Bank Lending Procyclicality of Islamic and Conventional Banks in Indonesia:

A System GMM Analysis

Muhammad Rizky Prima Sakti1 and Tami Astie Ulhiza2

1
PhD candidate. Universiti Teknologi Malaysia, Kuala Lumpur Campus.
Researcher in ISEFID (Islamic Economic Forum for Indonesian Development).
(Corresponding author: sakti.finance@hotmail.com)

2
Master candidate. Kulliyah of Engineering. International Islamic University Malaysia.

ABSTRACT

It is widely believed in the literature that procyclicality of banking system is one of important
causes of financial instability. This study examines bank lending behavior over the business
cycle in a dual banking system in Indonesia. The aims of this study are ascertaining whether
Islamic banks have a role in stabilizing credit. We also aim to test procyclicality of Islamic and
conventional banks in Indonesia using a dynamic panel regression. The study use unbalanced
panel data of 50 conventional banks and 10 Islamic banks covering the period 2001-2015. By
employing two-step dynamic GMM estimators, the study shows that the aggregate loans by
banks tend to be procyclical in line with existing studies. Nonetheless, when we categorize the
lending/financing behavior of conventional and Islamic banks, the cyclicality of bank lending
apparently applied only for conventional banks. As for the Islamic banks, interestingly, the
business cycle does not seem to influence their financing decision. This fact can be understood as
indication that Islamic banks can be counter-cyclical in their financing decisions. Specifically,
large size Islamic bank is counter-cyclical in their financing behavior as compared to small size
and medium size Islamic banks. To confirm this, we use a set of robustness check with different
loan measure and alternative model specifications. All in all, ours affirm to the view that Islamic
bank is more stable in that they have ability to stabilize credit.

JEL Classification: E59, E69, G29

Keywords: Procyclicality, Bank Lending, Dual Banking System, GMM, Indonesia

1
1. Introduction

Financial instability has direct consequences to the economy that will lead to economic
crisis or even recession. Historically speaking, since the collapse of Bretton Wood Agreement 1,
more than 400 episodes of financial crises have been recorded (Laeven & Valencia, 2008). In
more recent, the global financial meltdown which is viewed to be the worst since the great
depression of 1930, financial intermediaries have received a great attention of both regulators
and academicians. The procyclicality of banking system, means the interactions between banking
system and real economy which are mutual support and probably strengthen the amplitude of the
business cycle, has been accused as one of important causes of financial instability (Ascarya,
Rahmawati, & Karim, 2016).

In view that the shocks in banking system have greater economic costs and illness to the
economy, a bulk of studies have examined bank-specific characteristics that make them immune
or resilient using a set of bank performance proxies, for instance bank loans, profitability, and
non-performing loans (Ibrahim, 2016; Louati & Boujelbene, 2015; Maghyereh & Awartani,
2014; Zakaria, 2015). With regard with bank lending behavior, the literature has evolved to
examine the nexus between bank lending procyclicality to bank-specific characteristics, for
instance bank capital and bank ownership (Ascarya et al., 2016; A. Aysan, Disli, & Ozturk,
2016) . It is important note the main inquiry is which type of banks (i.e. private against public
banks and Islamic versus conventional banks) tend to show stability in their lending behavior
during cyclical downturns?

In the context of Indonesia, it has been recorded that Indonesia has experienced two
major financial crises, namely the Asian financial crisis of 1997 and global financial crisis of
2009. The former crisis had made Indonesia trapped into the acute condition, while Indonesia has
shown a resilient performance in the latter crises amid uncertainty in the global economic
downturn. Nonetheless, some challenges remain such as a number of shocks faced by financial
institutions. This probably will spread out quickly due to the interconnectedness of financial
system. This condition is further exacerbated by procyclical behavior of financial institutions in
the economy. When there are changes in the financial market, financial institutions can emit
similar common reactions, creating a collective behavior that amplifies the business cycle
fluctuations (Utari, Arimurti, & Kurniati, 2010). The key in preserving financial stability is not
only in controlling domestic and external imbalances, but also financial imbalances such as
credit growth, asset prices, and risk-taking behavior in the financial system. Thereof, controlling
financial market as a whole system is very crucial in managing the economy.

1
Bretton Wood Agreement (BWA) was collapsed in August 1971, the gold standard system collapsed and replace
by a modern financial system.

2
 In prior literature, most of studies on Islamic banks have paid a lot of attention on issues
related to performance and economic impact of Islamic banks (A. F. Aysan, Dolgun, & Turhan,
2013; Johnes, Izzeldin, & Pappas, 2014; Mirzaei & Moore, 2015). Some studies raise the issues
of risk and stability in Islamic banks and how they become more resilient than conventional
banks (ihk & Hesse, 2010; Hasan & Dridi, 2011). As for financial stability, an analysis of
behavior of Islamic banks becomes important due to several reasons; (1) Islamic banks are
closely interact with conventional ones in dual-banking system, (2) Islamic banks have limited
hedging instruments to protect their risk-exposure due to a small number of size relative to
conventional one. Additionally, it is believed that Islamic banking system is relatively stable and
resilient to financial meltdown and economic disruptions on the basis of inherent features. As
Farooq & Zaheer (2015) explains, the resilience and stability of Islamic banks stem from their
distinctive features, that is, the ban of interest, speculative transactions, and excessive risk taking
activities.

Research on procyclicality and financial stability has gained momentum and popularity
among policy makers and academia particularly in the aftermath of global financial crisis. The
amount of research in the forms of journal articles, conference papers, books, and reports has
increased gradually (see table 1). It is worth to note that research on procyclicality has been
recorded in the renowned citation index, such as Thomson Reuters and Scopus. However, the
literature on this topic have mainly devoted to conventional banking. Only few attempts carried
out to explore these issues in Islamic banking and financial institutions. Given this fact, the
objective of this research is to examine bank lending behavior over the business cycle in a dual
banking system in Indonesia. We also aim to ascertaining whether Islamic banks have a role in
stabilizing credit. The choice of Indonesia is straightforward. According to Thomson Reuters
(2015), Indonesia has a vast potential to become a next hub of global Islamic finance. The
sleeping giant, that is how the media labeling Indonesia, is considered as the top 10 developed
Islamic finance globally. This has been supported with well-planned government strategies and
comprehensive regulatory framework in enhancing the growth of Islamic economics and finance.

