Review of Middle East Economics and Finance

Transcription

1 Review of Middle East Economics and Finance Volume 5, Number Article 4 Bank Efficiency and Foreign Ownership in the Lebanese Banking Sector Ali Awdeh, Lebanese International University Chawki El Moussawi, Lebanese University Recommended Citation: Awdeh, Ali and El Moussawi, Chawki (2009) "Bank Efficiency and Foreign Ownership in the Lebanese Banking Sector," Review of Middle East Economics and Finance: Vol. 5: No. 2, Article 4. DOI: / Available at: Berkeley Electronic Press. All rights reserved.

2 Bank Efficiency and Foreign Ownership in the Lebanese Banking Sector Ali Awdeh and Chawki El Moussawi Abstract We have compared the efficiency of banks with majority domestic ownership, banks with majority foreign ownership, and the subsidiaries of foreign banks operating in the Lebanese market between 1996 and We have implemented the DEA methodology to calculate the yearly scores for technical, allocative, and cost efficiencies for the three groups of banks. Moreover, we have extended our study to reveal the factors that shape bank efficiency, by proposing several micro and macroeconomic variables and testing their correlation with this efficiency. Firstly, our results did not prove significant differences between the efficiency of the three groups of banks. However, the evolution of the efficiency scores over the period under study shows an improvement in the performance of banks with majority foreign ownership, but some deterioration in the performance of banks with majority domestic ownership and the subsidiaries of foreign banks. Secondly, and with respect to the factors affecting bank efficiency, we found that the employed variables have different impacts according to groups, which may confirm that bank efficiency is differently determined, according to bank ownership. KEYWORDS: foreign banks, bank efficiency, emerging markets Author Notes: Corresponding author: Ali Awdeh.

3 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership 1 INTRODUCTION Banks expand their operations internationally by establishing subsidiaries and branches or by taking over established foreign banks. This internationalisation of banking systems has been encouraged by the liberalisation of international financial markets. The increased presence of foreign banks in host markets raises two issues: (1) the effect of this presence on domestic banking systems and (2) the competition inequalities and differences in performance between foreign and domestic banks. The entrance of foreign banks may improve the quality and availability of financial services in the host market by increasing competition, enabling better application of modern banking skills and technologies, encouraging the development of bank supervision and legal framework, and enhancing a country s access to international capital markets. A considerable body of literature (mainly on developed markets) has focused on foreign bank performance, its determinants and how it differs from domestic bank performance. The different structure and characteristics of foreign and domestic banks on one hand, and the different influence of external factors on these banks on the other, could lead to performance differences between the two categories. Empirical studies on foreign and domestic bank performance in developing markets have captured less attention than those in developed markets. Thus, it is interesting to study the performance of foreign banks operating in developing countries, how this performance differs from that of domestic ones, in addition to knowing why the groups of banks perform differently. All this may help in clarifying the necessary conditions for successful entry of multinational banks into emerging foreign markets, and may assist in developing a regulatory framework for foreign bank entry and expansion. The Lebanese banking market has a substantial foreign bank presence. Foreign banks represent a significant portion of the number of banks operating in the Lebanese market, and thus provide a good case study for understanding how and why bank performance varies according to ownership. The Lebanese banking sector has a long history of foreign participation. After the First World War, and until the independence in 1943, the banking system in Lebanon was dominated by the presence of foreign banks, and a great proportion of funds was deposited with foreign banks. Those banks focused on financing foreign trade, and left domestic financing to domestic banks. Starting with the independence era, and continuing with the establishment of the central bank (Banque du Liban) in 1964, the banking system in Lebanon has witnessed continuous progress and rising prosperity. This development has promoted the establishment of more domestic banks, which has lead to a reduction of differences between them and foreign banks. The latter lost their overwhelming monopoly power and the domestic banks became important players in the market. Published by Berkeley Electronic Press,

