) (0.2498)CONCjt 0.0066*** 0.0067***
(0.0019) (0.0019)RELG jt 0.0178 0.0186
(0.0124) (0.0122)Fully Islamic banking country (dummy) 0.7149***
0.7286***
(0.2238) (0.2202)NSFR SIZE 0.0674***
(0.0184)NSFR EFF 0.0480
(0.1536)Constant 3.0961*** 3.6964*** 2.8430***
(0.1630) (0.2664) (0.3482)Observations 1226 1180 1180Number of
banks 136 133 133
Wooldridge test for autocorrelation 250.031*** 242.726***
207.705***
Modified Bhargava et al. Durbin-Watson 0.740*** 0.774***
0.783***
Baltagi-Wu LBI test for autocorrelation 1.121*** 1.147***
1.151***
*
D. Ashraf et al. / Journal of Fi
o control for adjustment costs that banks may face when
movingoward target regulatory requirements (Ayadi et al., 2009;
Dahert al., 2013; Elizalde and Repullo, 2007; Naceur and Omran,
2011).he economic reasoning for using the dynamic model is that
bankehavior displays persistence. Bank management change their
poli-ies based on the current financial and regulatory environment
thatould potentially affect their future stability implying that
past sta-ility does affect present and future stability (Jahn and
Kick, 2012).he persistence in banking stability can be
incorporated, econo-etrically, by using a dynamic panel model.
To test the effect of the NSFR on the financial stability of
Islamicanks, we modeled the stability of Islamic banks using a
dynamicanel model assuming that lagged stability values may
partiallyxplain the subsequent behavior of variables over time. We
assumehat the stability of bank i is a function of its funding
stability, itsundamentals, and a host of country-level control
variables includ-ng economic, constitution, and competition
condition in country. The basic econometric specification is:
TBLit = + oSTBLit1 + 1NSFRit + Bit + Cjt + Dt + i + it(4)here
STBLit is a measure of financial stability of an Islamic bank
s measured by Z-score, NSFRit is the net stable funding ratio
cal-ulated using Eq. (2), Vector Bit, Cjt and Dt are observable
bank,ountry-specific, and dummy control variables respectively
andi+itis the error (idiosyncratic) terms so that:
it = it1 + it (5)here || < 1 and it is independent and
identically distributed
i.i.d.) with mean 0 and variance2 . Due to the presence of
time-nvariant covariates i is assumed to be a realization of an
i.i.d.rocess with mean 0 and variance 2 .
The dynamic panel data allows for possible endogeneityetween the
dependent and the explanatory variables character-
zed by autocorrelation due to the presence of a lagged
dependentariable. Bond (2002) suggests that a dynamic model is
preferredor panel data even if the coefficient of the lagged
dependent vari-ble is not of direct interest; allowing for dynamic
in the underlyingrocess may be crucial for recovering consistent
estimates of otherarameters. The endogeneity problem associated
with dynamicodels is dealt with in this paper using the Baltagi and
Wus (1999)
ynamic panel model following an AR(1) distribution. Further-ore,
standard errors are clustered at bank level to account for
eteroscedasticity and serial correlation of error terms.
. Regression results and discussion
Table 4 reports the results of Eq. (4) using the dynamicanel
model following an AR(1) disturbance explained earlier. Theynamic
panel model is suitable if the data series exhibit a
serialorrelation to the AR(1) order. To test for serial correlation
in thediosyncratic errors of a linear panel-data model we
performedhree different tests:
The Wooldridge (2002) test for serial correlation with the
nullhypothesis of no autocorrelation was conducted.
A modified Bhargava et al. (1982) test that provides the upper
andlower bounds for the Durbin Watson statistic. For a large
numberof cross sections, if it is less than 2 then it would
indicate a positiveserial correlation.
Baltagi and Wu (1999)s recommended autocorrelation test
forunequally spaced panel data.
Test results are reported in the final row of Table 4. All
threeests indicate the presence of first order autocorrelation or
in otherords, the existence of persistence in financial stability.