Table 1. Publication of procyclicality and financial stability research

No Database or publisher Total no. of journals a No. of procyclicality articles b
1 Thomson Reuters (ISI) 439 25
2 Scopus 1,166 35
3 Emerald Insight 481 39
4 Springer 36 95
5 Taylor & Francis 264 191
6 Wiley-Blackwell 429 248
7 Science Direct 3,876 443
a
All journals are categorized under the subject of business, economics, finance, and accounting.
b
Using keyword procyclicality and financial stability
Source: authors compilation (2016)

3
 This paper is then structured as follows. Section 2 provides a review of literature
pertaining to procyclicality and lending behavior. Section 3 describes the data and methodology
used in this study. Section 4 provides the findings and critical analysis. Finally, section 5 is the
conclusion.

2. Literature Review

2.1. A brief overview of Indonesian Islamic banks

Indonesia, as the world most populous Muslim country, has been assumed to have
tremendous potential in Islamic banking and finance. The blossom of Islamic banking industry
has been started in in the early 1990s with the establishment of Bank Muamalat Indonesia (BMI).
Although the late adoption to Islamic banks relative to neighbor Malaysia, at present, Indonesia
is considered as the top 10 developed Islamic finance globally (Thomson Reuters Report, 2015).
According to Thomson Reuters Report (2015), the size of Indonesias Islamic finance industry in
2015 reaching to Rp 617 trillion which equivalent to 3% of total finance industry assets. Since
2010, the growth rate of Indonesias Islamic finance surpassed its conventional peers, accounted
for 139% and 42% respectively. Accordingly, under the purview of Indonesias Financial
Services Authority/Otoritas Jasa Keuangan (OJK), Indonesias Islamic banking and finance are
expected to continue its growth and to show improvements. Indonesias banking sector is
dominating with vast commercial banks. Nowadays, there are 85 conventional commercial banks
and 12 Islamic commercial banks. Besides, the number of Islamic banking costumers has
incredibly boosted to 13 million customers across 3,000 office network in Indonesia (Bank of
Indonesia, 2015). Nevertheless, the Islamic banking industry still have small share in the range
of 5% relative to its conventional peers (see figure 1).

Figure 1. The comparison between Islamic and conventional banks in Indonesia

Panel A
INDONESIAN BANKING INDUSTRY TOTAL
ASSETS
5.705.028
6.000.000
5.031.843
5.000.000 4.329.984
Rp Billion

3.798.631
4.000.000
3.054.595
3.000.000
2.000.000
1.000.000 148.987 199.717 248.110 278.917
100.258
-
2010 2011 2012 2013 2014
Islamic Banking Conventional Banking

Source: Bank of Indonesia (2016)

4
 In addition, when viewed from the growth of assets, financing, the amount of third party
fund, the trend is decline significantly. Figure 2 below depicts the development of Islamic banks
in terms of asset, financing, and third party fund. Although Islamic banking still remain sound as
supported by strong capital, however, structural problem remain unresolved, such as higher
financing pricing relative to conventional peers. The cost of funds of Islamic banks is less
competitive relative to its conventional counterpart. In addition, Islamic banks have to use the
correct business model to minimize the costs and, at the same time, to comply with sharia
principles. All of which can hinder the potential of Indonesias Islamic banking in the future.

Figure 2. The development of Islamic banks in terms of asset, financing, and third party fund

60,00%
50,00%
40,00% Pertumbuhan aset
Assets growth
30,00%
Pertumbuhan
Financing growth
20,00%
pembiayaan
10,00% Third party DPK
Pertumbuhan fund
growth
0,00%

Source: Financial Services Authority/Otoritas Jasa Keuangan (2015)

2.2. Prior studies on procyclicality and financial stability

It is widely believed in literature that procyclicality is an inherent feature of current

financial architecture which probably influencing financial stability. Many scholars have accused
this procyclicality as one of the main causes of financial imbalances in the economy. As Landau
(2009) explains, procyclicality is the tendency of financial variables to swing around a trend of
business cycle. As the procyclicality increases, the fluctuations probably will have greater
amplitude. In similar vein, Abdellah (2009) argues that procyclicality as the nexus between the
financial and real economy system which are mutually amplify. These interactions between the
two tend to reinforce the amplitude of business cycle. Additionally, Landau (2009) further
explains that a simple description rarely cope with the behavior of financial system in real life as
this system is characterized by feature of complexity. In most cases, the trajectory of asset prices
will exhibit various and highly volatile.

Many scholars has believed that Islamic financial system is relatively stable due to its
inherent features and moral values enshrined in the sharia tenets (Buiter, 2014; Galati &

5
Moessner, 2013; Husman, 2015). These features involving the ban of interest in deposit-lending
activities, the condemnation of leverage, and excessive speculation that trigger financial shocks
in conventional peers (Buiter, 2014). Prohibition of interest resembles the features of riba or
usury. In Islamic jurisprudence, any level of interest is viewed as usurious and hence prohibited.
As for replacement, the sharia scholars approve foe equity-based financing based on the concepts
of profit and loss sharing (PLS) and risk and the trade-based (al-bay) financing. Moreover, the
profit-loss sharing (PLS) instruments that have a crucial impact to the stability of Islamic
financial system. In this sense, Chapra (2009) argues that PLS contract will ensure the greater
discipline by making Islamic bank more vigilant in lending activities, at the same time, the
depositors more cautious with the health of Islamic banks. Ultimately, such discipline carries the
greater stability and even greater efficiency in Islamic financial system. Not just hat, Islamic
finance promotes the concept of justice and fairness. As stipulated in sharia law of contract,
parties that are involved in financial transaction have to share their associated profits and risks to
meet the concepts of justice and fairness

Concerning with this, it has been acknowledge that Islamic finance could be the solution
to mitigate financial instability problems. Buiter (2014) encourages all stakeholders in financial
sector should adopt Islamic finance. The excessive debt in the public sector, banking sector, and
households sector should be either write down or converted into equity. Notice that he
encourages households sector to convert excessive mortgage into Islamic-style mortgage. For
banking sector, he suggests them to convert excessive leverage into Islamic equity. Given the
facts about the salient features of Islamic financial system, one may conclude that Islamic
financial system has intrinsic stability rather than conventional-based system. With particular
reference to Islamic banks, a number of empirical studies have examined the relationship among
procyclicality and financial stability (Ascarya et al., 2016; Husman, 2015; Ibrahim, 2016; Louati
& Boujelbene, 2015)

