Examining Banking Productivity Drivers in MENA Banks After Financial Liberalisation in 1990s

The study investigates whether deregulation and economic reforms transformed the MENA banking sector into a more productive and efficient market. This is the first study to cover a large sample of 11 MENA countries for an extended and recent period (1999-2012). Initially this paper estimates the productivity and efficiency of MENA commercial banks using Malmquist DEA to estimate productivity (TFP), technological and technical efficiency, and scale efficiency change in order to investigate to what extent banking productivity in MENA economies has improved during the study period. Then, Tobit model is employed to examine the impact of bank and macroeconomic variables on the total factor productivity of MENA commercial banks. The obtained MPI results suggest that commercial banks operating in the Gulf countries, namely Saudi Arabia, Kuwait, Qatar, Oman, Bahrain and United Arab Emirates have exhibited productivity progress mainly due to the technological progress rather than efficiency change. Results also suggest that expenses preference behaviour leads to be more productive banks in the examined period and study countries. Whilst banking productivity is enhanced by financial reforms and technological progress, our findings overall do not indicate that foreign participation or state ownership lead to enhance productivity of banks, whilst suggesting that a different mix of policies should be adopted depending on the characteristics of the banking system in the examined countries.


Introduction
In the absence of significant role of efficient and well-developed capital market, the banking sector in MENA region plays a leading role in the economic development via providing funds to private and public sectors investments and by financing government deficits. Whilst also engaged in major economic reforms, which were required by the World Trade Organisation (WTO), to which these countries either belong or plan to join 1 . It is argued that these reforms are fundamental parameters for the performance of national financial systems, and the development of the economy in general (Gattoufi et al., 2009). During the last three decades, MENA economies have been witnessed major developments in terms of liberalisation of national economy, elimination of either capital or ownership barriers, transferring the control of ownership of large shares of the banking sector from the state to private local investors and to foreign investors (Turk-Ariss, 2009;Farazi et al., 2011;.The increasing number of bank failures in the 1980s and liberalisation of the banking industry in 1990s resulted in increasing academic interest in the examination of banking productivity and efficiency (Beger & Humphry, 1997;Jaffry et al., 2007;. The productivity and source of productivity for commercial banks are highly important, since the existence of productive and efficient commercial banks contributes in lowering operational costs, enhancing profitability which in turns will be reflected on the stability of the financial sector and improves the overall economy and growth. Economic liberalisation and financial reforms may contribute in improving total factor productivity of banks in terms of technological change and efficiency change. In this matter, the amount to which financial institutions such as banks are capable to achieve the best allocation of funds is one of concerns raised by policymakers and regulators in MENA countries In this context, the importance and contribution of this paper is threefold. Firstly, investigating productivity and its sources and determinants of MENA banks is crucial not only in terms of policy implications to optimally structure the banking system of these countries, but, most importantly, for the efficient allocation of funds and sustainable development of these regional economies in the long run. In particular, MENA region economies have been witnessing an unprecedented transformation from being solely petroleum producers to having diversified economies (Market based-economy). To succeed, this transformation requires an efficient institutions, something which, in other economies, is traditionally performed by both capital markets and banking sector. But, the fact is the capital markets in MENA economies are neither efficient nor well-developed, so banking institutions almost monopolise the role of financial intermediation. Thus, the existence of a wellfunctioned banking sector is crucial, not only for the development of the financial sector in these countries, and the optimal allocation of funds, but for the future of these economies as a whole. Additionally, the productive and efficient banks would contribute in reaching cost efficiency and improving performance and profitability which in turns leads to maintain the stability and the overall economic productivity and growth of a country. The role of state and foreign ownership in the banking system of the examined countries had not been adequately assessed in past studies. To the best of my knowledge, this is the first study in MENA region to cover a large sample of 11 MENA countries for an extended and recent period (14 years) using Data Envelopment Analysis (DEA) approach and filling a significant gap in research using this approach for MENA economies (Ramanathan ,2006;Ramanathan ,2007;Gattoufi et al. 2008;Gattoufi et al., 2009); Ben Naceur et al. 2009;Al Hammadi ,2013). Secondly, findings of this paper would help to draw, for the first time, reliable conclusions about how ownership, macroeconomic and bank-specific variables affect the total factor productivity in MENA banks after the liberalisation of the national economy. This paper is also the first in MENA that examines if state and foreign ownership contributed to an increase in productivity of banks using Tobit model. Moreover, results of the Tobit model contribute by evaluating the impact of, bank-specific variables, macroeconomic and bank ownership, year effects for financial liberalisation on productivity of MENA banks. These findings can be helpful in forming government policies intended to facilitate optimum foreign participation in a way which contributes to improve performance of banks and promote the competitive environment in MENA economies. This paper is structured as follows: Section 2 reviews past literature and generates hypotheses that are examined in this study. Section 3 presents the methodology and model used, while section 4 introduces the empirical results of DEA analysis. The empirical findings of Tobit regression analysis are discussed in section 5 and finally implications are undertaken in section 6. Berger and Humphrey (1997) outlined that there are two major approaches for measuring efficiency: the parametric approach and nonparametric approach. In general, the parametric approach specifies a functional form for the cost, revenue, profit or production relationship among inputs, outputs and other factors such as the environment. The most common method used in the parametric approach is Stochastic Frontier Approach (SFA). In contrast, the bestknown nonparametric approach is Data Envelopment Analysis (DEA), which is a linear programming formulation that outlines a nonparametric relationship between multiple outputs and multiple inputs used, or it can be defined as a mathematical model for making production frontiers and measuring the relative efficiency of these frontiers (Humphrey and Beger, 1997). According to Berger and Humphrey (1997), the main disadvantage of the nonparametric approach (DEA) that assumes there is no random error in constructing the frontier. In this regard, Vassiloglou and Giokas (1990) pointed out that DEA measures efficiency by estimating an empirical production function, which represents the highest value of outputs/benefits that could be generated by inputs/resources as given by a range of observed input/output measures. The original idea of DEA is to present a methodology whereby, within a set of comparable decision making (DMUs), those exhibiting best practice could recognised, and would produce an efficient frontier (Cook and Seiford, 2009). Therefore, DEA can help to determine an efficient frontier that contains the majority of efficient decision making units (DMUs) such as banks. Linear programming methods tolerate the construction of best practice cost and production frontiers from these data and performance of a bank is judged relative to this frontier (Al Shamsi et al., 2009. There is no lack that temporarily provides a decision making unit better measured one year from the next, and no inaccuracies generated by accounting rules that would make measured outputs and inputs differ from economic outputs and inputs. In this section DEA applications in the banking industry will be discussed as several studies have examined the efficiency and productivity of banks by employing the DEA approach.

