Global Finance Journal: Sciencedirect

i An update to this article is included at the end
Global Finance Journal 46 (2020) 100485
Contents lists available at ScienceDirect
Global Finance Journal

journal homepage: www.elsevier.com/locate/gfj
Hyperbolic distance function, technical efficiency and stability to

T
shocks: A comparison between Islamic banks and conventional
banks in MENA region
Mohamed Chaffai
⁎
ARTICLE INFO ABSTRACT
JEL classifications: In this paper, we compare banking performance and resiliency between Islamic banks and
G21 conventional banks in MENA region over the period 2002–2014. Stochastic hyperbolic and radial
D22 output distance functions are employed to evaluate banking performances. Furthermore, hy-
D24 perbolic distance function methodology is extended to evaluate bank's business risk through si-
P52
mulations. Historical extreme observed situations in terms of profit loss from the industry are
used in order to simulate their impact on the overall profit distribution within the industry. More
Keywords:
precisely, we compare the observed profit and the simulated one followed by an abrupt fall in
Technical efficiency
Distance functions bank lending and non-lending activities. Results find evidence of technical efficiency differences
Business risk with bank type, some evidence with bank size. It is found that conventional banks are much more
Banks vulnerable to an important drop on their lending activities than non-lending activities, while
Islamic banks are equally vulnerable to any drop of the activity. Furthermore, Islamic banks are
found to be less vulnerable than their counterparts when they are exposed to shocks on their
lending activities. Very large banks are much more resilient than small banks whatever is the
bank stream. It is also found that bank vulnerability is more important when the banks are not
able to adjust their costs in the short run, and increases with the exposure to higher shocks in
magnitude.
1. Introduction
An important wave of the financial institutions efficiency literature has been interested in the comparison of the efficiency of
Islamic and conventional banks. Both by considering the experience of one specific country, a group of homogeneous countries or
even a larger sample of countries in the world, most of the scholars were motivated by the comparison of their business model
orientation. Several arguments are provided, Islamic financial institutions are still growing1 their market shares within some
emerging countries and also in Middle Eastern countries, several western banks offer to their customers Islamic financial products and
Islamic banks seem to be better capitalized and have higher assets quality which render them less vulnerable during the financial
crisis. While there is no theoretical argument favoring Islamic bank against conventional banks in terms of performance, for example,
Beck, Demirgüç-Kunt, and Merrouche (2013) argue that Islamic banks might have lower costs due to lower monitoring and screening
costs, but in the same time, their younger age, and the complexity of Islamic banking products may increase their costs. Most of this
⁎
Faculté des Sciences Economiques et de Gestion de Sfax, BP 1088, 3018 Sfax. Sfax University, Tunisia
E-mail address: Lep.chafai@fsegs.rnu.tn.
1
According to several studies, Islamic banking and finance is still growing compared to conventional banks, with an amount of assets close to 2
trillion in 2016, of which 80% is accounted by Islamic banks, Abedifar et al. (2015). The annual growth rate of Islamic finance is rapid 10–12%
annually over the last two decades compared to 6% for conventional bank assets, Bitar et al. (2017).
https://doi.org/10.1016/j.gfj.2019.100485
Received 6 June 2018; Received in revised form 25 June 2019; Accepted 25 June 2019
Available online 20 July 2019
1044-0283/ © 2019 Elsevier Inc. All rights reserved.
M. Chaffai Global Finance Journal 46 (2020) 100485
empirical literature uses either some financial ratios or focuses on more synthetic indexes related to bank efficiency measures in order
to compare the performance of the two bank groups. As we will see later and for several reasons, there is no consensus in this
empirical literature, whether one bank group outperform the other one. More recently, a handful of papers were concerned by the
link between bank resiliency and their performances, and seem to find evidence that Islamic banks are much more resilient because
they have higher assets quality or are better capitalized, Beck et al. (2013). Furthermore, Cihak and Hesse (2010) and Abedifar,
Molyneux, and Tarazi (2013) using the Z-score bank measure of soundness, found some link between size and bank status, Islamic
banks being more stable when operating on a small scale, but less stable when operating on larger scale. The Z-score bank risk
measure has the limit to be one-dimensional risk indicator, more recent literature construct multidimensional indicator based on
principal components analysis and compare it to bank profitability of Islamic and conventional banks, Bitar, Hassan, and Walker
(2017), Bitar, Hassan, Pukthuanthong, and Walker (2018) or its link with market structure, Ariss (2010) and capital structure, Bitar
and Tarazi (2019). Furthermore, within the empirical literature on Islamic banks and financial stability, Hasan and Dridi (2010) and
Beck et al. (2013) found evidence that Islamic banks are more resilient in terms of risk insolvency compared to conventional banks
during the last financial crisis. More closely to our work, Ghosh (2016) evaluates the impact of the Arab Spring political event on the
MENA bank performance and risk, he estimates its impact on bank profitability loss by 0.2%, while bank stability has declined by
0.4% on average. However, this study did not establish significant differences of these political events on the profitability or the
stability between Islamic and conventional banks.
The main objective of this empirical study is twofold. First, to conduct an empirical comparison of economic performances
between Islamic banks and commercial banks for a large sample including 16 countries in MENA region ranging from 2002 to 2014,
using radial and non-radial distance functions. In our comparison we consider several measures of bank performance, measures of
profit efficiency and revenue efficiency which will provide a more global view of the performance of these banks. It has been shown
that conventional banks which are efficient in controlling revenues are less efficient in controlling costs, Berger, Hunter, and Timme
(1993). Conventional measures like productive efficiency, cost efficiency, provide only a certain view of the efficiency of the bank
managers, while the profit efficiency measure is more adequate because it takes into account the ability of the managers to control
both costs and revenues. Unfortunately this last measure imposes a strong assumption on the behavior of the banks, i.e. profit
maximization. This assumption is problematic if it is not shared by all the banks within the sample under study.2 For example, the
case of public banks, which do not necessarily have this behavioral objective, as they may also have other stated social and economic
development objectives. Also, the estimation of cost or profit frontier assumes a precise measure of input and output prices, this is
particularly difficult to obtain for heterogeneous samples, particularly those based on Bankscope data. Within this framework,
Koetter (2006) compare cost and profit efficiency sensitivity to three prices definitions. He found that both bank efficiency levels and
also German banks' ranking are affected when employing alternative input prices.
Second, to provide some measures of banks resiliency associated to an abrupt shock on banks activities by bank status through
simulations. More precisely, we compare the resiliency of these banks to an adverse depreciation in their activities by evaluating and
comparing their business risk. Based on a recent methodology using directional distance function developed by Chaffai and Dietsch
(2015) for French retail banks, we build on a simulation model extended to hyperbolic distance function, in order to evaluate the
business risk of each bank type, i.e. conventional banks versus Islamic banks. The main idea of the model employed is to use historical
extremes abrupt fall within the sample in terms of profit loss in order to predict their potential impact on bank profitability when
these shocks entail a reduction in banks activities. It has been established in the empirical literature that several important shocks
may impact the profitability and the stability of the MENA banking sector. For example, Hasan and Dridi (2010) and Beck et al.
(2013) analyze the impact of the shocks coming from the global finance crisis on bank performance and stability, Poghosyan and
Hesse (2009) consider the direct and indirect effect of oil prices shocks on bank profitability in 11 oil-exporting MENA countries.
More recently Ghosh (2016) focus on the impact of an important political shock (the Arab Spring Events) on the profitability and the
stability of MENA banks. Even if all these studies compare conventional banks to their Islamic peers, the conclusion is usually based
on some regressions models, where the related shocks are proxied by dummy variables, and the conclusions are based on their
significance. The methodology used in this paper is quite different from these previous studies, like a stress test, we evaluate the
impact of an important shock on the banking system through simulations. Several scenarios are considered, first we evaluate bank
business risk when banks are allowed to adjust their costs or not. In addition, we evaluate the business risk by considering either the
impact of a sharp abrupt deterioration in lending, non-lending activities or both, this exercise is important to evaluate bank vul-
nerability with respect to each bank activity. Finally, we also consider the link between business risk and bank size.
We use distance function methodology which do not necessitate data on prices to measure banks performances. Distance function
methodology has been extensively used in the empirical literature evaluating organizations performance, but much more limited in
banking, see for example Cuesta and Orea (2002), Feng and Serletis (2010), Park and Weber (2006), Koutsomanoli Filippaki,
Margaritis, and Staikouras (2012). Most of the empirical studies evaluating bank performances use cost frontier models or to a lesser
extent the non-standard profit model. However, several distance functions have been used in the literature on efficiency measure-
ment, (i) input distance functions, (ii) output distance functions or (iii) directional and hyperbolic distance functions. The first model
provide an efficiency measure much more related to the cost efficiency, by measuring by how much the inputs could be deflated to
reach the efficient frontier while producing the same level of outputs. The second model provides an efficiency measure related to
revenue, and measure the expansion of the activities while using the same bank resources. Both models, are called radial Shephard's
2
Non-standard profit function could also be used when banks have some market power; in this particular case all the firms in the sample should
verify this assumption, which is a very restrictive assumption too. This model is the most commonly used in evaluating bank's performances.
2
(1970) distance functions. By contrast, directional and hyperbolic distance functions, much more recent compared to the previous
models, evaluate firms performance while allowing simultaneously both outputs expansion and inputs reduction, which offer a more
complete measure of performance related to profit. Directional distance function has been recently used to evaluate bank dynamic
performances by Park and Weber (2006) for Korean banks, Koutsomanoli Filippaki et al. (2012) for European banks, while Cuesta
and Zofio (2005) employ hyperbolic distance function to evaluate the efficiency of Spanish saving banks, Chaffai and Dietsch (2015)
use directional distance function to evaluate business risk of French retail banks. However, radial distance functions (i) and (ii)
remain the most frequently employed models by scholars to evaluate banks' efficiency. Studies using this methodology to evaluate the
performance of Islamic and conventional banks include Abdul-Majid, Saal, and Battisti (2010) and Johnes, Izzeldin, and Pappa
(2014).3 To our knowledge, using an efficiency measure related to both cost and revenues (iii) to evaluate bank efficiency between
Islamic and conventional banks is much more limited. Bader, Shamsher, Ariff, and Taufiq (2008)compare cost efficiency, revenue
efficiency and profit efficiency between Islamic and conventional banks for a panel data including 21 countries. Employing DEA
model to construct the frontier envelop, they find no evidence of overall efficiency difference between Islamic banks and conven-
tional banks. The main limit of this study is that DEA model uses proxy for price data in order to evaluate profit efficiency. So, using
hyperbolic distance function model to compare the profit efficiency between Islamic and conventional banks in the MENA region
without using price data would contribute to this empirical literature.
Our findings suggest significant differences in the technology used by the two bank groups, but there is evidence of the superiority
of conventional banks to Islamic banks in terms of profit and revenue efficiency, the difference is very small but statistically sig-
nificant. However, very large banks seem to be more efficient than small or middle size banks. Additionally, Islamic banks are found
to be much less resilient in terms of business risk than conventional banks to an important drop on their activities. The low resiliency
of these banks is found to be much more linked to non-lending activities compared to conventional banks, while large banks seems to
be more resilient than small banks.
The remainder of the paper is organized as follows: Section 2 presents a brief overview of the empirical literature and the
motivations of the paper, Section 3 deals with the methodology of the efficiency measure and the evaluation of business risk. Section
4 discusses the empirical findings. Section 5 reports various robustness checks of the results and Section 6 concludes.
2. Related comparative literature
There is a vast comparative empirical literature on bank performances. Most of these studies focus on bank ownership structure,
public banks, private banks, foreign banks. The comparison is made within a common country or across countries. The comparative
performance literature between Islamic banks and conventional banks follow this wave of banking comparisons and can be divided
into three groups according to the methodology used, i.e. financial ratio analysis (FRA), parametric4 and non-parametric frontier
models and principal components analysis (PCA). FRA, are much more popular, focus on some financial ratios, for example cost to
revenue ratio, return on assets, return on equity, etc. Using this framework, Bader et al. (2007) explore a larger sample including 43
Islamic banks and 37 conventional banks in three regions. Over the period 1990–2005 they compare banking efficiency through
several ratios dealing with cost revenue and profit, but they do not find evidence for the superiority of Islamic banks in terms of
performances. More recently, Beck et al. (2013) use FRA, cost to total assets ratios and cost to income ratios, they find strong evidence
that Islamic banks are less efficient than conventional banks even if they have higher intermediation ratio, higher assets quality and
are better capitalized which explain why Islamic banks performs better during crises, Beck et al. (2013).
In the same vein of research Hasan and Dridi (2010), also find superior performance of Islamic Banks during the crisis. These two
studies relate efficiency to bank resiliency. The second group of literature employing PCA analysis to compare financial performance