Contrary however, a group of scholars have undermine the distinctive features of Islamic
banks that make them immune and resilience to economic disruptions (Abdul-Rahman, Abdul
Latif, Muda, & Abdullah, 2014; Azmat, Skully, & Brown, 2015; Chong & Liu, 2009). Chong &
Liu (2009) argue that there is no difference between Islamic and conventional banks since the
profit-loss sharing (PLS) based assets highlighted as a peculiarity of Islamic banks constitute
only a small portion of Islamic banks assets. Similarly, the recent research by Abdul Rahman et
al. (2014) and Azmat et al. (2015) question the ability of Islamic banks to uplift the PLS-based
activities in the current-setting. Moreover, Islamic banks tend to be more affected by adverse
shocks as their Islamic money markets are less developed and thus limited means by the Islamic
banks on funding sources other than deposits (Farooq & Zaheer, 2015). Additionally, the high
concentration of Islamic banks assets and the limited hedging instruments due to restrictions by
sharia principles, all of which add to potential instability in the face of adverse shocks (Beck,
Demirg-Kunt, & Merrouche, 2013). Some studies have challenged the stability view of
Islamic banks by posit insignificant difference between the two banking system (Nabi, 2013).

6
Similar line of research by Hasan and Dridi (2011) reveal that the profitability of Islamic banks
is more negatively affected when the financial crisis hit the real sector. Cihak & Hesse (2010)
also support the view that large Islamic bank tend to be less stable as compared to large
conventional banks. Thus, all of which have feel to skepticism to the stability view of Islamic
banks.

Similar with conventional-based system, the literature in Islamic finance has also
highlighted the importance separation between tools that address time series dimension of
financial stability, i.e. procyclicality in financial system (Ascarya et al., 2016), and cross-section
dimension, i.e how risk is distributed within financial system (A. Aysan et al., 2016; Ghosh,
2014, 2016). Ascarya et al (2016) have examined procyclicality of Islamic and conventional
banks in Indonesia by using a set of econometric techniques. The paper shows that Islamic bank
is more procyclical rather than conventional peers. Nonetheless, this procyclicality of Islamic
bank can be regarded as good procyclicality since it does not create credit bubbles, so that it can
bring benefits for economic growth in the long-run period. Concerning with this, the literature
has highlighted several instruments to address procyclicality, for example loan-loss provisions
(Lee, Asuncion, & Kim, 2016; Zhang & Zoli, 2016). A loan-loss provision2 is an important
instrument through which the mis-assessment of risk can strengthen financial cycle (Galati &
Moessner, 2013).

Other research by Nursechafia and Abduh (2014), attempted to address the long run
association of Islamic banks sustainability in response to volatility of macroeconomic variables
using time series analysis. Using calendar time series data, from 2005 until 2012 (monthly
basis), the results from variance decomposition (VDC) and impulse response function (IRF)
reveals that there is long-run association between credit risk ratio with the selected
macroeconomic variables. Negative relationship can be found among exchange rate, supply side-
inflation and growth with credit risk whereas money supply and interbank money market have
positive relationship with credit risk rate. Even though this paper clearly illustrated the key
relationship between macroeconomic variables and credit risk rate but it seems that this paper do
not utilizes the specific factors of banks in assessing the credit risk such as loan to deposit ratio,
third party fund, financing to deposit ratio which may have direct impact on credit buffer. This
informations are clearly absence from this paper and it subject to criticism. In contrast,
Imaduddin (2007) concludes that the performance of Islamic banking in Indonesia relatively low
compared to conventional system when it comes to credit default management. This study
adopted calendar time-series monthly basis from January 2003 till April 2006 and econometric
modeling. This study probably subject to certain condemnation because the time horizon used
are less and there will be possibility where it can influencing the robustness of outcome.

2
Banks have considerable policy of their loan-loss provisioning for bad loans. When comes to mounting losses,
banks will hold back on the provisioning for bad debt to preserve their capital.

7
 Another important line of research on financial stability examines the systemic risk
problems of individual Islamic financial institution or Islamic financial system as a whole
(Blundell-wignall & Roulet, 2014; Ghosh, 2016). This research highlights interconnectedness
among financial institutions and the availability substitutes. Systemic risks remain need to be
addressed by regulators as this leads to financial crisis and worse economic outcomes. Blundell-
wignall & Roulet (2014) argue systemic risks arise because financial institutions are actively
engaged with three main activities: (1) credit intermediation, (2) maturity transformation, and (3)
leverage. The paper examines the impact of macroprudential policy on bank systemic risk in
developed and emerging countries, including some Muslim nations such as Malaysia, Indonesia,
Turkey, Egypt, Kazakhstan, Pakistan, Morocco, and Tunisia 3. The measureres of bank riskiness
in this paper is distance-to-default (DTD)4, while taking into account systemic importance,
leverage, and business model aspects in banking sector. The paper suggests that there is high
complexity and interdepence in the financial system. The calibraton of macroprudential policy
seems relatively difficult than counter-cyclical rules. On top of that, much consideration should
be taken in adopting macroprudential policy in the face of bank systemic risk. In similar vein,
Ghosh (2016) posits that capital adequacy ratios and reserve requirements are effective in
reducing the potential of systemic risks.

3. Research Methodology

3.1.Data

All data to construct the indicators for bank lending procyclicality between Islamic and
conventional banks were retrieved from Bank Scope database from Bureau Van Dijk Company.
Meanwhile, the macroeconomic information was retrieved from Bank of Indonesia website. We
include 60 banks covering both conventional and Islamic banks in Indonesia, which consists of
50 conventional banks and 10 Islamic banks. Our dataset spans from 2001 until 2015.