4 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art. 4 The Lebanese banking system remains very open to the entry of foreign banks and the acquisition of domestic banks by foreign banks/investors is permitted. Foreign banks can receive deposits from the public and perform credit operations, fiduciary operations and portfolio management on the behalf of other parties. Moreover, foreign banks can carry out brokerage activities, without intermediation, on the floor of the Beirut Stock Exchange. 1 Currently, there are 54 commercial banks operating in the Lebanese market. 2 These banks can be classified into three categories: (i) banks with majority domestic control (33 banks), (ii) banks with majority foreign control (11 banks), and (iii) foreign banks (10 banks). Moreover, there are representative offices of another 16 foreign banks. The remaining of the paper is as follows: in section 2 we shed light on the literature regarding the performance of foreign banks in the developed and the developing countries and its determinants. Section 3 explains the empirical methodology implemented in the paper. We illustrate the data in section 4. The empirical results of the study are presented in section 5. 2 FOREIGN BANK EFFICIENCY AND PROFITABILITY The structural and organisational differences between foreign (FBs) and domestic banks (DBs) may have implications on differences in cost structures and scale and scope economies. These differences result from different management strategies, differences in the markets they serve (i.e. retail, corporate, etc), knowledge of the local market, international synergies, and regulation. A parent company of a FB can provide its subsidiary with access to international markets, international diversification, and access to low-cost funds. These factors represent benefits for the foreign subsidiaries with cost advantages. However, a lack of familiarity of the FB s parent company with the nature and features of the host markets, and difficulties with adaptation of home country strategy into the local markets may obstruct its development. According to the market studied, there are two types of evidence on FBs efficiency in comparison with DBs. Studies that had focused on developed markets found that FBs do not have efficiency superiority: they have less or the same efficiency as DBs. For instance, DeYoung and Nolle (1996) argued that FBs operating in the U.S. were significantly less efficient than DBs. Although there 1 On the other hand, foreign banks are prohibited from: (1) carrying out industrial or commercial activities or any activity other than banking, (2) participating, in any form, in industrial, commercial or agricultural institutions or any other institutions except within the limits of the private funds, (3) carrying out, on its behalf, any operation on derivatives, and (4) reducing the capital assigned for its investment, or buying back any part of it. 2 Source: Association of Banks in Lebanon, DOI: /

5 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership was little difference between the two sets of banks in terms of output efficiency, FBs had a distinct disadvantage in terms of input efficiency, primarily driven by excess expenditures on purchased funds. Elyasiani and Mehdian (1997) studied the production efficiency of DBs and FBs in the U.S. and found that FBs and DBs were operating under different technologies; however, FBs were as efficient as DBs. Elyasiani and Rezvanian (2002) investigated the efficiency difference between the cost structures and the cost economy characteristics of FBs and DBs in the U.S. Their results indicated that although the cost structure of the two categories of banks are different, scale and scope economy measures derived for the two groups are not significant. Kosmidou et al. (2006) found that FBs in the UK operate with lower return on equity, net interest revenue-to-total earning assets, loans to customer and short-term funding as compared to DBs. Their results support the home advantage hypothesis under which domestic institutions are generally more efficient than foreign institutions. With respect to less developed countries, the results were different and sometimes conflicting. Sturm and Williams (2004) compared the efficiency of foreign owned-banks operating in Australia with Australian DBs. They found that FBs were more input efficient than DBs, mainly due to superior scale efficiency; however, this did not result in superior profitability. Bongini et al. (2001, 2002) used foreign ownership data to predict financial distress and closure of financial institutions in the East Asian financial crisis. They found that foreign-owned institutions tend to be more efficient and less risky than domestic institutions due to their corporate governance and operational structures, or sometimes because they were more diversified. Unite and Sullivan (2003) studied the effect of foreign ownership on Philippines banks and found that the entry of FBs led to a decline in operating expenses and an increase in domestic banks risk. They claim that due to the FBs entry, DBs were forced to take on less creditworthy customers due to the increased competition. Detragiache and Gupta (2006) examined the experience of Malaysia during the crisis of 1997 and provided evidence on the performance of foreign banks during extreme financial fragility. They found that FBs (particularly those with operations not concentrated in Asia) had relatively low non-performing loans, and their profitability and capitalisation even improved during the crisis. In addition, foreign bank lending and deposits contracted less for DBs. Sensarma (2006) find that both efficiency and productivity of FBs have been consistently lower than those of DBs. The author argues that the reasons for this could be that FBs incur huge expenditures in paying high salaries to employees and the use of technology. Lensink et al. (2008) also found that on average FBs are less efficient than DBs. Besides, when the home country governance is a higher quality, this reduces FBs efficiency and if the institutional distance between the host and the home country governance becomes smaller, FBs inefficiency decreases. Sturm and Williams (2008) came with the conclusion Published by Berkeley Electronic Press,

