Is Bigger Better for Egyptian Banks? An Efficiency Analysis of the Egyptian Banks during a Period of Reform 2000–2006

1 Introduction

The banking sector plays a pivotal role in national economies; consequently continuous assessment of the performance of banks is essential for the soundness of the economy. For Egypt, which has undergone frequent banking sector reforms to cope with internal and global challenges, this assessment is crucial. In 2004, the country embarked on a comprehensive financial sector reform program (FSRP) to build a strong and competitive financial sector. The five main pillars of this reform included: enhancing the supervisory role of the Central Bank of Egypt (CBE), financial and managerial restructuring of state-owned banks, addressing the non-performing loans (NPLs) problem, privatizing and divesting state-owned banks’ stakes in private and joint-venture banks, and consolidating the banking sector.

By 2009, many of the reform goals had been achieved. For example, the government had successfully divested the shares of public banks in 13 joint-venture banks. Furthermore, the banking sector witnessed a very large number of mergers and acquisitions: almost a quarter of all independent banks operating in Egypt have disappeared between the years 2000 and 2008. The reforms have certainly changed the face of the country’s financial services industry.

Even though Egypt is the second largest economy in Africa and has one of the oldest financial systems in the Middle East since the nineteenth century, the number of studies examining the efficiency of Egyptian banks is small. Similar to most developing countries, lack of firm-level data has prevented a significant stream of research output. Only two studies have assessed the performance of Egyptian banks from an efficiency perspective, namely El-Shazly (2011) and Poshakwale and Qian (2011). In addition, Reda (2012), as well as Badreldin and Kalhoefer (2009), focused on the effect of mergers on bank performance using Data Envelopment Analysis (DEA) and financial analysis, respectively. On the macroeconomic level, Abu-Bader and Abu-Qarn (2008) found evidence for a bidirectional causality between the efficiency of the financial sector and growth in Egypt, concluding that further efficiency increases of the financial sector are necessary to stimulate growth via the saving-investment relation.

This study follows the stream of banking literature – for an overview see Amel et al. (2004) – that attempts to gauge banks’ performance through assessing their efficiency. The study was conducted using a panel data set of 34 commercial banks representing about 75% of the Egyptian banking industry. The panel includes observations for the period from 2000 to 2006, therefore including the big banking reform of 2004. ^[1] Our data set includes a variety of variables that go beyond the scope of the few earlier studies on the Egyptian banking sector. Specifically, three inputs and five outputs of the 34 banks are used. The study estimates banks’ efficiency utilizing a stochastic frontier cost function of the translog type. This allowed for the estimation of individual bank cost and scale efficiency scores, as well as technological progress.

The research questions addressed are: Do Egyptian banks suffer cost inefficiency? Do banks witness technological progress? Do big or small banks enjoy cost advantages? Is there a relationship between ownership and cost efficiency? Is the merger policy suitable for Egyptian banks? To answer those questions, individual bank X-efficiency scores and optimal bank size are estimated. In addition, the existence of economies of scale is tested. Further, Ordinary Least Squares (OLS) regressions analyzed the effect of different variables including the effect of mergers on bank cost efficiency. The last issue is especially important for weighing the potential benefits of consolidation against possible market power effects or the systemic risk of “too big to fail” institutions.

The rest of the paper is organized as follows: Section 2 presents the theoretical background. The methodology describing the model and estimation method is in Section 3. The data is explained in Section 4, and Section 5 reports the empirical results. Finally, Section 6 concludes.

2 Theoretical Background

This paper assesses the efficiency of the Egyptian banks from a triad of perspectives: cost efficiency, ^[2] scale efficiency, and technical change. In a first step, we estimate the cost frontier of the Egyptian banking industry, which is defined as the best-practice costs necessary to operate at given output levels and input prices. An important characteristic of cost frontiers is their ability to allow for firm-level inefficiency. The deviation of actual cost, corrected by pure randomness, from its minimum possible level as defined by the cost frontier captures the cost inefficiency. This inefficiency may be both allocative (wrong input mix) and productive (waste of inputs) inefficiency. Another important characteristic of cost functions is their duality ^[3] to production functions, which allows the determination of economies of scale and the measurement of technical change. Finally, in contrast to production functions, it is easy to consider the multi-output character of banks within a cost function framework.