Data Envelopment Analysis (DEA) Literatures and Hypotheses Development
However, it should be noted that there is a problem specification of DEA inputs and outputs; Sathye (2003) pointed out that inputs and outputs could be divided into two approaches, which are an intermediation approach and production approach. The intermediation approach, banks are regarded as financial intermediates which transfer financial assets between surplus parties and deficit parties. Therefore, outputs can be containing loans and deposits, whereas inputs may be workforce, fixed assets and loanable moneys. On the other hand, the production approach, banks are considered to be producers of banking services for customers, which are associated with transformation of deposit accounts and process loans. Outputs in this model may consist of interest income and non-interest income, while inputs can include labour and physical assets.
Liberalisation policies undertaken in developing countries are essentially intended at promoting competition in the domestic market and enhance efficiency and productivity of financial firms. Isik & Hassan, (2003) argued that enhancing productivity and performance can be achieved via resource management and locating them in a position where their survival and success will rely on their ability to adapt and operate efficiently in the new liberalised market. However, in finance and banking literatures, there are several different techniques used to estimate productivity change such as the Fisher index, the Tornqvist index and the Malmquist index. The most common method for measuring productivity is (MPI) Malmquist Productivity Index (Fare et al. 1994). Studies on productivity using MPI indices have been carried out in a number of different industries in well-developed and developing economies (Isik & Hassan, 2003;Jaffry et al. 2007;Gattoufi et al. 2009;. Achieving productivity and efficiency of banks is matter for different parties; it is very essential to obtaining appropriate resource allocations, which in turns benefits the whole society and ensuring better innovations, enhancing profitability, and creating suitable environment conditions of competition as well as safeguarding the stability of the financial sector. Nevertheless, the evidence of the effect of liberalisation on productivity of financial institutions and other industries are mixed. For instance, in well-developed economies, Fare et al. (1994) analysed productivity growth in seventeen OECD countries and suggested that US productivity growth is relatively higher than average. They suggest that the growth has resulted from technological progress, whereas In Japan, productivity growth was due to efficiency change. In EU economies,  observed that commercial banks in Spain have had to some extent lower rate of productivity growth, but a to some extent higher rate of potential productivity growth. Such development is linked to differences in both managerial efficiency and institutional efficiency, to differences in technological progress, and to the adverse effect of diseconomies of scale in the commercial banking industry.
However, studies in other markets, in particular emerging and developing markets have been looked at investigating banking productivity in Malaysia, India, China, Turkey and MENA countries. Isik and Hassan (2003) employed DEA using MPI to examine how financial reforms affect total factor productivity, efficiency and technology of Turkish banks over the period [1981][1982][1983][1984][1985][1986][1987][1988][1989][1990]. Findings imply that performance of all types of banks exhibited major improvements after liberalisation and the productivity was mainly driven by increasing in technical efficiency attributed to improvements of management practices rather than technological progress. In the same context, in Malaysia, Krishnasamy et al. (2003) examined the nature and extent of the productivity change of ten commercial banks operating in Malaysia for the period 2000-2001, finding that the total factor productivity increased, but two banks which showed a decline in productivity. Overall, the total factor productivity growth for Malaysian commercial banks was attributed to technological progress rather than technical efficiency change reflecting investment in technology. On the other hand,  provided empirical evidence about the impact of economic globalisation on total factor productivity in Malaysian banks. In the first stage, they employed the MPI to estimate the productivity of the Malaysian banking industry during the period 1998-2007. The empirical results showed that banks have experienced productivity growth, mainly due to efficiency increases rather than progress in technology. The decomposition of the efficiency change index into pure technical and scale efficiency components suggests that the increase in efficiency of Malaysian banks was related to the rise in scale efficiency. The second stage of this study was to obtain empirical results from the panel regression analysis. Findings report that banks with higher amounts of income originating from non-interest income are more productive. Whilst, credit risk appears to exert deteriorating effects on a bank's total productivity levels.
In India, Pakistan and Bangladesh, there have been a number of studies focusing on the impact of deregulation on banking sector. Howcroft and Ataullah (2006) examined the productivity of banks operating in India and Pakistan, and they found both countries have witnessed improvements in total factor productivity which was higher when the government's policy objective was used. In addition, they confirmed that state banks exhibited very little improvement in total factor productivity as a result of their inability to adopt new technology and the presence of high non-performing loans. In contrast, foreign banks showed the highest growth in productivity due to an improvement in their efficiency and technological innovation. Moreover, Jaffry et al. (2007) aimed at estimating changes in productivity and technical efficiency level within banking sectors of the Indian sub-continent: specifically India, Pakistan and Bangladesh for the period 1993-2001. This study was done in the context of a number of sweeping reforms across sub-continent in the early 1990s. They found that technical efficiency increases and converges across the Indian sub-continent. Also, it was observed that India and Bangladesh experienced an immediate and sustained increase in technical efficiency, while Pakistan suffered from a decline in efficiency over the years of the examined period.
However, in MENA economies, this market has witnessed varying levels of economic development over the past few decades: from the 1970s to the present date, as the economic performance of this region has experienced major developments due to changes in oil crude prices and other political factors. The banking sector in MENA economies is committed to sustainable development and is engaged in major economic reforms as required for the adhesion to the World Trade Organisation (WTO). In addition, as previously mentioned at the beginning of this paper, due to the absence of efficient and well-developed stock exchange markets, the banking sector in this region still plays a leading role in the intermediation process between users and suppliers of funds. From this point, the existence of productive and efficient banking sector would help to maintain the stability of the financial system and improve the overall productivity and growth of the economy. Ramanathan (2006) employed MPI to assess the comparative performance of selected MENA economies over the period 1980-1999, findings telling that technological progress have contributed more to improvements of total factor productivity than changes in technical efficiency. In another study, Ramanathan (2007) also suggested that banks operating in four of the six GCC countries (Bahrain, Kuwait, Saudi Arabia and UAE) experienced productivity growth during 2000-2004. The selected banks in Bahrain have revealed the highest productivity improvements during the examined period; whereas, the selected banks in Qatar have shown the highest decline in the productivity.