6 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art. 4 that banks from more sophisticated nations (as measured by home nation GDP per capita) are able to operate more efficiently in the host nation. They find that FBs are on average less efficient than the domestic incumbents due to increasing expenditure on inputs to produce the same level of outputs, which result in lower profits and efficiency. Finally, Van Horne (2007) claimed that developing country banks have a competitive advantage dealing with countries with weak institutional climate. He added that developing country FBs realise higher interest margin (but are less profitable) than FBs from high-income countries. On the determinants of FBs profitability, Williams (1998,a) found a negative effect of the size of the parent bank on FBs profitability, whereas home net interest margin and home GDP show some positive effect on it. Williams (1996, 1998,b) found that FBs profitability in Australia is a positive function of the Australian net interest margin and fees. Molyneux and Seth (1998) modelled the determinants of FBs profitability in the U.S. They found that capital strength, assets composition, commercial and industrial loan growth, and U.S. GDP growth were important factors in determining FBs ROA. Minh and Tripe (2002) investigated the determinants of FBs profitability in New Zealand and focused on the effect of their home factors. They found that: firstly, the size and the profitability of the parent bank have a positive effect on the profitability of its subsidiary operating in New Zealand. Secondly, the specific experience of operating in New Zealand is the most important factor determining a foreign bank s profitability. Third, the relative capital scarcity in New Zealand compared to the home countries (as reflected in interest rate differentials) affects negatively bank s ROA. 3 METHODOLOGY 3.1 Model Specification In order to detect the impact of ownership upon the efficiency of banks operating in Lebanon and the identification of the variables affecting this efficiency, we proceed in two stages. Firstly, we construct of a nonparametric frontier by using the DEA method to calculate the various components of the productive efficiency. The second stage consists of identifying (some of) the factors behind the differences in efficiency. The performance evaluation of a production unit is based on the measures of efficiency and productivity in order to make comparisons over time for the same economic entity and to compare among similar entities. The first indicators of performance were based on measuring the average productivity, which analyzes the output of factors of production. We distinguish usually between total productivity of the factors that relates the sum of the outputs to the sum of the DOI: /

7 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership inputs, and the partial productivity, which relates a quantity of outputs to a quantity of inputs. In general, the first method is preferable over the second because the partial productivity is based on an (implicit) assumption that the quantity of outputs is the result of only one input without any other factors entering in the production. The concept of total productivity could be used in a bank because of the difficulty of aggregation of outputs and inputs. In a pure economic view, the measurement of the performance exceeds the simple analysis of ratios and refers to the concept of production frontier. This frontier, which represents the best practice, is based on the comparison of practices observed in an industry and allows studying the productive efficiency of the producers by assuming that they have the same objectives and the same constraints. The economic theory does not only give an intuitive interpretation of the concept of efficiency by interpreting the production function as the relation between inputs and outputs, but also as the frontier of the production alternatives which maximizes the level of outputs when the factors of production are given. Thus, the term of efficiency indicates the success with which the inputs are used. This explains the measure in which a Decision Making Unit (DMU) uses inputs in order to produce the maximum potential outputs. The efficiency is then a relative concept and measuring it requires a standard performance with which the success of the DMU is evaluated (Forsund and Hjalmarsson, 1974). The distinction between DMUs which are or are not on the frontier allows determining the efficient and inefficient DMUs. A DMU that is on the production frontier is regarded as efficient. Conversely, a DMU that is below the frontier is regarded as inefficient. The first studies on the productive efficiency were carried out by Koopmans (1951), Debreu (1951) and Farrel (1957). Koopmans (1951) defined a feasible input-output vector to be technically efficient if it is technologically impossible to increase any output and/or to reduce any input without simultaneously reducing at least one other output and/or increasing at least one other input. By using this definition, he was able to prove that an input-output vector is efficient if, and only if, it possesses a positive normal to the production possibilities set, and he provided an economic interpretation to this normal as a set of shadow prices. Debreu (1951) provided a measure of the degree of technical efficiency with his coefficient of resource utilization. This coefficient is computed as the maximum proportionate reduction in all inputs consistent with continued production of existing outputs, and from this, Debreu obtained measures of the magnitude and the cost of technical inefficiency. But, Farrel (1957) was the first to propose a decomposition of the efficiency in two components: technical efficiency and allocative efficiency. The technical efficiency, inspired by the coefficient of resource utilization proposed by Debreu, exists when a given output is produced with the smallest quantities of available factors. It measures the way in which the Published by Berkeley Electronic Press,