The empirical implementation of the cost efficiency is based on the pioneering work of Aigner, Lovell, and Schmidt (1977) and Meeusen and Van den Broeck (1977), which led to the development of several empirical approaches to assess the efficiency of firms. Broadly these approaches can be divided into: non-structural (accounting) approaches and structural (frontier) approaches (Hughes and Mester 2008). Structural approaches can be subdivided into parametric and non-parametric approaches. The parametric approaches encompass three techniques: the Stochastic Frontier Analysis (SFA), the Thick Frontier Analysis (TFA), and the Distribution Free Approach (DFA); while the non-parametric approaches encompass the DEA and the Free Disposable Hull (Berger and Humphrey 1992). The two most common approaches for investigating overall bank efficiency are the non-parametric DEA and the parametric SFA (Coelli et al. 2005, 161). Considerable debate remains concerning the choice between DEA and SFA due to the fact that each approach possesses its own merits and disadvantages (Ahmed 2008).

This paper opted for the SFA technique to estimate the efficiency of Egyptian banks. The choice of SFA over DEA is justified on the grounds that the SFA technique is more adequate to the nature of bank studies. The SFA allows for the separation of random noise from inefficiency. This characteristic of the SFA is particularly important in the case of bank data where measurement error is a main problem (Kasman 2002, 3). Also, this technique generates firm-specific efficiency estimates which are essential for bank managers’ operational efficiency (Sarsour and Daoud 2015, 60). In contrast, the usage of the DEA would not allow for separating between inefficiency and data problems, perhaps seriously overstating the degree and fluctuation of the inefficiency term.

The SFA approach is a parametric technique that necessitates the specification of a functional form. In the literature, a broad range of functional forms – from the Cobb-Douglas to the transcendental-logarithmic (translog) to the Fourier flexible forms (e.g., Altunbas and Chakravarty (2001);Vennet (2002)) and Box-Cox transformations (Pulley and Humphrey 1993) – can be found. However, the cost of using very flexible forms is the higher number of parameters to be estimated, consequently, a rich data set is required which is often rarely available. Actually, many studies ^[4] concur that the translog specification is the most appropriate functional form and well-suited to characterize the banking sector characteristics. With about 140 observations available, we follow these predecessors and apply the translog functional form.

3 Methodology

4 Description of the Data

The quantitative analysis is based on a panel data set of 34 Egyptian banks covering the period 2000–2006. ^[8] According to the CBE, there were 58 banks active in 2000 and 41 banks active in 2006, but some banks were not included in our sample because of missing input prices or output quantities. Also, some of the banks in the CBE list are specialized banks or branches of foreign banks. These were not included because, specialized banks are technology outliers not representative of Egypt’s banking industry, whereas branches of foreign banks were excluded due to lack of sufficient data. Accordingly, the analysis is applied to commercial banks (public and private and joint-venture), which covers about 76% of the banking sector in Egypt. To be more specific, the sample consists of the four public commercial banks: BanqueMisr (BM), Banque du Caire (BC), National Bank of Egypt (NBE), and Bank of Alexandria (BoA) plus 30 private and joint-venture commercial banks listed in Table 5. The commercial banks used for our estimations do not only cover the largest share of Egyptian banks but were also heavily involved in the merger activities that were responsible for the disappearance of almost a quarter of the banks between 2000 and 2006 (see Tables 6 and 7 for details). Summarizing, our data allowed for the construction of an unbalanced ^[9] panel data set consisting of 147 observations used for estimating the cost frontier in eq. [1].