between Islamic banks and conventional banks is much more limited. Bitar et al. (2017) employ PCA to check the financial soundness
of conventional and Islamic banks of a large sample including 33 developing countries for the period 1999–2013. They find evidence
that Islamic banks are more efficient and more profitable, but less stable due to more volatile earnings compared to their counter-
parts. The third group of empirical literature employs more synthetic indexes of performances based on efficiency models to compare
Islamic and conventional banks. The empirical literature being voluminous, we only provide a sample of these studies related to the
Middle East region, see Johnes et al. (2014), Abedifar, Ebrahim, Molyneux, and Tarazi (2015) and Hassan and Aliyu (2018) for
relevant reviews. The pioneering empirical work is Al-Jarrah and Molyneux (2004) who find evidence for Islamic Banks superiority in
terms of cost efficiency in 4 countries but for an old period, 1992–2000 and a very limited sample of banks. In the same vein Ariss
(2007) compare cost efficiency on a larger sample and more recent data in three Gulf Cooperation Council Countries (GCC), she
establishes that Islamic banks are more cost efficient than conventional banks during the (1993–2003) period, the cost efficiency
being (88% and 74%) respectively. The comparison is based on a stochastic cost frontier model where off balance sheet is retained as
output. However, Srairi (2010) find that Islamic banks are much less efficient than conventional bank in the (GCC) over the period
1999–2007, the difference in efficiency being between 11% and 12%, on average. Both cost efficiency and profit efficiency (derived
from non-standard profit function) have been evaluated. Johnes et al. (2014) consider the case of a large sample including MENA and
5 Asian countries over the period 2004–2009. Using a non-parametric Meta frontier approach, technical inefficiency is then de-
composed into net technical inefficiency and technology inefficiency, they find no significant differences in gross efficiency between
3
For an exhaustive list of studies dealing with the comparison of the efficiency between Islamic and conventional banks, see Table 1, pages 28–29
in Abedifar et al. (2015), see also Table 5, pages 24–25 in the recent survey of Hassan and Aliyu (2018).
4
We will follow this approach in this paper.
3
Islamic and conventional banks but there are significant inefficiency differences between the bank types in the components. Ac-
cording to their results, Islamic banks perform better than conventional banks in terms of net technical inefficiency, but their
inefficiency linked to the technology is worse. Chaffai and Hassan (2019) attain recently similar conclusion when they estimate a
parametric meta cost frontier for 15 MENA country banks and decompose total cost inefficiency into managerial inefficiency and
technology gap. They show that Islamic banks are less efficient than conventional banks over the studied period 2002–2014, and that
technology plays a more important role in the global cost inefficiency of Islamic banks than does managerial inefficiency. Mobarek
and Kalonov (2014) consider a large sample, including 14 MENA countries over the period 2004–2009, they also find that Islamic
banks perform poorly in terms of cost efficiency compared to conventional banks, a difference which varies between 2% and 5% over
the studied years. In two other recent studies restricted to the Gulf Council Countries (GCC) and very close sampled period, Mohanty,
Lin, Aljuhani, and Bardesi (2016) compare cost and profit efficiency using stochastic frontier models but they do not find significant
differences in the efficiency across banks type. However, partially different conclusions are found by Alqahtani, Mayes, & Brown,
2017who employ DEA model to compare profit and cost efficiency in the GCC, they do not find significant differences between
Islamic banks and their counterparts in terms of cost efficiency but some differences in terms of profit efficiency in favor of con-
ventional banks. An important issue related to most of the empirical studies is that they did not establish a robust result on the
superiority of one type of banks against the other, a conclusion evidenced by most of the surveying empirical literature, Abedifar et al.
(2015) and Hassan and Aliyu (2018). It seems that the conclusions depend on the methodology used to evaluate bank's performance,
financial ratios or synthetic efficiency measures. The conclusions are also sensitive to the sample of countries considered or to the
efficiency measure employed i.e. parametric versus non parametric. Even with synthetic measure which is commonly used in the
empirical literature to evaluate bank performance, the results may depend on which type of efficiency is evaluated, technical effi-
ciency, scale efficiency, cost efficiency, profit efficiency and also on the definition of the outputs and the inputs retained. So many
considerations need to be kept in mind before taking a definitive answer on which bank category is performing better.
As it has been mentioned by Berger et al. (1993), some banks could be highly efficient in making profits, but are not necessarily
efficient in terms of costs. It has also been evidenced that banks which are using extra costs are those who accept to increase their
costs in order to compensate for extracting higher revenues or additional profits. Furthermore, comparing profit efficiency using
frontier models, assumes that banks have a common objective of maximizing profit in a competitive market. Efficiency is then
measured by estimating a standard profit frontier model. Sometimes, banks may have market power; Berger and Mester (1997)
suggest estimating a non-standard profit frontier which become very popular in estimating profit efficiency in banking, but
Kumbhakar (2006) warned researchers against using the nonstandard profit function due to some theoretical problem and econo-
metric model specification.
Another strand of literature, examines the links between the bank status and risk. This literature is much more limited compared
to bank performance comparison.5 Focusing on insolvency risk, Cihak and Hesse (2010) and Abedifar et al. (2013) employ Z-score
index to compare bank risk of Islamic and conventional banks. Cihak and Hesse (2010) show that the Z-score is on average sig-
nificantly much higher for Islamic banks compared to conventional banks and they conclude that Islamic banks are more stable. Some
differences across bank size are evidenced while small Islamic banks are found to be more stable than small conventional banks, the
conclusions are similar in the two studies. In another strand of the empirical literature there is some evidence of the high resiliency of
Islamic banks compared to conventional banks during the crisis. In this respect, Beck et al. (2013) consider a sample of 22 countries
where both Islamic and conventional banks operate over the period 1995–2009 and compare their business orientation. By focusing
on their efficiency and asset quality, they find no evidence of one bank type superiority against the other. However, when they
compare their business orientation during the historical crisis, by considering both country specific financial crisis and the global
financial crisis, they find strong evidence that Islamic banks are better capitalized, have lower non-performing loan ratios and have
higher assets quality compared to conventional banks. In the same strand of research, Hasan and Dridi (2010) examine the impact of
the last global financial crisis on the profitability, assets and credit growth, and external agency rating of Islamic banks and con-
ventional banks in 7 MENA countries plus Malaysia over the period 2007–2010. They find evidence that Islamic banks are on average
more resilient than conventional banks during the crisis. Finally, by considering the case of only one MENA country, and by applying
a stress test, Elsiefy (2012) assesses the resilience of the Qatari banking sector by considering three risk types, credit risk, interest rate
risk and foreign exchange rate risk by considering the case of 5 conventional banks and 3 Islamic banks over the period 2006–2010.
He finds evidence that Islamic banks are less resilient compared to conventional banks in this country. Moreover, his results show that
credit risk is the major source of vulnerability compared to the other two aforementioned risk types.
This paper builds on these two brands of literature by considering both efficiency and resiliency comparison between Islamic
banks and conventional banks. We consider a large sample of banks including 15 countries within MENA over the period 2002–2014.
First, bank efficiency measures are based on distance function models, hyperbolic distance function and radial output distance
functions. The first model measure by how much an inefficient bank could increase its activities while in the same time it could also
decrease its resources, a kind of profit efficiency measure. The second model allows only the improvement of the activities while still
using the same level of bank resources, a common measure related to revenue efficiency. These two models have the advantage by
providing measures of bank performances without imposing specific bank behavior assumption i.e. profit maximization or revenue
maximization and using dual profit or revenue functions. Moreover, distance function models require only data on input and output
quantities to evaluate banks performance. Unfortunately most of the aforementioned banking studies using cost frontier, non-
5
For an exhaustive list of studies dealing with insolvency and risk of Islamic and conventional banks, see Table 4 pages 21–23 in the recent survey
of Hassan and Aliyu (2018).
4
standard profit frontier or even DEA models to estimate profit efficiency use only proxy for inputs or outputs prices which render the
efficiency measures questionable. Second, resiliency measure retained for the comparison is a kind of business risk measure recently
proposed by Chaffai and Dietsch (2015) to evaluate the bank resiliency of French retail banks. These authors evaluate business risk
through simulations. Using frontier methodology, business risk is a measure of profit loss which would result from an important
adverse shock on lending and non-lending bank activities. More precisely, we compare the observed profit distribution and the
simulated one followed by an abrupt fall in bank activities. We evaluate the impact of the shocks according to two scenarios, when
the banks are able to contract their costs, or not.
3. Distance function methodology
We use two distance functions, the output distance function, Shephard (1970) and the hyperbolic distance function,6 Cuesta and
Zofio (2005) to model the production banking process, measure bank efficiency and check the resiliency of MENA banks. The two
proposed models offer a vision of the bank efficiency on conducting their business and measure by how much they could increase
their activities while using the same level of their resources (output distance function) and an alternative measure which allow the
contraction of the resources (hyperbolic distance function).
Fig. 1, illustrates these measures according to the two distance functions retained, taking the particular case of one output and one
input. Bank A inside the production possibility set below the curved line which represent the production frontier envelop, is tech-
nically inefficient. This bank could be projected on the frontier to derive a measure of its efficiency score in several ways. First, it can
be projected in A', by increasing its level of output while using the same input xA. Technical efficiency of this bank is yA/yA', this is
the output orientation measure of technical efficiency. Second, we can project the bank on A" along the hyperbolic path AA". In this
case, it will increase its efficiency by expanding its output and in the same time by contracting its input. Technical efficiency is
measured by the ratio yA/yA". Each distance function derives a specific level of inefficiency score, unless the bank is on the frontier
(100% efficient), the case of bank B. For the hyperbolic distance function, any improvement in technical inefficiency of bank A would
results in an increase of its profit coming from both the expansion of its revenues and the decline in its costs, while for the output
distance function, bank A will just improve its revenue, input being fixed. Notice that the output distance function is dual to the
revenue function while the hyperbolic distance function is dual to the profit function, Färe and Primont (1995) and Cuesta and Zofio
(2005).
Two commonly used methodologies are proposed to construct the frontier, parametric and non-parametric DEA methods. The first
approach allows considering random errors terms which are not under the control of the managers but the parameter estimates have
standard errors and statistical tests on the technologies can be derived. The second methodology uses linear programming methods
but do not allow such investigations. However, the parametric method needs to assume a particular functional form for the distance
function, while the non-parametric method is functional form free. We will retain the parametric approach in this study because it has
the merit to distinguish between random phenomena which are not under the control of the bankers from bank inefficiency which is
under their control. O'Donnell and Coelli (2005) notice that inefficiency level tend to be overestimated with DEA models because the
frontier is biased. This issue is particularly important in this paper when we construct our bank business risk measure. Also, it is much
easier to conduct usual statistical inference with the parametric approach and panel data structure.
3.1. The hyperbolic and the output distance function
The distance function is an interesting methodology to measure the efficiency of banking multiproduct outputs technology.
Consider a vector of M outputs produced Y = (y1, y2, …, yM) by a sample of banks which use a vector of K inputs X = (x1, x2, …, xK).
The production possibility set denoted T, is the set of all the combinations of outputs and inputs for which X can produce Y. For a
particular bank which belongs to this set, the distance function projects this point to the frontier and is defined by:
Y
D (Y , X ) = inf > 0, , X T
(1)
where δ is equal 1 for the hyperbolic distance function model, 0 for the output distance function model.
Following Cuesta and Zofio (2005) the stochastic distance along a particular path to the frontier can be represented by a flexible
translog functional form:
M M M k k k M k
1 1
Log (D ) = a0 + jLog (yj ) + jj Log (yj ) Log (yj )+ i Log (x i ) + ii Log (xi ) Log (x i ) + ji Log (yj ) Log (x i ) +v
j=1
2 j= 1 j =1 i=1
2 i=1 i = 1 j =1 i=1 (2)
v is the random error term, and Dδ is the unknown inefficiency term. The hyperbolic distance function should verify some
regularity conditions, the most important is the almost homogeneous property which suggests, that if the outputs are multiplied by a
certain proportion λ and the inputs deflated by the same proportion the distance to the frontier, i.e. the efficiency score remains
unchanged. For the output distance function case (δ = 0) linear homogeneity with respect to the outputs imply that the efficiency
score is the same if the outputs are multiplied by λ. According to this property, if we deflate the outputs and inflate the input by the
6
We do not use the directional distance model because it needs to fix a specific direction in order to construct the efficient frontier, hyperbolic
distance function does not impose such assumption.
5
Fig. 1. Distance functions and efficiency measures.
first output y1 retained as the numeraire:
Log (D / y1) = ln TL (yi / y1 , xj . y1 ) (3)
Given the homogeneity property, the estimable form of Eq. (3) is:
M M M k
1
Log (y1) = a0 + jLog (yj / y1 ) + jj Log (yj / y1 ) Log (yj /y1 ) + i [Log (x i ) + Log (y1 )]
j =2
2 j=2 j =2 i=1
k k
1
+ ii [Log (x i ) + Log (y1 )][Log (x i ) + Log (y1)]+
2 i=1 i =1
M k
ji Log (yj / y1 )[Log (x i ) + Log (y1 )] + v + u