In the same matter, the impact of mergers and acquisitions on the efficiency of commercial banks in MENA countries is investigated by Gattoufi et al. (2008). To examine such development, the MPI has been utilised to analyse the evolution of efficiency over time. In this study, two inputs are considered namely interest expenses and operating expenses, whereas outputs are interest income and operating income which makes this model in line with the intermediation approach. Results of the study exhibited the positive impact that mergers and acquisitions had on the banking industry, particularly in the MENA economies.
Another study (Gattoufi et al., 2009) examined the impact which change in ownership has had on the efficiency of commercial banks in MENA economies. This study considered two inputs, namely operating expense and interest expense. Outputs were considered to be interest income and operating income. To make this model consistent with the intermediation approach model, deposits, interest expense, and non-interest expense were used as inputs, while loans, interest income and non-interest income were used as outputs. This study also used a second model which included an additional input, namely loan loss provisions which can be used as proxy for non-performing loans. Findings emphasise the positive impact, in spite of limited change. Regarding the decomposition of the MPI and the technical efficiency scores, one can indicate that the impact of change in ownership has affected in the scale efficiency rather than in pure technical efficiency.
Ben Naceur et al. (2009) investigate the impact of deregulation policies on the performance of selected MENA commercial banks over the period 1993-2006. They assessed bank efficiency using a nonparametric model (DEA), then employed Tobit regression model to examine the impact of institutional, financial and bank characteristics variables on banks' efficiency obtained by DEA. This study implemented the intermediation approach as inputs used are total costs (personnel expenses + other administrative expenses +interest paid + noninterest expenses), while outputs are total loans and non-lending activities (earning assets). In respect to (DEA), results indicate that Morocco and Tunisia have demonstrated efficient banking systems when compared to other selected MENA countries. Banks in Jordan, however, seem to catch up with best practice from 2003. For the second stage, the empirical findings present a robust relationship between some environmental measures and cost efficiency. In this regard, results indicate that well capitalised and liquid banks recorded higher efficiency scores. Additionally, results revealed that banking sector development, measured by credit provided by banks to the private sector in a low regulated business, is more likely to reduce bank efficiency. Further, highly concentrated banking sectors tends to decrease banks' efficiency and financial reforms enhance efficiency of commercial banks in MENA, with Egyptian banks exhibiting the lowest efficiency in the region.
The impact of the 2007-2008 financial crisis on Islamic financial Institutions (IFIs) in Gulf Countries Council (GCC) is examined by Al Hammadi (2013). This study investigated the differences between GCC and non-GCC IFIs' efficiency and whether or not the IFIs' efficiency was enhanced during the post and pre-financial crisis period. DEA and MPI were employed on a balanced panel of 22 GCC and 19 non-GCC IFIs to examine the link between IFI size (total assets) and annual total factor productivity change. This study selected total assets and total equity as inputs and return on average assets, return on average equity and net income as outputs. The results stated that generally, GCC and non-GCC IFIs' efficiency was not significantly impacted by the financial crisis. Moreover, MPI results indicate that both GCC and non-GCC IFIs revealed a progress in efficiency during the study period, whereas scale efficiency was the least source of efficiency. The study also reports that efficiency and size of IFI is positively related but insignificant. Also, Al Hammadi suggested some policies and managerial implications as he pointed out that regulators and policymakers should pay more attention to bank efficiency when they pass new policies and regulations. Overall, the above arguments would lead me to generate the first hypothesis for this paper as follows: Hypothesis 1 (H1): Banking productivity in MENA economies has improved during the examined period.
However, this paper also provides empirical evidence on the impact of financial liberalisation using year effects and determinants of total factor productivity in MENA banking economies. Therefore, in the second stage of the analysis, the Tobit model is employed to investigate whether the financial liberalisation and other variables such as size of banks, risk, market structure and macroeconomic variables have an effect on the productivity of commercial banks. In this matter, in EU market, Rezitis (2004) investigated the productivity growth and technical efficiency in the Greek banking sector using MPI. The empirical results from the Tobit model reveal that size and specialisation have had positive impacts on both pure and scale efficiency. However, In Asia, Das and Kumbhakar (2012) investigated the impact of banking deregulation on efficiency and productivity change in the Indian banking sector. Empirical findings suggest that banks experienced a growth in their efficiency (from 61% in 1996 to 72% in 2005) during the post-deregulation period.
Within the context of developing and MENA banking sector, previous studies (Isik and Hassan, 2003;Krishnasamy et al. 2003;Howcroft and Ataullah, 2006;Ben Naceur et al. 2009;Das and Kumbhakar, 2012) suggested that performance of all types of banks exhibited major improvements after liberalisation, and such improvements were attributed to improvements of management practices, more investment in technology and positive impact of mergers and acquisitions on the banking sector. Ben Naceur et al. (2009) examine the impact of deregulation policies on the performance of selected MENA commercial banks over the period 1993-2006, employing Tobit model to examine the impact of institutional, financial and bank characteristics variables on banks efficiency. The empirical findings present a robust relationship between some environmental measures and cost efficiency, implying that well capitalised and liquid banks recorded higher efficiency scores. Results also revealed that banking sector development, measured by credit provided by banks to the private sector in a low regulated business, is more likely to reduce bank efficiency. Further, highly concentrated banking sectors tends to decrease banks' efficiency and financial reforms enhance efficiency of commercial banks in MENA, with Egyptian banks exhibiting the lowest efficiency in the region. Such literature and past studies in different emerging and developing economies would lead to formulate the second hypothesis of this study as follows:

Hypothesis 2 (H2): The productivity of banks in MENA economies is influenced by a range of bank, market structure and macroeconomic variables
The empirical findings of this paper are interesting from the policy makers' perception and bank managements. They will be more motivated to find what factors to achieve the optimal utilisation of capacities and ensuring that they are being optimised over the production of banking products and services. Taking such factors into account, the empirical results of this paper will demonstrate considerable policy implications.

The Data Envelopment Analysis (DEA) Methodology
The DEA approach was suggested by Charnes et al. (1978) in response to the needs for satisfactory procedures to measure the relative efficiency of multiple inputs-outputs production units. The DEA approach mainly aims at providing a methodology to create a set of comparable decision making units (DMUs) to identify those which have best practice and efficient frontier. The methodology of this paper encompasses measurements of the efficiency performance and productivity of MENA commercial banks for the period 1999-2012.

Specification of DEA Inputs and Outputs
Using DEA requires specifying inputs and outputs to estimate productivity and efficiency for banks (Berger and Humphrey, 1997). In the banking literature, there are two different perspectives for using DEA (intermediation approach and production approach). With respect to intermediation approach, banks are considered to be financial intermediates that aim at converting financial resources between surplus firms to deficit firms. Based on this view, outputs can be loans and deposits, whereas inputs may comprise labour, fixed assets, and loanable funds. On the other hand, the production approach considers banks to be producers of financial services for their customers that seek to execute transactions on deposits accounts and process loans. Therefore, outputs under this view may consist of interest income and noninterest income, while, inputs can be physical capital and the number of employees (Luo, 2003). Accordingly, MENA banks are treated as intermediaries between savers and borrowers, producing three outputs namely total loans (y1), interest income (y2) and noninterest income (y3), by using total deposits (x1), total fixed assets (x2), non-interest expense (x3) and non-interest expense (x4) see table 1.

DEA using the Malmquist Productivity Index (MPI)
This paper examines the productivity and efficiency of MENA banks using Malmquist DEA to estimate total factor productivity change (TFPCH), technological and technical efficiency, and to scale efficiency change in order to investigate first hypothesis. This index was primarily developed by Malmquist in 1953. The Malmquist productivity index (MPI) utilises panel data to compute indices of total factor productivity change and scale efficiency change (Krishnasamy et al., 2003;Coelli, 1996). Following Fare et al (1994) and Jaffry et al. (2007) the output orientation is more suitable given the objectives of developing economies' banking industry. This approach measures how much units' outputs can be regularly improved given the observed levels of inputs. However, the framework of the production technology is supposed to reveal constant return to scale (CRS) and the structure of production technology with the output distance is described as follows: Equation (1) measures the output technical efficiency of bank j at time t relative to the technology at time t (Sufian and Habibullah, 2013). As technical efficiency is measured relative to the contemporaneous technology, ( , ) ≤ 1, and ( , ) = 1 indicating that bank j is on the production frontier and technically efficient, while ( , )< 1 emphasising that bank is below the frontier and is technically inefficient, requiring action to fix such matter.
In order to run the MPI, distance functions have to be allocated regarding the two different periods (times).
Fare et al (1994) defines the MPI as: Electronic copy available at: https://ssrn.com/abstract=3309649 or + ( + , + , , ) = + ( + , + ) A measure of productivity change (TFPCH) is provided by equation (3); it grows or declines due to changes in technical efficiency (Eff) and technological change (Tech) and decomposition of the technical efficiency index are pure technical efficiency ( ) and scale efficiency( ).
To avoid selecting an arbitrary benchmark, two continuous MPIs are integrated into a single index by calculating the geometric mean and then multiplicatively decomposed into subindices measuring changes in technical efficiency and technology as follows (Fare et al., 1989;Fare et al., 1994).
, and (4) Equation (4) is an index of technical efficiency change between period t and t+1, as it measures whether bank j witnessed improvements or went away from best practices for the time period. The value of ∆ , +1 can be greater than, equal to, or less than based on whether the relative efficiency of bank j improved, unchanged, or decreased respectively through the period. Whereas, ∆ ℎ , +1 in equation (5) reflects technological change as it provides the geometric mean of two ratios. A value of ∆ ℎ , +1 greater than 1 indicates progress, equal to 1 refers that there is no change and less than one indicates decline or regress in technology for the period t and t+1. , Therefore, changes in productivity are the decomposition of changes in efficiency and technology as , +1 can be greater than, equal to, or less than 1 reflecting progress, no change, regress in total factor productivity between periods t and t+1.
Regarding the technical efficiency, The ∆ , +1 index is disaggregated into its mutually As Electronic copy available at: https://ssrn.com/abstract=3309649 The subscripts v and c represent VRS and CRS technologies. When ∆ , +1 > 1 implies that there is an increase in pure technical efficiency, while ∆ , +1 < 1 indicates decline or regress and ∆ , +1 = 1 shows that there is no change in pure technical efficiency. Likewise, ∆ , +1 > 1 suggests that there is an increase in the most efficient scale and hence, scale efficiency is improving, whereas ∆ , +1 < 1 indicates a decline and ∆ , +1 = 1 implies no change in scale efficiency.

The determinants of banking productivity
In the second stage of this methodology, this paper uses the Tobit regression model to test Hypothesis 2 to obtain the impacts of bank-specific variables and macroeconomic variables and year and country effects (representing the financial liberalisation) on the total factor productivity of MENA banks. Since total factor productivity indices obtained from the DEA, in this case, they are truncated data for which ordinary least squares (OLS) is not suitable for such purpose. In this model, the dependent variable is the total factor productivity (TFPCH) which is a measure of productivity growth calculated by MPI using DEA (Fare et al., 1989;Fare et al., 1994;Rezitis, 2004). Table 2 summarises the definitions of variables used in investigating the determinants of productivity of MENA commercial banks.

Table 2 Definitions of variables used in assessing productivity
Variable Descriptive TFPCH Total factor productivity change index derived from Malmquist Index (MPI) SIZE Log of total assets represents bank size including earning assets + cash and due from banks + foreclosed real estate + fixed assets +goodwill. EQAS Equity to total assets. This variable to measure capital adequacy, computed as equity to total assets. High capital-asset ratios indicate low leverage and therefore lower risks COST The cost to income ratio. LOANAST This is a measure of risk computed as loans to average total assets. Higher ratios imply lower liquidity and more interest revenues because of higher risks. However, loans also have higher operational costs due to monitoring, originating and serving of loans.