The measurement of output and productivity is not straightforward for a bank due to the multi-output, intangible nature of banks’ products, coupled with the difficulty to account for the quality of bank services (Heffernan 2005, 473). Even though several studies have emerged to examine bank productivity, they failed to provide a harmonized definition of inputs and outputs. To overcome this definitional problem, authors adopt either the production or the intermediation approach (Mlima and Hjalmarrson, 2002, 13). Given that the study focuses on the efficiency of the banking industry as banks are financial intermediaries, the study therefore follows the “intermediation approach” ^[10] that treats loans as outputs and deposits as inputs. Description of the data is illustrated in Table 1 where the dependent variable of the cost function [1] is total costs (C) defined as operating costs measured as follows:

The three input prices w_i and five outputs y_i representing the multi-output character of banks are defined respectively as follows:

1. w₁ labor price, measured as administration and general expenses divided over the number of employees at the annual average.

2. w₂ capital price, generated as the opportunity costs of capital divided by the equity of the bank. Opportunity costs are measured as the ratio of interest earned from loans to the volume of loans, reflecting the forgone interest in the use of equity.

3. w₃ deposit price, calculated by dividing interest paid by the amount of the deposits.

4. y₁ loans to other banks;

5. y₂ loans to non-banks, i.e., loans to firms and households;

6. y₃ securities, e.g., corporate or government bonds;

7. y₄ fees and commissions for banking services;

8. y₅ assets not covered by y₁ to y₄, e.g., reserves held at the central bank. It is determined as a residual by taking the difference between TA and outputs one to four.

All the data concerning output quantities were extracted from the banks’ balance sheets except for fees and commissions which were extracted from the banks’ income statements. The GDP deflator (World Bank 2009) is used to deflate all nominal values, i.e., outputs, TA, and input prices.

Our data, as described in Table 1, provide means, minimum, maximum, and standard deviation of each variable. Significant discrepancies between Egyptian banks in terms of total costs with a minimum of EGP 38.8 mn to a maximum of around EGP 8.4 bn can be noted. This suggests disparities in the cost management efficiency between banks. Also, there are significant size differences among Egyptian banks, for example, loans to non-banks range from EGP 122 mn to around EGP 50 bn. In fees and commissions, small banks earn less than 1% compared to the largest player (NBE). Similarly, input prices are also quite heterogeneous, where the interest rates paid on deposits range from 1% to 16%. Moreover, Egyptian banks’ employability varies vastly from a few hundred to thousands of employees. Likewise, the number of bank branches varies from four to 450 branches.

5 Empirical Results

This section presents our estimations of the cost frontier [1], generated by numerically maximizing the log-likelihood function [3]. We used Gauss software for all estimations and calculations of X-EFF, RSCE, and η. Table 2 provides the parameter estimates showing that most are statistically significant. The use of a stochastic frontier is clearly supported by highly significant σ2- and γ -parameters, rejecting the hypothesis that only random error occurs. Likelihood ratio tests rejected the null hypothesis of constant economies of scale technology in the Egyptian banking industry.

In line with similar bank efficiency studies, the degree of cost inefficiency turned out to be economically relevant for Egyptian banks. According to our estimations, the mean cost efficiency of all Egyptian banks is at 88.2%, implying an average cost reduction potential of 12%. For an international comparison, Table 8 displays the results of some recent studies on cost efficiency of banks in other developing countries. According to these estimates, the mean efficiency of South African banks is at 92%, making them the closest to the Egyptian banks. Estimates for the other countries are significantly below these figures, indicating a higher variation in cost efficiency because all banks are ranked relative to the best-practice technology. Note that cost savings could be realized with the given technology, outputs, input prices, and the current network of branches.

The derived X-efficiency, RSCE, and technical change measures specific to each bank are shown in Table 3. For the bank-specific X-efficiency values, our estimations range from 0.691 to 0.998. This indicates that the least efficient bank (Alexandria Commercial and Maritime Bank, ACMB) could reduce its costs by approximately 31% relative to the actual position if it operates on the frontier, while the most efficient bank (National Bank of Egypt) is more or less operating on the frontier. As for the rest of the banks, 15 institutions are below the average efficiency score of 0.882, whereas 18 institutions are above the average.