3.2.Methodology

A panel data regression could reduce the biased estimations stems from aggregating
individual units into a broad one. As Baltagi (2008) explains, panel data regression is a
measurement on the pooling of observations on cross sections of households, firms, countries,

3
In total, there 29 countries are included in the final sample; Argentina, Brazil, Bulgaria, Chile, China, Czech
Republic, Egypt, Hungary, Iceland, India, Indonesia, Israel, Kazakhstan, Korea, Latvia, Malaysia, Mexico,
Morocco, Pakistan, Peru, Philippines, Romania, Russia, South Africa, Sri Lanka, Thailand, Tunisia, Turkey, and
Venezuela
4
Distance-to-default is derived from Black-Scholes options pricing model (BSOPM)

8
and so forth over several time periods. According to Gujarati (2003), there are three major
advantages of panel data regression, namely as follows:

i. The combination of time series and cross-section observations allow having more
information, variability, and degree of freedom. At the same time, this combination provide
less collinearity among intercept dummy variables;
ii. The linkage in a panel data regression is examined by repeating the cross sectional
observations, accordingly, this provide better link to investigate the dynamic change or
adjustment;
iii. Panel data regression can discover some nexus that probably cannot be found either in time
series or cross sectional regression.

Generally speaking, a panel regression can be categorized into two major structures, that
is (i) static panel, and (ii) dynamic panel regression. In static panel regression, one could estimate
the coefficients in the model specification based upon random effect (RE) estimator and fixed
effect (FE) estimator (Baltagi, 2008). Under the dynamic panel, the coefficients could be
predicted through the generalized method of moments (GMM) and seemingly unrelated
regressor (SUR). In this regard, this study does not employ SUR estimator since the panel data
used is relatively short. Based on the rule, the SUR is based on a larger number of periods, that
is, T that approaches infinity compared to the number of cross sections. The FE is the preferred
estimator because the RE probably can be invalid or inconsistent estimator when several
regressors are associated with the unobserved heterogeneity effect. Nonetheless, this FE model
cannot compute the time invariant coefficients.

Albeit the FE estimator could be adjusted to compute the time-invariant variables through
Hausman instrumental variable model, however, it still cannot solve the correlation problem with
the unobserved heterogeneity effect. In this study, there is possibility of the presence of
unobserved bank specific effects. Ignoring them may produce a bias and inconsistent estimates
given that bank specific effects are likely to be correlated with the explanatory variables. In the
presence of any correlation between explanatory variables and the bank specific effect,
estimation methods will no longer produce best linear unbiased estimation. This is essentially
due to the violation of the assumption of exogeneity in the explanatory variables. In addition to
that, the orthogonality condition between the error term and the explanatory variables is unlikely
to be fulfilled for either the FE estimator or the Generalized Least Square (GLS) to produce
efficient and consistent estimates. One can be met the orthogonality condition through proper
differencing of the data. Yet, since the model contains endogenous explanatory variables and the
effects of lagged endogenous variables, thus the error term in the differenced equation is
associated with the lagged dependent variable. Thereof, neither the FE nor the GLS estimator
will produce efficient and consistent estimates under these circumstances.

Thanks to Arellano and Bond (1991) that provide seminal work to solve the presence of
endogeneity problem in panel data. They proposed using a dynamic panel estimator based on

9
Generalized Method of Moments (GMM) estimation that effectively uses the linear moment
restrictions implied by dynamic panel model. The dynamic GMM estimator is essentially an
instrumental variable estimator that uses lagged value of all endogenous regressors as well as
lagged and current values of all strictly exogenous regressors as their instruments. Equations can
be estimated through the level or first differences of variables. Moreover, Arellano and Bond
(1991) proposed two estimators, that is one-step and two-step estimators.

The one-step estimator, also known as first-differenced GMM, use the difference of each
variable for both dependent and independent variables in the regressions and create instrument
variables from the lagged levels of the independent variables. Yet, these lagged levels could be a
weak instrument if there is a presence of serial correlation in the error term. Because of this, first
difference GMM might give inconsistent or even biased estimators. To solve this problem,
Arellano and Bover (1995) and Blundell and Bond (1998) were introduced the system GMM.
This system GMM, to certain extent, can provide efficient and consistent estimators especially
when the time period is smaller than the number of cross-sections. There are two types of
simultaneous equations in the System GMM, namely (i) equation in levels that comprise the
lagged difference of the dependent variable as instrument, (ii) equation in first difference that
comprises the lagged levels of dependent variables as instruments. With regard to time-invariant
variables, the system GMM take out the effect of time-invariant variables in first difference but
estimates in levels.

The instrumental variables are essentially the lagged variables among the explanatory
variables in the GMM estimation. It is unlikely that this lagged variable would be associated with
the unobserved heterogeneity effect, thus the dynamic panel regression solves the presence of
endogeneity problem. For this reason, this study opts for the dynamic panel regression as it is
more robust, unbiased and efficient estimator, and particularly in solving endogeneity problem in
a panel data. Obviously, this study employs the system Generalized Method of Moments (GMM)
estimator proposed by Arellano & Bond (1995) and Blundell & Bond (1998), which is viewed to
be superior in dealing with dynamic panel modelling 5. Blundell and Bond (1998) show that the
system GMM has relatively small variances and more consistent and efficient, hence improving
the precision in the estimator. On top of that, the dynamic panel based on GMM addresses the
problems of endogeneity, heteroscedasticity, and autocorrelation in the panel data.

It is also important to note some necessary assumption before proceeding with the GMM
estimator. The assumption of there is no second order serial correlation in the first differences of
error term is necessary. In order to produce the efficient and consistent estimates, the GMM

5
According to Baum (2006), the system GMM estimator can solve the important issues concern like fixed effects,
endogeneity problem, and avoiding panel bias. In this regard, we use the system GMM since this model can handles
our modeling concern with regard to fixed effect. As we discussed earlier, the static panel based on fixed effect (FE)
estimator cannot predict the time invariant coefficients such as individual banking system used in the model of this
study. Specifically, two-step system GMM can predict the time-invariant coefficients.

10
estimator requires that assumption be satisfied. Having the instruments as lagged variables, the
existence of second order correlation in the model will render such instruments invalid. As for
the specification tests, the GMM estimator use the Sargan test of over identifying restrictions and
the test of lack of residual serial correlation. As the sargan test is based on the sample analog of
the moment conditions, it can use to determine the validity of predeterminacy assumption,
endogeneity, and exogeneity.