Under such
p < 0.1.** p < 0.05.
*** p < 0.01.
circumstances OLS estimates are not only inefficient but also
under-estimate the error variance resulting in inflated
t-statistics that maycause the erroneous rejection of null
hypothesis.
For empirical estimations, we used two different
specificationsof the model in Eq. (4). The first specification
includes bank-specificvariables and for the second specification we
added country-specific variables to the first specification. The
estimation results,reported in Table 4, are in line with
expectations. The null hypoth-esis of no association between NSFRit
and Z SOREit is rejected at5% significance level suggesting that
the maintenance of the NSFRrequirements has a positive impact on
the stability of Islamic banks.These results are in line with the
findings of Jiraporn et al. (2014)who argued that if the NSFR
requirements were implemented dur-ing 20052009 it would have
positively affected the stability of thebanking sector.
The coefficient of SIZEit is significantly negative which
suggeststhat Islamic banks smaller in size are more stable compared
tolarger Islamic banks. This finding is in line with Cihk and
Hesse(2010) who suggested that small Islamic banks were stronger
thanlarge Islamic banks in 19 banking jurisdictions from 1993 to
2004.This result may also reflect the challenge of risk-management
thatIslamic banks may face with the growth in bank size. From
theperspective of conventional banking this result is also in line
withthe findings of Demirgc -Kunt and Huizinga (2010) who foundthat
larger banks exhibited lower risk aversion over the period
19952007. Similarly Maudos and De Guevara (2011), after
exam-ining a large sample of EU, American, and Japanese banks
from20012008, concluded that size may have a negative, but not
linear,relationship with stability. This implies that for very
large banks an
-
54 D. Ashraf et al. / Journal of Financial Stability 25 (2016)
4757
terms
iemtw
aheutscI
abesctBrTi
Ifib
Fig. 2. The relationship of interaction
ncrease in bank size decreases the probability of bankruptcy.
How-ver, Hakenes and Schnabel (2011) argued that larger banks
areore stable due to their access to sophisticated
risk-management
ools and access to the lender of last resort. Our results are in
lineith the literature; larger banks demonstrate lower risk
aversion.
Profitability, as measured by the return-on-assets (ROA), is
neg-tive and significant indicating that Islamic banks, while
seekingigher profit margins, engage in high risk-taking operations.
How-ver, Hong et al. (2014) who linked profitability with failure
hazardsing call report data from US banks from 2001 to 2011,
concludedhat banks with higher profitability are more resilient to
short-termhocks and have less failure hazard. The differences in
our resultsan be attributed to the different nature of banks
(traditional versusslamic) and a different sample.
The measure of diversification, non-interest income to
totalssets, was significantly positive providing robust evidence
thatanks with diverse income sources (other than profit or
returnsarned from traditional intermediation activities) become
moretable. These results are in line with Busch and Kick (2009) who
con-luded that banks enjoy stability benefits with fee income as
thisype of income is more stable when compared to interest
income.ank efficiency, measured by taking payments to depositors as
aatio of total income from loans, positively affects bank
stability.his suggests that Islamic banks that pay higher returns
from their
ncome to their depositors are more stable.
The global financial crisis did not have any significant impact
on
slamic banks. Also, GDP growth has no significant impact on
thenancial stability of Islamic banks. Market competition,
measuredy 5-banks assets to total banking assets of a country, is
signifi-
of NSFR with SIZE and EFF with STB.
cantly negative in affecting the financial stability of banks.
Uhdeand Heimeshoff (2009), using aggregate data from the
EuropeanUnion banking sector, highlights the negative impact of
marketconcentration (a proxy for market competition) on financial
sta-bility. These results are in line with Vives (2011) who argues
thatthere are two possible ways in which higher levels of
competitioncan lead to banking instability. Firstly, by aggravating
the coordi-nation problem of depositors/investors on the liability
side andfostering runs/panics. Second, by increasing incentives to
engagein high risk activity ultimately results in an increased
probabil-ity of failure. Boyd and De Nicolo (2005) concur
suggesting thathigher competition within the banking sector may
lead to higherrisk-taking.