The optimal bank size was analyzed by determining the cost-output elasticity (RSCE), a measure for economies of scale. Our results, provided in Table 3, show strong evidence of increasing economies of scale, where the mean value of RSCE is at about 0.87. That is, a proportional growth by 1% of all outputs will increase total costs on average by only 0.87%, which in turn will decrease average costs. Table 3 demonstrates that almost all banks exhibit economies of scale with the exception of two. The first bank suffering from diseconomies of scale is NBE, which is the biggest public sector commercial bank in the data set. Our estimate of RSCE for NBE is close to 1.10, implying that NBE could increase its margin by reducing its output: A reduction of all output levels by 1% would decrease total costs by 1.10%, i.e., more than proportional. Even though NBE might slightly suffer from diseconomies of scale, this institution seems to be pretty well managed because X-EFF is estimated at 0.998, indicating that there is barely any waste of resources. In short, NBE is close to the estimated optimal cost frontier, but not at the best position along the frontier (suboptimal size). The second bank with diseconomies of scale is BM, which is the second largest public bank.

In contrast, there are two public banks operating close to the optimal scale, namely Banque due Caire and Bank of Alexandria (acquired by the Italian bank Sanpaolo in 2006). The Suez Canal Bank, the National Bank for Development, and the Commercial International Bank (CIB) are somewhat below the optimal scale with RSCE estimates of 0.98, 0.96, and 0.96, respectively. At the lower end of the RSCE results, a group of mainly foreign subsidiaries could benefit the most from an aggressive growth strategy. BNP Paribas, Audi Bank, Misr America International Bank, Credit Agricole Indosuez, Cairo Barclays, Cairo Far East Bank, and Societe Arabe Internationale de Banque are all clustering around an RSCE value of 0.77. Accordingly, the majority of the Egyptian banks should operate at a much larger scale. For many Egyptian banks, the advantage of growth is even more significant than – to take an example from a developed country – for small German co-operative banks, where RSCE values of around 0.84 were found (Lang and Welzel 1999, 1017). Roughly speaking, the optimal bank size is at around EGP30 bn, which is equivalent to $5 bn.

Finally, the results on technological change η are also shown in Table 3. All relevant parameters, i.e., the time trend, the quadratic time trend, and the interaction between the time trend and input prices, are – with one exception – highly significant. Our results suggest the existence of technological recess for all banks over time, where the mean value is about 0.078 (7.8% year over year). The bank-specific interval ranges from 0.04 to 0.11. In addition, it was found that cost control of large banks over time (0.0736) is better than their smaller counterparts, where costs grow by approximately 0.08% in medium and small banks. These results emphasize the argument that failure to catch up in adopting more advanced new technologies and financial services may slow down the small banks’ performance in the longer term. Whereas such a result would be very surprising for developed countries, it is not unusual for developing countries (c.f. El-Shazly (2011) for Egypt; Kasman (2002) for Turkey). It has to be kept in mind that the Egyptian banks are incurring high costs to adapt to modern information technologies and are operating within an ever tighter regulatory framework requiring a lot of information (e.g., a continuous valuation of all assets at market prices). However, in the long run, in a time frame much larger than seven years, the results on the dynamics of technological progress may differ.

Further hypothesis testing to investigate the relationship between cost efficiency and potential explanatory variables is shown in Table 4. First, several regressions relating X-efficiency to bank size variables were conducted, all of them turning out to be significant, and all of them confirming our surprising result: The larger the institution, the higher is the degree of cost efficiency. Obviously, larger Egyptian banks are better organized than their smaller rivals, which may be a consequence of the better qualified staff in larger institutions. Small banks are thus not benefiting from the advantages of higher flexibility and lower information costs of smaller institutions. In this respect, RSCE results have shown that small banks are not operating very efficiently as they operate below the optimal scale. Other potential reasons for this interesting result may be a lower educational level of the employees, lack of the ability to attract high managerial skills, the lack of an incentive system, or a relatively weak endowment of information technology.

