Using partial least square discriminant analysis to distinguish between Islamic and conventional banks in the MENA region

• Author: Sami Ben Jabeur,Asma Sghaier,Boutheina Bannour
• Published date: 01 April 2018
• Date: 01 April 2018
• DOI: http://doi.org/10.1002/rfe.1018

ORIGINAL ARTICLE

Using partial least square discriminant analysis to distinguish

between Islamic and conventional banks in the MENA region

Asma Sghaier

1

|

Sami Ben Jabeur

1

|

Boutheina Bannour

2

1

ISG Sousse, LAMIDED Research

Center, University of Sousse, Sousse,

Tunisia

2

FSEG, University of Sousse, Sousse,

Tunisia

Correspondence

Asma Sghaier, ISG Sousse, LAMIDED

Research Center, University of Sousse,

Sousse, Tunisia.

Email: xasma_sghaier1983@yahoo.fr

Abstract

The deterioration of bank profitability poses a threat not only to the interests of

consumers and internal staff members but also affects investors who may equally

suffer from significant financial losses. It is important to establish an effective

system which assists investors in their investment choices. In prior literature, tra-

ditional models have been developed, but achieved short-term performances such

as logistic regression and discriminant analysis. This paper applies a partial least

squares discriminant analysis (PLS-DA) to distinguish between conventional and

Islamic banks in the Middle East and North Africa (MENA) region based on the

financial information for the period 2005–2011. This method can successfully

identify the non-linearity and correlations between financial indicators. The results

demonstrate superior performance of the proposed method. On one hand, our

model can select all financial ratios to distinguish between banks and at the same

time identify the most important variables in the distinction process. On the other

hand, the proposed model has high levels in terms of accuracy and stability.

KEYWORDS

discriminant analysis, Islamic banks, logistic regression, partial least squares, principal component

analysis

1

|

INTRODUCTION

The recent financial crisis has led to difficulties in many conventional banks in the United States and, by a domino effect,

around the world. Islamic banks, however, were largely spared the repercussions of the crisis (Johnes, Izzeldin, & Pappas,

2014; Maghyereh & Awartani, 2014; Mollah & Zaman, 2015). It appeared that, in accordance with the principles of Sharia,

their highly regulated operating guidelines prohibit investments in risky financial products that are at the root of the finan cial

crisis (Hassan, Ngene, & Yu, 2015). Consequently, the attention of policy-makers and investors was largely drawn to Islamic

financing in recent years (Johnes et al., 2014). In fact, there are more than 300 Islamic financial institutions worldwide. Sev-

eral factors may explain this interest-free financing pattern including strong demand for sharia-compliant products, improved

legal and regulatory framework for Islamic finance, the increasing demand from conventional investors, and the ability of the

industry to innovate and to develop financial instruments that meet the needs of investors. Indeed, given the international

spread of Islamic banking practices, a study comparing the performance of Islamic and conventional banks is of general inter-

est. Previous studies that focus specifically on the performance of Islamic banks compared with conventional banks are incon-

clusive in their findings. In theory, there are many differences between Islamic and conventional banks. For example,

contracts for interest in conventional banks are replaced in the Islamic bank. According to contracts where profits and losses

as well as the risks are shared between the creditor and the borrower. Several other studies (Abedifar, Molyneux, & Tarazi,

2013; Beck, Demirg€

ucß-Kunt, & Merrouche, 2013; Gamaginta & Rokhim, 2011; Pappas, Izzeldin, & Fuertes, 2012) compared

Received: 24 January 2017

|

Revised: 24 November 2017

|

Accepted: 9 February 2018

DOI: 10.1002/rfe.1018

Rev Financ Econ. 2018;36:133–148. wileyonlinelibrary.com/journal/rfe ©2018 The University of New Orleans

|

133

the relative stability of Islamic and conventional banks at different times and in different countries. Some of this literature

concludes that Islamic banks are more stable; it is important to consider whether the Islamic and conventional banks behave

similarly or differently. Specifically, we test the financial ratios to see if they can distinguish between the two types of banks.

This article proposes using PLS-DA regression as a new alternative to distinguish between the two types of banks. This

method can successfully identify the non-linearity and correlations between financial indicators. Investors as well as analysts

require tools that are capable of predicting which categories these banks are placed in and this with extremely high precision.