8 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art. 4 DMU chooses the quantities of inputs used in the production process when the proportions of usage of the production factors are given. Allocative efficiency measures the ability of DMU to combine its inputs in optimal proportions taking into account their prices, and of the budget allocated to acquire them. The combination of technical efficiency and allocative efficiency determines the total efficiency. Two approaches are proposed by the literature to measure the productive efficiency: the nonparametric approach and the parametric approach. The first approach uses the linear programming and precisely, the data envelopment analysis (DEA) to construct the frontier, but imposes the assumption of convexity of all production alternatives. 3 It is not necessary to impose a priori a particular specification of profit, cost, and production functions, which constitutes an advantage for this approach. On the other hand, the scores of efficiency obtained by the nonparametric approach are sensitive to the errors, which may affect the results. The second approach, the econometric approach, takes account of the errors in the data, by introducing two types of risks in the specification of profit and cost production functions. The first risk is the usual statistical error, whereas the second error is an asymmetrical risk which represents the inefficiency. On the other hand, it is necessary to impose a particular specification for the parametric frontier as well as a particular distribution for the error terms which constitutes a weakness in this approach. As we can see, the weakness of one approach represents and advantage for the other, and vice versa. 3.2 Constructing the Efficiency Frontier Using DEA Charnes, Cooper and Rhodos (1978) generalized the approach of Farrel (1957) on the context of multi-outputs and multi-inputs by supposing the case of a technology with constant returns-to-scale. They constructed a mathematical program of optimization whose solution provides a measure of the relative efficiency of DMUs similarly to Farrel. The word relative means that the DMU is compared to all DMUs operating in a similar industry where the inputs and the outputs are homogeneous. More precisely, the DEA method measures the efficiency of a DMU by calculating the relative variation separating the point representing the values of inputs and outputs observed compared to a hypothetical point on the production frontier. We can in this manner measure the degree of efficiency of each DMU compared to this frontier which determines the best 3 Deprins et al. (1984) proposed the method called Free Disposal Hull (FDH) which has the advantage of relaxing the assumption of convexity of the frontier but its disadvantage is that the production scale is not regarded as source of inefficiency. Moreover, this method does not allow studying the various forms of productive efficiency particularly allocative efficiency and cost efficiency. DOI: /

9 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership practice. The DEA method is distinguished from an analysis of central tendency (like the regression technique) to the fact that the frontier is given from the point of view of best practice. The role of this frontier is to envelop the productive activities in such a way that the set of production alternatives is convex. The DEA method uses these data to construct an efficiency frontier, which joins the best practice i.e. the DMUs that cannot reduce their usage of production factors taking into account their production volume. The inefficiency of other DMUs is measured by their distance from this frontier. The larger the distance is, the more the DMU is inefficient. The results obtained by this deviation can be used in a procedure of benchmarking the banking sector. The model Charnes, Cooper and Rhodos (CCR), which we use, is based on the maximization of the weighted sum of outputs divided by the weighted sum of inputs (or the minimization of the weighted sum of inputs divided by the weighted sum of outputs). It is a question of maximizing the score of efficiency for each DMU while respecting the constraint of an efficiency score less than or equal to the unit for all observed DMUs, and that all weights are positive. This model shows that the efficiency of the DMU will be obtained as a ratio between outputs and inputs under the constraint that this ratio is equal to or lower than 1 for all DMUs. The optimization of the model is non-convex, nonlinear and provides an infinite number of solutions. However, it can be transformed into a problem of dual standard linear programming (see Charnes and Cooper, 1962). The evaluated DMU is represented by its vector of output and by its vector of inputs. With the resolution of the program, the latter is compared with a linear combination of efficient DMUs, those constituting a reference. The CCR programs measure exclusively the total technical efficiency while assuming constant returns-to-scale. Banker et al. (1984) extended the measure of efficiency and developed a model with variable returns-to-scale by introducing an additional constraint: the constraint of convexity that guarantees that the evaluated DMU is compared only with DMUs of similar size. By solving this model for all DMUs, the DEA method determines a production frontier which permits to evaluate the efficiency of each DMU by generating an efficiency score ranging between 0 and 1. A score equals to the unit (lower than 1) means efficiency (inefficiency) of the evaluated DMU. The DEA model accounts only for physical quantities of inputs and physical quantities of outputs and does not allow calculating allocative efficiency of different DMUs. The measure of allocative efficiency with the capacity of the producer to choose the best combination of inputs with given outputs or conversely to its capacity to choose the best combination of outputs with given inputs and this in the light of the prices at the time of production. From a theoretical point of view the basic model of this theory is based on the rational behavior of the producer. The minimization assumption of costs allows to Published by Berkeley Electronic Press,