Regressions analyzing the relationship between cost efficiency and mergers showed that faster growing banks are more efficient than banks with a low growth rate. Bank mergers have a positive impact on the efficiency score, where mergers trigger a jump of 0.074 closer to the frontier. It is worth mentioning that a study conducted by Badreldin and Kalhoefer (2009) investigated the effect of mergers on bank performance from a financial perspective which showed mixed results concerning the merger effect. It is also found that more efficient banks tend to acquire less efficient ones in the majority of cases. In 67% of the merger cases, the acquiring bank was more efficient than the acquired one as can be seen in Figure 1. This result is in line with the majority of research concerned with efficiency differences between acquiring and acquired institutions. However, the difference in the efficiency score between the acquirer and the acquired one is not very large, signaling that merger motives other than enhancing cost efficiency might be important. Probably, NPLs, avoiding insolvency, and conforming to higher minimum capital requirements of the CBE were the main reasons behind the consolidation process.

Figure 1:

Difference in X-efficiency between acquiring and acquired bank.

Finally, the results concerning efficiency and ownership do not also follow conventional wisdom: Public banks are found to be closer to the frontier than private banks. The mean cost efficiency score of public sector banks is 0.92, while the mean cost efficiency score of private banks is 0.87. According to Pasiouras, Tanna, and Zopounidis (2007), higher cost efficiency scores of public banks relative to private banks are due to a higher willingness of the latter to incur higher costs in return for better quality and higher revenues. The previous argument is in line with El-Shazly (2011) and Poshakwale and Qian (2011), who also find state-owned banks more cost efficient than their private counterparts in Egypt.

6 Conclusion

In the past few years and with the initiation of the FSRP in 2004, the banking sector in Egypt has witnessed dramatic developments that changed the landscape of the industry. Using an unbalanced panel of 34 commercial banks covering around 75% of all banks over the periods 2000–2006, this study attempted to gauge the efficiency and other core characteristics of the Egyptian banking industry. To avoid predetermined results, a flexible multi-output translog cost function is specified and estimated using SFA. The empirical results of this setup provided detailed insights into the Egyptian banking sector which was – to some degree – unexpected.

Addressing the question of X-efficiency, our results are similar to what has been found for many other countries: Cost inefficiency turned out to be statistically and economically significant. On average, commercial banks could reduce their costs by approximately 12% without a decrease in input prices, output levels, or number of branches. Furthermore, strong evidence for economies of scale for the majority of banks is found. The optimal bank size is estimated to be at about EGP30–35 bn, which is – using the exchange rate of 2006 – equivalent to about $6 bn. The time trend, often interpreted as technological change, was surprisingly found to have a cost-increasing effect.

In order to analyze potential determinants of cost efficiency, further regression analysis was conducted. The analysis confirmed a significant positive relationship between bank size and cost efficiency. Evidence showed that the large public sector banks in Egypt tend to be more cost efficient than their private (relatively smaller) counterparts. Finally, both growth and consolidation were found to have a significant positive impact on X-efficiency, where more efficient banks tended to acquire less efficient ones.

The big picture of our empirical results is that “bigger is better,” because larger banks tend to be more cost efficient than their smaller competitors. Therefore, Egyptian banks are recommended to continue focusing on growth strategies, supplemented by mergers & acquisitions. However, bank managers and the CBE should continuously monitor the impact of consolidations on bank performance. If banks grow too big, they might suffer diseconomies of scale, or worse, banks might have increased market power in setting interest rates.

Further research should more explicitly address the reasons for this surprisingly good performance of larger public institutions in comparison to the smaller private competitors. Another interesting question is the “wrong” sign of the time trend variable: Is our result only an artifact of the relatively short observation period and the policy reforms during that time? The answer can only be found once a study is done for longer data sets that would allow the modeling of long-run technological effects.