However, it is very hard to select an appropriate model in the course of processing linear data. A possible solution is to use a

regression analysis on the main components, which are orthogonal. This article aims at determining whether Islamic and con-

ventional banks in the MENA region are distinguishable on the basis of financial information. More precisely, we examine

whether researchers or regulatory bodies are able to come up with an accurate classification for Islamic or conventional banks

through the use of 19 financial ratios. While many studies have used data mining methods in predicting bankruptcy and credit

risk, few studies are interested in the value of accounting information, however, distinguishing Islamic and conventional banks.

Our primary aim will be to study these novel methods in this context. Two research questions are addressed in this paper: what

ratios discriminate between the two types of banks? Are non-parametric methods more efficient than traditional? The strength

of the PLS Discriminant Analysis arises from its ability to deal with multicollinearity, since it turns original variables into

orthogonal components (Barker & Rayens, 2003; Bastien, Vinzi, & Tenenhaus, 2005). It is also able to function with missing

data and able to function even when numbers of variables exceeds that of the observations, in contrast with the regression

method on main components or stepwise regression, we withhold at the end of iterations of all explanatory variables.

The remainder of the paper is structured in the following manner. Section 2 presents a literature review on comparing

Islamic banks to conventional banks. Section 3 describes the methodology followed in the paper, presenting the data sam-

ple and explanatory variables selected for inclusion in the model. Section 4 reports the empirical results and discussion.

The last section concludes and suggests directions for future research.

2

|

LITERATURE REVIEW

Many empirical studies focused on comparing the profitability between Islamic and conventional banks. Some works indi-

cate that Islamic banks are more profitable compared with their conventional counterparts (Azhar Rosly & Afandi Abu

Bakar, 2003; Iqbal, 2001; Olson & Zoubi, 2008). However, other empirical studies found different results (Ahmed, 2007;

Kader, Asarpota, & Al-Maghaireh, 2007; Metwally, 1997). Iqbal’s study revealed that Islamic banks report better equity

profitability as well as asset profitability compared with a reference sample from the conventional banks from 1900 to

1998. A study conducted by Azhar Rosly and Afandi Abu Bakar (2003) also showed that the asset profitability and the

profit margin of Islamic banks are higher in Malaysia from 1996 to 1999. However, many other empirical studies yielded

varying results (Ahmed, 2007; Kader et al. 2007; Metwally, 1997). Metwally (1997) compared a sample of 30 banks where

15 of them were Islamic in the period between 1992 and 1994. His results indicate that Islamic banks do not seem much

different from their conventional counterparts in terms of asset profitability and return rate on deposits. Based on this

empirical literature, the difference in profitability between Islamic and conventional banks can be explained by the differ-

ence in these profitability determinants between these two types of banks. 

Cih

ak and Hesse (2010) found that smaller banks

have proven to be more stable than those operating on a larger scale. The suggested explanation would be that Islamic

banks tend to show difficulties in monitoring credit risk when the undertaking becomes larger and thus more complex.

Smaller banks tend to focus on low-risk investments, whereas larger Islamic banks extend their financing to higher risk pro-

jects. A considerable number of authors such as Abreu and Brunnermeier (2002); Athanasoglou, Delis, and Staikouras

(2006); Demirg€

ucß-Kunt and Huizinga (1999); Goddard, Molyneux, and Wilson (2004) and Molyneux and Thornton (1992),

and analyzed performance based on data from many countries. Analyses on the determinants of banking performance wer e

a fundamental component of research in many countries and hinged on two categories of banks: Islamic on one hand

(Bashir & Hassan, 2003; Sanusi & Ismail, 2005; Srairi, 2008) and conventional on the other hand (Athanasoglou, Bris-

simis, & Delis, 2008; Dietrich & Wanzenried, 2011; Olson and Zoubi, 2011; Pasiouras & Kosmidou, 2007; Srairi, 2008).

According to Dridi and Hasan (2010), the crisis provided Islamic banks with the golden opportunity to showcase their resi-

lience; nevertheless, it also revealed major difficulties that needed to be addressed in order to create a more sustainable pro-

ductive base and growth trajectory. Dietrich and Wanzenried (2011) and Rouissi (2010) show that the size of the banking

organization affects profitability positively. Bashir and Hassan (2003) found out the same result for Islamic banks.

Pasiouras and Kosmidou (2007) showed that size can have an adverse impact on profitability. By contrast, a study by

Athanasoglou et al. (2008) showed that the size of the bank has no significance with regard to its profitability. Empirical

134

|

SGHAIER ET AL.