As Baltagi (2008) explains, the existence of a lagged dependent variable is a unique

characteristic of a dynamic panel model. The model in this study follows a one-way error
component model and can be written as follows:

= 1 + + (1)

= + (2)

Where:

= Level of deflated gross loan of bank i in period t

1 = The lagged of deflated gross loan of bank i in period t
= A scalar
= The explanatory variables of bank i in period t
= A random error term which consists of two components
= The unobservable time-invariant individual or bank specific effects
= The remainder disturbance

In this study, the empirical model is designed to examine bank lending procyclicality and
check whether Islamic banks are less or more procyclical. In doing so, we taking leads from
Micco & Panizza (2006) and Ibrahim (2016), we then specify the equation as follows:

= 1 + + 1 + + + (3)

Where:

= Natural logarithm of CPI-deflated gross loans of bank i in period t

1 = The lagged of CPI-deflated gross loans of bank i in period t
= Natural logarithm of real GDP
= A vector of bank-specific variables
Inf = Inflation rate
= The first difference of operator
= Bank-specific effects
= A random error term

11
4. Findings and Discussion

4.1.Descriptive statistics

Figure 1 below exhibits Indonesia macroeconomic performance, such as real GDP

growth and inflation rate from 2001-2015. In average, Indonesia recorded an average annual
growth rate of 5.33% over the study period. Noticeably, Indonesia recorded a low growth of
3.6% in 2001 and the highest growth of 6.35% in 2007. It is also interesting to note that
Indonesia is relatively immune to global financial crisis in 2008-2009. Indonesia, together with
China and India, is one of the countries that performed a positive growth when the financial
crisis hit the global economy. During the global financial downturn, Indonesia recorded an
average annual growth at 6% and 4% in 2008-2009 respectively. Lack exposure of toxic assets in
financial sector and strong domestic demands in real sector, to name a few, are the key factors
driving Indonesia economy into a positive real growth in global financial crisis. Nonetheless,
while Indonesia recorded a positive growth during global economic disruptions, its growth
performance has started to slow again in recent years, particularly in 2013-2015.

Subsequently, Indonesia inflation record has been relatively high of 7.65% per year over
2001-2015. It witnessed a swing trend from 2001 to 2009 before it relatively stable in 2010
onwards. The lowest inflation was recorded by 4.28% in 2012, while the peak inflation rate of
13% hit in 2006. From these facts, apparently Indonesia was relatively immune to the global
financial disruptions but it was well-preserved in maintaining the general price level. As the
global crisis hit the Indonesia economy, Indonesia real performance is not much affected. Yet,
high inflation rate is obviously affecting the purchasing power of Indonesian.

Figure 1. Real GDP growth and inflation over 2001-2015

Table 1 exhibits descriptive statistics of bank loans and bank-specific characteristics,

such as real assets, equity to assets ratio, and loans to deposit ratio, for both Islamic and

12
 conventional banks. It is apparent that conventional banks are significantly larger than Islamic
banks, as indicated by their greater size and loans. While the growth rate of conventional bank
lending was estimated at 32% per year over 20001-2015, the corresponding growth for Islamic
bank financing was relatively better over 40% per year in the same period. As we expected,
Islamic banks are better capitalized as indicated with higher equity to asset ratio of 17.4%
relative to 11% for conventional ones. It is also true that the funding ratio, as indicated by loans
to deposit ratio, is relatively higher for Islamic banks.

Table 1. Descriptive Statistics

Variables All samples Conventional banks Islamic banks

Mean Std.Dev Min Max Mean Std.Dev Min Max Mean Std.Dev Min Max
Loans measures
Gross loans 40,300,000 77,100,000 11,300 590,000,000 44,900,000 81,700,000 11,300 590,000,000 10,400,000 13,700,000 97,930 50,000,000
% growth 32.77 54.71 -83.53 666.79 31.59 56.76 -83.53 666.79 41.06 36.40 -23.61 140.91
Net loans 38,800,000 74,200,000 7,600 570,000,000 43,200,000 78,600,000 7,600 570,000,000 10,100,000 13,200,000 63,662 49,000,000
% growth 33.61 55.66 -42.42 667.55 32.52 57.71 -42.42 667.55 41.30 37.66 -31.25 147.62
Bank-specific variables
Real assets (log) 16.90 1.63 12.61 20.63 17.10 1.58 12.61 20.63 15.63 1.40 12.74 18.07
Equity-asset ratio (%) 12.21 7.64 -7.15 63.23 11.41 5.28 -7.15 51.07 17.42 15.10 5.48 63.23
loans-deposits ratio (%) 91.84 73.28 1.23 712.41 81.57 37.19 1.23 326.42 164.23 166.85 66.75 712.41

With regard to these backdrops of Indonesian dual banking system and current
macroeconomic situations, we then attempt to check whether bank lending show cyclical swings
of real economic activities by the way of examining procyclicality of bank loans. In detail, we
also aim to examine whether there are differences prevail between Islamic financing and
conventional lending over the economic cycle.

4.2.System GMM - Baseline Results

Table 2 below presents the results of the dynamic panel data estimation using the two-
step system GMM analysis. In model 3, we have out baseline regression model, which includes
variables such as bank-specific variables and inflation rates. The odd regression numbers (in
regression 1 & 3) incorporate only bank specific variables. Meanwhile, the even regression
numbers (in regression 2 & 4) add the inflation rate as a proxy for controlled variables. The
specification tests show the appropriation of the GMM estimators in all models. The Sargan test
fails to reject the over-identification restrictions, indicating that the validity of instruments.
Moreover, the serial correlation test (autocorrelation test) does not reject the null of second-order
autocorrelation. Eventually, the residual of level equation (prior differencing) do not suffer from
the autocorrelation problems. All in all, both Sargan and autocorrelation tests tend to affirm the
model estimated using the GMM estimation approach.