j=2 i =1 (4)
where u = − Log(Dδ) ≥ 0, is the inefficiency term. Technical details are provided in Cuesta and Zofio (2005). Notice that Eq. (4) is a
common stochastic frontier model which could be estimated by maximum likelihood method, assuming a specific distribution for the
inefficiency component. Technical inefficiency component u is estimated according to the conditional method of Jondrow, Lovell,
Materov, and Schmidt (1982), see details in Kumbhakar and Lovell (2000). Notice also, that whatever is the numeraire retained to
deflate the outputs, the efficiency scores are the same unless the likelihood function does not converge.
3.2. The hyperbolic distance function and the evaluation of bank business risk
The methodology used by Chaffai and Dietsch (2015) for the directional distance function to measure bank business risk is
adapted to the hyperbolic distance function case. The main idea according to this methodology is to use historical extreme observed
situations in terms of profit loss from the industry in order to simulate their impacts on the overall profit within the industry. As
mentioned in Chaffai and Dietsch (2015), “shocks imply a decrease in profits which corresponds to an increase of the distance value.
In other words, for a given bank, and a given output's volume shock, the increase in distance represents the decrease in profits owing
to the decline in the outputs' volume”.
We illustrate in Fig. 2 the steps used to construct the business risk measure for the case of a frontier model with one output and
one input. Step 1 construct the initial frontier by projecting each firm on the frontier along the hyperbolic curve, bank A is projected
to A* on the frontier and efficiency scores are derived. In a second step, we apply a random shock on the activity of each bank, let for
example a reduction of the output by 20%, point A will move to point As, while bank B is moved to point Bs according to another
random shock. The shocks distribution is obtained by sample drawing in the first percentile (10%) of the inefficiency profit dis-
tribution obtained in Step 1, in other words we consider extreme historical and exceptional situations which impacted important loss
on the observed banks profit during the observed period 2002–2014. Step 2, the shocked points which are plotted with hollow circles
in the Fig. 2, are then used as new production possibility sets to estimate a new frontier, called shocked frontier. In Step 3, we
measure the distance between the two frontiers as a measure of business risk, the distance A* As⁎, the higher is the distance between
the two frontiers, the higher is the vulnerability of the bank to a hypothetical reduction in its activities. This exercise in Step2 and
Step3 are replicated several times by bank, the business risk measure being the upper percentile of this profit loss distribution (1% or
5%), which can be interpreted as an earning at risk measure, Chaffai and Dietsch (2015). This resiliency measure can be seen as a
kind of stress test which assesses the vulnerability of each bank to exceptional events.
6
Fig. 2. Hyperbolic distance function and bank resiliency measure.
Summing up, the simulations we perform proceed in the following steps:
Step 1
Estimate the hyperbolic distance function and derive the distance to the frontier for each bank. This step is also used to derive the
first decile of the inefficiency bank scores, within this sub-sample, only extreme historical shocks on the activities are retained for the
simulations. Notice, at this step we will pool the two bank categories in the MENA region, which mean that simulations of banking
shocks is as large as possible within the studied period 2002–2014, including the financial banking crisis period in 2008 and also the
economy drop of the countries involved in the Arab spring events of 2011.
Step 2
For each simulation experience, b = 1,2,…B, we draw a random sample of shocks in the first decile (Step 1), and for each bank we
reduce the outputs by the same amount. After constructing the new vector of shocked outputs we re-estimate the frontier. The
number of replications used is B = 1000.7
Step 3
The business risk is obtained by evaluating for each drawing and for each bank the distance between the two frontiers (the initial
frontier, estimated in Step 1, and the shocked outputs frontier, in Step 3), which provide a measure of business risk. More precisely we
retain in each simulation the upper percentile, (1%, 5%), the average values of the B percentiles is the value of the business risk
(earning-at-risk) measure.
Let us mention that we consider three cases: the first one is to apply a shock on total loans only; the second scenario considers the
shock on the other non-lending services (commissions, investments…), the business model retained considers two outputs produced.
Finally we evaluate the impact of the same shock on the two activities. This exercise is interesting because it allows estimating the
magnitude of a sudden deterioration of a specific activity on the business risk of each bank type so its vulnerability. Moreover, we
consider two illustrative scenarios: the first one allows banks which face a deterioration of its activities to adjust costs (hyperbolic
distance), the second scenario is to evaluate the bank resiliency where adjusting costs is not possible (output distance).
4. Empirical results
To conduct a comparative analysis, we consider a sample of 2743 observations in MENA region using Bankscope data. The sample
is very large and includes 325 banks in 16 countries over the period is 2002–2014. The country list (with the number of banks
between parentheses) is: Algeria (17), Bahrain (44), Egypt (37), Iraq (14), Iran (18), Jordan (15), Kuwait (17), Lebanon (57),
Morocco (13), Oman (10), Qatar (15), Saudi Arabia (19), Syria (15), Tunisia (14), United Arab Emirates (36) and Yemen (9). In total,
the sample is an unbalanced panel of banks which are classified in two categories according to their activities: conventional com-
mercial banks (2080 observations, 231 banks), and Islamic banks (663 observations, 94 banks). Whatever their stream, these banks
compete in the same markets, and have different market shares within each country. Also we retain the intermediation approach for
the definition of the outputs and the inputs to evaluate the efficiency performances and the resiliency of these banks. This choice is
7
In fact the number 1000 is the total net number of simulations for which the likelihood function converge. Notice even if we consider a much
higher number of replications 2000, the results prove robust.
7
dictated by the lack of more detailed data on the number of accounts or the number of loans by bank, and also by the fact that most of
the comparative studies between banks use this approach, so our results could be compared to most of these studies. Furthermore, we
assume that the banks use three inputs, labor measured by personnel expenses,8 physical input measured by book value of fixed
assets, and financial input measured by interest expenses to produce two kinds of outputs, lending activities measured by total loans,
and all other financial services (commissions, trading, investments), measured by other earning assets. All the monetary variables
have been deflated by each country price index, 2000 being the base year. Table 1, provides the descriptive statistics pooled for all
years and banks by bank type. On average sampled Islamic banks produce more loans but less other non-lending services compared to
conventional banks. However, there are no significant differences in terms of physical input but average labor costs are lower for
Islamic banks. The average size in terms of total assets of Islamic banks is lower (40%) than commercial banks. Preliminary efficiency
ratios and financial ratios, income ratio, average cost, and profit to total assets ratio suggest that Islamic banks have quite similar
profitability ratio and revenue ratio, but they have higher costs in average. In the same time, Islamic banks in the sample have higher
equity ratio and higher Z-score which suggests that the sampled banks have on average lower risk of insolvency. Overall, even if our
sample is smaller (16 countries), these results are in line with Beck et al. (2013), (their sample includes 22 countries, of which 8
countries are not from MENA).
In order to compare bank efficiency according to their type, we estimate both output oriented distance function and hyperbolic
distance. In other words, banks which are efficient and active in providing maximum services are not necessarily producing them
with great vigilance with respect to their costs. The two definitions are retained in particular to evaluate the impact of gross shocks on
banks activities with respect to their income and also their profit. So it is important to test for the assumption of a common tech-
nology according to bank type.
We have estimated stochastic hyperbolic and output distance functions according to Eq. (4), the model includes country dummy
variables to take account of environmental and regulation differences and the trend to take account of the shift of the frontier due to
technical progress. In the stochastic specification, we assume that the inefficiency term follows a half normal distribution which is
one of the most commonly used distributions in stochastic frontier modeling. Likelihood ratio test is conducted in order to test for the
assumption of a common technology. Under the null hypothesis, the two banking streams are assumed to share a common tech-
nology, i.e. have the same frontier, while under the alternative the technologies of each bank type are different. We estimate a
stochastic frontier for the pooled sample, (under the null), and then a stochastic frontier by banking stream, then we derive the LR test
statistics. Results presented in Table 2, reject the common technology assumption at the 99% level of significance, so the technologies
used are likely to be different across the two bank groups. In other words the technology used to transform resources into financial
services differ between Islamic and conventional banks. Moreover, this result is robust to the model retained to represent the
technology, i.e. the specification of the distance function, and also when we replace the time trend by temporal dummies. Meta
frontier models can be used to decompose global efficiency scores into the technology gaps and net efficiency measures, Johnes et al.
(2014), Bos and Schmiedel (2007), Chaffai and Hassan (2019), but this decomposition is not an objective of this paper.
4.1. Efficiency comparison
Estimating separate frontiers is not a useful tool to compare efficiency scores across bank categories since the reference sets are
not comparable. So a common frontier is estimated while dummy bank category is introduced to capture potential technology
differences, country dummies variables are also included to capture country regulations and environmental differences. We propose
estimates of the efficiency scores according to the two aforementioned different orientations, i.e. contraction of inputs and expansion
of outputs, the hyperbolic distance function model (profit function), and by only allowing expansion of the outputs, the output
distance model (revenue function). For both orientations, a stochastic translog frontier is estimated, and a half normal for the
inefficiency distribution is assumed.
4.1.1. Technical efficiency measure from the profit function