10 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art. 4 incorporate the traditional behavioral assumptions of the microeconomic theory and thus to consider the technical efficiency, allocative efficiency, and cost efficiency. The assumption of profit maximization allows establishing the dual relations between the profit function and the distance functions of input and output. It measures the distance between the current profit of a DMU and the maximum potential profit given the prices of inputs and outputs. Since, the data concerning various outputs are not all available, we decided in the calculation of allocative efficiency to adopt the minimization assumption of costs of banks operating in Lebanon. A last methodological point regarding the orientation of efficiency measurements is presented in the following statement. The passage from the assumption of constant returns-to-scale to the assumption of variable returns-toscale cannot be achieved without affecting the measures of efficiency; two efficiency measurements are always possible according to whether one chooses output oriented or input oriented. The measurement of efficiency output-oriented gives the maximum increase in output which can be obtained with an unchanged level of inputs. The measurement of efficiency input-oriented gives the maximum proportional reduction of the vector of input allowing producing the same level of output. With the assumption of constant returns, the two measurements of efficiency are equivalent. However, with the assumption of variable returns-toscale, the results are different according to the selected orientation. Our study on the performance of domestic and foreign banks is based on the assumption of variable returns-to-scale and the input-orientation. The choice of this model is justified by the fact that the assumption of variable returns-to-scale is certainly the most suitable assumption in the case of banks; the input-orientation has the advantage of insisting on the reduction of quantity of inputs used in the production process in order to increase the efficiency, which corresponds to the behavior of the majority of banks in deregulation and competition. 4 DATA 4.1 Source of Data The empirical analysis of this study uses a sample of unbalanced panel data of 46 commercial banks operating in Lebanon (28 banks with majority domestic ownership, 9 banks with majority foreign ownership, and 9 foreign banks). Some of the banks operating in Lebanon were excluded from our sample due to missing data for some variables. Information about banks is extracted from the international banking database Bankscope of BVD-IBCA that provides individual time series (i.e. by bank). The sample is formed from commercial banks only in order to have a homogeneous sample. We use annual accounting data (balance DOI: /

11 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership sheet and income statement) for banks for the period We have dropped the period before 1996 from our analysis since it was a transition period where banks in Lebanon were implementing modernization programs and adopting Basel requirements for capital adequacy. Other information is extracted from BilanBanques. The macroeconomic data are taken from the International Financial Statistics, IMF. We have decomposed our sample into three categories: the first category includes banks with majority domestic ownership (MDO), the second includes banks with majority foreign ownership (MFO), and the third category includes the subsidiaries of foreign banks (FBs). 4.2 Variables Specification To measure the outputs and the inputs, we use the intermediation approach suggested by Sealey and Lindley (1977), which assumes that the bank collects deposits and transforms them into loans by using the labor factor and the capital factor. The alternative approach is the production approach, where the bank is assumed to use the labor and capital factors to produce loans and deposits. Certain studies showed that the chosen approach for the definition of bank inputs and outputs has an impact on the levels of the scores of efficiency, but does not imply major change in the classifications of the scores of efficiency (e.g. Wheelock and Wilson, 1995; Berger et al., 1997). The bank production is measured by three outputs: earning assets, other earning assets and off-balance sheet activities. The three factors of production selected are the bank deposits, the physical capital (measured by the fixed assets) and labor (measured by the number of bank employees). To estimate a cost frontier, we use the price of physical capital (measured by the general expenditures of exploitation divided by fixed assets), the price of the labor factor (measured by the ratio of staff costs divided by the number of employees), and the price of deposits (measured by the interest expenses divided by total deposits). Regarding the factors affecting the efficiency scores, the literature shows that the productive efficiency of banks can be influenced by two types of factors: (i) macroeconomic factors reflecting the economic environment, and (ii) microeconomic factors reflecting the strategy of the bank. Following Berger (1993), Mester (1993), Allen and Rai (1996) and Mester (1996), we will test the influence of certain internal and external factors on bank efficiency which, do not have to be considered when estimating the production frontier. Thus, we implement the following variables. To proxy for the effect of economic development on bank performance, we implement the growth rate of the Gross Domestic Product (GDP) and the inflation rate measured by the growth rate of the Consumer Price Index (INF). To detect the relationship between bank Published by Berkeley Electronic Press,