13
Table 2. System GMM Estimation-Baseline Results

Variables (1) (2) (3) (4)

L1it-1 0.6505*** 0.6831*** 0.6481*** 0.6771***
(0.0000) (0.0000) (0.0000) (0.0000)
yit 0.147*** 0.131* 0.199*** 0.553**
(0.090) (0.0791) (0.0000) (0.0278)
yit x IBi - - -0.331*** -0.629***
(0.0000) (0.0000)
LnSIZEit-1 0.3029*** 0.2512*** 0.3149*** 0.2690***
(0.0000) (0.0000) (0.0000) (0.0000)
CAPit-1 -0.02723*** 0.0281*** -0.0274*** -0.0285***
(0.000) (0.000) (0.000) (0.000)
FUNDit-1 0.0003** 0.0002** 0.0003** 0.0002*
(0.034) (0.031) (0.039) (0.078)
Inft - -0.0114*** - -0.0102***
(0.000) (0.000)

P-values
AR(2) 0.1476 0.2500 0.1565 0.2521
Sargan test 0.2151 0.2461 0.217 0.2258

Note: number in parentheses ( ) are p-values

***, **, * indicates significance at 1%, 5%, and 10% alpha respectively

The result from model (1) and (2) as given in table 2 affirming that bank size and
capitalization ratio, and funding ratio are significant. Similarly, these results prevail to other
regressions shown in the table 2. As for our main concern, we find evidence affirming
procyclicality of bank lending in Indonesia, which is in line with Ibrahim (2016), and Ascarya et
al., (2016). The findings from regression (1) and (2) indicate that a 1 percentage point increase in
GDP growth is associated to approximately 0.13 to 0.14 percentage point increase in the growth
of real gross loans. Moreover, in model (3) and (4) we add the Islamic bank interactive dummy
and show the significance difference of conventional bank loans and Islamic bank financing.
This is explained by the significant estimated coefficient of the Islamic bank interactive dummy.
Additionally, the result indicates that Islamic bank financing will increase more than two times
than the conventional loan when real GDP increases. A 1 percentage point increase in real GDP
growth is related with roughly 0.19 percentage point increase in the growth of real gross loans of
conventional banks, and 0.55 percentage point increase in the growth of real gross financing of
Islamic banks. It is important to note that the coefficient of Islamic bank interactive dummy in
model (3) and (4) to be negative and significant at 1% alpha. The magnitude is substantially
larger than the coefficients of the real GDP growth. This fact is essentially suggests that Islamic

14
banks apparently are more counter-cyclical than conventional ones. Our result is against with
Ascarya et al. (2016) reveal that Islamic bank is more procyclical than conventional banks.
According to the paper, this procyclicality of Islamic bank is viewed as good procyclicality since
it does not resulting credit bubbles. Meanwhile, conventional bank is considered to have bad
procyclicality as it triggers credit bubbles which eventually resulting in systemic risk and
financial imbalances.

4.3. System GMM Results based on different size groups

In this study, we also consider the size of the banks to explain the bank lending
procyclicality of Islamic and conventional banks. We use the loan of total assets (LnSIZE) as a
proxy to determine the classification of size groups. We disaggregate the sample into three
subsamples: large size, medium size, and small size. The larger size is indicated by the banks
above 75th percentile. Their lnSIZE is above 17.9. The medium size is the banks between 25th
and 75th percentile. Their lnSIZE ranges from 16.9 to 17.9. The small size is the banks below the
25th percentile. Their lnSIZE is below 15.900. Table 3 below show the different size groups of
the banks with regard to bank lending procyclicality between Islamic and conventional banks.

Similar with baseline results, the specification tests also show the appropriation of the
GMM estimators in all groups. The Sargan test fails to reject the over-identification restrictions,
indicating that the Sargan test supports the GMM procedure. Moreover, the serial correlation test
(autocorrelation test) does not reject the null of second-order autocorrelation. The test also
validates the instruments of GMM analysis.

Table 3. System GMM Estimation Results (Based on different size groups)

Variables Model 1 Model 2 Model 3

(small size) (medium size) (large size)
Lit-1 0.2696* 0.3408*** 0.6857***
(0.074) (0.0000) (0.0000)
yit 0.1076*** 0.660*** 0.186***
(0.002) (0.000) (0.0000)
yit x IBi -0.531 -0.674 -0.217*
(0.461) (0.296) (0.076)
LnSIZEit-1 0.7660*** 0.609*** 0.2905***
(0.000) (0.0000) (0.000)
CAPit-1 -0.0154*** -0.0023*** -0.0278***
(0.000) (0.001) (0.000)
FUNDit-1 0.012** 0.0002** 0.005**

15
 (0.011) (0.034) (0.002)
Inft -0.0005* -0.0043*** -0.0092***
(0.087) (0.000) (0.000)

P-values
AR(2) 0.4349 0.4126 0.7496
Sargan test 0.4528 0.4374 0.756

Note: number in parentheses ( ) are p-values

***, **, * indicates significance at 1%, 5%, and 10% alpha respectively

Apparently, most of preceding results in baseline models prevail under the different size
groups category, albeit there are few exceptions. Similar with baseline results, the result from
model (1), (2), and (3) as given in table 3 confirming that bank size and capitalization ratio, and
funding ratio are significant. Turning to our key focus, we find evidence affirming procyclicality
of bank lending in Indonesia, which is in line with Bertray et al. (2015), Ibrahim (2016), and
Ascarya et al., (2016). The findings from table 3 indicate that a 1 percentage point increase in
GDP growth is associated to approximately 0.11 to 0.66 percentage point increase in the growth
of real gross loans. It is interesting to note that the medium size group of Islamic banks has a
greater Islamic financing than lower and large size groups, albeit it is not statistically significant.
A 1 percentage point increase in real GDP growth is related with 0.66 percentage point increases
in the growth of Islamic financing for medium size group. Meanwhile, the increase in Islamic
financing for small and large size group of Islamic are only accounted for 0.10 and 0.18
percentage point respectively.

Interestingly, we find only the coefficient of the large size Islamic bank interactive
dummy to be negative and statistically significant. Meanwhile, the interaction between
small/medium size Islamic bank and economic growth is not significant. This fact essentially
suggests that large Islamic bank tend to be favorable to the stability view of Islamic banks in
that they have ability to stabilize credit. Based on the estimated coefficient, the large Islamic
banks in particular can even be counter-cyclical in their financing activities.

4.4.Robustness check

To ascertain the aforementioned results, we then perform a set of robustness checks. At

first, we use net loan growth replacing a gross loan growth. Both of the Sargan and
Autocorrelation tests validate the instruments of GMM analysis. The results provided in table 4
using the system GMM further affirming the role of Islamic banks in smoothing the credit
activities. Notice that the coefficient of Islamic bank interactive dummy in model (3) and (4) to
be negative and significant at 1% alpha. Similar with prior results, their magnitude are

16
substantially larger as compared to the coefficients of the real GDP growth. In essence, the
evidence indicates on the counter-cyclicality of Islamic bank financing. As for controlling
variables, all bank specific characteristics such as asset size, capitalization ratio, and funding
ratio remains significant under the specifications of net loan growth models. Similarly,
macroeconomic factor as indicated by inflation rate carry negative and significant coefficients.
All in all, these evidences support the baseline results of bank lending procyclicality.