To check how efficient are banks in adjusting both outputs and inputs in their production process, we estimate a stochastic
hyperbolic distance function and report efficiency scores. The general pattern that emerges in Table 3 is that commercial banks are
slightly more efficient (88.8%) than Islamic (88.2%), but the t-test of the mean difference in efficiency scores is significant.9 The mean
technical efficiency score by bank is close to 88–89% whatever is the bank type. So, during the studied period, MENA banks could on
average decrease their resources by about 16–15% and in the same time expand their activities by (13%–13.6%)10 without altering
the technology used, which could render them more profitable. We have also evaluated bank efficiency according to their size, some
empirical literature have evidenced higher efficiency of large banks which has been explained by scale economies, scope economies
or market power. This issue is particularly important since bank categories in MENA differ across countries and size. We construct
four subgroups according to the distribution of total assets in the sample, small (< 1 Billion), moderate (1–5 Billion), large (5–10
Billion), very large (> 10 Billion).
8
Most of the sampled banks do not report information on the total number of their employees staff, we use salary as proxy for labor input.
9
The difference in the efficiency score is the same on average even when Iran which has only Islamic banks is excluded.
10
As shown by Cuesta and Zofio (2005) the proportional change for an inefficient bank to reach the frontier is not the same for inputs or the
outputs. According to the level of the efficiency score DH, the proportional contraction of inputs is (1- Dδ) and the expansion level of outputs is (1/
Dδ).
8
Table 1
Descriptive statistics.
Variables Conventional banks Islamic banksa t-Test
(p-value)
Mean Med sd Mean Med sd
Loans (y1) 4213.5 1198.0 8137.5 3048.8 650.9 5968.5⁎,⁎⁎ 0.000⁎⁎⁎

Other earnings (y2) 2937.4 793.6 4964.8 1441.0 544.1 2411.4 0.000⁎⁎⁎
Personnel expenses (x1) 62.8 27.6 98.1 47.2 15.9 85.9 0.000⁎⁎⁎
Book value (x2) 83.1 34.4 124.3 97.3 20.0 297.0 0.23
Interest (x3) 165.8 66.3 290.4 92.6 33.7 165.0 0.000⁎⁎⁎
Total assets 8100.8 2641.7 13,905. 4891.7 1442.3 9221.1 0.000⁎⁎⁎
Total income/total assets 0.41 0.41 0.20 0.41 0.37 0.24 0.94
Total cost/total assets 0.04 0.03 0.03 0.06 0.04 0.06 0.000⁎⁎⁎
Profit/total assets 0.01 0.01 0.03 0.01 0.01 0.07 0.67
Equity Ratio 0.13 0.11 0.12 0.32 0.19 0.29 0.000⁎⁎⁎
Z-score 19.11 15.96 16.19 42.40 17.32 25.96 0.000⁎⁎⁎
Number of obs. 2161 642
(All monetary variables are in Billions of US dollars.)

a
Iran which has only Islamic banks is excluded in the comparison.
⁎
Suggests that the mean difference test is significant at the 90%.
⁎⁎
Suggests that the mean difference test is significant at the 95%.
⁎⁎⁎
Suggests that the mean difference test is significant at 99% respectively.
Table 2
Likelihood ratio test for common frontier assumption.
Hyperbolic distance (Log L) Output distance (Log L)
Pooled sample banks 455.07 −1506.16

Commercial banks 791.67 −709.67
Islamic banks 35.78 −405.12
LR test statistics 744.76⁎⁎⁎ 762.74⁎⁎⁎
Degrees of freedom 66 66
⁎⁎⁎
Significant at 99% level, country specific effects are included and time trend in the frontier, Log L is the log like-
lihood function.
Some differences in efficiency by bank size prove to be significant in particular for medium and very large banks for conventional
and Islamic banks (0.5% and 2% difference respectively). Within these two size classes which include 60.2% and 58.2% of the
sampled commercial and Islamic banks respectively, conventional banks outperform Islamic banks. This result suggests than the
challenge of Islamic banks in MENA in terms of profit efficiency is with very large conventional banks which seem to have an
advantage.
We also report the efficiency for the 16 countries by bank type and pair-wise t-test comparisons in Table 4. Islamic banks are the
less efficient in Algeria, Bahrain, Kuwait, Oman,11 Bahrain, and Saudi Arabia with an average efficiency score close to 86.4%
compared to 89.5% for conventional banks within this group of countries. In contrast, Islamic banks perform better in Egypt, Iraq,
Jordan, Syria, Tunisia and Yemen. Within these countries, average efficiency of Islamic banks is equal to 90.6% versus 88.3% for
conventional banks. Furthermore, Islamic banks dominate commercial banks in 6 countries, Yemen (+5.4%), Iraq (+5.3%), Tunisia
(+3.88%), Egypt (+1.2%) and Jordan (+1.1%). However, conventional banks outperform Islamic banks in 5 countries, Oman
(+6.6%), Kuwait (+4.6%), Bahrain (+3.9%), Algeria (+3.5%) and Saudi Arabia (+0.9%). There are no significant differences in 3
countries, Lebanon, Qatar and United Arab Emirates. This result suggest that there is no evident conclusion in favor of the superiority
of one bank type over another, due to country heterogeneity and the market share of Islamic banks. Even if we consider a more
homogeneous country group, the (GCC) group of countries for example, the dominance of conventional banks in terms of efficiency is
mitigated; compare for example Bahrain or Kuwait and Saudi Arabia to Qatar and United Arab Emirates.
4.1.2. Technical efficiency measure from the revenue function

For the output oriented model, the efficiency scores represent the percentage by which a bank expands its activities while using
the same level of inputs. On average the average efficiency score is equal to 73.6%, in the overall region, banks activities could be
expanded by 26.4% with the same resources and the technology whatever is their bank group. As reported in Table 3 commercial
banks are again slightly more efficient than Islamic banks (+1.3%), the difference being higher for medium size banks (+1.6%) but
much important for very large banks (+4.9%). Some differences in the efficiency scores are found across country and bank type, see
Table 4. Overall we have qualitatively the same conclusions, with respect to bank size and bank type.
11
Oman was the last GCC country to introduce Islamic finance. Islamic banks began operating in 2012.
9
Table 3
Technical efficiency by bank category and size in%.
Efficiency measure Commercial Islamic t-Test
Small (< 1 Billion)

Hyperbolic D. (Profit) 88.1 88.3 ns
Output D. (Revenue) 70.5 71.6 ns
z-score 23.8 48.9 ***
# obs 566 181
Medium (1–5 Billion)

Hyperbolic D. (Profit) 88.8 88.2 *
Output D. (Rev) 74.2 72.6 *
z-score 16.4 26.1 ***
# obs 772 203
Large (5–10 Billion)

Hyperbolic D. (Profit) 88.7 89.2 ns
Output D. (Rev) 74.8 75.3 ns
z-score 15.6 19.9 ***
# obs 261 96
Very Large (≥10 Billion)