The coefficient of religiosity is not statistically significant
albeitpositive. The insignificance of the RELG is a little
surprising. How-ever, this can be explained by the fact that
countries with higherlevels of religiosity in their constitution
are still working under aconventional banking system and regulating
Islamic banks underthe same regulatory framework. Such an
environment does not sig-nificantly channel the potential positive
impacts of constitutionalreligiosity into the Islamic banking
system.
The dummy used to measure the impact of a fully Islamicfinancial
system is significantly negative. This result suggests thatIslamic
banks operating in a fully Islamic financial system are unsta-ble;
this may be because of limited investment avenues and Sharah
restrictions on doing business. We ask that these findings be
con-sidered cautiously as countries with fully Islamic banking
systems(Iran and Sudan) were subject to political and economic
difficulties
-
nancial Stability 25 (2016) 4757 55
dt
fittiametHvNsralmnosw
NmTfitort
iblo
6
6
atttiTCcb
svrocN
Table 5Estimation results based on modified z-scores.
Definitions are in column 2 of Table 2.Data is sourced from the
Islamic banking information system (IBIS). The time periodis from
2000 to 2013. Standard errors in parentheses.
(1) (2) (3)Variables STBLit STBLit STBLit
NSFRit 0.1962*** 0.2395*** 1.5982***
(0.0713) (0.0732) (0.3921)SIZEit 0.0466*** 0.0565*** 0.0313
(0.0156) (0.0158) (0.0299)NITAit 1.1368*** 1.1765***
0.9358***
(0.3233) (0.3249) (0.3286)NONIIit 8.0813*** 9.2564***
8.7601***
(2.2991) (2.3265) (2.3130)EFFit 1.2356*** 1.2624***
1.2277***
(0.1299) (0.1340) (0.3085)Global financial crisis (dummy) 0.0220
0.0185
(0.0410) (0.0407)GDPjt 0.7951 0.7218
(0.5157) (0.5112)CONCjt 0.0073** 0.0074**
(0.0032) (0.0032)RELG jt 0.0782*** 0.0795***
(0.0180) (0.0179)Fully Islamic banking country (dummy) 1.0061***
1.0308***
(0.3521) (0.3519)NSFR SIZE 0.1056***
(0.0308)NSFR EFF 0.0220
(0.2410)Constant 2.7940*** 2.8927*** 1.7373***
(0.2577) (0.4178) (0.5308)Observations 585 548 548Number of
banks 72 69 69
Wooldridge (2002) test for autocorrelation 157.870*** 136.620***
95.545***
Modified Bhargava et al. Durbin-Watson 0.881*** 0.945***
0.957***
Baltagi-Wu LBI test for autocorrelation 1.252*** 1.1303***
1.307***
D. Ashraf et al. / Journal of Fi
uring the time period under examination that might have biasedhe
estimation results.
The interaction between covariates may pose a challenge to
thendings of this study. Royston and Sauerbrei (2008) recommend
he use of a multivariable fractional polynomials interaction
(MFPI)echnique to cater for the interaction between pairs of
covariatesn the model. The MFPI is designed to investigate the
interactionnd statistical significance between each pair of
covariates in theodel whether continuous, binary or categorical.
Specifically, we
mployed Royston and Sauerbrei (2012) and identified the
interac-ion of the NSFR with SIZE and the NSFR with EFF as
significant at 1%.owever, the relationship of interaction pairs
with the dependentariable is not linear as depicted in Fig. 2. For
example, when theSFR is below the required 100% level smaller banks
show higher
tability as compared to larger banks. However, this relationship
iseversed once the NSFR reached the 100% levellarger banks show
greater level of stability as compared to smaller banks. The
non-inear relationship suggests that there is a possibility of a
negative
oderating role of SIZE with the NSFR-stability relationship.
Erro-eously assuming that the effect of SIZE is linear on the
stabilityf Islamic banks while estimated slopes of SIZE and NSFR
indicatestrong interaction between SIZE and the NSFR, and may lead
to arong inference.