Table 4. System GMM Estimation results (Net loans)

Variables (1) (2) (3) (4)

Lit-1 0.4203*** 0.4479*** 0.4575*** 0.4823***
(0.0000) (0.0000) (0.0000) (0.0000)
yit 0.431*** 0.281*** 0.393*** 0.465***
(0.0000) (0.0000) (0.0000) (0.0000)
yit x IBi - - -0.426*** -0.526***
(0.0000) (0.0000)
LnSIZEit-1 0.5143*** 0.4672*** 0.4820*** 0.4382***
(0.0000) (0.0000) (0.0000) (0.0000)
CAPit-1 -0.0194*** -0.0206*** -0207*** -0.0215***
(0.000) (0.000) (0.000) (0.000)
FUNDit-1 0.0005** 0.0005** 0.0005** 0.0005***
(0.000) (0.000) (0.000) (0.000)
Inft - -0.0079*** - -0.008***
(0.000) (0.000)

P-values
AR(2) 0.1456 0.2069 0.1665 0.1632
Sargan test 0.2246 0.2077 0.2116 0.2145

Note: number in parentheses ( ) are p-values

***, **, * indicates significance at 1%, 5%, and 10% alpha respectively

We also perform the robustness check for the dynamic regressions based on different size
groups. Table 5 below show the GMM estimation results under the specification of net loan
growth model. We repeat the estimation of the baseline results in table 3 by replacing the gross
loan with net loan growth. The results add the credence of this study in that Islamic banks have a
crucial role in smoothing the credit activities. Both diagnostic tests (Sargan and autocorrelation
test) reported on that table indicates the adequacy of model estimated with GMM procedure.
Apparently, all preceding results in table prevail under the specification of net loan growth
model based on size groups. Most of bank specific characteristics carry significant coefficients,

17
except for funding ratio. As for other control variables, such as inflation rate remains negative
and statistically significant, except form small size group. Similar with baseline results, it is
found that only large size Islamic bank interactive dummy have a negative and significant
coefficient. While the interaction between small/medium size Islamic bank and real GDP growth
is not significant. This result ascertains the stability view of Islamic banks particularly applied
to large size group of Islamic banks.

Table 5. System GMM Estimation results (Net loans, based on different size groups)

Variables Model 1 Model 2 Model 3

(small size) (medium size) (large size)
L1it-1 0.2686*** 0.3408*** 0.4761***
(0.0074) (0.0000) (0.0000)
yit 0.1076*** 0.661*** 0.169*
(0.002) (0.000) (0.091)
yit x IBi -0.153 -0.694 -0.2102*
(0.461) (0.296) (0.0607)
LnSIZEit-1 0.7660*** 0.6093*** 0.3936***
(0.0000) (0.0000) (0.0000)
CAPit-1 -0.0154*** 0.0022** -0.0113***
(0.000) (0.001) (0.000)
FUNDit-1 0.0012 -0.0002 0.0002**
(0.119) (0.314) (0.003)
Inft 0.0005 -0.004*** -0.0112***
(0.874) (0.000) (0.000)

P-values
AR(2) 0.4349 0.4126 0.231
Sargan test 0.4629 0.8153 0.278

Note: number in parentheses ( ) are p-values

***, **, * indicates significance at 1%, 5%, and 10% alpha respectively

From these robustness checks, the evidence tends to show that Islamic banks have a role
in smoothing credit during economic imbalances. In addition, we have no support whatsoever
that Islamic bank is procyclical in their financing activities. In fact, there is ample evidence to
say that Islamic bank financing can be counter-cyclical. All in all, our results ascertain to the
stability view of the Islamic banks in smoothing their lending behavior.

18
5. Conclusion and policy recommendations

We investigate the main inquiry of which type of banks (i.e Islamic versus conventional
banks) tend to show stability in their lending behavior during economic shocks. Ours is among a
few studies in literature that examine the lending behavior of Islamic banks particularly in
emerging countries. By using dynamic system GMM approaches and bank-level panel data set
of 60 banks over the period of 2001-2015, we affirm the bank lending procyclicality for both
Islamic and conventional banks in Indonesia. Interestingly however, when we categorize
between conventional loans and Islamic financing, the evidence tends to shows that bank
procyclicality only applied for conventional banks. As for Islamic banks, we find no support that
Islamic bank is more procyclical in their financing. In fact, this study suggests that Islamic banks
in general and large size Islamic banks in particular can even be counter-cyclical in their
financing activities. We arrive at these conclusions under a set of empirical investigations
including an alternative loan measure model.
Accordingly, the study unveils the tip of iceberg of the role played by Islamic banks in
smoothing their credit during the time of economic downturns. We document the significance
coefficient of the Islamic bank interactive dummy (see table 2 and 4) and of large size Islamic
bank interactive dummy (see table 3 and 5). In all cases, Islamic banks are tend to be counter-
cyclical than conventional ones. This study supports the view of Farooq and Zaheer (2015) and
Ibrahim (2016) that Islamic banks tend to be more stable. Islamic banks in Indonesia have been
able to stabilize their financing during the recent economic recessions. Because of this, it is not
surprising that Indonesian Islamic banks still perform well during the global financial crisis
albeit the crisis hit hard the global economy.
As for the regulators, procyclicality as one the major causes of systemic risk should be
well understood. Islamic banks in Indonesia tend to be counter-cyclical, while conventional ones
is more procyclical in their lending behavior. As a consequence, it is required to established a
sound framework and effective instruments to address the procyclical issues between the two
banking system. Additionally, Islamic macroprudential policy is necessitated to develop a sound
framework and effective instruments in addressing financial instability. Since both Islamic and
conventional banks are operated in same horizon, so that macroprudential policies and
framework for Islamic and conventional banks should be unique and effective to prevent
systemic risk and financial imbalances.

19
References

Abdellah, B. (2009). Pro-cyclicality of the Islamic financial system: theoretical essay and
comparative study. Komar University of Science and Technology, Kurdistan.