Hyperbolic D. (Profit) 89.7 87.7 ***
Output D. (Rev) 77.1 72.2 ***
z-score 16.5 17.4 ns
# obs 481 183
All Banks
Hyperbolic D. (Profit) 88.8 88.2 ***
Output D. (Rev) 73.9 72.6 ***
z-score 18.3 29.0 ***
# obs 2080 663
Spearman-Correlation(Output-Hyperbolic) 95.9*** 98.1 ***
t-Test is the equality of means test result (one tailed test)/*,**,*** significance level, (90%, 95% and 99% respectively; ns, not significant).
If we compare the two efficiency measures, revenue inefficiency is much higher than profit inefficiency, which means that on
average banks revenues could be much more expanded than profit for inefficient banks. Large but also very large conventional and
Islamic banks, are more efficient than small banks. Moreover, by size class, there are significant differences in technical efficiency
between Islamic and conventional banks for only large and very large banks. This result is consistent with our assumption that
comparing Islamic and conventional banks depends on the efficiency measure retained and also on bank type and their size. However,
the correlation coefficient between the efficiency scores is quite high, which suggests that bank managers which are efficient in terms
of maximizing revenue are also efficient in maximizing profit.
However, there are significant differences between the two efficiency measures across countries, and also between Islamic and
conventional bank as reported in Table 4 and Figs. 3 and 4.
To sum up, size seems to have a positive and a significant impact on bank efficiency for both efficiency models. Mobarek and
Kalonov (2014) report similar findings in MENA using a cost model. This result suggests that in MENA region regulators should
encourage small and middle sized banks to increase their size in order to benefit from scale economies and increase their efficiencies.
4.2. Bank resiliency comparison
A preliminary assessment of bank stability is to follow the standard comparison based on Z-score12 statistics. According to the
results of Table 3, Islamic banks have on average higher Z-scores than conventional banks and the difference is statistically sig-
nificant, but some qualitative differences exists with bank size. Small and medium size banks are more stable than large or very large
banks, the Z-score decreases with bank size, the result being robust with ownership structure. This result is online with the finding of
Cihak and Hesse (2010) who based on the Z-score, found that small Islamic banks are financially stronger than small conventional
banks, and that small Islamic banks are more stronger than large Islamic banks. In contrast, in our sample very large banks have the
same risk insolvency whatever is the banking group Islamic or conventional, the mean difference t-test being not significant. This
result contradicts Cihak and Hesse (2010) who found evidence that large conventional banks are more resilient than large Islamic
banks in MENA. One possible explanation is that their sample includes 20 countries and is not restricted to MENA region but also, the
sampled period is older 1993–2004. Another explanation is that the threshold used to define size, large (small) banks is defined as
having assets larger than (smaller than or equal to) 1 billion USD. According to their size scale, large banks in their study regroup
12
Z-score measures the risk of insolvency, i.e. the risk that a bank runs out of its capital and reserves. It is defined by the sum of two ratios equity
and return to total assets ROA, divided by the standard error of ROA. The denominator being considered as a measure of return volatility. The higher
is the Z-score the more resilient is the bank.
10
Table 4
Technical efficiency across country.
Country Bank-type Hyperbolic D. (Profit) Output D. (Revenue)
Mean min max N Mean min max N
Algeria Commercial 0.89 0.66 0.96 141 0.74 0.24 0.90 141
Islamic 0.85 0.73 0.81 18 0.64 0.35 0.76 18
Bahrain Commercial 0.90 0.71 0.96 98 0.77 0.35 0.91 98
Islamic 0.86 0.65 0.95 103 0.68 0.26 0.90 103
Egypt Commercial 0.89 0.82 0.95 268 0.74 0.58 0.89 268
Islamic 0.90 0.85 0.94 27 0.77 0.63 0.88 27
Iran Commercial – – – – – – – –
Islamic 0.88 0.60 0.96 163 0.73 0.24 0.92 163
Iraq Commercial 0.86 0.60 0.96 50 0.67 0.26 0.93 50
Islamic 0.91 0.84 0.97 16 0.78 0.54 0.94 16
Jordan Commercial 0.89 0.82 0.92 133 0.74 0.55 0.83 133
Islamic 0.90 0.82 0.95 26 0.76 0.53 0.89 26
Kuwait Commercial 0.91 0.88 0.94 52 0.79 0.70 0.87 52
Islamic 0.86 0.59 0.93 84 0.69 0.22 0.86 84
Lebanon Commercial 0.90 0.70 0.95 411 0.74 0.30 0.90 411
Islamic 0.88 0.85 0.92 4 0.67 0.63 0.79 4
Morocco Commercial 0.89 0.78 0.96 137 0.74 0.51 0.90 137
Islamic – – – – – – – –
Oman Commercial 0.89 0.86 0.93 69 0.76 0.67 0.85 69
Islamic 0.83 0.81 0.88 4 0.65 0.60 0.71 4
Qatar Commercial 0.89 0.83 0.95 79 0.73 0.55 0.90 79
Islamic 0.89 0.81 0.95 39 0.75 0.51 0.91 39
Saudi Arabia Commercial 0.89 0.85 0.93 104 0.75 0.62 0.84 104
Islamic 0.88 0.76 0.93 44 0.74 0.55 0.86 44
Syria Commercial 0.88 0.78 0.95 79 0.71 0.43 0.91 79
Islamic 0.91 0.87 0.96 9 0.78 0.66 0.91 9
Tunisia Commercial 0.89 0.76 0.95 201 0.74 0.42 0.90 201
Islamic 0.93 0.87 0.95 15 0.82 0.67 0.88 15
United Arab Emirates Commercial 0.89 0.77 0.98 217 0.74 0.47 0.91 217
Islamic 0.89 0.83 0.94 76 0.75 0.58 0.86 76
Yemen Commercial 0.86 0.75 0.91 41 0.67 0.45 0.79 41
Islamic 0.91 0.86 0.96 35 0.80 0.62 0.91 35
Fig. 3. Profit and Revenue Efficiency in MENA, 2002–2014 for Conventional banks.
11
Fig. 4. Profit and Revenue Efficiency in MENA, 2002–2014 for Islamic banks.
medium, large and very large banks in our study. To sum up, this first aspect of the superiority of Islamic banks in terms of bank
stability, is only based on comparative statistics of the Z-score indicator. Another measure of bank resiliency, similar to stress test is
employed here to provide another aspect of bank's vulnerability.
We compare the business risk of the two bank groups to a potential reduction in their activities.13 The methodology is based on
applying some important shocks on the banks activities, derived from the historical gross shocks through simulations in order to
evaluate the business risk by bank type. The gross shocks distribution is obtained from estimating a common frontier, (profit
function), and considering the 10% least inefficient situations. Replications within this distribution tail, are then used to evaluate the
impact of these potential shocks on each bank category. Remind, that our previous results find differences in the technologies used, so
for all the simulations conducted, specific frontiers have been estimated for each bank category in order to evaluate their business
risk. So any differences in the bank resiliency will be attributed to the bank type of business instead of the impact of the shock itself.
Table 5 reports the results of the proposed resiliency measure for the overall sampled banks and also by size class.
Fig. 5, provides the initial profit efficiency distributions by bank type and country. The first decile varies between 0.592 (the
lower level obtained by a commercial bank in Iran) and 0.848 (a score obtained by several banks and countries, indicated by the
vertical line in Fig. 5). So we randomly impact a reduction in the bank activity between 40.8% and 15.2% at each replication in the
simulation process. Let us mention that these exceptional situations within this decile have been observed both by conventional and
Islamic banks, which means that the historical shocks used for the simulations are not restricted to a specific type of bank group.
Moreover, most of the countries belong to the first decile of the extreme situation for which the business risk measure is evaluated
(Fig. 5), which suggests that the business risk measure is based on most of the exceptional historical events which have affected the
sampled MENA banks region as a whole, including the international financial crisis period of 2008 and also the economy drop of the
countries involved by the Arab spring events of 2011.
Several scenarios are considered with two cases in order to evaluate banks business risk, the difference focus on the possibility to
the banks which face an abrupt reduction in its business activities to adjust costs or not. We evaluate the business risk by bank type,
conventional versus Islamic with three possible scenarios, an abrupt reduction in loans activities alone, an abrupt reduction in all
other non-loan activities alone and finally an abrupt reduction in both activities.14 This exercise, similar to a stress test, will permit us
to evaluate the business risk and bank vulnerability by activity.
To evaluate bank resiliency we first consider the general case where banks are able to adjust costs (the first scenario), the
simulations reported in Table 5, evaluate the business risk to (9.82%) for conventional banks, (10.47%) for Islamic banks. Remind
that the higher is the score the lower is the resiliency. These scores correspond to the potential decrease in total profit evaluated at the
5% percentile of the profit loss distribution. Accordingly, Islamic banks are slightly less resilient than commercial banks for a global
13
We will not report estimates of business risk by country, the objective here is just to compare the overall bank vulnerability by ownership
structure in MENA region.
14
In the simulation program, for each re-sampling case the shock is the same within the three scenarios.
12
Table 5
Bank business risk measures to gross shocks in MENA (2002–2014).
Scenario Bank size Bank type Percentile Resiliency to lending Resiliency to other non-lending Resiliency to all
services services services
With cost adjustment Small Commercial 5% 9.81 5.73 9.85