INF
The real inflation rate GDPGR The real gross domestic product (GDP) growth CR The Herfindahl-Hirschman Index (HHI In respect to independent variables, Bank size (SIZE) is measured by using average total assets (Smirlock, 1985;Lloyd-Williams, 1994;Demirguc-Kunt and Huizinga, 1999;Samad, 2008;Tu and Yuan Chen, 2000;Dietrich and Wanzenried, 2014). The bank size variable takes into consideration differences derived by size in terms of economies of scale. Compared to smaller banks, larger banks are expected to experience economies of scale by having superior investments opportunities. The cost to income ratio (COST) is defined as operating costs (staff salaries, property costs, administrative expenditures etc.) over total generated revenues. It measures the overheads or expenses required to run a bank or to measure the effect of efficiency in managing expenditures (Pasiouras and Kosmidou, 2007;Kosimdou, 2008;Obamuyi, 2013;Dietrich and Wanzenried, 2014). The key component of which wages and administrative expenses as percentage of income can provide information on variation of bank expenditures over banking system. The equity to total assets (EQAS) variable is included in this model as a measure of capital strength. In line with previous studies (including Demirguc-Kunt and Huizinga, 1999;Pasiouras and Kosmidou, 2007;Kosimdou, 2008;Samad, 2008;Obamuyi, 2014;Dietrich and Wanzenried, 2014), Sufian (2013) suggest that well-capitalised bank is vital in developing and developed economies because it provides additional strength to withstand financial crisis and increased safety for depositors during unstable economic conditions. Moreover, the bank with a higher capital ratio is regarded relatively safer and protected in the event of loss or liquidation, and lower risk increases a bank's creditworthiness and therefore reduces costs of funding. Loans over assets ratio (LOANAST) representing the loans portfolio and the coefficient of this variable is expected to be positive as more loans indicate more productivity.
In respect to macroeconomic and market structure variables, the Herfindahl-Hirschman Index (HHI) is used to denote market concentration (CR), which is employed by policy makers and regulators in the banking sector computing by squaring the market share of each bank competing in a defined geographic banking industry and then summing the squares. The second measure to estimate the impact of market structure is the market share (MS) as it takes the market share of each bank in a market measured by total assets. Given the fact that the economic condition of a country may affects the performance of banking industry, following (Demirguc-Kunt and Huizinga, 1999;Pasiouras and Kosmidou, 2007;Kosimdou, 2008;Samad, 2008;Obamuyi, 2014;Dietrich and Wanzenried, 2014) the gross domestic product growth (GDPGR) is considered the most commonly macroeconomic factor to measure the total economic activity within an economy. GDPGR is expected to have a positive effect on bank's productivity. Another important factor that affects both costs and revenues of banks is inflation (INF). The relationship between inflation and bank productivity depends on whether inflation is anticipated or unanticipated. With respect to anticipated inflation, banks are able to adjust interest rates which will lead to increased revenue than costs (Demirguc-Kunt and Huizinga, 1999). Finally, to investigate whether ownership influences bank profitability, in this study I categorise a bank as state-owned bank if the government owns more than 50%. Foreign ownership is also regarded to exhibit an impact on productivity. A bank is considered to be a foreign bank if foreigners own more than 50% of its shares are owned by foreign investors. Earlier studies (Demirguc-Kunt and Huizinga, 1999;Pasiouras and Kosmidou, 2007;Micooo et al. 2007;Kosimdou, 2008;Samad, 2008;Obamuyi, 2014;Dietrich and Wanzenried, 2014). Finally, YEAR and COUN are used to capture the year and country effects respectively.
The following regression model (10) is estimated, the total factor productivity change (TFPCH) is used as dependent variable. Whilst, for independent variables, Size represents the bank size measured by log of total assets, equity to assets ratio (EQAS) is employed to capture capital adequacy and loans to assets ratio (LOANAST) to account for bank-specific risk. The cost to income ratio (COST) is the cost management efficiency in MENA banks (Pasiouras and Kosmidou, 2007;Kosimdou, 2008;Obamuyi, 2014;Dietrich and Wanzenried, 2014).
The gross domestic product (GDPGR) and the inflation rate (INF) are employed to control differences in macroeconomic; whilst FORE, STAT are used in the model to investigate the effect of foreign ownership, state control as dummy variables, whilst, COUN and YEAR are employed to capture the country effects and year effects respectively on MENA banks productivity.

DEA and the Malmquist Index (MPI) Findings
In this section, DEA is employed to calculate the distance functions of (MPI) using DEAP version. This software utilises panel data to compute indices of total factor productivity change, technological change, technical efficiency change, pure technical efficiency change and scale efficiency change.

Empirical results of Malmquist Index in Saudi Arabia
Estimates of MPI for commercial banks in Saudi Arabia are presented in table 3. Given that the total factor productivity is a multiplicative compound of technical efficiency change and technological change, productivity improvements are determined by comparing the values of efficiency change and technological change indices. In other words, productivity improvement is due to results of efficiency gains (loss), technological progress (decline) or both.  Table 3 shows the annual means of MPI over 14 years. There is an increase in total factor productivity for banks operating in Saudi Arabia of 8.1%, suggesting that total factor productivity of the commercial banks in Saudi Arabia have regressed during the years 2000, 2007, 2008, 2011 and 2012, while years of 2001, 2002, 2003, 2004, 2005, 2006, 2009 and 2010 witnessed productivity growth. The 8.1% increase in total factor productivity in Saudi commercial banks could be due to the 8.1% increase in technological progress .The results for this country are expected and support the hypothesis (1). It can be observed from the table in respect to technological change the overall rise in total factor productivity was mainly determined by technological progress rather technical efficiency. I can link this development with the fact that the Saudi banking sector has a share of foreign presence in its commercial banks which helps them to move much faster in terms of investing in technological innovations. Table 4 presents the results of total factor productivity for commercial banks operating in the United Arab Emirates (UAE). The annual means of MPI are presented in table 4. The results indicate that over the examined period, there was annual mean increase in total factor productivity for all banks of 4.4%. In more details, the commercial banks in the UAE exhibited a rise in total factor productivity during the years 2001, 2002, 2003, 2004, 2008, 2011 and 2012, telling that the overall improvement in total factor productivity was attributed to technological progress (upward shift of frontier) of 3.9%. The major increase in total factor productivity was in 2011 of 68% mainly attributed to technological progress of 65.4% and then increase in technical efficiency of 1.7%. In respect to annual results of technological change, results show that over the study period there was a technological progress for years 2001, 2002, 2003, 2008, 2011 and 2012. On the other hand, it can be seen that there is no massive development in efficiency (technical efficiency) as the annual mean of efficiency was 4% which confirms that technological change has contributed more in improving productivity of UAE banks rather than technical efficiency.

Table 4 Annual Means of Malmquist Indices of Commercial banks in United Arab Emirates
Year Overall, the total productivity changes for commercial banks in the UAE appear to be determined by technological progress rather than technical efficiency. Such results indicate that most of the banks operating in the UAE tend to be investing more in retail banking technologies such as ATMs, internet banking, smart cards and wireless services confirming the first hypothesis. Since the advent of the global business, it has become obvious for the banking industry in the UAE to pursue technological progress, leading domestic banks to benefit from ranges of technological items brought by the foreign investment.