12 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art. 4 capitalization and efficiency, we implement the equity-to-asset ratio (CAP). To test the effect of scale and scope economies on bank performance, we exploit the logarithm of total assets (LnAssest). The effect of credit risk on efficiency is studied by implementing the ratio of provisions for doubtful loans-to-total assets (RISK). The liquidity of the bank measured by liquid assets divided by total assets (LIQ) is used to test the effect of reserve requirements. Finally, we utilize the return on asset ratio (ROA) to perceive the interaction between bank profitability and efficiency. 4.3 Descriptive Statistics Table 1 presents the descriptive statistics of the values of outputs, the prices of inputs and the other variables used, for the entire period under study. The purpose is to test the homogeneity of the data. We notice that the data for the three sub-samples of banks are generally homogeneous since the coefficients of variation for the three categories of banks are close and contained in a narrow interval. By looking at the variables, we notice that the MDO are the largest banks, followed by the MFO, and the FBs are the smallest in terms of total assets. This is also the case of deposits, and the off-balance sheet. The deposit price is found to be equal for MDO and MFO, and lower for FBs. This may suggest that foreign banks enjoy some pricing power over other banks, or depositors accept lower interest rate from the subsidiaries of foreign banks as they consider them less risky than other banks. On the other hand, the MDO have the lower physical capital price, followed by FBs, whereas the MFO have the highest costs. Finally, we notice that FBs have the highest labour price, followed by MFO, and the lowest is for the MDO. This may be due to the fact that foreign banks pay higher wages than the other banks operating in Lebanon. 5 EMPIRICAL RESULTS In order to achieve the two objectives of this paper knowing the effect of ownership on the efficiency of banks, and the identification of factors affecting this efficiency we proceed in two stages. In the first stage, we construct a nonparametric efficiency frontier by using the DEA method to compute the different components of productive efficiency. The ownership structure (domestic vs. foreign) is presented in the analysis of productive performance of banks proceeding in separate estimations for each category of banks. In the second stage we will try to explain these performance differences. DOI: /

13 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership Table 1 Descriptive Statistics for Banks operating in Lebanon between 1996 and 2005 (LBP millions) MDO MFO FBs Total Asset Mean 2,045, ,129, , SD 3,080, ,346, , CV Total Earning Asset Mean 1,960, , , SD 2,965, ,124, , CV Deposit Mean 1,665, , , SD 2,533, , , CV Other Earning Asset Mean 1,465, , , SD 2,352, , , CV Staff expenses Mean 15, , , SD 20, , , CV Off-balance sheet Mean 302, , , SD 896, , , CV Fixed assets Mean 55, , , SD 88, , , CV General expenses Mean 26, , , SD 34, , , CV Interest paid Mean 111, , , SD 151, , , CV Deposit price (%) Mean SD CV Physical capital price (%) Mean SD CV Labour price (%) Mean SD CV Source: The international banking database Bankscope, BVD-IBCA. Published by Berkeley Electronic Press,

14 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art Comparing the Efficiency of Domestic and Foreign Banks Table 2 provides all scores of productive efficiency of banks operating in Lebanon. These results are obtained through the estimation of a nonparametric frontier (DEA) input-oriented and under the assumption of variable returns-toscale. The total efficiency is split into technical efficiency and allocative efficiency. The application of the DEA method reveals a high total efficiency for all groups of banks (MDO, MFO, and FBs). Besides, the (grand) average technical efficiency is very close between the three categories of banks. The average technical efficiency of the MDO is 96% and 98% for the MFO and FBs, with a slight difference of 2%, in favor of the MFO and FBs. Additionally, this component of productive efficiency lays within the interval [94% ; 97%] for MDO, [96% ; 99%] for MFO, and [96% ; 100%] for FBs. Our results indicate that an increase from 4% to 6% of the outputs would have allowed the group of banks under study to reduce their technical inefficiency. The dispersion of technical efficiency scores is very close between the groups since the coefficient of variation during the period remains contained in narrow intervals. Within each group, we notice that the performance variation is higher among the MFO and FBs especially after the year 2002, whereas it is stable for MDO. This suggests a stable performance of Lebanese banks with majority domestic ownership, and a divergence in foreign bank efficiency. Another source of the studied cost inefficiency is the allocative inefficiency. The (grand) average allocative efficiency during the period under study is equal to 87%, 88% and 90% for the MDO, the MFO and the FBs respectively, which shows a slight difference of 2% to 3% in favor of the FBs. These results indicate that the mismanagement in the allocation of resources and the quantity of employed factors results in an allocative inefficiency of about 13% for the MDO, 12% for the MFO, and of 10% for the FBs. In of other terms, the allocative inefficiency observed in the Lebanese bank sector is explained by the fact that certain banks use the production factors in wrong proportions, which does not allow them to minimize their production costs. It may be interesting to compare the observed results over the entire period under study. The allocative efficiency of the MDO fall slightly between 1996 and In fact, the average value of allocative efficiency of the MDO dropped from 89% in 1996 to 87% in On the other hand, the results show that the average allocative efficiency of the MFO and the FBs witnessed an improvement during the same period, since the score of allocative efficiency increased from 82% to 87% for the MFO and from 95% to 98% for the FBs. The dispersion of average allocative efficiency is lower for the MDO than that for the MFO and FBs since the coefficient of variation remains in DOI: /