To incorporate the impact of identified interaction termsSFR
SIZE and NSFR EFF we re-estimated the last empirical esti-ation by
incorporating these two additional interaction terms.
he empirical results are reported in column 3 of Table 4. The
coef-cient of NSFR SIZE is negative and significant suggesting
that
he marginal impact of the NSFR on stability diminishes as the
sizef the bank increases. This clearly indicates a cautious
approach isequired for cost-benefit analysis purposes when moving
above thehreshold level.
The coefficient of the second interaction term, NSFR EFF,
isnsignificant suggesting that the effect of the NSFR on financial
sta-ility is not influenced by a change in bank efficiency levels.
The
evel of significance and direction generally remain the same
amongther covariates.
. Robustness checks
.1. Alternative stability measure
Cihk and Hesse (2010), when comparing the stability of Islamicnd
conventional banks, suggested that more interesting informa-ion
lies in the downward spikes of ROAs (and Z-scores) than inhe
overall sample. However, the standard deviation underlyinghe
standard Z-score based on the overall sample gives only
partialnformation about the behavior of Z-scores as a measure of
stability.o take into account the downward spikes in ROAs (and
Z-scores),
ihk and Hesse (2010) suggested an alternative method for
theomputation of Z-score wherein capitalization and ROA are
dividedy the absolute value of the downward volatility of ROA.9
To further confirm our results, we calculated the modified
Z-core as suggested by Cihk and Hesse (2010) using the
downwardolatility of ROA. We estimated the dynamic panel model
andesults are reported in Table 5. Results remained robust in
termsf direction and significance. In fact, the magnitude of the
coeffi-
ient increased which suggests that the stabilizing function of
theSFR is more pronounced when considering downside risk.
9 Hesse and Cihk (2007) also used a modified z-score.
*p < 0.1.** p < 0.05.
*** p < 0.01.
6.2. Endogeneity
To address for endogeneity, we turn to instrumental
variabletechniques using a two stage least square (2SLS) estimator.
Weemploy the deposit-to-loan ratio, leverage ratio, and variability
inincome (standard deviation of return-on-assets) as instrumentsto
explain the NSFR in the first stage. All of these variables
areimportant measures for the funding stability of banks. A
higherdeposit-to-loan ratio determines the funding gaps and could
harmthe funding stability of banks in the case of bank runs. A
higherleverage ratio (long term liabilities to total assets)
determines howmuch funding is provided by long term investors.
Similarly, theprofitability of banks can provide a buffer to the
equity of banks.
The empirical results based on the 2SLS estimator are reportedin
Tables 6 and 7. The dependent variable in the case of Table 6is the
Z-score while in the case of Table 7, the dependent variableis
modified Z-score as discussed above. The main results
remainunchanged even after controlling for endogeneity. The NSFR
ispositive and significant in both tables while the interaction
term(NSFR SIZE) is negative and significant suggesting that the
NSFR,on average, enhances the stability of Islamic banks. However,
thestability enhancement function of the NSFR is more pronounced
insmaller banks.
Our results offer empirical support for the new regulatory
mea-sure of a NSFR introduced by the IFSB for Islamic banks as it
has apositive effect on bank stability. Our findings support the
view that
funding stability positively affects the soundness of Islamic
banks.However, the positive stability impact of the NSFR is not
uniform.Islamic banks that are smaller in size benefit more from
the newfunding requirements than larger banks. We ask that a
cautious
-
56 D. Ashraf et al. / Journal of Financia
Table 6Estimation results based on an Instrumental variable
model. Definitions are in col-umn 2 of Table 2. Data is sourced
from the Islamic banking information system (IBIS).The time period
is from 2000 to 2013. Standard errors in parentheses.