Empirical results of Malmquist Index in Oman
Results presented in table 5 show the MPI estimates for annual means of banks in Oman. The annual total factor productivity exhibited an increase of 14.3%, which seems to suggest that commercial banks in Oman witnessed improvements in total factor productivity during years 2000, 2001, 2002, 2003, 2004, 2007, 2008 and 2012. The 14.3% increase in total factor productivity of the Omani commercial banking sector is related as shown to 14.2% increase in technological change (technological progress), which reflects that commercial banks have benefited from expending their capital investments on technology. Such figures reflect that Omani banks have invested more in better capabilities systems and equipment due to the financial liberalisation and financial reforms that have been taken place in that period.

Table 5 Annual Means of Malmquist Indices of Commercial banks in Oman
Year On the other hand, the annual efficiency of commercial banks seems to be increased only by 1% as the highest increase of technical efficiency was 33% in 2009. The improvement of efficiency as it is shown in the table is mainly due to scale efficiency rather than pure technical efficiency, so that, Omani banks are operating with efficient level of outputs (optimal scale of efficiency). Regional developments are highlighted as one of the objectives within Omani 'Vision of 2020' (Tarawenh, 2006). That vision seeking diversification of Oman's economy, aims for a greater role of private sector, particularly the banking sector.
Based on the above discussion, the productivity growth in the banking sector in Oman showed that for the examined period, improvement in total factor productivity was 14.3%. Overall, the rise in total factor productivity was essentially determined by technological change rather than technical efficiency which in line with hypothesis (1). This result implies that most banks tended to increase their spending on banking technology items such as ATMs, smart cards, internet banking, etc. to improve cost efficiency.

Empirical results of Malmquist Index in Qatar
In the State of Qatar, table 6 presents MPI annual results for Qatari banks. Results indicate that commercial banks in Qatar received on average total factor productivity of 7% increase for the study period. The overall improvement in productivity was associated with technological progress of 7% and decline in technical efficiency (efficiency) of 1%. As previously discussed, the efficiency can be decomposed into pure technical efficiency and scale efficiency. During the period of study, pure technical efficiency remained unchanged while the decrease in technical efficiency was merely the product of scale efficiency deterioration of 1%, suggesting that commercial banks in Qatar have been operating at the operation optimal scale, but also that they have not been efficient in controlling their operating costs.
In more details, total factor productivity revealed a rise during years 2001,2002,2006,2008 and 2011 as the most massive improvement was in 2011 of 88.1% as a result of increasing in technological change (technological progress) of 86% for the same year supporting hypothesis (1). Paying attention to technological change, commercial banks in Qatar exhibited progress in years 2001, 2002, 2006, 2008 and 2011. According to Qatar Central Bank, a plausible reason for the increase in the technological change is associated with continuing investment in technological banking items and human resources development with strong capital base provides a solid foundation for further growth.
Overall, the empirical findings confirm that the total factor productivity growth, which originates solely from the technological change, is higher in a number of years and is attributed to the rapid adoption of new technology by Qatari banks. The deterioration in efficiency during the study period can be attributed to the presence of adjustments costs related to investing in new technology.  Table 7 revealed an annual increase of 10%, the overall improvement in productivity for Bahraini banks over the study period was attributed to an average increase in technological change of 10% while efficiency (technical efficiency) remained unchanged. Findings in this table suggest that Bahraini banks recorded a rise of productivity during years 2000, 2002, 2003, 2006 and 2007, and decrease during 2001, 2004, 2005. Regarding technical efficiency (efficiency), it can be noted that the annual mean of efficiency is unchanged but yearly it observed a decline during 2000, 2004, 2005 and showed an increase in 2001, 2006, and 2007. The decline in efficiency in some years is related mainly to a decrease in scale efficiency than to pure technical efficiency. The results suggest that Bahraini banks have been operating at optimal scale of operation. The overall increase in total factor productivity was attributed solely to technological progress and hypothesis 1 is supported. A possible reason for the increase in technological change may be related to investment decisions in sophisticated systems and equipment as well as banking financial liberalisation which have been taking place for that period. Bahraini banking has witnessed a presence of foreign banks as those banks have a crucial effect on domestic banks to move further in investing in technological innovations.

Empirical results of Malmquist Index in Bahrain
The Bahraini banking sector has also shown a reaction to the changes in the global nature of business which leads banks in Bahrain to pursue technological progress and invest more in retail banking technologies such as internet banking (E-Banking), ATMs and wireless banking. Therefore, in order to maintain market share in the market, commercial banks should invest more in new technology to avoid losing customers because of some factors attributed to easy access and competitive prices. Table 8 presents the annual means of MPI results for banks operating in the State of Kuwait. Findings tell that over the study period, there was a mean annual growth in total factor productivity for Kuwaiti banks of 50%. Such improvement in productivity during that period was attributed solely to an increase in technological change (progress) of 50%. The overall increase in factor productivity is driven by technological progress only, according to results year by year; it seems to indicate that Kuwaiti banks experienced a growth in productivity during years 2000, 2001, 2002, 2003, 2007, 2010, 2011 and 2012. The highest improvement in productivity was in 2011 of 40%, due to the highest technological change of 44%. A probable reason for the progress in technology during those years could be associated with financial reforms and deregulation that have occurred in the last two decades which led to growth of assets of commercial banks with more expertise to invest in sophisticated system and equipment. Such a growth continued rapidly as a result of adoption of new information technology by Kuwaiti banking sector, and hence they moved faster for technological innovation. Such findings are in line with hypothesis (1).

Empirical results of Malmquist Index in Egypt
As depicted in table 9, the MPI findings show that Egyptian banks have observed a decrease in total factor productivity by 7.8%. The results seem to suggest that Egyptian commercial revealed a total factor productivity decline during the years 2000, 2001, 2002, 2004, 2006, 2007, 2008, 2009, 2011 and 2012 which in turns does not support the hypothesis (1), whereas, total productivity was recorded to increase during the years 2003, 2005 and 2010. Over the examined period, technical efficiency and technological change exposed an annual mean decline of 1.4% and 6.6% respectively. On the other hand, it can be seen that technical efficiency (efficiency) of commercial banks operating in Egypt were observed to have a decrease by 1.4%. According to the results, this decline was attributed to regress in pure technical efficiency and scale efficiency as there is no significant difference between both of them (pure technical efficiency and scale efficiency). The results suggesting that banks operating in Egypt have not been operating at optimal scale of operations and have been managerially inefficient in managing their operating expenditures. Although the financial services and banking reforms have a critical element of economic reform presented in 1990s in Egypt, results for this case implies that regulators, the central bank, and bank managers in Egypt should pay more attention to technical efficiency (efficiency) of banks as results confirm that banks have not been operating with efficient level of operation and inefficient in controlling their operating costs. Furthermore, regulators should go further to encourage banks investing in technology such as ATMs, interment banking and wireless in order to reach cost efficiency and to avoid losing customers. It can be argued that banks will be able to expand their productivity by investing more in advance technology.