1% 11.94 6.40 10.31
Islamic 5% 8.39 8.30 10.14
1% 10.79 9.81 10.99
Medium Commercial 5% 9.23 5.28 9.79
1% 11.58 5.87 10.31
Islamic 5% 6.99 6.40 10.29
1% 8.20 7.19 10.63
Large Commercial 5% 8.67 4.73 9.78
1% 9.33 5.03 10.07
Islamic 5% 7.28 6.37 10.44
1% 7.72 6.91 10.71
Very large Commercial 5% 8.73 4.27 9.81
1% 9.09 4.79 10.08
Islamic 5% 7.59 5.01 10.66
1% 8.43 5.48 11.22
All banks Commercial 5% 8.82 4.79 9.82
1% 10.31 5.74 10.20
Islamic 5% 7.52 6.85 10.47
1% 9.28 8.71 11.04
Without cost adjustment Small Commercial 5% 18.48 12.20 19.06
1% 22.64 13.59 19.83
Islamic 5% 17.87 18.10 20.62
1% 22.23 20.97 22.91
Medium Commercial 5% 17.34 11.09 18.76
1% 22.33 12.34 19.53
Islamic 5% 13.67 13.59 19.58
1% 16.29 15.48 20.39
Large Commercial 5% 14.18 9.80 18.68
1% 17.20 10.39 19.10
Islamic 5% 14.02 12.83 19.46
1% 14.85 14.69 20.06
1% 17.08 9.89 19.17
Islamic 5% 14.05 10.85 19.40
1% 15.80 11.71 20.29
All Banks Commercial 5% 16.53 10.05 18.82
1% 19.51 12.15 19.44
Islamic 5% 14.58 14.94 19.85
1% 19.14 18.92 21.53
Notes: simulations conducted with 1000 replications, the higher is the score the lower is the resiliency. Shocks are re-sampled in the first decile of
the profit efficiency distribution.
shock on the two banks activities, i.e. lending and non-lending activities. However, when the shock impacts only one activity, both
bank groups seem to be much more vulnerable when a given choc affects lending activities (8.82%–7.52%) than the non-lending
activities (4.79%–6.85%). Furthermore, some differences in banks resiliency by bank activity and across bank groups are found.
Indeed, when shocks impact only lending activities the business risk measure is even higher for conventional banks (8.82%) com-
pared to Islamic banks being more resilient (7.52%). However, the resiliency to non-lending activities (commissions, investment,
trading) suggests that Islamic banks are more vulnerable than conventional banks (6.85%) versus (4.79%) when an important drop
impact this activity. Summing up, conventional banks are more resilient when shocks impact non lending activities, while Islamic
banks are more resilient when the shocks affect lending activities. The conclusions are robust when we take the lower percentile 1%
to evaluate the business risk. Moreover, according to the second scenario, the banks business risk increases when the costs are more
rigid, i.e. quasi fixed, whatever is the bank group. There is evidence that if banks do not have the possibility to adjust rapidly their
costs when they face a strong reduction in their activities, their business risk increase. In term of economic policy, cost adjustments
could be facilitated for example through, a staff redundancy arrangement, by closing some non-profitable branches, by selling illiquid
assets or when the interest rates are fixed. As shown in Table 5, and compared to the first scenario, both banks types are highly less
resilient (18.82%) for conventional banks and (19.85%) for Islamic banks, about two times more important compared to the first
scenario with cost adjustment possibilities. To sum up, Islamic banks are less resilient than conventional banks when a choc impacts
the two bank activities, even if they are much more resilient when a potential choc affects only lending activities. Abedifar et al.
(2013) using regression analysis, also found that Islamic banks face less credit risk than conventional banks.
Our findings seems consistent with the one obtained by Chaffai and Dietsch (2015) who, using the same methodology to evaluate
bank's business risk with directional distance function, found that a shock on lending activities causes the highest decrease in profit
13
Fig. 5. Box plot of the profit efficiency distribution by bank type and country.
for the French retail banks. They also found that the business risk is sustainable when the banks are not able to adjust their costs in the
short run. Moreover, our results are therefore in line with those of Elsiefy (2012) who found evidence that credit risk is the major
source of bank vulnerability for both Islamic and conventional banks in Qatar, applying the stress tests methodology. He also found
that Islamic banks are less resilient than conventional banks in this country.
Another important issue investigated is the question of bank resiliency and its link to the size of the banks. We reconsider the
simulation exercise and evaluate the business risk by bank size based on the definition of the size classes used previously. The results
reported in Table 5 according to the first scenario with the possibility of cost adjustments, find evidence that bank vulnerability
decreases with bank size whatever is the bank stream. This result could be explained by the fact that size allows better diversification
which reduces risks and allow banks to operate with lower capital and less stable funding, Laeven, Ratnovski, and Tong (2014).
Furthermore, small and medium size Islamic banks are equally vulnerable when a shock impacts their lending or non-lending ac-
tivities. The situation differs for large and very large Islamic banks where bank vulnerability is much more linked to lending activities.
In contrast, conventional banks seems to be much more vulnerable to important shocks on their lending activities compared to similar
shocks on non-lending activities whatever is their size which require a greater amount of scrutiny. Overall, the conclusions are
qualitatively robust under the second scenario with no cost adjustment.
5. Robustness checks
To check the robustness of our results, we conduct further investigations for both the frontier model estimation and also the
simulation exercise regarding the business risk measure under its two scenarios. First with respect to the efficiency comparison by
bank type we employ six different models, results are reported in Table 6 in the appendix. We compare the efficiency by employing
two additional more familiar models,15 a stochastic cost frontier model and a non-standard profit frontier model. Both models are
estimated by stochastic frontier method assuming a Translog specification, the same functional form used for our distance functions
base models. Other robustness checks of the results incorporate equity to total asset as a control variable in the base models to control
for credit risk16 while evaluating bank performances. Finally, since the stochastic frontier estimated is a standard model and to take
into account of the panel data structure we also estimate a panel data frontier model by using the true fixed frontier model, Greene
(2005). This model takes into account unobservable individual effects and the frontier is estimated by maximum likelihood method,
the inefficiency component is assumed to follow a half normal distribution as in the base model. The results of the efficiency
components by banks stream are presented in Table 6 in the appendix. A comparison of the results in Table 6 with those of Table 3
reveals that the major conclusions are qualitatively valid. Overall, conventional banks are significantly much more efficient than their
15
I would like to thank an anonymous referee for making this observation.
16
Notice here that it was not possible to include non-performing loans as a control variable for risk, because this variable is not disposable for
many sampled banks.
14
counterparts, whatever is the efficiency model retained. Notice here, that cost efficiency is the lowest whatever is the bank stream,
compared to profit or revenue efficiency which suggests that banks in MENA are less efficient in controlling their costs than in making
higher revenue. Profit efficiency is much more important than revenue efficiency by bank stream. Even if banking efficiency is
evaluated according to different models and behavioral assumptions, the correlation coefficients of the different models are positive
and statistically significant. With respect to the link between bank efficiency and bank size, the conclusions are also robust, large
banks are more efficient than small banks. Notice also that, the efficiency scores are robust to the two frontier estimation methods
employed, we consider further issues to evaluate business risk.17
Second we check the robustness results of the business risk proposed measure, when we exclude particular countries or when we
consider higher shocks in amplitude. Notice that our MENA sample include a country which has a banking system completely Islamic,
we exclude this country i.e. Iran from the sample and check whether the overall business risk for Islamic banks in the region is
sensitive to this particular country exclusion. Furthermore Lebanese banks represent 17.2% of bank observations, so it would be also
interesting to check the robustness of the results after excluding this country from the sample. In addition, the business risk proposed
method need to re-sample in the distribution of the first decile of the efficiency distribution, the profit model being previously used as
a base for banking system shocks exposure. We check the robustness of the results by re-sampling the shocks in the first decile of the
revenue function efficiency distribution. This exercise is interesting since the magnitude of the shocks are much more important than
those derived from the profit function model, i.e. revenue inefficiency in MENA banks is much higher than profit inefficiency. The
results of the new business risk measures are reported in Tables 8, 9 and 10 in the appendix.18 Excluding the country which has only
Islamic banking system has no impact on the conclusions reported before, namely that Islamic banks are equally vulnerable to a
potential shock which affect one of their banking activities and small banks are less resilient than very large banks, see Table 8.
Furthermore, the conclusion of the base model remains also unchanged when we exclude Lebanon, see Table 9. However, when we
consider a much larger shock in magnitude, the banks becomes much more vulnerable, compare for example the reported estimates
for all banks in Tables 10 and 5. The overall business risk for conventional banks is 22.59% versus 9.81% and 27.17% versus 10.66%
for Islamic banks (a measure at 5% level). However, the previous results remain qualitatively robust, conventional banks are much
more vulnerable to an important drop on their lending activities than with a similar drop on their non-lending activities while Islamic
banks are equally vulnerable to any drop of the activity. In addition, very large banks are much more resilient than small banks
whatever is the bank stream. Islamic banks are much more resilient than conventional banks to any drop on their lending activities,
but less resilient if a shock impacts their non-lending activities. Finally, banks vulnerability to business risk increase if banks costs are
more rigid in the banking system.
6. Conclusion
Our aim in this paper is to compare the performance of Islamic banks to conventional banks in MENA, during the period
2002–2014 and to evaluate their business risk. Performance is evaluated according to parametric stochastic frontier modeling using
distance function methodology, where both profit and revenue efficiencies are estimated. Efficiency comparisons are made by al-
lowing banks to improve their activities with or without possible contraction of their resources, which provide two different measures
of technical efficiency. First, we test for the common technology assumption which has been rejected, suggesting that Islamic banks
are using different technologies compared to conventional banks. Significant weak differences in technical efficiencies by bank
streams have been evidenced, but some differences are found with size, in particular large commercial banks outperform small and
medium sized banks in the region, and also some differences have been evidenced in some countries. It is also found that most of the
bank inefficiencies in MENA are coming from revenue inefficiencies instead of profit inefficiencies. In other words to improve their
efficiencies, these banks need to reinforce and develop their activities instead of reducing their costs. The comparison also in-
corporates an important issue related to the bank resiliency comparison through evaluating their business risk. Two scenarios are
assumed, the banks have the possibility to adjust their costs or not when they are exposed to a sudden drop in their activities.
Through simulations, by evaluating the impact of historical shocks on their business, Islamic banks are found much more vulnerable
than conventional banks to important drop on their non-lending activities. However, they are less vulnerable to a similar drop on
their lending activities compared to their counterparts. Large and very large banks seem to be more stable than smaller banks
whatever is the bank stream. More research is needed to confirm or reverse these findings for other Islamic banking systems.
However, it seems that the issue of the reverse resiliency of Islamic banks to a gross shock on their lending or non-lending activity
compared to conventional banks remains an open issue for further research. Finally, comparing business risk resiliency bank measure
used in this paper to stress tests could be an avenue for further research on bank resiliency.
17
Estimating the frontier by the true fixed effect method is much more time consuming than the standard stochastic frontier model. So it would
take more time in the simulation exercise which is highly time consuming. For this reason we did not conduct simulations employing the true fixed
effect method.
18
For these three Tables and to save space, we just report the business risk measure for small and very large banks. For the other two size classes
the conclusions are still robust and could be obtained upon request.
15
Appendix A
Table 6
Technical efficiency by bank category and size in%.
Efficiency measure Commercial Islamic t-Test
Small (< 1 Billion)

Cost Frontier 66.5 63.7 *
NS-Profit Frontier 80.2 78.0 ns
Hyperbolic DF. (Profit)b 88.2 88.3 ns
Output DF. (Revenue)b 71.4 72.6 ns
Hyperbolic DF. (Profit) 86.8 85.8 ns
Output DF. (Revenue) 76.1 73.9 ns
# obs 553 175
Medium (1–5 Billion)

Cost Frontier 68.7 61.8 ***
NS-Profit Frontier 80.6 79.6 ns
Hyperbolic DF. (Profit)b 88.9 88.2 *
Output DF. (Revenue)b 74.8 73.4 **
Hyperbolic DF. (Profit) 90.2 86.8 ***
Output DF. (Revenue) 83.0 78.1 ***
# obs 771 199
Large (5–10 Billion)