Abdul-Rahman, A., Abdul Latif, R., Muda, R., & Abdullah, M. A. (2014). Failure and potential
of profit-loss sharing contracts: A perspective of New Institutional, Economic (NIE)
Theory. Pacific Basin Finance Journal, 28, 136151.
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations. The review of economic
studies, 58(2), 277-297.

Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of error-
components models. Journal of econometrics, 68(1), 29-51.

Ascarya, Rahmawati, S., & Karim, A. A. (2016). Testing the Procyclicality of Islamic and
Conventional Banks in Indonesia. In Macroprudential Regulation and Policy for the Islamic
Financial Industry.
Aysan, A., Disli, M., & Ozturk, H. (2016). Financial crisis, macroprudential policies and
depositor discipline. The Singapore Economic Review, 62(1).
Aysan, A. F., Dolgun, M. H., & Turhan, M. I. (2013). Assessment of the Participation Banks and
Their Role in Financial Inclusion in Turkey. Emerging Markets Finance and Trade, 49(s5),
99111.
Azmat, S., Skully, M., & Brown, K. (2015). Can Islamic Banking Ever Become Islamic?
Pacific-Basin Finance Journal.
Baltagi, B. (2008). Econometric analysis of panel data. John Wiley & Sons.

Baum, C. F. (2006). An introduction to modern econometrics using Stata. Stata press.

Beck, T., Demirg-Kunt, A., & Merrouche, O. (2013). Islamic vs. conventional banking:
Business model, efficiency and stability. Journal of Banking & Finance, 37(2), 433447.
Bertay, A. C., Demirg-Kunt, A., & Huizinga, H. (2015). Bank ownership and credit over the
business cycle: Is lending by state banks less procyclical?. Journal of Banking &
Finance, 50, 326-339.

Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel
data models. Journal of econometrics, 87(1), 115-143.

Blundell-wignall, A., & Roulet, C. (2014). Macro-prudential policy, bank systemic risk, and
capital controls. OECD Journal: Financial Market Trends, 2013(2), 122.

20
Buiter, W. H. (2014). The Role of Central Banks in Financial Stability: How Has It Changed?
World Scientific, 30, 1156.

Chapra, M. U. (2009). The Global Financial Crisis: Some suggestion from Islamic Finance.
Kyoto.
Chong, B. S., & Liu, M.-H. (2009). Islamic banking: Interest-free or interest-based? Pacific-
Basin Finance Journal, 17(1), 125144.
ihk, M., & Hesse, H. (2010). Islamic Banks and Financial Stability: An Empirical Analysis.
Journal of Financial Services Research, 38(2-3), 95113.

Farooq, M., & Zaheer, S. (2015). Are Islamic Banks More Resilient During Financial Panics?
Pacific Economic Review, 20(1), 101124.
Galati, G., & Moessner, R. (2013). Macroprudential policy - a literature review. Journal of
Economic Surveys, 27(5), 846878.
Ghosh, S. (2014). Risk, capital and financial crisis: Evidence for GCC banks. Borsa Istanbul
Review, 14(3), 145157.

Ghosh, S. (2016). Macroprudential policies, crisis and risk-taking: Evidence from dual banking
systems in GCC countries. Journal of Islamic Accounting and Business Research, 7(1), 6
27.
Gujarati, D. N. (2003). Basic Econometrics. 4th. New York: McGraw-Hill.

Hasan, M., & Dridi, J. (2011). the Effects of the Global Crisis on Islamic and Conventional
Banks: a Comparative Study. Journal of International Commerce, Economics and Policy,
02(02),
Husman, J. (2015). Essays on Banking and Monetary Policy in the Presence of Islamic Banks.
University of Warwick.
Ibrahim, M. H. (2016). Business cycle and bank lending procyclicality in a dual banking system.
Economic Modelling, 55, 127134.
Imaduddin, M. (2007). Determinants of Banking Credit Default in Indonesia: A Comparative
Analysis. Working Paper

Johnes, J., Izzeldin, M., & Pappas, V. (2014). A comparison of performance of Islamic and
conventional banks 20042009. Journal of Economic Behavior & Organization, 103, S93
S107.
Laeven, L., & Valencia, F. (2008). Systemic banking crises: a new database.IMF working
papers, 1-78
Landau, J. P. (2009). Procyclicality: what it means and what could be done. In remarks at the
Bank of Spains conference on Procyclicality and the Role of Financial Regulation,
Madrid, May.

21
Lee, M., Asuncion, R. C., & Kim, J. (2016). Effectiveness of Macroprudential Policies in
Developing Asia: An Empirical Analysis. Emerging Markets Finance and Trade,
forthcomin(June), 923937.
Louati, S., & Boujelbene, Y. (2015). Banks stability-efficiency within dual banking system: a
stochastic frontier analysis. International Journal of Islamic and Middle Eastern Finance
and Management, 8(4), 472490.
Maghyereh, A. I., & Awartani, B. (2014). Bank distress prediction: Empirical evidence from the
Gulf Cooperation Council countries. Research in International Business and Finance,
30(1), 126147.
Micco, A., & Panizza, U. (2006). Bank ownership and lending behavior.Economics
Letters, 93(2), 248-254.

Mirzaei, A., & Moore, T. (2015). Banking performance and industry growth in an oil-rich
economy: Evidence from Qatar. The Quarterly Review of Economics and Finance, 112.
Nabi, M. S. (2013). Access to Finance and Investment: Does Profit Sharing Dominate Debt?,
(49815).
Nursechafia, N., & Abduh, M. (2014). The Susceptibility of Islamic Banks Credit Risk Towards
Macroeconomic Variables. Journal of Islamic Finance,3(1).
Thomson Reuters (2015). State of Global Islamic Economy Report 2015. Dubai: Thomson
Reuters & Dinar Standard.

Utari DGA, Arimurti T, Kurniati IN (2011) Prosiklikalitas Sektor Perbankan dan Assessment
Risk Taking Channel. Bank Indonesia Working Paper WP/09/2011

Zakaria, F. (2015). Systemic Risk and Financial Contagion in Morocco: New Approaches of
Quantification. In Overlaps of Private Sector with Public Sector around the Globe (pp.
141171).
Zhang, L., & Zoli, E. (2016). Leaning against the wind: Macroprudential policy in Asia. Journal
of Asian Economics, 42, 3352.

22