Empirical results of Malmquist Index in Jordan
Estimation of annual mean of MPI results for Jordanian banks are presented in table 10. Results for the study period imply that commercial banks in Jordan have revealed an average increase of total factor productivity of 5.6% and hence supporting hypothesis 1. Findings seem to suggest that the overall improvement in total factor productivity was attributed to a technological change of 5.1% rather than technical efficiency of 0.5%, which reflects that the increased in technological change and its positive effects on productivity are due to deregulation and the liberalisation of the banking system and the General Agreement on Trade in Services (GATS) signed by the Jordon government in 2001. According to the literature (Al-Fayoumi, 2009) highlighted that the banking sector in Jordan is introducing new banking products and investing more in technological banking items and human resources. Additionally, foreign banks that operate in Jordan have has put a pressure on domestic banks to move much faster in investing in technological innovation to achieve the cost efficiency and gaining a competitive edge.
It can be noted from the table that the technical efficiency (efficiency) of banks in Jordan has increased by 0.5%, but this is attributed mainly to scale efficiency rather than pure technical efficiency. The results of this case indicate that banks in Jordan have been operating at optimal scale or with efficient level of outputs, but that they did not do so efficiently by managing their operating costs. Generally, due to the (GATS), it is imperative for Jordanian banks to pursue technological progress as more domestic banks offer the products introduced by foreign investors. Results also suggest that commercial banks have been moving toward investing in retail banking technologies such as ATMs, software, internet banking.

Empirical results of Malmquist Index in Lebanon
The annual means of MPI of Lebanese banks are presented in table 11. It shows that over the study period, the annual mean of total factor productivity for commercial banks operating in Lebanon have experienced regressed by 2% and therefore hypothesis (1) is not supported. These findings suggest that commercial banks in Lebanon regressed in total productivity during the years 2000, 2001, 2002, 2003, 2004, 2008 and 2011 while total productivity increased during the years 2005, 2006, 2007, 2009, 2010, 2012. According to the table, the growth in productivity can be attributed to technological change rather than technical efficiency (efficiency). These results are in line with Turk-Ariss (2008) as he highlighted that the government of Lebanon plans to join the World Trade Organisation (WTO) which justifies the need for an efficient and more competitive banking sector. As a result of such matter, Lebanon has made a heavy investment in information technology and technological innovations. However, the efficiency of commercial banks in Lebanon have decreased for annual mean of banks by 7%, which can be attributed to a decrease in the annual mean of pure technical efficiency of 3% and scale efficiency of 5%. Results suggest that commercial banks operating in Lebanon have not been working at the optimal scale of operation and controlling their operating expenses. Overall, the previous results show that the monetary authority, regulators and bank managers should reconsider their policies to improve efficiency of banks in terms of operating at optimal outputs and controlling their operating expenses efficiently.

Empirical results of Malmquist Index in Morocco
As introduced in table 12, the MPI results show that commercial banks operating in Morocco have, on average, revealed a growth in total factor productivity of 4% to confirm hypothesis (1). Such growth resulted from the increase in technological change of 4% and annual technical efficiency remained unchanged. This result suggests that banks Moroccan banks in have not operated at constant return to scale and they have not efficiently selected their inputs combinations.
However, technological change which is the main source of productivity growth of 4% improved during the years 2001, 2002, 2004, 2006, 2007, 2008, 2009 and 2011. The plausible reason for the increase in technological change could be related to a mega-merger programme which has led to large banks with good abilities to improve their productivity through capital investment in information technology. Furthermore, Morocco witnessed a comprehensive financial reform, particularly in the banking industry before 1990s, as the state owns merely 29% of banking assets (Ben Naceur, 2011).

Empirical results of Malmquist Index in Tunisia
The results of the annual means of MPI for commercial banks operating in Tunisia are reported in table 13. According to the results, Tunisian banks have an average record growth in total factor productivity of 1.4% and this result supports hypothesis (1). The overall improvement for the whole period was attributed to a rise in technological change of 0.8% and technical efficiency increase of 0.6%. Findings suggest that commercial banks in Tunisia observed total productivity growth over the years 2004, 2005, 2006, 2008, 2010 and 2011, whereas productivity is shown to have decreased during the years 2007, 2009 and 2012. It can be noted from the table that efficiency of banks has obtained an annual mean growth of 0.6%, mostly related to scale efficiency rather than to pure technical efficiency, indicating that commercial banks operating in Tunisia have been operating at optimal level of outputs but have not been efficient in managing their operating expenses.