(1) (2) (3)Variables STBLit STBLit STBLit
NSFRit 0.7677*** 0.9265*** 1.3887***
(0.1495) (0.1541) (0.2871)SIZEit 0.0514*** 0.0700*** 0.0054
(0.0107) (0.0117) (0.0216)NITAit 0.4806 0.4793 0.3537
(0.3158) (0.3185) (0.2851)NONIIit 3.3079* 5.1771*** 3.9422**
(1.8253) (1.8909) (1.6947)EFFit 1.0253*** 1.0824***
0.8961***
(0.0976) (0.1018) (0.2046)Global financial crisis (dummy)
0.0582** 0.0432*
(0.0266) (0.0238)GDPjt 0.4759 0.0985
(0.3444) (0.2985)CONCjt 0.0088*** 0.0090***
(0.0020) (0.0018)RELG jt 0.0230* 0.0175
(0.0124) (0.0124)Fully Islamic banking country (dummy) 0.8570***
0.7848***
(0.2262) (0.2276)NSFR SIZE 0.0855***
(0.0202)NSFR EFF 0.0051
(0.1696)Constant 2.6547*** 3.3437*** 3.0280***
(0.2241) (0.3306) (0.3881)Observations 1226 1180 1180Number of
banks 136 133 133
* p < 0.1.** p < 0.05.
*** p < 0.01.
Table 7Estimation results based on an Instrumental variable
model. Dependent variable ismodified z-score. Definitions of other
variables are in column 2 of Table 2. Data issourced from the
Islamic banking information system (IBIS). The time period is
from2000 to 2013. Standard errors in parentheses.
(1) (2) (3)Variables STBLit STBLit STBLit
NSFRit 1.6352*** 1.6725*** 1.8659***
(0.3267) (0.3137) (0.4330)SIZEit 0.0070 0.0289 0.0240
(0.0234) (0.0235) (0.0336)NITAit 0.2011 0.0726 0.5004
(0.4655) (0.4561) (0.3436)NONIIit 9.5527*** 12.0512***
12.2696***
(3.1712) (3.1981) (2.5136)EFFit 1.6989*** 1.7638***
1.1351***
(0.1844) (0.1870) (0.3243)Global financial crisis (dummy)
0.0895* 0.0368
(0.0532) (0.0404)GDPjt 1.5729** 0.8542
(0.7424) (0.5621)CONCjt 0.0085** 0.0121***
(0.0036) (0.0030)RELG jt 0.0936*** 0.0772***
(0.0196) (0.0195)Fully Islamic banking country (dummy) 1.2353***
1.1842***
(0.3818) (0.3862)NSFR SIZE 0.1139***
(0.0331)NSFR EFF 0.1873
(0.2631)Constant 0.9250* 1.1148 2.0318***
(0.5498) (0.6973) (0.5825)Observations 585 548 548Number of
banks 72 69 69
* p < 0.1.** p < 0.05.
*** p < 0.01.
l Stability 25 (2016) 4757
approach be adopted when interpreting these results as a high
NSFRmay potentially affect the size of banks balance sheets as
banksmight forgo good investment opportunities trying to maintain
ahigh NSFR ratio.
7. Summary and conclusion
In the aftermath of the recent financial crisis (20072009)the
BCBS incorporated new changes in their regulatory frame-work and
proposed new checks on funding stability namely therequirement to
meet a Net Stable Funding Ratio. However, thestructural framework
and product nature of Islamic banks differsfrom the traditional
interest-based banking system. Therefore, itis not appropriate to
calculate the NSFR of Islamic banks in thesame way as traditional
banks. The Islamic Financial Services Board(IFSB), the standard
setting body for the Islamic banking indus-try, while endorsing the
Basel III accord modified the Net StableFunding Ratio (NSFR) to
cater for the unique aspects of the Islamicbanking industry.
We calculated the NSFR of Islamic banks according to the
IFSBsrequirements and investigated its potential link to the
financial sta-bility of Islamic banks from 30 countries. This study
used Z-score asa measure of a banks stability and the NSFR as a
tool to increase andstrengthen a banks stability. Our research
found robust evidenceto suggest that the IFSBs requirement of a
NSFR has a significantpositive effect on the stability of the
Islamic banking industry andthus adds to the growing body of
literature in favor of regulatoryframeworks introduced after the
global financial crisis.