15 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership the interval [9% ; 13%] for the MDO versus [11% ; 23] and [3% ; 20%] for the MFO and the FBs respectively. The (grand) average cost efficiency reaches 83% for the MDO, 86% for the MFO and 88% for the FBs. By looking at the evolution of the cost efficiency, we note that it has decreased, and then increased for all groups of banks under study. But, overall, the annual average cost efficiency remains contained in a narrow interval: [83% ; 86%] for the MDO, [82% ; 89%] for the MFO and [84%-95%] for the FBs. The dispersion of cost efficiency of MDO is lower than that of MFO and FBs, since the coefficient of variation ranges between 10% and 14% for the MDO and between 11% and 23% for the MFO and the FBs. This increase of the dispersion of the cost efficiency scores of FBs is due to the fact that the least performing banks in terms of cost belong to the MFO and FBs groups present a handicap by more than 46% compared to the best practices. They can therefore reduce their costs by 46% by adopting the techniques of those with the highest performance. So generally, we find a slight difference in efficiency between the three categories of banks operating in Lebanon. These findings match those of Ariss (2008) who also compared the efficiency of domestic and foreign banks operating in Lebanon. Ariss (2008) found that the average cost inefficiency of Lebanese banks appears to be small (around 12%), and that domestic banks are as efficient as foreign banks. On the other hand, and regarding the evolution of efficiency scores, Djoundourian and Raad (2008) evaluated the efficiency of the Lebanese banking sector and found that bank inefficiency decreased with time. We have found that the efficiency of MFO has enhanced between 1996 and 2005, but a deterioration in MDO allocative efficiency and FBs technical efficiency. 5.2 Factors Affecting the Level of Productive Efficiency of Banks Operating in Lebanon After calculating the efficiency scores (technical, allocative, and cost) in the previous section, we will try to identify some macro- and microeconomic factors that shape those efficiency scores for the three groups of banks. The variables used are those cited in section 4.2. Table 3 presents the results obtained from the estimates of the ordinary least squares method. A general overview of Table 3 shows that the models have high explanatory power shown by their high coefficients of determination (adjusted R-square). Furthermore, the F-statistics for all models show their overall significance. Thus, we may claim that the variables included in our estimations explain reasonably the efficiency scores. Turning to the individual variables, we observe the following. Published by Berkeley Electronic Press,

16 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art. 4 Regarding the coefficients of the explanatory variables, we notice firstly a positive correlation between the capital ratio (CAP) and the efficiency scores of the three categories of banks. This relation is statistically significant in seven out of the nine estimations. This positive relationship between CAP and bank efficiency suggests that the higher the bank capital, the higher its efficiency in exploiting its available funds. Another possible explanation for this positive relation is that efficient banks are those with higher solvency. Table 2 Technical efficiency, allocative efficiency, and cost efficiency ( ) MDO MFO FBs TE AE CE TE AE CE TE AE CE Mean Min Max SD CV Mean Min Max SD CV Mean Min Max SD CV Mean Min Max SD CV Mean Min Max SD CV Mean Min Max SD CV Mean Min DOI: /

17 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership 2002 Max SD CV Mean Min Max SD CV Mean Min Max SD CV Mean Min Max SD CV Grand average The results also show that the correlation between efficiency and credit risk (RISK) is statistically different from zero in seven estimations. The coefficients are positive in MFO's estimations and negative in MDO's and FBs' estimations. These signs do not allow determining the effect of this variable on bank efficiency. The positive correlation between risk and efficiency suggests that the most efficient banks are those with the highest level of risk. Thus, an increase in risk would be compensated with a higher loan pricing, which allows realizing a higher interest margin. However, this contradicts the findings of Berger and DeYoung (1997) and Dietsch (1996) who found an inverse relationship between the level of risk and efficiency. The negative relationship found for MDO and FBs could be explained by the fact that banks that limit their risk have to improve their solvency, profitability, and thus their productive efficiency. In other words, the decrease in economic activity, which is often accompanied by an increase in the probability of default of borrowers, affects banks and increases the amount of doubtful loans. This situation involves an increase in total cost for the banks (because of the need to increase loan monitoring for example), an increase in the provisions for loan losses, a reduction in capital, and therefore, a reduction of the productive performance of banks. The empirical results show that, with exception of FBs, the liquidity is negatively correlated with the efficiency of banks and this is shown by the negative coefficients of MDO and MFO. An increase in liquidity in a bank could be interpreted as a sign of inefficiency in the exploitation of funds. Another possible explanation is that smaller banks maintain usually high liquidity ratios. Published by Berkeley Electronic Press,