Total Factor Productivity Tobit Regression Findings
In this stage, this paper uses the Tobit regression model to obtain the impacts of bankspecific, macroeconomic variables and year and countries effects on the total factor productivity of MENA banks. Since total factor productivity results obtained from the DEA, in this case, they are truncated data for which ordinary least squares (OLS) is not appropriate for such purpose.
Starting by the variable measuring bank risk (LOANAST) net loans to total assets, it exhibits a negative relationship and statistically significant at 10% level in tables 14 and 15. Also recorded negative and significantly relationship at 5% level in table 5.16. Such result is in line with Sufian and Habibullah (2013) who indicates a positive relationship between the productivity of banks and the level of liquidity held by banks. The ratio is considered high if banks are less liquid (lending more), so that results emphasise the more productive banks are the more likely they are to be more liquid. One reason which could explain why banks with less liquidity are less productive is related to monitoring cost increases for higher amounts of loans in terms of originated, serviced and monitored as suggested by Ben Naceur (2011). The negative impact of liquidity risk on bank productivity is explained by the fact that less liquid banks are more involved in financing risky loans which in turns lead to have nonperforming loans. Yes Dependent variable TFPCH= Total factor productivity change. Independent variables: SIZE =average total assets, EQAS= equity to average total assets, LOANAST= net loans to average total assets, NIE/TA =non-interest expense to average total assets CR= the Herfindahl-Hirschman Index (HHI), MS= Market share, GDPGR= real gross domestic growth, INF = real inflation rate, FORE= dummy for foreign ownership, STATE= dummy for state ownership. *significant at the 10% level **significant at the 5% level ***significant at 1% level Regarding the effect of banks size (size) measured by total assets, the empirical results show a negative coefficient of the size variable but insignificant. Such a finding is supported by Kosmidou (2008) and Sufian and Habibullah (2013) as they pointed out that growing in size could have a negative impact on the performance of a bank, as a result of more bureaucratic procedures. Concerning the level of capital adequacy (EQAS), results are mixed as some regressions exhibited a negative relationship and were statistically insignificant with total factor productivity of banks. Based on these results there is no evidence for the relationship between productivity of MENA banks and their capital adequate.
In respect to the impact of market concentration and market share on banking productivity, results show a positive but insignificant sign: the empirical findings of this variable do not support the structure-conduct-performance (SCP) hypothesis. On the other hand, using market share as the market power indicator exhibited a negative value but also insignificant, so that the results do not provide strong evidence that banks with higher market share or more productive. Turning to the effect of a bank's cost management on its productivity; it is interesting to observe that the coefficients of COST revealed a positive and significant impact on banks' total productivity at the 5% and 10% level in all regressions. Findings indicate that an increase (decrease) in costs improves (reduce) productivity of banks operating in MENA economies. It seems to suggest that expenses preference behaviour in this case leads banks to be increase their productivity. A reasonable justification for this result is that higher remunerations package would be required by highly qualified and professional management as well as employee's incentives programmes which can encourage or promote a bank's employees to work efficiently and produce quantifiable products and services, and therefore a positive relationship with productivity of banks in line with (Sathye, 2003). Yes Dependent variable TFPCH= Total factor productivity change. Independent variables: SIZE =average total assets, EQAS= equity to average total assets, LOANAST= net loans to average total assets, NIE/TA =non-interest expense to average total assets CR= the Herfindahl-Hirschman Index (HHI), MS= Market share, GDPGR= real gross domestic growth, INF = real inflation rate, FORE= dummy for foreign ownership, STATE= dummy for state ownership *significant at the 10% level **significant at the 5% level ***significant at 1% level Referring to the impact of macroeconomic indicators on the productivity of banks, results of (GDPGR) growth in gross domestic product are not consistent with the theory that banks tend to be more productive in a growing economy, since all results in all regressions are negative and insignificant as well. In respect to inflation (INF), results show that the inflation variable has a negative and significant relationship with the productivity of commercial banks. It could be explained that banks during the study period, have not anticipated a level of inflation which allowed them to adjust interests rates and consequently improve productivity. Also, during times of growth, banks are more encouraged to provide loans which in turns require additional costs to monitor and screen. Therefore higher proportions of loans are observed highest operational costs. Finally, to account market changes, financial reforms and technological changes on productivity of MENA banks, year and country effects have been introduced in the models for this study. All years in all regressions are compared to the basic year 1999, and all countries are compared to Bahrain. The general findings from the table below indicate that total factor productivity of commercial banks are mostly higher than those of the basic year implying that productivity has improved during the examined period of this study. These improvements could be linked to the fact that banking sector in the MENA countries has witnessed major financial reforms programmes during this period which led MENA banks to be engaged in allocating more capital investments in technology including ATMs ,internet banking services, and increasing the availability of debit and credit cards.

Conclusion
This paper provides empirical evidence on the drivers of total factor productivity in MENA banking economies and how it is influenced by bank-specific variables, market structure and macroeconomic variables. Firstly, I employed the non-parametric frontier DEA using MPI to estimate the total factor productivity for commercial banks operating in eleven MENA countries over the period 1999-2012. Using the MPI allows the determination not only the total factor productivity of banks, but also the frontier growth (technological change) and the optimal resource utilisation (technical efficiency change). The obtained MPI results suggest that commercial banks operating in the Gulf countries, namely Saudi Arabia, Kuwait, Qatar, Oman, Bahrain and United Arab Emirates have exhibited productivity progress mainly due to the technological progress rather than efficiency change confirming what past studies found in the MENA and other emerging economies (Ramanathan, 2006;Gattoufi et al. 2009;Al Hammadi (2013); Krishnasamy et al. 2013). Such findings imply that banks in these countries are moving toward spending huge investments on retail banking technologies such as ATMs, internet banking, wireless banking and smart cards to achieve cost efficiency. However, other MENA countries have also experienced productivity progress, due to technological changes and technical efficiency for Jordanian commercial banks and from technological changes alone for Moroccan commercial banks. The source of increase in the technical efficiency of the Jordanian banks was generally related to scale rather than pure technical efficiency, suggesting that Jordanian banks have been operating at relatively optimal scale of operations. On the other hand, banks in Egypt and Lebanon have recorded decline in their productivity as a result of regress in technological changes and efficiency. This might highlighted the relative ineffectiveness of social and economic policies which were not well directed in those countries, and thus appropriate actions would be necessary to reverse such a trend.
Secondly, the Tobit model is employed to investigate whether the financial liberalisation which has taken place in that period and other variables such as size of banks, risk, market structure and macroeconomic variables have had an effect on the productivity of MENA banks. The empirical results from the Tobit model suggest that the financial liberalisation and enable MENA banks to be more productive for the period 1999-2012. Within the context of MENA banking sector, previous studies in emerging markets (Isik and Hassan, 2003;Krishnasamy et al. 2003;Howcroft and Ataullah, 2006;Ben Naceur et al. 2009;Das and Kumbhakar, 2012) have suggested that performance of all types of banks exhibited major improvements after liberalisation, and such improvements were attributed to improvements of management practices, more investment in technology and positive impact of mergers and acquisitions on the banking industry.
In respect to bank-specific factors, size of commercial banks and their liquidity measures seem to exert a regressive impact on total factor productivity, meaning growing in size could create a negative impact on the performance of a bank, as a result of more bureaucratic procedures. Regarding the bank risk variable (loanast) also has a negative association with total factor productivity of MENA banks and a plausible reason for such a matter can be found in the increased costs of monitoring required by a higher proportion of loans (Ben Naceur and Omran, 2011). However, the impact of a bank's cost suggests that expenses behaviour leads banks to be more productive confirming that highly qualified and professional managers may require a higher remuneration package as well as employees incentives programmes which can encourage or promote a bank's employees to produce good quality of banking services Finally, the empirical findings of this paper are interesting from the policy makers' perception. Despite improvements in productivity in banks operating in Gulf countries, Morocco, Jordan and Tunisia, I suggest that further reforms may be desired in order to obtain the optimal utilisation of capacities as well as making the greatest use of resources. Overall, different mix of policy should be adopted depending on the characteristics of the banking system on the examined countries.