Our research findings indicate that the NSFR has the capabil-ity
to increase the financial stability of Islamic banks by
reducingmaturity mismatch of assets and liabilities resulting in
improvedfinancial stability. However, the impact of stability is
not uniformamong all banks. This indicates a cautious approach is
needed whenviewing the NSFR as a tool to increase and strengthen
the stabil-ity of all Islamic banks. The results remained robust
after usingalternative stability measures and controlling for
endogeneity.
This study also opens up new avenues for future research.
Thestability-enhancing function of the NSFR is welcome however,
thequestion what does compliance with the NSFR mean for the
bal-ance sheet and the efficiency of Islamic banks still needs to
beaddressed.
Acknowledgments
We would like to thank Professor Kabir Hassan, Professor
Obiy-athullah Ismath Bacha, Professor Mansor Ibrahim, Dr.
AndreasJobst, the Managing Editor (Professor Iftekhar Hasan), and
twoanonymous referees for their helpful comments and
suggestionsthat enabled us to improve the quality of this paper
considerably.Useful comments by the participants of the Islamic
Finance Bankingand Business Ethics conference held in Lahore,
Pakistan on March2627, 2016 are also gratefully acknowledged. We
also wish tothank Mr. Zahid ur Rehman Khokher, Assistant
Secretary-General,Islamic Financial Services Board (IFSB) for his
valuable feedback onan earlier draft.
We also wish to thank Dr. Anis Ben Khedher, Information
Solu-tions Specialist, IRTI for his efforts in collecting and
ensuring theaccuracy of the data.
The authors acknowledge and are grateful for financial
supportand encouragement from the Islamic Research and Training
Insti-
tute, Jeddah, Saudi Arabia and the Deanship of Graduate Studies
andResearch at Prince Mohammad Bin Fahd University, Saudi
Arabia.
The views expressed in this paper are those of the authors and
donot necessarily reflect the views of the Islamic Research and
Train-
-
nancia
it
R
A
A
A
A
B
B
B
B
B
B
B
B
B
C
D
D
D
D
D
D
E
G
H
H
H
H
I
II
479497.Wooldridge, J.M., 2002. Econometric Analysis of Cross
Section and Panel Data. MIT
Press, Cambridge, MA.Yan, M., Hall, M.J., Turner, P., 2012. A
cost-benefit analysis of Basel III: some
evidence from the UK. Int. Rev. Financ. Anal. 25, 7382.
D. Ashraf et al. / Journal of Fi
ng Institute or the Islamic Development Bank Group. All errors
arehe responsibility of the authors.
eferences
hmed, D.I., Gouda, M., 2014. Measuring Constitutional
Islamization: The IslamicConstitutions Index. Hastings
International and Comparative Law Review,Available at SSRN
http://ssrn.com/abstract=2523337.
shraf, D., Goddard, J., 2012. Derivatives in the wake of
disintermediation: asimultaneous equations model of commercial and
industrial lending and theuse of derivatives by US banks. Int. J.
Bank. Account. Finance 4 (3), 250271.
shraf, D., Ramady, M., Albinali, K., 2016. Financial fragility
of banks: ownershipstructure and income diversification: empirical
evidence from the GCC region.Res. Int. Bus. Finance 38, 5668.
yadi, R., Llewellyn, D., Schmidt, R., Arbak, E., De Groen, W.P.,
2009. InvestigatingDiversity in the Banking Sector in EuropeKey
Performance and Role ofCooperative Banks. Center for European
Policy Studies, Brussels.
altagi, B.H., Wu, P.Z., 1999. Unequally spaced panel data
regressions with AR(1)disturbances. Econom. Theory 15 (6),
814823.
arrell, R., Davis, E.P., Karim, D., Liadze, I., 2010.
Calibrating macroprudentialpolicy. NIESR (September), 10.
eck, T., Demirgc -Kunt, A., Merrouche, O., 2013. Islamic vs.
conventional banking:business model, efficiency and stability. J.