NS-Profit Frontier 77.9 72.4 ***
Hyperbolic DF. (Profit)b 88.8 89.2 ns
Output DF. (Revenue)b 75.5 75.7 ns
Hyperbolic DF. (Profit) 90.3 89.6 ns
Output DF. (Revenue) 83.2 82.7 ns
# obs 261 95
Very Large (≥10 Billion)

Hyperbolic DF. (Profit)b 89.7 87.8 ***
Output DF. (Revenue)b 77.7 73.0 ***
# obs 481 183
All banks
Hyperbolic DF. (Profit)b 88.9 88.3 ***
Output DF. (Revenue)b 74.6 73.4 ***
# obs 2066 663
***, **, and * indicate statistical significance at the 99%, 95% and 90% confidence levels respectively.
Cost frontier, Non-standard profit frontier, Hyperbolic DF. (Profit)b: Hyperbolic distance function with equity added as a control
variable in the model, Output DF. (Revenue)b: Output distance function with equity added as a control variable in the model.
Hyperbolic DF. (Profit), Output DF. (Revenue), the frontiers are estimated by the true fixed effect Greene (2005) model.
Table 7
Correlation matrix of the different efficiency models.
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8
Model 1 1.000
Model 2 0.970⁎⁎⁎ 1.000
Model 3 0.092⁎⁎⁎ 0.065⁎⁎⁎ 1.000
Model 4 0.089⁎⁎⁎ 0.090⁎⁎⁎ 0.075⁎⁎⁎ 1.000
Model 5 0.999⁎⁎⁎ 0.970⁎⁎⁎ 0.090⁎⁎⁎ 0.089⁎⁎⁎ 1.000
Model 6 0.971⁎⁎⁎ 0.998⁎⁎⁎ 0.067⁎⁎⁎ 0.095⁎⁎⁎ 0.971⁎⁎⁎ 1.000
Model 7 0.436⁎⁎⁎ 0.450⁎⁎⁎ 0.091⁎⁎⁎ 0.095⁎⁎⁎ 0.436⁎⁎⁎ 0.449⁎⁎⁎ 1.000
Model 8 0.423⁎⁎⁎ 0.463⁎⁎⁎ 0.075⁎⁎⁎ 0.132⁎⁎⁎ 0.423⁎⁎⁎ 0.462⁎⁎⁎ 0.914⁎⁎⁎ 1.000
Model 1: Hyperbolic distance function estimated by maximum likelihood (base model).

Model 2: Output distance function estimated by maximum likelihood (base model).
16
Model 3: Stochastic cost frontier (Translog model).

Model 4: Non-standard profit frontier (Translog model).
Model 5: Hyperbolic distance function estimated by maximum likelihood (base model), with equity ratio added as a control variable.
Model 6: Output distance function estimated by maximum likelihood (base model), with equity ratio added as a control variable.
Model 7: Hyperbolic distance function estimated by maximum likelihood (the true fixed Greene (2005) model).
Model 8: Output distance function estimated by maximum likelihood (the true fixed Greene (2005) model).
⁎⁎⁎
Significant at the 99% level.
Table 8
Bank business risk measures to gross shocks in MENA with Iran excluded (2002–2014).
Scenario Bank size Bank type Percentile Resiliency to lending ser- Resiliency to other non-lending ser- Resiliency to all ser-
vices vices vices

1% 30.17 17.32 23.74
Islamic 5% 23.29 22.80 26.67
1% 28.38 26.53 28.55
1% 21.39 11.61 23.17
Islamic 5% 19.67 13.69 27.50
1% 21.59 14.56 28.77
1% 26.48 16.18 23.51
Islamic 5% 19.82 18.80 27.17
1% 24.97 23.63 28.41
Commercial 5% 42.90 30.15 41.33
1% 51.61 33.18 42.61
Without cost adjust- Small Islamic 5% 44.93 44.66 48.20
ment 1% 52.22 50.27 52.27
Commercial 5% 36.99 20.77 41.89
1% 38.33 22.66 42.64
Very large Islamic 5% 34.80 28.33 45.97
1% 37.90 30.24 47.74
Commercial 5% 38.84 26.52 41.78
1% 45.64 30.88 42.78
All banks Islamic 5% 36.50 37.96 46.76
1% 46.75 46.10 49.70
Notes: simulations conducted with 1000 replications, shocks are re-sampled in the first decile of the profit efficiency distribution.
Table 9
Bank business risk measures to gross shocks in MENA, with Lebanon excluded (2002–2014).
vices vices vices

1% 10.44 5.18 8.30
Islamic 5% 10.47 11.28 11.34
1% 17.37 18.43 21.12
1% 7.43 3.19 7.88
Islamic 5% 8.24 5.00 11.36
1% 11.09 6.01 13.75
1% 9.03 4.59 8.02
Islamic 5% 8.20 7.57 11.07
1% 16.13 16.17 20.40
Without cost adjust- Small Commercial 5% 19.70 11.52 19.64
ment 1% 23.37 13.00 20.68
Islamic 5% 22.03 24.73 25.60
1% 31.77 35.00 39.44
1% 17.19 8.59 19.71
Islamic 5% 16.61 10.77 21.33
1% 21.00 11.71 24.69
1% 20.54 11.71 20.04
(continued on next page)
17
Table 9 (continued)
vices vices vices
Islamic 5% 16.89 17.31 22.07

1% 29.85 30.83 36.30
Notes: simulations conducted with 1000 replications, the higher is the score the lower is the resiliency. Shocks are re-sampled in the first decile of
the profit efficiency distribution.
Table 10
Bank business risk measures to larger gross shocks in MENA (2002–2014).b
vices vices vices

1% 30.17 17.32 23.74
Islamic 5% 23.29 22.80 26.67
1% 28.38 26.53 28.55
1% 21.39 11.61 23.17
Islamic 5% 19.67 13.69 27.50
1% 21.59 14.56 28.77
1% 26.48 16.18 23.51
Islamic 5% 19.82 18.80 27.17
1% 24.97 23.63 28.41
Without cost adjust- Small Commercial 5% 42.90 30.15 41.33
ment 1% 51.61 33.18 42.61
Islamic 5% 44.93 44.66 48.20
1% 52.22 50.27 52.27
1% 38.33 22.66 42.64
Islamic 5% 34.80 28.33 45.97
1% 37.90 30.24 47.74
1% 45.64 30.88 42.78
Islamic 5% 36.50 37.96 46.76
1% 46.75 46.10 49.70
Notes: simulations conducted with 1000 replications, shocks are re-sampled in the first decile of the revenue efficiency distribution.
b
To save space, the detailed results for the other two size classes (medium, large) are not reported in this paper. These results confirm the
previous findings and are available upon request.
References
Abdul-Majid, M., Saal, D. S., & Battisti, G. (2010). Efficiency in Islamic and conventional banking: An international comparison. Journal of Productivity Analysis, 34(1),
25–43.
Abedifar, P., Ebrahim, S. M., Molyneux, P., & Tarazi, A. (2015). Islamic banking and finance: Recent empirical literature and directions for future research. Journal of
Economic Surveys, 29(4), 637–670.
Abedifar, P., Molyneux, P., & Tarazi, A. (2013). Risk in Islamic banking. Review of Finance, 17(6), 2035–2096.
Al-Jarrah, I., Molyneux, P. (2004) Efficiency in Arab banking. In: Islamic perspectives on wealth creation, M. Iqbal and R. Wilson, (2004).
Alqahtani, F., Mayes, D. G., & Brown, K. (2017). Islamic bank efficiency to conventional banks during the global crisis in the GCC region. Journal of International
Financial Markets, Institutions & Money, 51, 58–74.
Ariss, R. T. (2007). On the rise of Islamic banking among giant commercial banks: A performance and efficiency analysis. Journal of International Business and
Economics, 7(1), 70–83.
Ariss, R. T. (2010). Competitive conditions in Islamic and conventional banking: A global perspective. Review of Financial Economics, 19, 101–108.
Bader, M. K., Shamsher, M., Ariff, M., & Taufiq, H. (2007). Cost, revenue and profit efficiency of conventional versus Islamic banks: Financial ratios approach. Review of
Islamic Economics, 11(1), 89–106.
Bader, M. K., Shamsher, M., Ariff, M., & Taufiq, H. (2008). Cost, revenue and profit efficiency of Islamic versus conventional banks: International evidence using data
envelopment analysis. Islamic Economic Studies, 15(2), 23–75.
Beck, T., Demirgüç-Kunt, & Merrouche, O. (2013). Islamic vs. conventional banking: Business model, efficiency and stability. Journal of Banking and Finance, 37,
433–447.
Berger, A. N., Hunter, W., & Timme, S. (1993). The efficiency of financial institutions: A review and preview of research past, present, and future. Journal of Banking
and Finance, 17, 221–249.
Berger, A. N., & Mester, L. J. (1997). Inside the black box: What explains differences in the efficiencies of financial institutions. Journal of Banking and Finance, 21,
895–947.
Bitar, M., Hassan, M. K., Pukthuanthong, K., & Walker, T. (2018). The performance of Islamic vs. conventional banks: Evidence on the suitability of capital ratios. Open
Economies Review, 29(5), 1003–1038.
18
Bitar, M., Hassan, M. K., & Walker, T. (2017). Political systems and the financial soundness of Islamic banks. Journal of Financial Stability, 31, 18–44.
Bitar, M., & Tarazi, A. (2019). Creditor rights and bank capital decisions: Conventional vs. Islamic banking. Journal of Corporate Finance, 55, 69–104.
Bos, J. W. B., & Schmiedel, H. (2007). Is there a single frontier European banking market? Journal of Banking and Finance, 31, 2081–2102.
Chaffai, M., & Dietsch, M. (2015). Modelling and measuring business risk and the resiliency of retail banks. Journal of Financial Stability, 16, 173–182.
Chaffai, M., & Hassan, K. (2019). Technology gap and managerial efficiency: A comparison between Islamic and conventional banks in MENA. Journal of Productivity
Analysis, 51, 39–53.
Cihak, M., & Hesse, H. (2010). Islamic banks and financial stability. Journal of Financial Services Research, 38, 95–113.
Cuesta, R. A., & Orea, L. (2002). Mergers and technical efficiency in Spanish savings banks: A stochastic distance function approach. Journal of Banking and Finance, 26,
2231–2247.
Cuesta, R. A., & Zofio, J. L. (2005). Hyperbolic efficiency and parametric distance functions: With application to Spanish saving banks. Journal of Productivity Analysis,
24, 31–48.
Elsiefy, E. (2012). Stress test for Islamic and conventional banks using sensitivity scenario test: Evidence from Qatari banking sector. International Journal of Economics
and Management Sciences, 1(N12), 44–63.
Färe, R., & Primont, D. (1995). Multi-output production and duality: Theory and applications. Kluwer Academic Publishers.
Feng, G., & Serletis, A. (2010). Efficiency, technical change, and return to scale in large US banks: Panel data evidence from an output distance function satisfying
theoretical regularity. Journal of Banking and Finance, 34, 127–138.
Ghosh, S. (2016). Political transition and bank performance: How important was the Arab Spring? Journal of Comparative Economics, 44, 372–382.
Greene, W. (2005). Reconsidering heterogeneity in panel data estimators of the stochastic frontier model. Journal of Econometrics, 126, 269–303.
Hasan, M., & Dridi, J. (2010). The effect of global crisis on Islamic and conventional banks. Journal of International Commerce, Economics and Policy, 2(2), 163–200.
Hassan, M. K., & Aliyu, S. (2018). A contemporary survey of Islamic banking literature. Journal of Financial Stability, 34, 12–43.
Johnes, J., Izzeldin, M., & Pappa, V. (2014). A comparison of performance of Islamic and conventional banks. Journal of Economic Behavior & Organization, S93–S107.
Jondrow, J., Lovell, C. A. K., Materov, I. S., & Schmidt, P. (1982). On the estimation of technical inefficiency in the stochastic frontier production function model.
Journal of Econometrics, (23), 269–274.
Koetter, M. (2006). Measurement matters: Alternatives input price proxies for bank efficiency analyses. Journal of Financial Services Research, 30, 199–227.
Koutsomanoli Filippaki, A., Margaritis, D., & Staikouras, C. (2012). Profit efficiency in the European Union banking industry: A directional technology distance
function approach. Journal of Productivity Analysis, 37, 277–293.
Kumbhakar, S. (2006). Specification and estimation of nonstandard profit functions. Empirical Economics, 31, 243–260.
Kumbhakar, S., & Lovell, K. (2000). Stochastic frontier analysis. Cambridge: Cambridge University Press.
Laeven, L., Ratnovski, L., & Tong, H. (2014). Bank size and systemic risk. IMF Staff Discussion Note 14/04.
Mobarek, A., & Kalonov, A. (2014). Comparative performance analysis between conventional and Islamic banks: Empirical evidence from OIC countries. Applied
Economics, 46(3), 253–270.
Mohanty, S. K., Lin, H. J., Aljuhani, E. A., & Bardesi, H. J. (2016). Banking efficiency in Gulf Cooperation Council (GCC) countries: A comparative study. Review of
Financial Economics, 31, 99–107.
O'Donnell, J. C., & Coelli, J. T. (2005). A Bayesian approach to imposing curvature on distance functions. Journal of Econometrics, 126(2), 493–523.
Park, K. H., & Weber, L. H. (2006). A note on efficiency and productivity growth in the Korean banking industry: 1992–2002. Journal of Banking and Finance, 30,
2371–2386.
Poghosyan, T., & Hesse, H. (2009). Oil prices and bank profitability: Evidence from major oil-exporting countries in the Middle East and North Africa. IMF working
paper, 09/220.
Shephard, R. W. (1970). Theory of cost and production functions. Princeton: Princeton University Press.
Srairi, S. A. (2010). Cost and profit efficiency of conventional and Islamic banks in GCC countries. Journal of Productivity Analysis, 34, 45–62.
19
Update
Volume , Issue , , Page
DOI: https://doi.org/10.1016/j.gfj.2020.100600
Global Finance Journal xxx (xxxx) xxx
Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/gfj
Erratum regarding missing Declaration of Competing Interest