18 Review of Middle East Economics and Finance, Vol. 5, No. 2 [2009], Art. 4 Those banks (as we will see later with the effect of the size) are less efficient than larger banks. Thus, smaller banks have high liquidity and suffer low efficiency. The relationship between bank profitability (measured by ROA) and the efficiency shows also opposing results. For MDO, ROA has a negative effect on technical efficiency and positive effect on both allocative and cost efficiencies. Whereas for MFO, this factor has the opposite sign: a positive effect on technical efficiency, but negative effect on both allocative and cost efficiencies. Regarding FBs, it has a significant (negative) effect on allocative efficiency only. The signs of these coefficients do not allow determining a definite effect of this variable on the productive efficiency of banks. It might be concluded that there is no direct relationship between efficiency and profitability. A bank may be efficient, but realizes low profitability, possibly due to large expenditures on staff, IT, real estate, or else. On the other hand, an inefficient bank may enjoy high profitability due to high margins charged by that bank. Two hypotheses can be proposed here. First is the X-inefficiency hypothesis (Leibenstein, 1966), which is based on the managerial theory. According to this hypothesis, inefficiency reveals problems of organization. Thus, the problems of organization could explain why certain banks (with high profitability) solve less than other banks the problems of reorganization. In addition, banks which would have reserve of profits or of market capacity would not be prompted as much as the others, to carry out productivity efforts and to control their production costs. The second hypothesis is inspired by the theory of imperfect competition. In particular, if competition is important, banks that have good control over costs can choose (or are forced to choose) an aggressive and costly marketing policy which does not allow them to realize high profitability. In other words, certain banks, which carry out efforts of productivity, choose their inputs effectively and control their costs better, and seem to have difficulties in increasing their margins, and thus increasing their profits. With respect to the size of banks, the results in Table 3 show a positive and significant effect of MDO bank size and the three efficiency scores. On the other hand, this variable does not affect any of MFO scores, and negatively FBs technical efficiency only. The positive correlation between the efficiency of MDO and bank size confirms the presence of increasing return-to-scale for MDO and shows that the larger the domestic bank size, the more the possibility of managing its production factors efficiently, and therefore, improving its productive performance. This result must be interpreted cautiously since, to our best knowledge, no other study has been carried out to reveal empirically the existence of scale economies in the Lebanese bank sector. DOI: /

19 Awdeh and El Moussawi: Bank Efficiency and Foreign Ownership Table 3 The determinants of allocative, cost, and total efficiency MDO MFO FBs C CAP RISK LIQ ROA LnAssets INF GDP AE CE TE AE CE TE AE CE TE *** *** *** 1.233*** 1.329*** 1.094*** 0.985*** 1.010*** 1.003* (5.95) (4.54) (12.01) (5.79) (6.12) (12.36) (11.57) (10.73) (7.98) 0.198** 0.283** 0.154*** 0.187** ** 0.359* 0.251** (1.89) (2.01) (3.16) (1.75) (-0.82) (2.09) (1.68) (1.78) (-0.15) *** ** *** 0.006*** 0.006*** ** * (-3.02) (-3.60) (-2.65) (5.80) (6.11) (0.20) (-1.53) (-1.81) (-2.15) ** ** ** *** *** *** (-1.86) (-1.93) (-1.71) (-7.71) (-10.10) (-2.79) (0.74) (0.54) (0.06) 0.887** 0.664** ** ** ** 1.946*** ** (1.69) (1.81) (-1.89) (-2.49) (-1.70) (2.50) (-1.87) (-1.39) (-0.67) 1.142** 1.566** 1.329*** * (1.76) (1.83) (2.88) (0.43) (0.33) (0.13) (-1.16) (-1.25) (-0.46) (-0.39) (-1.19) (0.99) (-0.73) (-0.68) (-0.25) (-1.48) (-1.11) (0.01) 0.275** 0.385** ** 2.582** (1.77) (1.973) (-1.50) (0.83) (0.89) (-0.53) (1.71) (1.81) (0.79) 2 Adjusted- R F-statistic Probability (F-statistic) Notes : t-satitics in parantheses. *** Significantly different from zero at the 1% level. **Significantly different from zero at the 5% level. * Significantly different from zero at the 10% level Published by Berkeley Electronic Press,