Bank. Finance 37 (2), 433447.
hargava, A., Franzini, L., Narendranathan, W., 1982. Serial
correlation and thefixed effects model. Rev. Econ. Stud. 49 (4),
533549.
ond, S., 2002. Dynamic panel data models: a guide to micro data
methods andpractice. Port. Econ. J. 1, 141162.
oyd, J.H., De Nicolo, G., 2005. The theory of bank risk taking
and competitionrevisited. J. Finance 60 (3), 13291343.
oyd, J.H., Runkle, D., 1993. Size and performance of banking
firms. J. Monet. Econ.31, 4767.
oyd, J., De Nicolo, G., Jalal, A.M., 2006. Bank risk taking and
competition revisited:new theory and evidence. IMF Working paper,
WP/06/297.
usch, R., Kick, T., 2009. Income diversification in the German
banking industry.Bank. Financ. Stud., Discussion Paper Series 2, No
09/2009.
ihk, M., Hesse, H., 2010. Islamic banks and financial stability:
an empiricalanalysis. J. Financ. Serv. Res. 38 (23), 95113.
aher, H., Masih, M., Ibrahim, M., 2013. The unique risk
exposures of Islamic bankscapital buffers: a dynamic panel data
analysis. J. Int. Financial Mark. Inst.Money 36, 3652.
e Haan, J., Poghosyan, T., 2012. Bank size, market
concentration, and bankearnings volatility in the US. J. Int.
Financ. Mark. Inst. Money 22 (1), 3554.
e Nicolo, G., 2001. Size, charter value and risk in banking: an
internationalperspective. In EFA 2001 Barcelona Meetings, Available
at:.
emirgc -Kunt, A., Huizinga, H., 2010. Bank activity and funding
strategies: theimpact on risk and returns. J. Financ. Econ. 98 (3),
626650.
emsetz, R.S., Strahan, P.E., 1997. Diversification, size, and
risk at bank holdingcompanies. J. Money Credit Bank., 300313.
istinguin, I., Roulet, C., Tarazi, A., 2013. Bank regulatory
capital and liquidity:evidence from US and European publicly traded
banks. J. Bank. Finance 37 (9),32953317.
lizalde, A., Repullo, R., 2007. Economic and regulatory capital
in banking: what isthe difference? Int. J. Cent. Bank. 3 (3),
87118.
oddard, J., Molyneux, P., Wilson, J.O., Tavakoli, M., 2007.
European banking: anoverview. J. Bank. Finance 31 (7),
19111935.
akenes, H., Schnabel, I., 2011. Bank size and risk-taking under
Basel II. J. Bank.Finance 35 (6), 14361449.
asan, M.M., Dridi, J., 2010. The effects of the global crisis on
Islamic andconventional banks: A comparative study. IMF Working
Papers, 146.
esse, H., Cihk, M., 2007. Cooperative Banks and Financial
Stability. IMF WorkingPaper No. 07/02 (Washington: International
Monetary Fund).
ong, H., Huang, J.Z., Wu, D., 2014. The information content of
Basel III liquidityrisk measures. J. Financ. Stab. 15, 91111,
http://www.federalreserve.gov/pubs/IFDP/2000/689/ifdp689.pdf.
nternational Monetary Fund, 2011. Global Financial Stability
Report, Available
at:https://www.imf.org/external/pubs/ft/ar/2011/eng/.
slamic Financial Services Board, 2014. Guidance Note Number 12,
www.ifsb.org.slamic Financial Services Board, 2015a. Comparative
Study on the Implementation
of Selected IFSB Standards. Working Paper, www.ifsb.org.
l Stability 25 (2016) 4757 57
Islamic Financial Services Board, 2015b. Islamic Financial
Services IndustryStability Report 2015, www.ifsb.org.
Jahn, N., Kick, T., 2012. Determinants of Banking System
Stability: AMacro-prudential Analysis. Finance Center Mnster,
University of Mnster.