statements in previously published articles
Declaration of Competing Interest statements were not included in the published version of the following articles that appeared in
previous issues of Global Finance Journal.
The appropriate Declaration/Competing Interest statements, provided by the Authors, are included below.
1. “The bank lending channel in the Malaysian Islamic and conventional banking system” [Global Finance Journal, 2019; 45:
100478] doi:https://doi.org/10.1016/j.gfj.2019.100478
Declaration of competing interest: The authors declare that they have no known competing financial interests or personal
relationships that could have appeared to influence the work reported in this paper.
2. “Internet search volumes of UK banks during the crisis: The role of banking structure and business model” [Global Finance
Journal, 2019; 45: 100472] doi:https://doi.org/10.1016/j.gfj.2019.05.001
3. “Investment efficiency: Dual-class vs. Single-class firms” [Global Finance Journal, 2020; 45: 100477] doi:https://doi.org/10.
1016/j.gfj.2019.100477
4. “VPIN, liquidity, and return volatility in the U.S. equity markets” [Global Finance Journal, 2019; 45: 100479] doi:https://doi.
org/10.1016/j.gfj.2019.100479
5. “On the financial market impact of euro area monetary policy: A comparative study before and after the Global Financial Crisis”
[Global Finance Journal, 2020; 45: 100480] doi:https://doi.org/10.1016/j.gfj.2019.100480
6. “Impact of the 2008–2009 financial crisis on the external and internal linkages of European frontier stock markets” [Global
Finance Journal, 2019; 46: 100481] doi:https://doi.org/10.1016/j.gfj.2019.100481
7. “Principles for financial regulatory reform” [Global Finance Journal, 2020; 39: 58–62] doi:https://doi.org/10.1016/j.gfj.2018.
01.016
Declaration of competing interest: The authors were contacted after publication to request a Declaration of Interest statement.
DOIs of original article: https://doi.org/10.1016/j.gfj.2018.02.002, https://doi.org/10.1016/j.gfj.2019.100479, https://doi.org/10.1016/j.gfj.

2018.01.005, https://doi.org/10.1016/j.gfj.2019.100478, https://doi.org/10.1016/j.gfj.2019.100481, https://doi.org/10.1016/j.gfj.2018.01.016,
https://doi.org/10.1016/j.gfj.2018.11.005, https://doi.org/10.1016/j.gfj.2019.100485, https://doi.org/10.1016/j.gfj.2019.100477, https://doi.
org/10.1016/j.gfj.2019.05.001, https://doi.org/10.1016/j.gfj.2019.100480, https://doi.org/10.1016/j.gfj.2018.01.004, https://doi.org/10.1016/
j.gfj.2018.02.001, https://doi.org/10.1016/j.gfj.2018.01.009, https://doi.org/10.1016/j.gfj.2018.07.003.
https://doi.org/10.1016/j.gfj.2020.100600
1044-0283/© 2020 Published by Elsevier Inc.
Please cite this article as: , Global Finance Journal, https://doi.org/10.1016/j.gfj.2020.100600

Global Finance Journal xxx (xxxx) xxx
8. “Hyperbolic distance function, technical efficiency and stability to shocks: A comparison between Islamic banks and conven
tional banks in MENA region” [Global Finance Journal, 2019; 39: 100485] doi:https://doi.org/10.1016/j.gfj.2019.100485
9. “Stock Market Valuation, Foreign Investment, and Cross-Country Arbitrage” [Global Finance Journal, 2019; 40: 74–84] doi:
https://doi.org/10.1016/j.gfj.2018.01.004
10. “Foreign acquisitions and firm performance: The moderating role of prior foreign experience” [Global Finance Journal, 2019;
42: 100415] doi:https://doi.org/10.1016/j.gfj.2018.02.001
11. “Are socially responsible firms less likely to restate earnings?” [Global Finance Journal, 2020; 38: 97–109] doi:https://doi.org/
10.1016/j.gfj.2018.02.002
12. “Regulatory responses to banking crisis: Lessons from Japan” [Global Finance Journal, 2019; 39: 10–16] doi:https://doi.org/
10.1016/j.gfj.2018.01.009
13. “Benchmark error and socially responsible investments” [Global Finance Journal, 2018; 38: 24–29] doi:https://doi.org/10.
1016/j.gfj.2018.01.005
14. “Stock market integration between the UK and the US: Evidence over eight decades” [Global Finance Journal, 2019; 41: 32–43]
doi:https://doi.org/10.1016/j.gfj.2018.11.005
15. “Informed trading in hybrid bond markets” [Global Finance Journal, 2018; 44: 100444] doi:https://doi.org/10.1016/j.gfj.
2018.07.003

Global Finance Journal: Sciencedirect

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Global Finance Journal: Sciencedirect

Uploaded by

Copyright:

Available Formats

i An update to this article is included at the end

Global Finance Journal 46 (2020) 100485

Contents lists available at ScienceDirect

Global Finance Journal

Hyperbolic distance function, technical efficiency and stability to

ARTICLE INFO ABSTRACT

2. Related comparative literature

3. Distance function methodology

3.1. The hyperbolic and the output distance function

Fig. 1. Distance functions and efficiency measures.

first output y1 retained as the numeraire:

Log (D / y1) = ln TL (yi / y1 , xj . y1 ) (3)

ji Log (yj / y1 )[Log (x i ) + Log (y1 )] + v + u

Fig. 2. Hyperbolic distance function and bank resiliency measure.

Summing up, the simulations we perform proceed in the following steps:

4.1. Efficiency comparison

4.1.1. Technical efficiency measure from the profit function

Loans (y1) 4213.5 1198.0 8137.5 3048.8 650.9 5968.5⁎,⁎⁎ 0.000⁎⁎⁎

(All monetary variables are in Billions of US dollars.)

Pooled sample banks 455.07 −1506.16

4.1.2. Technical efficiency measure from the revenue function

Small (< 1 Billion)

Medium (1–5 Billion)

Large (5–10 Billion)

Very Large (≥10 Billion)

4.2. Bank resiliency comparison

Mean min max N Mean min max N

With cost adjustment Small Commercial 5% 9.81 5.73 9.85

Efficiency measure Commercial Islamic t-Test

Small (< 1 Billion)

Medium (1–5 Billion)

Large (5–10 Billion)

Very Large (≥10 Billion)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

Model 1: Hyperbolic distance function estimated by maximum likelihood (base model).

Model 3: Stochastic cost frontier (Translog model).

With cost adjustment Small Commercial 5% 24.92 15.75 22.64

With cost adjustment Small Commercial 5% 8.60 4.49 7.69

Islamic 5% 16.89 17.31 22.07

With cost adjustment Small Commercial 5% 24.92 15.75 22.64

Contents lists available at ScienceDirect

Global Finance Journal

Erratum regarding missing Declaration of Competing Interest

DOIs of original article: https://doi.org/10.1016/j.gfj.2018.02.002, https://doi.org/10.1016/j.gfj.2019.100479, https://doi.org/10.1016/j.gfj.

1044-0283/© 2020 Published by Elsevier Inc.

Please cite this article as: , Global Finance Journal, https://doi.org/10.1016/j.gfj.2020.100600

You might also like