Jiraporn, P., Jumreornvong, S., Chalermchatvichien, P., 2014.
Basel III, capitalstability, risk-taking: ownership: evidence from
Asia. J. Multinatl. Financ.Manag. 28, 2846.
Jokipii, T., Monnin, P., 2013. The impact of banking sector
stability on the realeconomy. J. Int. Money Finance 32, 116.
King, M.R., 2013. The Basel III net stable funding ratio and
bank net interestmargins. J. Bank. Finance 37 (11), 41444156.
Khler, M., 2015. Which banks are more risky? The impact of
business models onbank stability. J. Financ. Stab. 16, 195212.
Laeven, L., Levine, R., 2009. Bank governance: regulation and
risk taking. J. Financ.Econ. 93, 259275.
Laker, J., 1999. Monitoring financial system stability. Reserve
Bank Aust. Bull.(October), 113.
Lepetit, L., Strobel, F., 2013. Bank insolvency risk and
time-varying Z-scoremeasures. J. Int. Financ. Mark. Inst. Money 25,
7387.
Lowe, P., Rohling, T., 1993. Agency costs, balance sheets and
the business cycle.Research Discussion Paper 9311, Reserve Bank of
Australia.
Maudos, J., De Guevara, J.F., 2011. Bank size, market power and
financial stability.In: Molyneux, P. (Ed.), Bank Performance, Risk
and Firm Financing. PalgraveMacmillan, Basingstoke, England, pp.
731.
Mergaerts, F., Vennet, V.R., 2016. Business models and bank
performance: along-term perspective. J. Financ. Stab. 22, 5775.
Naceur, S.B., Omran, M., 2011. The effects of bank regulations,
competition, andfinancial reforms on banks performance. Emerg.
Mark. Rev. 12 (1), 120.
Royston, P., Sauerbrei, W., 2008. Multivariable Model-building.
A PragmaticApproach Based on Fractional Polynomials for Modelling
Continuous Variables.John Wiley & Sons Ltd, London.
Royston, P., Sauerbrei, W., 2012. Handling interactions in
Stata, especially withcontinuous predictors. In United Kingdom
Stata Users Group Meetings 2012(No. 16) Stata Users Group.
Schaeck, K., Cihk, M., 2012. Banking competition and capital
ratios. Eur. Financ.Manag. 18 (5), 836866.
Schwerter, S., 2011. Basel IIIs ability to mitigate systemic
risk. J. Financ. Regul.Compliance 19 (4), 337354.
St. Clair, R., 2004. Macroeconomic Determinants of Banking
Financial Performanceand Resilience in Singapore, Monetary
Authority of Singapore Staff Paper, No.38.
Stiroh, K.J., 2004. Diversification in banking: is noninterest
income the answer? J.Money Credit Bank., 853882.
Stiroh, K.J., Rumble, A., 2006. The dark side of
diversification: the case of USfinancial holding companies. J.
Bank. Finance 30 (8), 21312161.
Thomson Reuters, 2015. State of the Global Islamic Economy
Report 2015/16,available at: http://halalfocus.
net/wp-content/uploads/2015/01/SGIE-Report-2014.
Uhde, A., Heimeshoff, U., 2009. Consolidation in banking and
financial stability inEurope: empirical evidence. J. Bank. Finance
33, 12991311.
Vives, X., 2011. Competition policy in banking. Oxford Rev.
Econ. Policy 27 (3),
http://ssrn.com/abstract=2523337http://ssrn.com/abstract=2523337http://ssrn.com/abstract=2523337http://ssrn.com/abstract=2523337http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0010http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0015http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0020http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0025http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0030http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0030http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0030http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0030http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0030http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0030http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0035http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0040http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0045http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0050http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0055http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0065http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0070http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0075http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0080http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0085http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0090http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0095http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0095http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0095http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0095http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0095http://refhub.elsevier.com/S1572-3089(16)30052-3/sbref0095http://refhub.elsevier.com/S1572-3089(16)30052-